@article{mohanty_2020,
title = {3 DOF Autonomous Control Analysis of an Quadcopter Using Artificial Neural Network},
author = {Mohanty S.; Misra A. },
booktitle = {Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough.},
year = {2020},
institution = {Defense Institute of Advanced TechnologyPuneIndia},
abstract = {The Quadcopter is an Unmanned Aerial Vehicle (UAV) which has turned out to be exceptionally mainstream among specialists in the recent past due to the advantages it offers over conventional helicopters. Quadcopter is extremely unique and interesting, however it is inherently unsteady from streamlined features perspective and aerodynamics point of view. In recent past scientists have proposed many control schemes for the stability of quadcopter, but Artificial Neural Network (ANN) systems provide us with the fusion of human intelligence, logic and reasoning. The research focuses on the use of ANN for the control plant systems whose plant dynamics are expensive to model, inaccurate or change with time and environment. In this paper, we explore the Linear Quadratic Regulator (LQR) and Sliding Mode Control (SMC) control is designed for an quadcopter with 3 Degree Of freedom (DOF) Hover model by Quancer. The main benefits of this approach are the model’s ability of adapt quickly to unmodeled aerodynamics, disturbances, component failure due to battle damage, etc. It eliminates the costs and time associated with the wind tunnel testing and generation of control derivatives for the UAV’s. },
keywords = {Quadcopter, Unmanned aerial vehicle (UAV), Artificial neural network (ANN), Linear quadratic regulator (LQR), Sliding mode control (SMC)},
language = {English},
series = {Studies in Computational Intelligence},
publisher = {Springer, Cham},
isbn = {978-3-030-38444-9},
pages = {39-57}
}

The Quadcopter is an Unmanned Aerial Vehicle (UAV) which has turned out to be exceptionally mainstream among specialists in the recent past due to the advantages it offers over conventional helicopters. Quadcopter is extremely unique and interesting, however it is inherently unsteady from streamlined features perspective and aerodynamics point of view. In recent past scientists have proposed many control schemes for the stability of quadcopter, but Artificial Neural Network (ANN) systems provide us with the fusion of human intelligence, logic and reasoning. The research focuses on the use of ANN for the control plant systems whose plant dynamics are expensive to model, inaccurate or change with time and environment. In this paper, we explore the Linear Quadratic Regulator (LQR) and Sliding Mode Control (SMC) control is designed for an quadcopter with 3 Degree Of freedom (DOF) Hover model by Quancer. The main benefits of this approach are the model’s ability of adapt quickly to unmodeled aerodynamics, disturbances, component failure due to battle damage, etc. It eliminates the costs and time associated with the wind tunnel testing and generation of control derivatives for the UAV’s.

This paper describes a novel stereo vision sensor based on deep neural networks, that can be used to produce a feedback signal for visual servoing in unmanned aerial vehicles such as drones. Two deep convolutional neural networks attached to the stereo camera in the drone are trained to detect wind turbines in images and stereo triangulation is used to calculate the distance from a wind turbine to the drone. Our experimental results show that the sensor produces data accurate enough to be used for servoing, even in the presence of noise generated when the drone is not being completely stable. Our results also show that appropriate filtering of the signals is needed and that to produce correct results, it is very important to keep the wind turbine within the field of vision of both cameras, so that both deep neural networks could detect it.

@article{wang3_2020,
title = {A dual adaptive fault-tolerant control for a quadrotor helicopter against actuator faults and model uncertainties without overestimation},
author = {Wang, B.; Yu, X.; Mu, L.; Zhang, Y.},
journal = {Aerospace Science and Technology},
year = {2020},
month = {4},
volume = {99},
institution = {Northwestern Polytechnical University, China; Beihang University, China; Xi'an University of Technology, China, Concordia University, Canada},
abstract = {This paper presents a dual adaptive fault-tolerant control strategy for a quadrotor helicopter based on adaptive sliding mode control and adaptive boundary layer. Within the proposed adaptive control strategy, both model uncertainties and actuator faults can be compensated without the knowledge of the uncertainty bounds and fault information. By virtue of the proposed adaptive control scheme, the minimum discontinuous control gain is adopted, which significantly reduces the control chattering effect. As compared to the existing adaptive sliding mode control schemes in the literature, larger actuator faults can be tolerated by employing the proposed control scheme while suppressing control chattering. Moreover, boundary layer is used to smoothen control discontinuity and further eliminate control chattering. Nevertheless, the choice of boundary layer thickness is a trade-off between system stability and tracking accuracy. By explicitly considering this fact, an adaptive boundary layer is developed and synthesized with the proposed adaptive control framework to ensure stability and tracking accuracy of the considered system. When the control parameter tends to be overestimated, the thickness of boundary layer can be appropriately adjusted to avoid control parameter overestimation. Simulation and experimental tests of a quadrotor helicopter are both conducted to validate the effectiveness of the proposed control scheme. Its advantages are demonstrated in comparison with a conventional adaptive sliding mode control scheme. },
keywords = {Actuator fault, Adaptive sliding mode control, Fault-tolerant control, Model uncertainty, Quadrotor helicopter},
language = {English},
publisher = {Elsevier Masson}
}

This paper presents a dual adaptive fault-tolerant control strategy for a quadrotor helicopter based on adaptive sliding mode control and adaptive boundary layer. Within the proposed adaptive control strategy, both model uncertainties and actuator faults can be compensated without the knowledge of the uncertainty bounds and fault information. By virtue of the proposed adaptive control scheme, the minimum discontinuous control gain is adopted, which significantly reduces the control chattering effect. As compared to the existing adaptive sliding mode control schemes in the literature, larger actuator faults can be tolerated by employing the proposed control scheme while suppressing control chattering. Moreover, boundary layer is used to smoothen control discontinuity and further eliminate control chattering. Nevertheless, the choice of boundary layer thickness is a trade-off between system stability and tracking accuracy. By explicitly considering this fact, an adaptive boundary layer is developed and synthesized with the proposed adaptive control framework to ensure stability and tracking accuracy of the considered system. When the control parameter tends to be overestimated, the thickness of boundary layer can be appropriately adjusted to avoid control parameter overestimation. Simulation and experimental tests of a quadrotor helicopter are both conducted to validate the effectiveness of the proposed control scheme. Its advantages are demonstrated in comparison with a conventional adaptive sliding mode control scheme.

@article{reyes-baez_2020,
title = {A family of virtual contraction based controllers for tracking of flexible-joints port-Hamiltonian robots: theory and experiments},
author = {Reyes-Baez, R.; van der Schaft, A.; Jayawardhana, B.; Pan, l.},
journal = {arXiv},
year = {2020},
institution = {University of Groningen, The Netherlands},
abstract = {In this work we present a constructive method to design a family of virtual contraction based controllers that solve the standard trajectory tracking problem of flexible-joint robots (FJRs) in the port-Hamiltonian (pH) framework. The proposed design method, called virtual contraction based control (v-CBC), combines the concepts of virtual control systems and contraction analysis. It is shown that under potential energy matching conditions, the closed-loop virtual system is contractive and exponential convergence to a predefined trajectory is guaranteed. Moreover, the closed-loop virtual system exhibits properties such as structure preservation, differential passivity and the existence of (incrementally) passive maps. },
keywords = {Flexible-joints robots, tracking control, port-Hamiltonian systems, contraction, virtual control systems},
language = {English}
}

In this work we present a constructive method to design a family of virtual contraction based controllers that solve the standard trajectory tracking problem of flexible-joint robots (FJRs) in the port-Hamiltonian (pH) framework. The proposed design method, called virtual contraction based control (v-CBC), combines the concepts of virtual control systems and contraction analysis. It is shown that under potential energy matching conditions, the closed-loop virtual system is contractive and exponential convergence to a predefined trajectory is guaranteed. Moreover, the closed-loop virtual system exhibits properties such as structure preservation, differential passivity and the existence of (incrementally) passive maps.

@article{kamal_2020,
title = {A New Class of Uniform Continuous Higher Order Sliding Mode Controllers},
author = {Kamal, S.; Ramesh Kumar, P.; Chalanga, A.; Kumar Goyal, J.; Bandyopadhyay, B.; Fridman, L.},
journal = {Journal of Dynamic Systems, Measurement, and Control},
year = {2020},
month = {01},
volume = {142},
number = {1},
institution = {Indian Institute of Technology (BHU) Varanasi, India; Government Engineering College Thrissur, India; University College London, UK; IIT Bombay, India; Universidad Nacional Aut ´onoma de M´exico (UNAM), Mexico },
abstract = {This paper proposes a new class of uniform continuous higher-order sliding mode algorithm (UCHOSMA) for the arbitrary relative degree systems. The proposed methodology is a combination of two controllers where one of the components is a uniform super-twisting control which acts as the disturbance compensator and the second part gives the uniform finite time convergence for the disturbance free system. This algorithm provides uniform finite time convergence of the output and its higher derivatives using an absolutely continuous control signal and thus alleviating the chattering phenomenon. The attractive feature of the proposed controller is that irrespective of the different initial conditions, the control is able to bring the states of the system to the equilibrium point uniformly in finite time. The effectiveness of the proposed controller has been demonstrated with both simulation and experimental results. },
language = {English},
publisher = {ASME}
}

This paper proposes a new class of uniform continuous higher-order sliding mode algorithm (UCHOSMA) for the arbitrary relative degree systems. The proposed methodology is a combination of two controllers where one of the components is a uniform super-twisting control which acts as the disturbance compensator and the second part gives the uniform finite time convergence for the disturbance free system. This algorithm provides uniform finite time convergence of the output and its higher derivatives using an absolutely continuous control signal and thus alleviating the chattering phenomenon. The attractive feature of the proposed controller is that irrespective of the different initial conditions, the control is able to bring the states of the system to the equilibrium point uniformly in finite time. The effectiveness of the proposed controller has been demonstrated with both simulation and experimental results.

@article{tosatto_2020,
title = {A Nonparametric Off-policy Policy Gradient},
author = {Tosatto, S.; Carvalho, J.; Abdulsamad, H.; Peters, J.},
journal = {arXiv},
year = {2020},
institution = {TU Darmstadt, Germany; Max Planck Institute for Intelligent Systems, Germany},
abstract = {Reinforcement learning (RL) algorithms still suffer from high sample complexity despite outstanding recent successes. The need for intensive interactions with the environment is especially observed in many widely popular policy gradient algorithms that perform updates using on-policy samples. The price of such inefficiency becomes evident in real-world scenarios such as interaction-driven robot learning, where the success of RL has been rather limited. We address this issue by building on the general sample efficiency of off-policy algorithms. With nonparametric regression and density estimation methods we construct a nonparametric Bellman equation in a principled manner, which allows us to obtain closed-form estimates of the value function, and to analytically express the full policy gradient. We provide a theoretical analysis of our estimate to show that it is consistent under mild smoothness assumptions and empirically show that our approach has better sample efficiency than state-of-the-art policy gradient methods. }
}

Reinforcement learning (RL) algorithms still suffer from high sample complexity despite outstanding recent successes. The need for intensive interactions with the environment is especially observed in many widely popular policy gradient algorithms that perform updates using on-policy samples. The price of such inefficiency becomes evident in real-world scenarios such as interaction-driven robot learning, where the success of RL has been rather limited. We address this issue by building on the general sample efficiency of off-policy algorithms. With nonparametric regression and density estimation methods we construct a nonparametric Bellman equation in a principled manner, which allows us to obtain closed-form estimates of the value function, and to analytically express the full policy gradient. We provide a theoretical analysis of our estimate to show that it is consistent under mild smoothness assumptions and empirically show that our approach has better sample efficiency than state-of-the-art policy gradient methods.

@article{kassem_2020,
title = {Active dynamic vibration absorber for flutter suppression},
author = {Kassem, M.; Yang, Z.; Gu, Y.; Wang, W.; Safwat, E.},
journal = {Journal of Sound and Vibration},
year = {2020},
month = {03},
volume = {469},
institution = {Military Technical College, Egypt; Northwestern Polytechnical University, China},
abstract = {A novel flutter suppression technique is proposed using an active dynamic vibration absorber (ADVA). The ADVA is introduced by adding an active element to a classical mass-spring-damper system. This active element drives the mass by a controlling force using feedback signal from the response of the aeroelastic system. An aeroelastic mathematical model of a 2 degrees of freedom (DOFs) airfoil equipped with the proposed ADVA is established based on unsteady aerodynamics theory. The proposed ADVA is experimentally realized by a cantilever beam with a bonded Macro-fiber Composite (MFC) and a lumped mass. MFC nonlinearities are compensated by applying Prandtl-Ishlinskii approach to design a proportional integral (PI)-based hysteresis compensator controller. A test rig is designed, fabricated and the parameters of the physical airfoil model and the connected ADVA are identified. The feasibility of the ADVA for flutter suppression is assessed via a wind tunnel testing. The experimental results show that applying the closed-loop ADVA increases the airfoil flutter speed by 42.9%. Whereas the open-loop ADVA results in 26.6% improvement in the flutter speed. Moreover, if the airfoil experiences a high amplitude limit cycle oscillation (LCO), the ADVA efficiently suppresses up to 93.3% of its amplitude when turned to the closed-loop control. },
keywords = {Flutter, LCO, Dynamic vibration absorber, Closed-loop control, Macro-fiber composite (MFC)},
language = {English},
publisher = {Elsevier Ltd.}
}

A novel flutter suppression technique is proposed using an active dynamic vibration absorber (ADVA). The ADVA is introduced by adding an active element to a classical mass-spring-damper system. This active element drives the mass by a controlling force using feedback signal from the response of the aeroelastic system. An aeroelastic mathematical model of a 2 degrees of freedom (DOFs) airfoil equipped with the proposed ADVA is established based on unsteady aerodynamics theory. The proposed ADVA is experimentally realized by a cantilever beam with a bonded Macro-fiber Composite (MFC) and a lumped mass. MFC nonlinearities are compensated by applying Prandtl-Ishlinskii approach to design a proportional integral (PI)-based hysteresis compensator controller. A test rig is designed, fabricated and the parameters of the physical airfoil model and the connected ADVA are identified. The feasibility of the ADVA for flutter suppression is assessed via a wind tunnel testing. The experimental results show that applying the closed-loop ADVA increases the airfoil flutter speed by 42.9%. Whereas the open-loop ADVA results in 26.6% improvement in the flutter speed. Moreover, if the airfoil experiences a high amplitude limit cycle oscillation (LCO), the ADVA efficiently suppresses up to 93.3% of its amplitude when turned to the closed-loop control.

@article{wang2_2020,
title = {Active fault-tolerant control for a quadrotor helicopter against actuator faults and model uncertainties},
author = {Wang, B.; Shen, Y.; Zhang, Y.},
journal = {Aerospace Science and Technology},
year = {2020},
month = {04},
volume = {99},
institution = {Northwestern Polytechnical University, China; Concordia University, Canada},
abstract = {This paper proposes an active fault-tolerant control strategy for a quadrotor helicopter against actuator faults and model uncertainties while explicitly considering fault estimation errors based on adaptive sliding mode control and recurrent neural networks. Firstly, a novel adaptive sliding mode control is proposed. In virtue of the proposed adaptive schemes, the system tracking performance can be guaranteed in the presence of model uncertainties without stimulating control chattering. Then, due to the fact that model-based fault estimation schemes may fail to correctly estimate fault magnitudes in the presence of model uncertainties, a fault estimation scheme is proposed by designing a parallel bank of recurrent neural networks. With the trained networks, the severity of actuator faults can be precisely estimated. Finally, by synthesizing the proposed fault estimation scheme with the developed adaptive sliding mode control, an active fault-tolerant control mechanism is established. Moreover, the issue of actuator fault estimation error is explicitly considered and compensated by the proposed adaptive sliding mode control. The effectiveness of the proposed active fault-tolerant control strategy is validated through real experiments based on a quadrotor helicopter subject to actuator faults and model uncertainties. Its advantages are demonstrated in comparison with a model-based fault estimator and a conventional adaptive sliding mode control. },
keywords = {Actuator fault estimation, Active fault-tolerant control, Adaptive sliding mode control, Model uncertainties, Recurrent neural network},
language = {English},
publisher = {Elsevier Masson}
}

This paper proposes an active fault-tolerant control strategy for a quadrotor helicopter against actuator faults and model uncertainties while explicitly considering fault estimation errors based on adaptive sliding mode control and recurrent neural networks. Firstly, a novel adaptive sliding mode control is proposed. In virtue of the proposed adaptive schemes, the system tracking performance can be guaranteed in the presence of model uncertainties without stimulating control chattering. Then, due to the fact that model-based fault estimation schemes may fail to correctly estimate fault magnitudes in the presence of model uncertainties, a fault estimation scheme is proposed by designing a parallel bank of recurrent neural networks. With the trained networks, the severity of actuator faults can be precisely estimated. Finally, by synthesizing the proposed fault estimation scheme with the developed adaptive sliding mode control, an active fault-tolerant control mechanism is established. Moreover, the issue of actuator fault estimation error is explicitly considered and compensated by the proposed adaptive sliding mode control. The effectiveness of the proposed active fault-tolerant control strategy is validated through real experiments based on a quadrotor helicopter subject to actuator faults and model uncertainties. Its advantages are demonstrated in comparison with a model-based fault estimator and a conventional adaptive sliding mode control.

@article{yang_2020,
title = {Adaptive Sliding Mode Fault-Tolerant Control for Uncertain Systems with Time Delay},
author = {Yang, P.; Liu, Z.; Wang, Y.; Li, D.},
journal = {International Journal of Automation Technology},
year = {2020},
volume = {14},
number = {2},
institution = {Nanjing University of Aeronautics and Astronautics, China; Chinese Flight Test Establishment, China},
abstract = {In this work, an adaptive sliding mode fault-tolerant controller is proposed for a class of uncertain systems with time delay. The integral term is added to the traditional sliding surface to improve the robustness of the control system, and then a type of special sliding surface is designed to cancel the reaching mode based on global sliding mode method. Without the need for fault detection and isolation, an adaptive law is proposed to estimate the value of actuator faults, and an adaptive sliding mode fault-tolerant controller is designed to guarantee the asymptotic stability of sliding dynamics. Finally, the presented control scheme is applied to the position control of a Qball-X4 quad-rotor UAV model to verify the effectiveness. },
issn = {1881-7629},
keywords = {fault-tolerant control, adaptive estimation, time delay, sliding mode control, uncertain systems},
language = {English},
publisher = {Fuji Technology Press}
}

In this work, an adaptive sliding mode fault-tolerant controller is proposed for a class of uncertain systems with time delay. The integral term is added to the traditional sliding surface to improve the robustness of the control system, and then a type of special sliding surface is designed to cancel the reaching mode based on global sliding mode method. Without the need for fault detection and isolation, an adaptive law is proposed to estimate the value of actuator faults, and an adaptive sliding mode fault-tolerant controller is designed to guarantee the asymptotic stability of sliding dynamics. Finally, the presented control scheme is applied to the position control of a Qball-X4 quad-rotor UAV model to verify the effectiveness.

@article{ren_2_2020,
title = {Advising reinforcement learning toward scaling agents in continuous control environments with sparse rewards},
author = {Ren, H.; Ben-Tzvi, P.},
journal = {Engineering Applications of Artificial Intelligence},
year = {2020},
month = {04},
volume = {90},
institution = {Virginia Tech, USA},
abstract = {This paper adapts the success of the teacher–student framework for reinforcement learning to a continuous control environment with sparse rewards. Furthermore, the proposed advising framework is designed for the scaling agents problem, wherein the student policy is trained to control multiple agents while the teacher policy is well trained for a single agent. Existing research on teacher–student frameworks have been focused on discrete control domain. Moreover, they rely on similar target and source environments and as such they do not allow for scaling the agents. On the other hand, in this work the agents face a scaling agents problem where the value functions of the source and target task converge at different rates. Existing concepts from the teacher–student framework are adapted to meet new challenges including early advising, importance of advising, and mistake correction, but a modified heuristic was used to decide on when to teach. The performance of the proposed algorithm was evaluated using the case study of pushing, and picking and placing objects with a dual arm manipulation system. The teacher policy was trained using a simulated scenario consisting of a single arm. The student policy was trained to handle the dual arm manipulation system in simulation under the advice of the teacher agent. The trained student policy was then validated using two Quanser Mico arms for experimental demonstration. The effects of varying parameters on the student performance in the advising framework was also analyzed and discussed. The results showed that the proposed advising framework expedited the training process and achieved the desired scaling within a limited advising budget. },
keywords = {Reinforcement learning, Advising framework, Continuous control, Sparse reward, Multi-agent},
language = {English},
publisher = {Elsevier Ltd.}
}

This paper adapts the success of the teacher–student framework for reinforcement learning to a continuous control environment with sparse rewards. Furthermore, the proposed advising framework is designed for the scaling agents problem, wherein the student policy is trained to control multiple agents while the teacher policy is well trained for a single agent. Existing research on teacher–student frameworks have been focused on discrete control domain. Moreover, they rely on similar target and source environments and as such they do not allow for scaling the agents. On the other hand, in this work the agents face a scaling agents problem where the value functions of the source and target task converge at different rates. Existing concepts from the teacher–student framework are adapted to meet new challenges including early advising, importance of advising, and mistake correction, but a modified heuristic was used to decide on when to teach. The performance of the proposed algorithm was evaluated using the case study of pushing, and picking and placing objects with a dual arm manipulation system. The teacher policy was trained using a simulated scenario consisting of a single arm. The student policy was trained to handle the dual arm manipulation system in simulation under the advice of the teacher agent. The trained student policy was then validated using two Quanser Mico arms for experimental demonstration. The effects of varying parameters on the student performance in the advising framework was also analyzed and discussed. The results showed that the proposed advising framework expedited the training process and achieved the desired scaling within a limited advising budget.

@inbook{de-abreu_2020,
title = {An Implementable and Stabilizing Model Predictive Control Strategy for Inverted Pendulum-Like Behaved Systems},
author = {de Abreu, O.S.L.; Martins, M.A.F.; Schnitman, L.},
booktitle = {Inverted Pendulum},
year = {2020},
institution = {Universidade Federal da Bahia, Brazil},
abstract = {In control theory, the inverted pendulum is a class of dynamic systems widely used as a benchmarking for evaluating several control strategies. Such a system is characterized by an underactuated behavior. It is also nonlinear and presents open-loop unstable and integrating modes. These dynamic features make the control more difficult, mainly when the controller synthesis seeks to include constraints and the guarantee of stability of the closed-loop system. This chapter presents a stabilizing model predictive control (MPC) strategy for inverted pendulum-like behaved systems. It has an offset-free control law based on an only optimization problem (one-layer control formulation), and the Lyapunov stability of the closed-loop system is achieved by adopting an infinite prediction horizon. The controller feasibility is also assured by imposing a suitable set of slacked terminal constraints associated with the unstable and integrating states of the system. The effectiveness of the implementable and stabilizing MPC controller is experimentally demonstrated in a commercial-didactic rotary inverted pendulum prototype, considering both cases of stabilization of the pendulum in the upright position and the output tracking of the rotary arm angle. },
keywords = {rotary inverted pendulum, model predictive control, nonlinear system, Lyapunov stability, feasible-optimization problem},
language = {English},
publisher = {IntechOpen}
}

In control theory, the inverted pendulum is a class of dynamic systems widely used as a benchmarking for evaluating several control strategies. Such a system is characterized by an underactuated behavior. It is also nonlinear and presents open-loop unstable and integrating modes. These dynamic features make the control more difficult, mainly when the controller synthesis seeks to include constraints and the guarantee of stability of the closed-loop system. This chapter presents a stabilizing model predictive control (MPC) strategy for inverted pendulum-like behaved systems. It has an offset-free control law based on an only optimization problem (one-layer control formulation), and the Lyapunov stability of the closed-loop system is achieved by adopting an infinite prediction horizon. The controller feasibility is also assured by imposing a suitable set of slacked terminal constraints associated with the unstable and integrating states of the system. The effectiveness of the implementable and stabilizing MPC controller is experimentally demonstrated in a commercial-didactic rotary inverted pendulum prototype, considering both cases of stabilization of the pendulum in the upright position and the output tracking of the rotary arm angle.

@article{silva_2020,
title = {An implementable stabilizing model predictive controller applied to a rotary flexible link: An experimental case study},
author = {Silva, B.P.M.; Santana, B.A.; Santos, T.L.M.; Martins, M.A.F.},
journal = {Control Engineering Practice},
year = {2020},
month = {06},
volume = {99},
institution = {Universidade Federal da Bahia, Brazil},
abstract = {This work addresses the application of implementable stabilizing model predictive control (MPC) strategies to a rotary flexible link (RFL). Despite their practice usefulness and design simplicity, the implementation of these stable MPC controllers with guaranteed feasibility in a real experiment, and dedicated to dynamic features of RFL systems, have not yet been documented by the literature. In contrast to conventional finite-horizon MPC strategies, infinite-horizon MPC (IHMPC) techniques ensure nominal closed-loop stability irrespective of the cost function parameters. Also, if these IHMPC strategies have feasible-optimization problem based formulations, such as implementable ones investigated here, their applicability naturally becomes attractive to the industry. Simulation and experimental results illustrate the usefulness of the implementable stabilizing MPC controllers when compared to non-feasible in practice ones, such as a classical infinite-horizon MPC and a conventional finite-horizon generalized predictive controller, thus ensuring their benefits in terms of performance and computational burden in the context of constrained RFL mechatronic systems. },
keywords = {Model predictive control, Rotary flexible link, Closed-loop stability, Feasible-optimization problem},
language = {English},
publisher = {Elsevier Ltd.}
}

This work addresses the application of implementable stabilizing model predictive control (MPC) strategies to a rotary flexible link (RFL). Despite their practice usefulness and design simplicity, the implementation of these stable MPC controllers with guaranteed feasibility in a real experiment, and dedicated to dynamic features of RFL systems, have not yet been documented by the literature. In contrast to conventional finite-horizon MPC strategies, infinite-horizon MPC (IHMPC) techniques ensure nominal closed-loop stability irrespective of the cost function parameters. Also, if these IHMPC strategies have feasible-optimization problem based formulations, such as implementable ones investigated here, their applicability naturally becomes attractive to the industry. Simulation and experimental results illustrate the usefulness of the implementable stabilizing MPC controllers when compared to non-feasible in practice ones, such as a classical infinite-horizon MPC and a conventional finite-horizon generalized predictive controller, thus ensuring their benefits in terms of performance and computational burden in the context of constrained RFL mechatronic systems.

@conference{halverson_2020,
title = {Attitude Control of a Spacecraft Flexible Appendage using Parallel Feedforward Control},
author = {Halverson, R.D.; Caverly, R.},
booktitle = {AIAA Scitech 2020 Forum},
year = {2020},
institution = {University of Minnesota, USA},
abstract = {In this paper, static and dynamic strictly positive real (SPR) parallel feedforward control methods are applied to a spacecraft with a large payload attached to the end of a flexible appendage. A dynamic model of this spacecraft is considered with a torque applied directly to the hub of the spacecraft, where the control objective is to track a desired angular velocity of the payload. This setup leads to a noncolocated relationship between the spacecraft hub torque input and the payload angular velocity output. Numerical simulation results demonstrate that the parallel feedforward controller successfully renders the system SPR, which simplifies the choice of a stabilizing feedback controller. It is shown that parallel feedforward control significantly increases the effective gain of the feedback controller within a specified frequency bandwidth. These results are expanded to an experimental rotary flexible joint manipulator, which is used as a physical analog to the flexible-appendage spacecraft. The experimental results confirm the findings from the numerical results, demonstrating the practical nature of the proposed control method. Both numerical and experimental results include a comparison to the state-of-the-art mu-tip control method that relies on a massive payload assumption, where it is shown that parallel feedforward control is implementable in situations where mu-tip control is not. },
language = {English},
publisher = {AIAA}
}

In this paper, static and dynamic strictly positive real (SPR) parallel feedforward control methods are applied to a spacecraft with a large payload attached to the end of a flexible appendage. A dynamic model of this spacecraft is considered with a torque applied directly to the hub of the spacecraft, where the control objective is to track a desired angular velocity of the payload. This setup leads to a noncolocated relationship between the spacecraft hub torque input and the payload angular velocity output. Numerical simulation results demonstrate that the parallel feedforward controller successfully renders the system SPR, which simplifies the choice of a stabilizing feedback controller. It is shown that parallel feedforward control significantly increases the effective gain of the feedback controller within a specified frequency bandwidth. These results are expanded to an experimental rotary flexible joint manipulator, which is used as a physical analog to the flexible-appendage spacecraft. The experimental results confirm the findings from the numerical results, demonstrating the practical nature of the proposed control method. Both numerical and experimental results include a comparison to the state-of-the-art mu-tip control method that relies on a massive payload assumption, where it is shown that parallel feedforward control is implementable in situations where mu-tip control is not.

@article{schleer_2020,
title = {Augmentation of haptic feedback for teleoperated robotic surgery},
author = {Schleer, P.; Kaiser, P.; Drobinsky, S.; Radermacher, K.},
journal = {International Journal of Computer Assisted Radiology and Surgery },
year = {2020},
institution = {Helmholtz Institute for Biomedical Engineering, Germany},
abstract = {Purpose: A frequently mentioned lack of teleoperated surgical robots is the lack of haptic feedback. Haptics are not only able to mirror force information from the situs, but also to provide spatial guidance according to a surgical plan. However, superposition of the two haptic information can lead to overlapping and masking of the feedback and guidance forces. This study investigates different approaches toward a combination of both information and investigates effects on system usability. Methods: Preliminary studies are conducted to define parameters for two main experiments. The two main experiments constitute simulated surgical interventions where haptic guidance as well as haptic feedback provide information for the surgeon. The first main experiment considers drilling for pedicle screw placements, while the second main experiment refers to three-dimensional milling tasks such as during partial knee replacements or craniectomies. For both experiments, different guidance modes in combination with haptic feedback are evaluated regarding effectiveness (e.g., distance to target depth), efficiency and user satisfaction (e.g., detectability of discrepancies in case of technical guidance error). Results: Regarding pedicle screw placements a combination of a peripheral visual signal and a vibration constitutes a good compromise regarding distance to target depth and detectability of discrepancies. For milling tasks, trajectory guidance is able to improve efficiency and user satisfaction (e.g., perceived workload), while boundary constraints improve effectiveness. If, assistance cannot be offered in all degrees of freedom (e.g., craniectomies), a visual substitution of the haptic force feedback shows the best results, though participants prefer using haptic force feedback. Conclusion: Our results suggest that in case haptic feedback and haptic assistance are combined appropriately, benefits of both haptic modalities can be exploited. Thereby, capabilities of the human–machine system are improved compared to usage of exclusively one of the haptic information. },
language = {English},
publisher = {Springer Nature Switzerland}
}

Purpose: A frequently mentioned lack of teleoperated surgical robots is the lack of haptic feedback. Haptics are not only able to mirror force information from the situs, but also to provide spatial guidance according to a surgical plan. However, superposition of the two haptic information can lead to overlapping and masking of the feedback and guidance forces. This study investigates different approaches toward a combination of both information and investigates effects on system usability.

Methods: Preliminary studies are conducted to define parameters for two main experiments. The two main experiments constitute simulated surgical interventions where haptic guidance as well as haptic feedback provide information for the surgeon. The first main experiment considers drilling for pedicle screw placements, while the second main experiment refers to three-dimensional milling tasks such as during partial knee replacements or craniectomies. For both experiments, different guidance modes in combination with haptic feedback are evaluated regarding effectiveness (e.g., distance to target depth), efficiency and user satisfaction (e.g., detectability of discrepancies in case of technical guidance error).

Results: Regarding pedicle screw placements a combination of a peripheral visual signal and a vibration constitutes a good compromise regarding distance to target depth and detectability of discrepancies. For milling tasks, trajectory guidance is able to improve efficiency and user satisfaction (e.g., perceived workload), while boundary constraints improve effectiveness. If, assistance cannot be offered in all degrees of freedom (e.g., craniectomies), a visual substitution of the haptic force feedback shows the best results, though participants prefer using haptic force feedback.

Conclusion: Our results suggest that in case haptic feedback and haptic assistance are combined appropriately, benefits of both haptic modalities can be exploited. Thereby, capabilities of the human–machine system are improved compared to usage of exclusively one of the haptic information.

@article{li_2020,
title = {AutoTrack: Towards High-Performance Visual Tracking for UAV with Automatic Spatio-Temporal Regularization},
author = {Li, Y.; Fu, C.; Ding, F.; Huang, Z.; Lu, G.},
journal = {arXiv},
year = {2020},
institution = {Tongji University, China; National University of Singapore, Singapore; Tsinghua University, China},
abstract = {Most existing trackers based on discriminative correlation filters (DCF) try to introduce predefined regularization term to improve the learning of target objects, e.g., by suppressing background learning or by restricting change rate of correlation filters. However, predefined parameters introduce much effort in tuning them and they still fail to adapt to new situations that the designer did not think of. In this work, a novel approach is proposed to online automatically and adaptively learn spatio-temporal regularization term. Spatially local response map variation is introduced as spatial regularization to make DCF focus on the learning of trust-worthy parts of the object, and global response map variation determines the updating rate of the filter. Extensive experiments on four UAV benchmarks have proven the superiority of our method compared to the state-of-the-art CPU- and GPU-based trackers, with a speed of ~60 frames per second running on a single CPU. Our tracker is additionally proposed to be applied in UAV localization. Considerable tests in the indoor practical scenarios have proven the effectiveness and versatility of our localization method. The code is available at https://github.com/vision4robotics/AutoTrack },
language = {English}
}

Most existing trackers based on discriminative correlation filters (DCF) try to introduce predefined regularization term to improve the learning of target objects, e.g., by suppressing background learning or by restricting change rate of correlation filters. However, predefined parameters introduce much effort in tuning them and they still fail to adapt to new situations that the designer did not think of. In this work, a novel approach is proposed to online automatically and adaptively learn spatio-temporal regularization term. Spatially local response map variation is introduced as spatial regularization to make DCF focus on the learning of trust-worthy parts of the object, and global response map variation determines the updating rate of the filter. Extensive experiments on four UAV benchmarks have proven the superiority of our method compared to the state-of-the-art CPU- and GPU-based trackers, with a speed of ~60 frames per second running on a single CPU.
Our tracker is additionally proposed to be applied in UAV localization. Considerable tests in the indoor practical scenarios have proven the effectiveness and versatility of our localization method. The code is available at https://github.com/vision4robotics/AutoTrack

@article{muratore_2020,
title = {Bayesian Domain Randomization for Sim-to-Real Transfer},
author = {Muratore, F.; Eilers, C.; Gienger, M.; Peters, J.},
journal = {arXiv},
year = {2020},
institution = {Technical University Darmstadt, Germany; Honda Research Institute Europe, Germany},
abstract = {When learning policies for robot control, the real-world data required is typically prohibitively expensive to acquire, so learning in simulation is a popular strategy. Unfortunately, such polices are often not transferable to the real world due to a mismatch between the simulation and reality, called 'reality gap'. Domain randomization methods tackle this problem by randomizing the physics simulator (source domain) according to a distribution over domain parameters during training in order to obtain more robust policies that are able to overcome the reality gap. Most domain randomization approaches sample the domain parameters from a fixed distribution. This solution is suboptimal in the context of sim-to-real transferability, since it yields policies that have been trained without explicitly optimizing for the reward on the real system (target domain). Additionally, a fixed distribution assumes there is prior knowledge about the uncertainty over the domain parameters. Thus, we propose Bayesian Domain Randomization (BayRn), a black box sim-to-real algorithm that solves tasks efficiently by adapting the domain parameter distribution during learning by sampling the real-world target domain. BayRn utilizes Bayesian optimization to search the space of source domain distribution parameters which produce a policy that maximizes the real-word objective, allowing for adaptive distributions during policy optimization. We experimentally validate the proposed approach by comparing against two baseline methods on a nonlinear under-actuated swing-up task. Our results show that BayRn is capable to perform direct sim-to-real transfer, while significantly reducing the required prior knowledge. },
keywords = {machine learning},
language = {English}
}

When learning policies for robot control, the real-world data required is typically prohibitively expensive to acquire, so learning in simulation is a popular strategy. Unfortunately, such polices are often not transferable to the real world due to a mismatch between the simulation and reality, called 'reality gap'. Domain randomization methods tackle this problem by randomizing the physics simulator (source domain) according to a distribution over domain parameters during training in order to obtain more robust policies that are able to overcome the reality gap. Most domain randomization approaches sample the domain parameters from a fixed distribution. This solution is suboptimal in the context of sim-to-real transferability, since it yields policies that have been trained without explicitly optimizing for the reward on the real system (target domain). Additionally, a fixed distribution assumes there is prior knowledge about the uncertainty over the domain parameters. Thus, we propose Bayesian Domain Randomization (BayRn), a black box sim-to-real algorithm that solves tasks efficiently by adapting the domain parameter distribution during learning by sampling the real-world target domain. BayRn utilizes Bayesian optimization to search the space of source domain distribution parameters which produce a policy that maximizes the real-word objective, allowing for adaptive distributions during policy optimization. We experimentally validate the proposed approach by comparing against two baseline methods on a nonlinear under-actuated swing-up task. Our results show that BayRn is capable to perform direct sim-to-real transfer, while significantly reducing the required prior knowledge.

@article{frohlich_2020,
title = {Bayesian Optimization for Policy Search in High-Dimensional Systems via Automatic Domain Selection},
author = {Fröhlich, L.P.; Klenske, E.D.; Daniel, C.G.; Zeilinger, M.N.},
journal = {arXiv},
year = {2020},
institution = {Bosch Center for Artificial Intelligence, Germany; ETH Zurich, Switzerland},
abstract = {Bayesian Optimization (BO) is an effective method for optimizing expensive-to-evaluate black-box functions with a wide range of applications for example in robotics, system design and parameter optimization. However, scaling BO to problems with large input dimensions (>10) remains an open challenge. In this paper, we propose to leverage results from optimal control to scale BO to higher dimensional control tasks and to reduce the need for manually selecting the optimization domain. The contributions of this paper are twofold: 1) We show how we can make use of a learned dynamics model in combination with a model-based controller to simplify the BO problem by focusing onto the most relevant regions of the optimization domain. 2) Based on (1) we present a method to find an embedding in parameter space that reduces the effective dimensionality of the optimization problem. To evaluate the effectiveness of the proposed approach, we present an experimental evaluation on real hardware, as well as simulated tasks including a 48-dimensional policy for a quadcopter. },
language = {English}
}

Bayesian Optimization (BO) is an effective method for optimizing expensive-to-evaluate black-box functions with a wide range of applications for example in robotics, system design and parameter optimization. However, scaling BO to problems with large input dimensions (>10) remains an open challenge. In this paper, we propose to leverage results from optimal control to scale BO to higher dimensional control tasks and to reduce the need for manually selecting the optimization domain. The contributions of this paper are twofold: 1) We show how we can make use of a learned dynamics model in combination with a model-based controller to simplify the BO problem by focusing onto the most relevant regions of the optimization domain. 2) Based on (1) we present a method to find an embedding in parameter space that reduces the effective dimensionality of the optimization problem. To evaluate the effectiveness of the proposed approach, we present an experimental evaluation on real hardware, as well as simulated tasks including a 48-dimensional policy for a quadcopter.

@article{wang_2020,
title = {Bounded UDE-based Controller for Input Constrained Systems with Uncertainties and Disturbances},
author = {Wang, Y.; Ren, B.; Zhong, Q.-C.},
journal = {IEEE Transactions on Industrial Electronics},
year = {2020},
institution = {Texas Tech University, USA; Illinois Institute of Technology, USA},
abstract = {The uncertainty and disturbance estimator (UDE)-based controller has emerged as an effective robust control method to handle systems with uncertainties and disturbances. However, similar to other controllers including an integral action, the UDE-based controller faces the integral windup issue when the system has input constraints. In this paper, a bounded UDE-based controller is proposed to deal with systems subject to uncertainties, disturbances and input constraint. An additional time-varying variable is introduced into the design of the error dynamics to naturally avoid integral windup. The boundedness design guarantees that both the final controller output and the additional time-varying variable dynamically move on an ellipse so that the controller output always satisfies the input constraint. The proposed bounded UDE-based controller inherits the robustness of the conventional UDE-based control method, and has a clear structure, with guidelines provided for parameter selections. Both theoretical analysis and experimental results are provided to validate the proposed design. },
issn = {0278-0046},
keywords = {Uncertainty and disturbance estimator (UDE), boundedness design, input constraint, integral windup},
language = {English},
publisher = {IEEE}
}

The uncertainty and disturbance estimator (UDE)-based controller has emerged as an effective robust control method to handle systems with uncertainties and disturbances. However, similar to other controllers including an integral action, the UDE-based controller faces the integral windup issue when the system has input constraints. In this paper, a bounded UDE-based controller is proposed to deal with systems subject to uncertainties, disturbances and input constraint. An additional time-varying variable is introduced into the design of the error dynamics to naturally avoid integral windup. The boundedness design guarantees that both the final controller output and the additional time-varying variable dynamically move on an ellipse so that the controller output always satisfies the input constraint. The proposed bounded UDE-based controller inherits the robustness of the conventional UDE-based control method, and has a clear structure, with guidelines provided for parameter selections. Both theoretical analysis and experimental results are provided to validate the proposed design.

@article{rsetam_2020,
title = {Cascaded Extended State Observer Based Sliding Mode Control for Underactuated Flexible Joint Robot},
author = {Rsetam, K.A.; Cao, Z.; Man, Z.},
journal = {IEEE Transactions on Industrial Electronics},
year = {2020},
institution = {Swinburne University of Technology, Australia},
abstract = {This paper presents a new cascaded extended state observer (CESO) based sliding mode control (SMC) for an underactuated flexible joint robot (FJR). The control of FJR has many challenges including coupling, underactuation, nonlinearity, uncertainties and external disturbances, and the noise amplification especially in the high order systems. The proposed control integrates CESO and SMC, in which the CESO estimates the states and disturbances, and the SMC provides the system robustness to the uncertainty and disturbance estimation errors. Firstly, a dynamic model of the FJR is derived and converted from an underactuated form to a canonical form via the Olfati transformation and flatness approach, which reduces the complexity of the controller design. Furthermore, by taking the advantage of available measurable states, the CESO is adopted to attenuate the noises and make SMC feasible for high order systems. Moreover, the CESO estimates the disturbances, which relaxes the upper bound of the disturbance in the SMC and reduces the chattering due to smaller switching gains. A stability analysis of the closed loop system is presented based on the Lyapunov method. The effectiveness of the proposed control is verified experimentally on a real time FJR system. },
issn = {1557-9948},
keywords = {Flexible Joint Robot (FJR), underactuation, cascade extended state observer (CESO), sliding mode Control (SMC)},
language = {English},
publisher = {IEEE}
}

This paper presents a new cascaded extended state observer (CESO) based sliding mode control (SMC) for an underactuated flexible joint robot (FJR). The control of FJR has many challenges including coupling, underactuation, nonlinearity, uncertainties and external disturbances, and the noise amplification especially in the high order systems. The proposed control integrates CESO and SMC, in which the CESO estimates the states and disturbances, and the SMC provides the system robustness to the uncertainty and disturbance estimation errors. Firstly, a dynamic model of the FJR is derived and converted from an underactuated form to a canonical form via the Olfati transformation and flatness approach, which reduces the complexity of the controller design. Furthermore, by taking the advantage of available measurable states, the CESO is adopted to attenuate the noises and make SMC feasible for high order systems. Moreover, the CESO estimates the disturbances, which relaxes the upper bound of the disturbance in the SMC and reduces the chattering due to smaller switching gains. A stability analysis of the closed loop system is presented based on the Lyapunov method. The effectiveness of the proposed control is verified experimentally on a real time FJR system.

@article{shahid_2020,
title = {Comparative Analysis of Different Model-Based Controllers Using Active Vehicle Suspension System},
author = {Shahid, Y.; Wei, M.},
journal = {Algorithms},
year = {2020},
institution = {Nanjing University of Aeronautics and Astronautics, China},
abstract = {This paper deals with the active vibration control of a quarter-vehicle suspension system. Damping control methods investigated in this paper are: higher-order sliding mode control (HOSMC) based on super twisting algorithm (STA), first-order sliding mode control (FOSMC), integral sliding mode control (ISMC), proportional integral derivative (PID), linear quadratic regulator (LQR) and passive suspension system. Performance comparison of different active controllers are analyzed in terms of vertical displacement, suspension travel and wheel deflection. The theoretical, quantitative and qualitative analysis verify that the STA-based HOSMC exhibits better performance as well as negate the undesired disturbances with respect to FOSMC, ISMC, PID, LQR and passive suspension system. Furthermore, it is also robust to intrinsic bounded uncertain dynamics of the model. },
issn = {1999-4893},
keywords = {higher-order sliding mode control; quarter-car; super twisting algorithm; active suspension system},
language = {English},
publisher = {MDPI AG}
}

This paper deals with the active vibration control of a quarter-vehicle suspension system. Damping control methods investigated in this paper are: higher-order sliding mode control (HOSMC) based on super twisting algorithm (STA), first-order sliding mode control (FOSMC), integral sliding mode control (ISMC), proportional integral derivative (PID), linear quadratic regulator (LQR) and passive suspension system. Performance comparison of different active controllers are analyzed in terms of vertical displacement, suspension travel and wheel deflection. The theoretical, quantitative and qualitative analysis verify that the STA-based HOSMC exhibits better performance as well as negate the undesired disturbances with respect to FOSMC, ISMC, PID, LQR and passive suspension system. Furthermore, it is also robust to intrinsic bounded uncertain dynamics of the model.

@article{guo_2020,
title = {Composite learning control of robotic systems: A least squares modulated approach},
author = {Guo, K.; Pan, Y.; Zheng, D.; Yu, H.},
journal = {Automatica},
year = {2020},
month = {01},
volume = {111},
institution = {Shandong University, China; Sun Yat-sen University, China; National University of Singapore, Singapore},
abstract = {Most current studies of adaptive robot control concentrate on parameter convergence in the steady state, while parameter convergence rates are rarely investigated. This paper proposes a least-squares modulated composite learning robot control based on Moore–Penrose pseudoinverse to improve the performance of parameter convergence. In the composite learning, a prediction error is constructed based on online historical data and regressor extension, and both the prediction and tracking errors are applied to update parameter estimates such that accurate and smooth parameter estimation is obtained under a weak excitation condition termed interval excitation (IE). The distinctive features of the proposed method include: (1) Asymptotic stability of the closed-loop system is proven without the IE condition; (2) exponential stability is proven and balanced and easily tunable rates of parameter convergence are achieved under the IE condition, where the rates are independent of unpredictable excitation levels in different regressor channels. These two features are generally not achievable with the existing adaptive robot control methods. Experimental results on an industrial manipulator have demonstrated the effectiveness and superiority of the proposed approach. },
keywords = {Adaptive control, Composite learning, Least squares, Moore–Penrose pseudoinverse, Robot manipulators},
language = {English},
publisher = {Elsevier Ltd.}
}

Most current studies of adaptive robot control concentrate on parameter convergence in the steady state, while parameter convergence rates are rarely investigated. This paper proposes a least-squares modulated composite learning robot control based on Moore–Penrose pseudoinverse to improve the performance of parameter convergence. In the composite learning, a prediction error is constructed based on online historical data and regressor extension, and both the prediction and tracking errors are applied to update parameter estimates such that accurate and smooth parameter estimation is obtained under a weak excitation condition termed interval excitation (IE). The distinctive features of the proposed method include: (1) Asymptotic stability of the closed-loop system is proven without the IE condition; (2) exponential stability is proven and balanced and easily tunable rates of parameter convergence are achieved under the IE condition, where the rates are independent of unpredictable excitation levels in different regressor channels. These two features are generally not achievable with the existing adaptive robot control methods. Experimental results on an industrial manipulator have demonstrated the effectiveness and superiority of the proposed approach.

@article{yildiz_2020,
title = {Constrained adaptive backstepping control of a semi‐active suspension considering suspension travel limits},
author = {Yildiz, A.S.; Sivrioglu, S.},
journal = {Asian Journal of Control},
year = {2020},
institution = {Gumushane University, Turkey; Gebze Technical University, Turkey},
abstract = {In this article, a new constrained adaptive backstepping control design is presented for vibration suppression of the quarter car suspension system equipped with a magnetorheological (MR) damper. In the controller design, parametric uncertainties related to suspension parameter and the semi‐active actuator are considered. Also, the proposed controller is investigated, along with the designed road profiles based on the International Standard Organization (ISO) 8608 classification and tested under different speed conditions. The effectiveness of the proposed controller has been validated through comparisons with uncontrolled, passive, and separately designed adaptive control cases. According to the simulation and experimental results, the proposed control approach improves ride comfort and ride safety without violating suspension rattle space constraint despite the uncertainties in suspension parameters. },
keywords = {Semi-active suspension, Barrier Lyapunov function, Backstepping controller, ISO 8608},
language = {English},
publisher = {John Wiley & Sons, Inc.}
}

In this article, a new constrained adaptive backstepping control design is presented for vibration suppression of the quarter car suspension system equipped with a magnetorheological (MR) damper. In the controller design, parametric uncertainties related to suspension parameter and the semi‐active actuator are considered. Also, the proposed controller is investigated, along with the designed road profiles based on the International Standard Organization (ISO) 8608 classification and tested under different speed conditions. The effectiveness of the proposed controller has been validated through comparisons with uncontrolled, passive, and separately designed adaptive control cases. According to the simulation and experimental results, the proposed control approach improves ride comfort and ride safety without violating suspension rattle space constraint despite the uncertainties in suspension parameters.

@article{kong_2020,
title = {Contraction analysis of nonlinear noncausal iterative learning control},
author = {Kong, F.H.; Manchester, I.R.},
journal = {Systems & Control Letters},
year = {2020},
month = {02},
volume = {136},
institution = {Universty of Sydney, Australia},
abstract = {Iterative learning control (ILC) is a method for learning input signals for repetitive control tasks. In this paper, we provide a new method based on convex optimization for certifying convergence and estimating convergence rate in ILC schemes involving a nonlinear plant and a noncausal update law, which are common in practice. Using sum-of-squares (SOS) optimization, we compute the convergence rate of an example nonlinear, noncausal ILC system and verify its accuracy in experiment. },
keywords = {Iterative learning control, Contraction analysis, Linear matrix inequalities},
language = {English},
publisher = {Elsevier B.V.}
}

Iterative learning control (ILC) is a method for learning input signals for repetitive control tasks. In this paper, we provide a new method based on convex optimization for certifying convergence and estimating convergence rate in ILC schemes involving a nonlinear plant and a noncausal update law, which are common in practice. Using sum-of-squares (SOS) optimization, we compute the convergence rate of an example nonlinear, noncausal ILC system and verify its accuracy in experiment.

@article{mehedi_2020,
title = {Controlling a Rotary Servo Cart System Using Robust Generalized Dynamic Inversion},
author = {Mehedi, I.M.; Ansari, U.; Al-Saggaf, U.M.; Bajodah, A.H.},
journal = {International Journal of Robotics and Automation},
year = {2020},
volume = {35},
number = {1},
institution = {King Abdulaziz University, Saudi Arabia},
abstract = {This paper presents the design and implementation of the Robust Generalized Dynamic Inversion (RGDI) based position and speed control of the rotary servo cart system. The RGDI control law comprises the two sub-parts, i.e., the equivalent control part and the robust control part. The equivalent control part represents the conventional form of the Generalized Dynamic Inversion control, in which dynamical constraints based on the deviation functions of angular position and its rate are formulated that encapsulates the control objectives. The control law is realized by inverting the constraint dynamics using dynamically scaled generalized inversion. The robust control part consists of an additional term based on the sliding mode principle that makes the control law robust with respect to the parametric fluctuation and outward disturbances while providing improved tracking performance. The closed-loop stability of RGDI control is ensured using the positive definite Lyapunov candidate function, which will guarantee globally practical stable tracking performance. Numerical simulations in Matlab/Simulink environment and the real-time experiments are performed on the Quanser’s SRV02 rotary servo cart platform. It exhibits better tracking performance through simulation and the experimental results. Robustness attributes to the proposed control law under the influence of the parametric uncertainties and external disturbances. },
keywords = {Generalized dynamic inversion, servo cart system, Lyapunov stability, sliding mode control, robust control},
language = {English},
publisher = {ACTA Press}
}

This paper presents the design and implementation of the Robust Generalized Dynamic Inversion (RGDI) based position and speed control of the rotary servo cart system. The RGDI control law comprises the two sub-parts, i.e., the equivalent control part and the robust control part. The equivalent control part represents the conventional form of the Generalized Dynamic Inversion control, in which dynamical constraints based on the deviation functions of angular position and its rate are formulated that encapsulates the control objectives. The control law is realized by inverting the constraint dynamics using dynamically scaled generalized inversion. The robust control part consists of an additional term based on the sliding mode principle that makes the control law robust with respect to the parametric fluctuation and outward disturbances while providing improved tracking performance. The closed-loop stability of RGDI control is ensured using the positive definite Lyapunov candidate function, which will guarantee globally practical stable tracking performance. Numerical simulations in Matlab/Simulink environment and the real-time experiments are performed on the Quanser’s SRV02 rotary servo cart platform. It exhibits better tracking performance through simulation and the experimental results. Robustness attributes to the proposed control law under the influence of the parametric uncertainties and external disturbances.

@article{jonnalagadda_2020,
title = {Current cycle feedback iterative learning control for tracking control of magnetic levitation system},
author = {Jonnalagadda, V.K.; Elumalai, V.K.; Agrawal, S.},
journal = {Transactions of the Institute of Measurement and Control},
year = {2020},
month = {2},
volume = {42},
number = {3},
institution = {Vellore Institute of Technology, India},
abstract = {This paper presents the current cycle feedback iterative learning control (CCF-ILC) augmented with the modified proportional integral derivative (PID) controller to improve the trajectory tracking and robustness of magnetic levitation (maglev) system. Motivated by the need to enhance the point to point control of maglev technology, which is widely used in several industrial applications ranging from photolithography to vibration control, we present a novel CCF-ILC framework using plant inversion technique. Modulating the control signal based on the current tracking error, CCF-ILC reduces the dependency on accurate plant model and significantly improves the robustness of the closed loop system by synthesizing the causal filters to counteract the effect of model uncertainty. To assess the stability, we present a maximum singular value based criterion for asymptotic stability of linear iterative system controlled using CCF-ILC. In addition, we prove the monotonic convergence of output sequence in the neighbourhood of reference trajectory. Finally, the proposed control framework is experimentally validated on a benchmark magnetic levitation system through hardware in loop (HIL) testing. Experimental results substantiate that synthesizing CCF-ILC with the feedback controller can significantly improve the trajectory tracking and robustness characteristics of maglev system. },
keywords = {ILC, PIV, intelligent control, magnetic levitation system, command following},
language = {English},
publisher = {SAGE Publications}
}

This paper presents the current cycle feedback iterative learning control (CCF-ILC) augmented with the modified proportional integral derivative (PID) controller to improve the trajectory tracking and robustness of magnetic levitation (maglev) system. Motivated by the need to enhance the point to point control of maglev technology, which is widely used in several industrial applications ranging from photolithography to vibration control, we present a novel CCF-ILC framework using plant inversion technique. Modulating the control signal based on the current tracking error, CCF-ILC reduces the dependency on accurate plant model and significantly improves the robustness of the closed loop system by synthesizing the causal filters to counteract the effect of model uncertainty. To assess the stability, we present a maximum singular value based criterion for asymptotic stability of linear iterative system controlled using CCF-ILC. In addition, we prove the monotonic convergence of output sequence in the neighbourhood of reference trajectory. Finally, the proposed control framework is experimentally validated on a benchmark magnetic levitation system through hardware in loop (HIL) testing. Experimental results substantiate that synthesizing CCF-ILC with the feedback controller can significantly improve the trajectory tracking and robustness characteristics of maglev system.

@article{oliveira_2020,
title = {Design and experiments on an inflatable link robot with a built-in vision sensor},
author = {Oliveira, J.; Ferreira, A.; Reis, J.C.P.},
journal = {Mechatronics},
year = {2020},
month = {02},
volume = {65},
institution = {Universidade de Lisboa, Portugal},
abstract = {Soft robots and manipulators with inflatable links possess inherent safety characteristics that make them a very interesting choice for tasks requiring human-robot interaction. However, they may face important challenges in their performance due to low structural stiffness. Accurate positioning of the end effector may prove a difficult task due to limitations in gravity compensation, or due to the occurrence of oscillations during motion, especially in the absence of precise information about the payload that is being manipulated. In this paper a novel approach to the design and control of a robot with an inflatable link is proposed, where the link assumes the triple role of compliant structural element, touch sensor, and active mechanism to adjust the performance of vibration control algorithms. For this, a vision sensor is placed inside the link to provide information about the structural deformation. The sensor is used to measure the tip displacement, and also to detect contact of the link surface with the surrounding environment. The concept was implemented on a two-joint rigid manipulator with a single inflatable link. Experiments on vibration control and contact detection of the inflatable link are reported where the control system was able to significantly reduce tip oscillations, including when a payload was added to the tip. The robot was also able to make binary detection of contact between the inflatable link and the user, and change its operation mode accordingly. },
keywords = {Inflatable robots, Soft material robotics, Flexible robots, Physical human-robot interaction},
language = {English},
publisher = {Elsevier B.V.}
}

Soft robots and manipulators with inflatable links possess inherent safety characteristics that make them a very interesting choice for tasks requiring human-robot interaction. However, they may face important challenges in their performance due to low structural stiffness. Accurate positioning of the end effector may prove a difficult task due to limitations in gravity compensation, or due to the occurrence of oscillations during motion, especially in the absence of precise information about the payload that is being manipulated. In this paper a novel approach to the design and control of a robot with an inflatable link is proposed, where the link assumes the triple role of compliant structural element, touch sensor, and active mechanism to adjust the performance of vibration control algorithms. For this, a vision sensor is placed inside the link to provide information about the structural deformation. The sensor is used to measure the tip displacement, and also to detect contact of the link surface with the surrounding environment. The concept was implemented on a two-joint rigid manipulator with a single inflatable link. Experiments on vibration control and contact detection of the inflatable link are reported where the control system was able to significantly reduce tip oscillations, including when a payload was added to the tip. The robot was also able to make binary detection of contact between the inflatable link and the user, and change its operation mode accordingly.

@article{shendge_2020,
title = {Disturbance observer based controller under noisy measurement for tracking of nDOF uncertain mismatched nonlinear},
author = {Chaudhari, S.; Shendge, P.D.; Phadke, S.B.},
journal = {IEEE/ASME Transactions on Mechatronics},
year = {2020},
institution = {College of Engineering Pune, India},
abstract = {This paper analyzes a recent result on tracking control of nonlinear, mismatched interconnected systems and points out its drawback with regard to noise amplification that originates from numerical differentiation of tracking error. One of the objectives of the paper is to propose two remedial measures and show their improvements by analysis for performance and stability. In the first remedy, the derivative of the tracking error is obtained without recourse to numerical differentiation while the second remedy offers a solution that does not need the derivative of the tracking error. Another objective of the paper is to propose an extension to decentralized sliding mode control of interconnected systems. Unlike other published results, the extension can assure local sliding in each subsystem. The proposed remedies are validated by numerical simulation, and experimentation on a laboratory coupled tank system. A discussion with a generalized extended state observer (GESO) based method and comparative simulations with other well known methods are carried out. },
keywords = {disturbance observer, sliding mode control, measurement noise, mismatched nonlinear systems, decentralized control},
language = {English},
publisher = {IEEE}
}

This paper analyzes a recent result on tracking control of nonlinear, mismatched interconnected systems and points out its drawback with regard to noise amplification that originates from numerical differentiation of tracking error. One of the objectives of the paper is to propose two remedial measures and show their improvements by analysis for performance and stability. In the first remedy, the derivative of the tracking error is obtained without recourse to numerical differentiation while the second remedy offers a solution that does not need the derivative of the tracking error. Another objective of the paper is to propose an extension to decentralized sliding mode control of interconnected systems. Unlike other published results, the extension can assure local sliding in each subsystem. The proposed remedies are validated by numerical simulation, and experimentation on a laboratory coupled tank system. A discussion with a generalized extended state observer (GESO) based method and comparative simulations with other well known methods are carried out.

@article{lin_2020,
title = {Earthquake early warning-enabled smart base isolation system},
author = {Lin, Y.-S.; Chan, R.W.K.; Tagawa, H.},
journal = {Automation in Construction},
year = {2020},
month = {07},
volume = {115},
institution = {RMIT University, Australia; Hiroshima University, Japan},
abstract = {Recent deployments of Earthquake Early Warning (EEW) system in Japan and some other earthquake-prone regions provide the vital warning signals prior to the arrival of destructive ground motions. The EEW system uses the different traveling speed of seismic P- and S-waves to achieve the goal of earthquake warning. This technology is primarily used to produce warning signals to alert the public to avoid potential risks, such as to evacuate from buildings. It is also used to mitigate other risks such as reducing train speed of Shinkansen trains. The present study suggests a new seismic-risk mitigation technique by connecting a base isolation system with the EEW. Base isolation is a mature technology which decouples structure from its base and lengthens its natural period of vibration. However, existing base isolation devices must possess certain lateral resistance to withstand service lateral forces such as wind, and such stiffness hinders the effectiveness of vibration isolation. In addition, supplementary damping devices are sometimes added to control excessive displacements in isolation level. This paper proposes a smart system which changes the property of a base isolation system upon EEW signal. In normal times when earthquake risk is not present, the base isolation system is locked by shear keys. When an earthquake is signalled by the EEW system, a mechanical system releases the base isolation system. When the earthquake ceases, the system resets and base isolation is locked again. On-board vibration sensors are added to activate the system in case of EEW fails to detect incoming waves. A crucial benefit of the system is that supplementary damper is no longer required to control excessive isolator displacements, and elastic stiffness of base isolation is no longer required to re-enter the base isolation system. The result is that it maximizes the vibration isolation effectiveness. A conceptual framework of proposed system is described and demonstrated by laboratory-scaled experiments. A 6-storey test frame is excited on a shake table subjected to historical earthquakes. Results indicates that the proposed system is effective in reducing earthquake responses on the building. It is an IoT-enabled earthquake-risk mitigation system. },
keywords = {Earthquake early warning; Base isolation; Smart structures; Internet of things},
language = {English},
publisher = {Elsevier B.V.}
}

Recent deployments of Earthquake Early Warning (EEW) system in Japan and some other earthquake-prone regions provide the vital warning signals prior to the arrival of destructive ground motions. The EEW system uses the different traveling speed of seismic P- and S-waves to achieve the goal of earthquake warning. This technology is primarily used to produce warning signals to alert the public to avoid potential risks, such as to evacuate from buildings. It is also used to mitigate other risks such as reducing train speed of Shinkansen trains. The present study suggests a new seismic-risk mitigation technique by connecting a base isolation system with the EEW. Base isolation is a mature technology which decouples structure from its base and lengthens its natural period of vibration. However, existing base isolation devices must possess certain lateral resistance to withstand service lateral forces such as wind, and such stiffness hinders the effectiveness of vibration isolation. In addition, supplementary damping devices are sometimes added to control excessive displacements in isolation level. This paper proposes a smart system which changes the property of a base isolation system upon EEW signal. In normal times when earthquake risk is not present, the base isolation system is locked by shear keys. When an earthquake is signalled by the EEW system, a mechanical system releases the base isolation system. When the earthquake ceases, the system resets and base isolation is locked again. On-board vibration sensors are added to activate the system in case of EEW fails to detect incoming waves. A crucial benefit of the system is that supplementary damper is no longer required to control excessive isolator displacements, and elastic stiffness of base isolation is no longer required to re-enter the base isolation system. The result is that it maximizes the vibration isolation effectiveness. A conceptual framework of proposed system is described and demonstrated by laboratory-scaled experiments. A 6-storey test frame is excited on a shake table subjected to historical earthquakes. Results indicates that the proposed system is effective in reducing earthquake responses on the building. It is an IoT-enabled earthquake-risk mitigation system.

@article{yazdani_2020,
title = {Estimating Human Teleoperator Posture Using Only a Haptic-Input Device},
author = {Yazdani, A.; Sabbagh Novin, R.; Merryweather, A.; Hermans, T.},
journal = {arXiv},
year = {2020},
institution = {University of Utah, USA},
abstract = {Ergonomic analysis of human posture plays a vital role in understanding long-term, work-related safety and health. Current analysis is often hindered due to difficulties in estimating human posture. We introduce a new approach to the problem of human posture estimation for teleoperation tasks which relies solely on a haptic-input device for generating observations. We model the human upper body using a redundant, partially observable dynamical system. This allows us to naturally formulate the estimation problem as probabilistic inference and solve the inference problem using a standard particle filter. We show that our approach accurately estimates the posture of different human users without knowing their specific segment lengths. We evaluate our posture estimation approach from a haptic-input device by comparing it with the human posture estimates from a commercial motion capture system. Our results show that the proposed algorithm successfully estimates human posture based only on the trajectory of the haptic-input device stylus. We additionally show that ergonomic risk estimates derived from our posture estimation approach are comparable to those estimates from gold-standard, motion-capture based pose estimates. },
keywords = {Human-Computer Interaction; Signal Processing},
language = {English}
}

Ergonomic analysis of human posture plays a vital role in understanding long-term, work-related safety and health. Current analysis is often hindered due to difficulties in estimating human posture. We introduce a new approach to the problem of human posture estimation for teleoperation tasks which relies solely on a haptic-input device for generating observations. We model the human upper body using a redundant, partially observable dynamical system. This allows us to naturally formulate the estimation problem as probabilistic inference and solve the inference problem using a standard particle filter. We show that our approach accurately estimates the posture of different human users without knowing their specific segment lengths. We evaluate our posture estimation approach from a haptic-input device by comparing it with the human posture estimates from a commercial motion capture system. Our results show that the proposed algorithm successfully estimates human posture based only on the trajectory of the haptic-input device stylus. We additionally show that ergonomic risk estimates derived from our posture estimation approach are comparable to those estimates from gold-standard, motion-capture based pose estimates.

@article{armenta_2020,
title = {Exact Takagi-Sugeno descriptor models of recurrent high-order neural networks for control applications},
author = {Armenta, C.; Bernal, M.; Estrada-Manzo, V.; Sala, A.},
journal = {Computational and Applied Mathematics},
year = {2020},
month = {03},
volume = {39},
institution = {Sonora Institute of Technology, Mexico; Universidad Politécnica de Pachuca, Mexico; Universitat Politecnica de Valencia, Spain},
abstract = {This work presents an exact Takagi-Sugeno descriptor model of a recurrent high-order neural network arising from identification of a nonlinear plant. The proposed rearrangement allows exploiting the nonlinear characteristics of the neural model for H∞-optimal controller design whose conditions are expressed as linear matrix inequalities. Simulation and real-time results are presented that illustrate the advantages of the proposal. },
issn = {2238-3603},
keywords = {Descriptor system, Linear matrix inequality, Takagi-Sugeno model, Recurrent high-order neural network},
language = {English},
publisher = {Springer International Publishing}
}

This work presents an exact Takagi-Sugeno descriptor model of a recurrent high-order neural network arising from identification of a nonlinear plant. The proposed rearrangement allows exploiting the nonlinear characteristics of the neural model for H∞-optimal controller design whose conditions are expressed as linear matrix inequalities. Simulation and real-time results are presented that illustrate the advantages of the proposal.

@conference{goyal_2020,
title = {Experimental Design of Robust Decentralized PI Controller for TRMS through Polytopic Modeling},
author = {Goyal J.K.; Aggarwal, S.; Ghosh, S.; Kamal, S.; Dworak, P. ; Pal, A.K.},
booktitle = {2020 IEEE International Conference on Industrial Technology (ICIT)},
year = {2020},
institution = {Indian Institute of Technology Varanasi, India},
abstract = {This paper presents a new approach to deal with the nonlinear dynamics present in a Twin Rotor MIMO System (TRMS) without any approximation of the system model. Conventionally, there are two approaches/options for adopting linear control design methods for a nonlinear system. The first one is to linearize the model around an operating point, neglecting the higher order dynamics. This could result in drastic changes in the transient behavior of the model for which the controller fails to ensure robustness and a good tracking response. The second approach is to consider the exact model and represent the nonlinear terms involved, in the form of structural or model uncertainties. The design technique adopted in this paper is based on the second approach. The nonlinear model of TRMS is represented in the form of a linear parameter-varying system, thereby facilitating the design of a linear robust controller, with a better tracking control of the desired states. Additionally, a robust Linear Matrix Inequality (LMI) based decentralized Proportional Integral (PI) controller is designed along with locating closed-loop poles in a desired region. A comparative study is done for experimental results of various existing designs and the proposed method. },
issn = {2641-0184 },
keywords = {Twin Rotor MIMO System, Linear matrix inequalities, Decentralized PI controller, Polytopic modeling},
language = {English},
publisher = {IEEE},
isbn = {978-1-7281-5755-9}
}

This paper presents a new approach to deal with the nonlinear dynamics present in a Twin Rotor MIMO System (TRMS) without any approximation of the system model. Conventionally, there are two approaches/options for adopting linear control design methods for a nonlinear system. The first one is to linearize the model around an operating point, neglecting the higher order dynamics. This could result in drastic changes in the transient behavior of the model for which the controller fails to ensure robustness and a good tracking response. The second approach is to consider the exact model and represent the nonlinear terms involved, in the form of structural or model uncertainties. The design technique adopted in this paper is based on the second approach. The nonlinear model of TRMS is represented in the form of a linear parameter-varying system, thereby facilitating the design of a linear robust controller, with a better tracking control of the desired states. Additionally, a robust Linear Matrix Inequality (LMI) based decentralized Proportional Integral (PI) controller is designed along with locating closed-loop poles in a desired region. A comparative study is done for experimental results of various existing designs and the proposed method.

@article{silva2_2020,
title = {Fast Nonsingular Terminal Sliding Mode Flight Control for Multirotor Aerial Vehicles},
author = {Silva Jr. A.L.; Santos, D.A.},
year = {2020},
institution = {Instituto Tecnologico de Aeronautica, Brazil},
abstract = {This paper is concerned with the robust flight control of multirotor aerial vehicles (MAVs) subject to bounded force and torque disturbances. The focus is on the entire class of MAVs containing an arbitrary even number (≥ 4) of fixed (not 4 vectoring) rotors. To deal with this problem, firstly, a ubiquitous hierarchical control architecture in which the attitude control loop is nested inside the position control loop is adopted and augmented with a control allocator which makes the design of the control laws themselves independent of the rotor arrangement. Especially, the control allocation problem is formulated as a quadratic program that minimizes the thrust commands and accounts for the thrust range and rate bounds. Secondly, geometric attitude and position control laws are designed separately using a multi-input fast nonsingular terminal sliding mode control (FNTSMC) strategy, which guarantees singularity-free finite-time stability and robustness. The main contributions are: 1) the augmentation of the hierarchical control scheme for extending its applicability to any fixed-rotor MAV; and 2) detailed geometric design and finite-time stability analysis of the position and attitude control loops using the FNTSMC theory. The system is evaluated on computational simulations as well as on a hardware-in-the-loop experiment, showing that it is effective, simple to implement and adjust, and reliable to operate in nonlinear regimes as well as under bounded disturbances. },
keywords = {Multirotor aerial vehicle, quadrotor, flight control, control allocation, terminal sliding mode control, hardware-in-the-loop simulation},
language = {English}
}

This paper is concerned with the robust flight control of multirotor aerial vehicles (MAVs) subject to bounded force and torque disturbances. The focus is on the entire class of MAVs containing an arbitrary even number (≥ 4) of fixed (not 4 vectoring) rotors. To deal with this problem, firstly, a ubiquitous hierarchical control architecture in which the attitude control loop is nested inside the position control loop is adopted and augmented with a control allocator which makes the design of the control laws themselves independent of the rotor arrangement. Especially, the control allocation problem is formulated as a quadratic program that minimizes the thrust commands and accounts for the thrust range and rate bounds. Secondly, geometric attitude and position control laws are designed separately using a multi-input fast nonsingular terminal sliding mode control (FNTSMC) strategy, which guarantees singularity-free finite-time stability and robustness. The main contributions are: 1) the augmentation of the hierarchical control scheme for extending its applicability to any fixed-rotor MAV; and 2) detailed geometric design and finite-time stability analysis of the position and attitude control loops using the FNTSMC theory. The system is evaluated on computational simulations as well as on a hardware-in-the-loop experiment, showing that it is effective, simple to implement and adjust, and reliable to operate in nonlinear regimes as well as under bounded disturbances.

@inbook{yazdjerdi_2020,
title = {Fault Tolerant Controller Schemes for Single and Multiple Mobile Robots},
author = {Yazdjerdi, P.; Meskin N. },
booktitle = {Soft Computing in Condition Monitoring and Diagnostics of Electrical and Mechanical Systems},
year = {2020},
institution = {Qatar University, Qatar},
abstract = {In this chapter, an actuator fault tolerant controller is developed for both single and multiple differential drive mobile robots. First, a fault tolerant controller is proposed for loss of effectiveness actuator faults in differential drive mobile robots while tracking the desired trajectory. The actuator loss of effectiveness fault is injected on the kinematic equation of the robot as a multiplicative gain in the left and right wheels angular velocity. Accordingly, the goal is to estimate and tolerate the injected actuator faults. A fault diagnosis method based on joint state and parameter estimation scheme is proposed to estimate the actuator loss of effectiveness gains as well as the states of the system. The estimated values of actuator faults are then used in the controller to compensate their effects on the mobile robots performance. Next, the proposed Fault Tolerant Controller (FTC) method is extended for the leader–follower formation control of mobile robots in the presence of actuator fault in a leader and followers robots. An Extended Kalman Filter (EKF) is utilized for each robot to obtain the parameters and states of the system and as the actuator fault is detected in any of the robots, the corresponding controller compensates the fault. Finally, an obstacle avoidance feature is added to the designed actuator fault tolerant controller, for single mobile robot. Toward this goal, a go-to goal controller is designed for the mobile robot to reach the desired destination. Meanwhile, the robot can detect any obstacle with a given color specification using Kinect sensor mounted on the robot and consequently avoid collisions. The efficacy of the proposed FTC framework is demonstrated for both single and multiple mobile robots by real-time simulation and implementation on Qbot-2 from Quanser. },
language = {English},
series = {Advances in Intelligent Systems and Computing},
publisher = {Springer, Singapore},
isbn = {978-981-15-1532-3}
}

In this chapter, an actuator fault tolerant controller is developed for both single and multiple differential drive mobile robots. First, a fault tolerant controller is proposed for loss of effectiveness actuator faults in differential drive mobile robots while tracking the desired trajectory. The actuator loss of effectiveness fault is injected on the kinematic equation of the robot as a multiplicative gain in the left and right wheels angular velocity. Accordingly, the goal is to estimate and tolerate the injected actuator faults. A fault diagnosis method based on joint state and parameter estimation scheme is proposed to estimate the actuator loss of effectiveness gains as well as the states of the system. The estimated values of actuator faults are then used in the controller to compensate their effects on the mobile robots performance. Next, the proposed Fault Tolerant Controller (FTC) method is extended for the leader–follower formation control of mobile robots in the presence of actuator fault in a leader and followers robots. An Extended Kalman Filter (EKF) is utilized for each robot to obtain the parameters and states of the system and as the actuator fault is detected in any of the robots, the corresponding controller compensates the fault. Finally, an obstacle avoidance feature is added to the designed actuator fault tolerant controller, for single mobile robot. Toward this goal, a go-to goal controller is designed for the mobile robot to reach the desired destination. Meanwhile, the robot can detect any obstacle with a given color specification using Kinect sensor mounted on the robot and consequently avoid collisions. The efficacy of the proposed FTC framework is demonstrated for both single and multiple mobile robots by real-time simulation and implementation on Qbot-2 from Quanser.

@conference{lim2_2020,
title = {Federated Reinforcement Learning for Controlling Multiple Rotary Inverted Pendulums in Edge Computing Environments},
author = {Lim, H.-K.; Kim, J.-B.; Kim, C.-M.; Hwang, G.-Y.; Choi, H. ; Han, Y.-H.},
booktitle = {2020 International Conference on Artificial Intelligence in Information and Communication (ICAIIC)},
year = {2020},
institution = {Korea University of Technology and Education, Korea},
abstract = {Reinforcement learning has recently been studied in various fields and also used to optimally control real devices (e.g., robotic arms). In this paper, we try to allow multiple reinforcement learning agents to learn optimal control policy on their own devices of the same type but with slightly different dynamics. For such multiple devices, there is no guarantee that an agent who interacts only with one device and learns the optimal control policy will also control another device well. Therefore, we may need to apply independent reinforcement learning to each device individually, which requires time-consuming effort. To solve this problem, we propose a new federated reinforcement learning architecture where each agent working on its independent device shares their learning experience with each other, and transfers a mature policy model parameters into other agents. We incorporate the Actor-Critic PPO algorithm into each agent in the proposed collaborative architecture, and propose an efficient procedure for the gradient sharing and the model transfer. We also use edge computing to solve network problems that occur when training multiple real devices at the same time. Using multiple rotary inverted pendulum devices, we demonstrate that the proposed federated reinforcement learning scheme can effectively facilitate the learning process for multiple devices, and that the learning speed can be faster if more agents are involved. },
keywords = {Actor-Critic PPO, Edge computing, Federated reinforcement learning, Multiple RIP control},
language = {English},
publisher = {IEEE},
isbn = {978-1-7281-4986-8}
}

Reinforcement learning has recently been studied in various fields and also used to optimally control real devices (e.g., robotic arms). In this paper, we try to allow multiple reinforcement learning agents to learn optimal control policy on their own devices of the same type but with slightly different dynamics. For such multiple devices, there is no guarantee that an agent who interacts only with one device and learns the optimal control policy will also control another device well. Therefore, we may need to apply independent reinforcement learning to each device individually, which requires time-consuming effort. To solve this problem, we propose a new federated reinforcement learning architecture where each agent working on its independent device shares their learning experience with each other, and transfers a mature policy model parameters into other agents. We incorporate the Actor-Critic PPO algorithm into each agent in the proposed collaborative architecture, and propose an efficient procedure for the gradient sharing and the model transfer. We also use edge computing to solve network problems that occur when training multiple real devices at the same time. Using multiple rotary inverted pendulum devices, we demonstrate that the proposed federated reinforcement learning scheme can effectively facilitate the learning process for multiple devices, and that the learning speed can be faster if more agents are involved.

@article{lim_20220,
title = {Federated Reinforcement Learning for Training Control Policies on Multiple IoT Devices},
author = {Lim, H.-K.; Kim, J.-B.; Heo, J.-S.; Han, Y.-H. },
journal = {Sensors},
year = {2020},
volume = {20},
number = {5},
institution = {Korea University of Technology and Education, Korea},
abstract = {Reinforcement learning has recently been studied in various fields and also used to optimally control IoT devices supporting the expansion of Internet connection beyond the usual standard devices. In this paper, we try to allow multiple reinforcement learning agents to learn optimal control policy on their own IoT devices of the same type but with slightly different dynamics. For such multiple IoT devices, there is no guarantee that an agent who interacts only with one IoT device and learns the optimal control policy will also control another IoT device well. Therefore, we may need to apply independent reinforcement learning to each IoT device individually, which requires a costly or time-consuming effort. To solve this problem, we propose a new federated reinforcement learning architecture where each agent working on its independent IoT device shares their learning experience (i.e., the gradient of loss function) with each other, and transfers a mature policy model parameters into other agents. They accelerate its learning process by using mature parameters. We incorporate the actor–critic proximal policy optimization (Actor–Critic PPO) algorithm into each agent in the proposed collaborative architecture and propose an efficient procedure for the gradient sharing and the model transfer. Using multiple rotary inverted pendulum devices interconnected via a network switch, we demonstrate that the proposed federated reinforcement learning scheme can effectively facilitate the learning process for multiple IoT devices and that the learning speed can be faster if more agents are involved. },
issn = {1424-8220},
keywords = { Actor–Critic PPO; federated reinforcement learning; multi-device control},
language = {English},
publisher = {MDPI AG}
}

Reinforcement learning has recently been studied in various fields and also used to optimally control IoT devices supporting the expansion of Internet connection beyond the usual standard devices. In this paper, we try to allow multiple reinforcement learning agents to learn optimal control policy on their own IoT devices of the same type but with slightly different dynamics. For such multiple IoT devices, there is no guarantee that an agent who interacts only with one IoT device and learns the optimal control policy will also control another IoT device well. Therefore, we may need to apply independent reinforcement learning to each IoT device individually, which requires a costly or time-consuming effort. To solve this problem, we propose a new federated reinforcement learning architecture where each agent working on its independent IoT device shares their learning experience (i.e., the gradient of loss function) with each other, and transfers a mature policy model parameters into other agents. They accelerate its learning process by using mature parameters. We incorporate the actor–critic proximal policy optimization (Actor–Critic PPO) algorithm into each agent in the proposed collaborative architecture and propose an efficient procedure for the gradient sharing and the model transfer. Using multiple rotary inverted pendulum devices interconnected via a network switch, we demonstrate that the proposed federated reinforcement learning scheme can effectively facilitate the learning process for multiple IoT devices and that the learning speed can be faster if more agents are involved.

@article{chang_2020,
title = {Fixed-Time Active Disturbance Rejection Control and Its Application to Wheeled Mobile Robots},
author = {Chang, S.; Wang, Y.; Zuo, Z.},
journal = { IEEE Transactions on Systems, Man, and Cybernetics: Systems},
year = {2020},
institution = {Tianjin University, China},
abstract = {This article is concerned with the fixed-time active disturbance rejection control approach for nonlinear systems subject to uncertainties and disturbances. First, a new type of extended state observer (ESO) which can attain fixed-time convergence is established to estimate the states and the total disturbance. Then a fixed-time controller based on the above ESO is developed to achieve high precision control performance. Finally, the proposed method is utilized for the wheeled mobile robot where the experiment validates the effectiveness of the theoretical results. },
issn = {2168-2216},
keywords = {Active disturbance rejection control (ADRC), fixed-time controller (FTC), fixed-time extended state observer (ESO), wheeled mobile robot (WMR)},
language = {English},
publisher = {IEEE}
}

This article is concerned with the fixed-time active disturbance rejection control approach for nonlinear systems subject to uncertainties and disturbances. First, a new type of extended state observer (ESO) which can attain fixed-time convergence is established to estimate the states and the total disturbance. Then a fixed-time controller based on the above ESO is developed to achieve high precision control performance. Finally, the proposed method is utilized for the wheeled mobile robot where the experiment validates the effectiveness of the theoretical results.

@conference{wiercigroch_2020,
title = {Force and torque feedback in endoscopic vessel harvesting},
author = {Wiercigroch, J.; Hashtrudi-Zaad, K.; Ungi, T.; Bisleri, G.; Fichtinger, G.},
booktitle = {Medical Imaging 2020: Image-Guided Procedures, Robotic Interventions, and Modeling },
year = {2020},
institution = {Queen's University, Canada},
abstract = {PURPOSE: Endoscopic vessel harvesting is the preferred minimally invasive approach to obtain grafts for coronary bypass surgeries, however it requires extensive practice to minimize vessel damage. We propose to create a surgical training simulation with visual and haptic feedback. In this study, we focus on analyzing the force and torque peaks on the surgical retractor during the procedure. METHODS: The original retractor handle was 3D scanned and modified to attach an ATI Mini40 force-torque transducer. The forces and torques in two radial artery and two saphenous vein procedures in human cadavers were recorded. The measurements, endoscopic video and surgical surface video were collected. The median and interquartile range of the force and torque peaks were calculated for the artery and vein harvesting procedures. RESULTS: The median and interquartile range for saphenous vein harvests was larger than radial artery harvests. The largest median force and torque generated in the vein was 11.654 N [posterior] and 0.661 Nm [- frontal], whereas in the artery was 6.163 N [anterior] and 0.381 Nm [+ frontal], respectively. CONCLUSION: The distribution of force and torque peaks in the retractor was found for endoscopic vessel harvests. This data can be used to design a haptic user interface, and to establish expert benchmarks for learning curve evaluation. },
language = {English},
series = {Proceedings SPIE 11315},
publisher = {SPIE}
}

PURPOSE: Endoscopic vessel harvesting is the preferred minimally invasive approach to obtain grafts for coronary bypass surgeries, however it requires extensive practice to minimize vessel damage. We propose to create a surgical training simulation with visual and haptic feedback. In this study, we focus on analyzing the force and torque peaks on the surgical retractor during the procedure.
METHODS: The original retractor handle was 3D scanned and modified to attach an ATI Mini40 force-torque transducer. The forces and torques in two radial artery and two saphenous vein procedures in human cadavers were recorded. The measurements, endoscopic video and surgical surface video were collected. The median and interquartile range of the force and torque peaks were calculated for the artery and vein harvesting procedures.
RESULTS: The median and interquartile range for saphenous vein harvests was larger than radial artery harvests. The largest median force and torque generated in the vein was 11.654 N [posterior] and 0.661 Nm [- frontal], whereas in the artery was 6.163 N [anterior] and 0.381 Nm [+ frontal], respectively.
CONCLUSION: The distribution of force and torque peaks in the retractor was found for endoscopic vessel harvests. This data can be used to design a haptic user interface, and to establish expert benchmarks for learning curve evaluation.

@article{lopez-pacheco_2020,
title = {Frequency domain CNN and disipate energy approach for damage detection in building structures},
author = {Lopez-Pacheco, M.; Morales-Valdez, J.; Yu, W.},
journal = {Soft Computing},
year = {2020},
institution = {CINVESTAV-IPN, Mexico},
abstract = {Recent developments tools and techniques for Structural Health Monitoring (SHM) allows the design of early warning systems for the damage diagnosis and structural assessment. Mosts method to damage detection involves vibration data analysis by using identification systems, that generally required a mathematical model and much information about the system, as parameters and states that are mostly unknown. In this paper, a novel Frequency Domain Convolutional Neural Network (FDCNN) proposed aims to design an identification system for damage detection based on Bouc-Wen hysteretic model. FDCNN, unlike other works, only requires acceleration measurements for damage diagnosis , that are very sensitive to environmental noise. In contrast with Neural Network (NN) and Time Domain Convolutional Neural Network (TDCNN), FDCNN reduces the computational time required for the learning stage and add robustness against noise in data. The FD-CNN includes random filters in the frequency domain to avoid measurement noise using a spectral pooling operation , which is useful when the system bandwidth is unknown. Inappropriate filtering can produce unwanted results, as a shifted and attenuation signal relative to the original. Moreover, FDCNN allows overcoming the parameterization problem in nonlinear systems, which is often difficult to achieve. In order to validate the proposed methodology, a comparison between two different architectures of convolutional neural networks are present, showing that proposed CNN in frequency domain brings better performance in the identification system for damage diagnosis in building structures. Experimental results from reducing scale two-storey building confirm the effectiveness of the proposed. },
keywords = {Structural health monitoring, Damage detection, Convolutional neural network, Identification system},
language = {English}
}

Recent developments tools and techniques for Structural Health Monitoring (SHM) allows the design of early warning systems for the damage diagnosis and structural assessment. Mosts method to damage detection involves vibration data analysis by using identification systems, that generally required a mathematical model and much information about the system, as parameters and states that are mostly unknown. In this paper, a novel Frequency Domain Convolutional Neural Network (FDCNN) proposed aims to design an identification system for damage detection based on Bouc-Wen hysteretic model. FDCNN, unlike other works, only requires acceleration measurements for damage diagnosis , that are very sensitive to environmental noise. In contrast with Neural Network (NN) and Time Domain Convolutional Neural Network (TDCNN), FDCNN reduces the computational time required for the learning stage and add robustness against noise in data. The FD-CNN includes random filters in the frequency domain to avoid measurement noise using a spectral pooling operation , which is useful when the system bandwidth is unknown. Inappropriate filtering can produce unwanted results, as a shifted and attenuation signal relative to the original. Moreover, FDCNN allows overcoming the parameterization problem in nonlinear systems, which is often difficult to achieve. In order to validate the proposed methodology, a comparison between two different architectures of convolutional neural networks are present, showing that proposed CNN in frequency domain brings better performance in the identification system for damage diagnosis in building structures. Experimental results from reducing scale two-storey building confirm the effectiveness of the proposed.

@article{junejo_2020,
title = {Fuzzy Logic Based PID Auto Tuning Method of QNET 2.0 VTOL},
author = {Junejo, M.; Kalhoro, A.N.; Kumari, A.},
journal = {International Journal of Information Technology and Computer Science (IJITCS)},
year = {2020},
month = {2},
volume = {12},
number = {1},
institution = {Mehran University of Engineering and Technology, Pakistan},
abstract = {Unmanned aerial vehicles (UAVs) have gained a lot of attention from researchers due to their hovering and vertical take-off and landing. Different techniques and methods are being employed to imple-ment UAVs. The QNET 2.0 VTOL board, specially de-signed for NI ELVIS II, is an important platform in the field of unmanned aerial vehicles (UAV). It is a helpful tool to demonstrate the essentials of vertical take-off and landing flight control (VTOL) at educational institutes. The PID controller installed in QNET 2.0 VTOL board is manually tuned is usually done by a skilled operator. This process of tuning is time-consuming and requires an expert’s knowledge. Although PID control of various sys-tems has been reported in the literature, its use is limited in nonlinear systems. For nonlinear systems. Fuzzy logic is suitable due to its nonlinearity capability. The purpose of this research is to study the dynamics of the QNET 2.0 VTOL model, simulate the flight control model in Lab-VIEW and to design an auto-tuned PID controller using Fuzzy logic for QNET 2.0 VTOL model in LabVIEW environment. This study shows that Fuzzy based auto-tuned PID controller controls the pitch angle of the QNET 2.0 VTOL model and gives promising results as compared to the existing PID controller in terms of auto-tuning in real-time and stability of the system. },
issn = {2074-9007},
keywords = {Vertical take-off and landing (VTOL); QNET 2.0 VTOL; Pitch angle; Fuzzy based PID controller; Rule base; fuzzy algorithm },
language = {English},
publisher = {MECS}
}

Unmanned aerial vehicles (UAVs) have gained a lot of attention from researchers due to their hovering and vertical take-off and landing. Different techniques and methods are being employed to imple-ment UAVs. The QNET 2.0 VTOL board, specially de-signed for NI ELVIS II, is an important platform in the field of unmanned aerial vehicles (UAV). It is a helpful tool to demonstrate the essentials of vertical take-off and landing flight control (VTOL) at educational institutes. The PID controller installed in QNET 2.0 VTOL board is manually tuned is usually done by a skilled operator. This process of tuning is time-consuming and requires an expert’s knowledge. Although PID control of various sys-tems has been reported in the literature, its use is limited in nonlinear systems. For nonlinear systems. Fuzzy logic is suitable due to its nonlinearity capability. The purpose of this research is to study the dynamics of the QNET 2.0 VTOL model, simulate the flight control model in Lab-VIEW and to design an auto-tuned PID controller using Fuzzy logic for QNET 2.0 VTOL model in LabVIEW environment. This study shows that Fuzzy based auto-tuned PID controller controls the pitch angle of the QNET 2.0 VTOL model and gives promising results as compared to the existing PID controller in terms of auto-tuning in real-time and stability of the system.

@article{palma_2020,
title = {H2 control and filtering of discrete-time LPV systems exploring statistical information of the time-varying parameters},
author = {Palma, J.M.; Morais, C.F.; Oliveira, R.C.L.F.},
journal = {Journal of the Franklin Institute},
year = {2020},
institution = {University of Campinas, Brazil},
abstract = {This paper introduces a new strategy to improve performance in gain-scheduled control and filtering for LPV systems exploiting statistical information about the time-varying parameters whenever available. The novelty of the technique, named sub-domain optimization heuristic (SDOH), is to design controllers or filters treating robust stability independently of performance. The performance is optimized only in a sub-domain of the time-varying parameters, where a higher frequency of occurrence is expected, while the robust stability is certificated for the whole domain. The problem of gain-scheduled design subject to inexact measurements is discussed in details as main motivation but any other feedback or filter strategy for LPV systems were statistical information about the time-varying parameters is known can be handled in a similar way. Still in the context of inexact measurements, a more complete modeling for the additive uncertainty is given, generalizing previous results from the literature for two types of uncertainties, polytopic and affine. A new design condition for full-order LPV filtering is also given as contribution. Several numerical examples are presented to illustrate the results. },
language = {English},
publisher = {Elsevier B.V.}
}

This paper introduces a new strategy to improve performance in gain-scheduled control and filtering for LPV systems exploiting statistical information about the time-varying parameters whenever available. The novelty of the technique, named sub-domain optimization heuristic (SDOH), is to design controllers or filters treating robust stability independently of performance. The performance is optimized only in a sub-domain of the time-varying parameters, where a higher frequency of occurrence is expected, while the robust stability is certificated for the whole domain. The problem of gain-scheduled design subject to inexact measurements is discussed in details as main motivation but any other feedback or filter strategy for LPV systems were statistical information about the time-varying parameters is known can be handled in a similar way. Still in the context of inexact measurements, a more complete modeling for the additive uncertainty is given, generalizing previous results from the literature for two types of uncertainties, polytopic and affine. A new design condition for full-order LPV filtering is also given as contribution. Several numerical examples are presented to illustrate the results.

@article{tran_2019,
title = {Implementing Homomorphic Encryption Based Secure Feedback Control for Physical Systems},
author = {Tran, .; Farokhi, F.; Cantoni, M.; Shames, I.},
journal = {Control Engineering Practice},
year = {2020},
month = {04},
volume = {97},
institution = {University of Melbourne, Australia},
abstract = {This paper is about an encryption based approach to the secure implementation of feedback controllers for physical systems. Specifically, Paillier's homomorphic encryption is used in a custom digital implementation of a class of linear dynamic controllers, including static gain as a special case. The implementation is amenable to Field Programmable Gate Array (FPGA) realization. Experimental results, including timing analysis and resource usage characteristics for different encryption key lengths, are presented for the realization of a controller for an inverted pendulum; as this is an unstable plant, the control is necessarily fast. },
keywords = {secure control, homomorphic encryption, Paillier encryption, digital design, FPGA},
language = {English},
publisher = {Elsevier B.V.}
}

This paper is about an encryption based approach to the secure implementation of feedback controllers for physical systems. Specifically, Paillier's homomorphic encryption is used in a custom digital implementation of a class of linear dynamic controllers, including static gain as a special case. The implementation is amenable to Field Programmable Gate Array (FPGA) realization. Experimental results, including timing analysis and resource usage characteristics for different encryption key lengths, are presented for the realization of a controller for an inverted pendulum; as this is an unstable plant, the control is necessarily fast.

@article{shen_2020,
title = {Improving Tracking Performance of Nonlinear Uncertain Bilateral Teleoperation Systems with Time-Varying Delays and Disturbances},
author = {Shen, H.; Pan, Y.-J.},
journal = {IEEE/ASME Transactions on Mechatronics},
year = {2020},
institution = { Dalhousie University, Canada},
abstract = {In this paper, an adaptive non-singular terminal sliding mode (ANTSM) method is proposed for the motion tracking control of a bilateral teleoperation system. Efforts in this paper seek to improve the position tracking performance of nonlinear systems subject to time-varying network delays, parametric uncertainties, and unknown external disturbances and frictions. Another issue addressed in this paper is the common delay-induced phase shift of tracking profiles in many control methods, which is greatly reduced by introducing a novel mixed-type of feedback signals in the ANTSM control design. Furthermore, the proposed adaptive control design with two online-estimated compensatory bounds removes the requirement of exact knowledge of network delays and disturbance bounds as a prior. In the master side, a force predictor is used to estimate the current environmental force for the reference signal generator. Therefore, the direct transmission of force signals is avoided. By comparing with the existing model-based and model-free methods, numerical simulation results with six degrees of freedom (6 DOFs) manipulators illustrate the merits of the developed robust and adaptive controllers. Experimental results with two Phantom Omni devices are also provided to demonstrate the effectiveness and the significant performance improvements of the proposed controllers. },
issn = {1941-014X },
keywords = {Nonlinear Bilateral Teleoperation, Non-Singular Terminal Sliding Mode, Adaptive Control, Time-Varying Network Delays, Phase Shift, Mixed-Type Feedback},
language = {English},
publisher = {IEEE}
}

In this paper, an adaptive non-singular terminal sliding mode (ANTSM) method is proposed for the motion tracking control of a bilateral teleoperation system. Efforts in this paper seek to improve the position tracking performance of nonlinear systems subject to time-varying network delays, parametric uncertainties, and unknown external disturbances and frictions. Another issue addressed in this paper is the common delay-induced phase shift of tracking profiles in many control methods, which is greatly reduced by introducing a novel mixed-type of feedback signals in the ANTSM control design. Furthermore, the proposed adaptive control design with two online-estimated compensatory bounds removes the requirement of exact knowledge of network delays and disturbance bounds as a prior. In the master side, a force predictor is used to estimate the current environmental force for the reference signal generator. Therefore, the direct transmission of force signals is avoided. By comparing with the existing model-based and model-free methods, numerical simulation results with six degrees of freedom (6 DOFs) manipulators illustrate the merits of the developed robust and adaptive controllers. Experimental results with two Phantom Omni devices are also provided to demonstrate the effectiveness and the significant performance improvements of the proposed controllers.

@conference{mantegh_2020,
title = {Integrated Collision Avoidance for Unmanned Aircraft Systems with Harmonic Potential Field and Haptic Input},
author = {Mantegh, I.; Liao, J.; He, T. },
booktitle = {2020 IEEE/SICE International Symposium on System Integration (SII)},
year = {2020},
institution = {National Research Council of Canada; University of Toronto Institute for Aerospace Studies (UTIAS), Canada; Concordia University,Canada},
abstract = {he growing interest and trend for deploying unmanned aircraft systems (UAS) in civil applications requires robust traffic management approaches that can safely integrate the unmanned platforms into the airspace. Although there have been significant advances in autonomous navigation, especially in the ground vehicles domain, there are still challenges to address for navigation in the dynamic 3D environment that airspace presents. An integrated approach that facilitates semi-autonomous operations in dynamic environments and also allows for operators to stay in the loop for intervention may provide a workable and practical solution for UAS integration in the airspace. In this paper, a new integrated approach is presented for the problem of collision avoidance that applies a variant of potential field method along with haptic-based tele-operation for added safety in high-risk situations. },
issn = {2474-2317 },
keywords = {System integration},
language = {English},
publisher = {IEEE},
isbn = {978-1-7281-6668-1}
}

he growing interest and trend for deploying unmanned aircraft systems (UAS) in civil applications requires robust traffic management approaches that can safely integrate the unmanned platforms into the airspace. Although there have been significant advances in autonomous navigation, especially in the ground vehicles domain, there are still challenges to address for navigation in the dynamic 3D environment that airspace presents. An integrated approach that facilitates semi-autonomous operations in dynamic environments and also allows for operators to stay in the loop for intervention may provide a workable and practical solution for UAS integration in the airspace. In this paper, a new integrated approach is presented for the problem of collision avoidance that applies a variant of potential field method along with haptic-based tele-operation for added safety in high-risk situations.

@article{ren_2020,
title = {Learning inverse kinematics and dynamics of a robotic manipulator using generative adversarial networks},
author = {Ren, H.; Ben-Tzvi, P.},
journal = {Robotics and Autonomous Systems},
year = {2020},
month = {02},
volume = {124},
institution = {Virginia Tech, USA},
abstract = {Obtaining inverse kinematics and dynamics of a robotic manipulator is often crucial for robot control. Analytical models are typically used to approximate real robot systems, and various controllers have been designed on top of the analytical model to compensate for the approximation error. Recently, machine learning techniques have been developed for error compensation, resulting in better performance. Unfortunately, combining a learned compensator with an analytical model makes the designed controller redundant and computationally expensive. Also, general machine learning techniques require a lot of data to perform the training process and approximation, especially in solving high dimensional problems. As a result, state-of-the-art machine learning applications are either expensive in terms of computation and data collection, or limited to a local approximation for a specific task or routine. In order to address the high dimensionality problem in learning inverse kinematics and dynamics, as well as to make the training process more data efficient, this paper presents a novel approach using a series of modified Generative Adversarial Networks (GANs). Namely, we use Conditional GANs (CGANs), Least Squares GANs (LSGANs), Bidirectional GANs (BiGANs) and Dual GANs(DualGANs). We trained and tested the proposed methods using real-world data collected from two types of robotic manipulators, a MICO robotic manipulator and a Fetch robotic manipulator. The data input to the GANs was obtained using a sampling method applied to the real data. The proposed approach enables approximating the real model using limited data without compromising the performance and accuracy. The proposed methods were tested in real-world experiments using unseen trajectories to validate the “learned” approximate inverse kinematics and inverse dynamics as well as to demonstrate the capability and effectiveness of the proposed algorithm over existing analytical models. },
keywords = {Inverse kinematics, Inverse dynamics, Generative adversarial networks},
language = {English},
publisher = {Elsevier B.V.}
}

Obtaining inverse kinematics and dynamics of a robotic manipulator is often crucial for robot control. Analytical models are typically used to approximate real robot systems, and various controllers have been designed on top of the analytical model to compensate for the approximation error. Recently, machine learning techniques have been developed for error compensation, resulting in better performance. Unfortunately, combining a learned compensator with an analytical model makes the designed controller redundant and computationally expensive. Also, general machine learning techniques require a lot of data to perform the training process and approximation, especially in solving high dimensional problems. As a result, state-of-the-art machine learning applications are either expensive in terms of computation and data collection, or limited to a local approximation for a specific task or routine. In order to address the high dimensionality problem in learning inverse kinematics and dynamics, as well as to make the training process more data efficient, this paper presents a novel approach using a series of modified Generative Adversarial Networks (GANs). Namely, we use Conditional GANs (CGANs), Least Squares GANs (LSGANs), Bidirectional GANs (BiGANs) and Dual GANs(DualGANs). We trained and tested the proposed methods using real-world data collected from two types of robotic manipulators, a MICO robotic manipulator and a Fetch robotic manipulator. The data input to the GANs was obtained using a sampling method applied to the real data. The proposed approach enables approximating the real model using limited data without compromising the performance and accuracy. The proposed methods were tested in real-world experiments using unseen trajectories to validate the “learned” approximate inverse kinematics and inverse dynamics as well as to demonstrate the capability and effectiveness of the proposed algorithm over existing analytical models.

@article{wiley_2020,
title = {Magneto-rheological actuator with permanent magnets for low-power activation},
author = {Wiley, B.; Gurocak, H.},
journal = {Journal of Intelligent Material Systems and Structures},
year = {2020},
institution = {Washington State University, USA},
abstract = {This research presents a new magneto-rheological brake that uses only permanent magnets to produce variable torque output. A new magnet array was developed that can be operated by a tiny motor allowing the device to apply 1.8 Nm torque using only 1.4 W of electrical power. In many applications, it is desirable to have actuators with high torque output, low volume, and low input power. In recent years, this has led to extensive study of magneto-rheological brakes since these devices can provide high torque-to-volume ratios when compared to electric motors of similar size. There have been many studies to optimize volume and torque, which usually leads to designs requiring complex internal magnetic flux paths and excessive power requirement resulting in joule heating and reduced performance in long-term use. Our brake can operate for 22 h using a small battery while cycling through full on/off torque output. It can also apply full torque output indefinitely with no power input required to keep it on. There is no joule heating since there are no coils in the design. },
issn = {1530-8138},
keywords = {Magneto-rheological brake, magneto-rheological fluid, MRF, MRB, permanent magnet array, Halbach array, haptics, robotics},
language = {English},
publisher = {SAGE Publications}
}

This research presents a new magneto-rheological brake that uses only permanent magnets to produce variable torque output. A new magnet array was developed that can be operated by a tiny motor allowing the device to apply 1.8 Nm torque using only 1.4 W of electrical power. In many applications, it is desirable to have actuators with high torque output, low volume, and low input power. In recent years, this has led to extensive study of magneto-rheological brakes since these devices can provide high torque-to-volume ratios when compared to electric motors of similar size. There have been many studies to optimize volume and torque, which usually leads to designs requiring complex internal magnetic flux paths and excessive power requirement resulting in joule heating and reduced performance in long-term use. Our brake can operate for 22 h using a small battery while cycling through full on/off torque output. It can also apply full torque output indefinitely with no power input required to keep it on. There is no joule heating since there are no coils in the design.

@article{wu_2020,
title = {Memristor Hardware-Friendly Reinforcement Learning},
author = {Wu, N.; Vincent,A.; Strukov, D.; Xie,Y.},
journal = {arXiv},
year = {2020},
abstract = {Recently, significant progress has been made in solving sophisticated problems among various domains by using reinforcement learning (RL), which allows machines or agents to learn from interactions with environments rather than explicit supervision. As the end of Moore's law seems to be imminent, emerging technologies that enable high performance neuromorphic hardware systems are attracting increasing attention. Namely, neuromorphic architectures that leverage memristors, the programmable and nonvolatile two-terminal devices, as synaptic weights in hardware neural networks, are candidates of choice to realize such highly energy-efficient and complex nervous systems. However, one of the challenges for memristive hardware with integrated learning capabilities is prohibitively large number of write cycles that might be required during learning process, and this situation is even exacerbated under RL situations. In this work we propose a memristive neuromorphic hardware implementation for the actor-critic algorithm in RL. By introducing a two-fold training procedure (i.e., ex-situ pre-training and in-situ re-training) and several training techniques, the number of weight updates can be significantly reduced and thus it will be suitable for efficient in-situ learning implementations. As a case study, we consider the task of balancing an inverted pendulum, a classical problem in both RL and control theory. We believe that this study shows the promise of using memristor-based hardware neural networks for handling complex tasks through in-situ reinforcement learning. },
keywords = {Artificial neural networks, Reinforcement learning, Memristor, ReRAM, In-situ training, Hardware implementation, Actor-Critic},
language = {English}
}

Recently, significant progress has been made in solving sophisticated problems among various domains by using reinforcement learning (RL), which allows machines or agents to learn from interactions with environments rather than explicit supervision. As the end of Moore's law seems to be imminent, emerging technologies that enable high performance neuromorphic hardware systems are attracting increasing attention. Namely, neuromorphic architectures that leverage memristors, the programmable and nonvolatile two-terminal devices, as synaptic weights in hardware neural networks, are candidates of choice to realize such highly energy-efficient and complex nervous systems. However, one of the challenges for memristive hardware with integrated learning capabilities is prohibitively large number of write cycles that might be required during learning process, and this situation is even exacerbated under RL situations. In this work we propose a memristive neuromorphic hardware implementation for the actor-critic algorithm in RL. By introducing a two-fold training procedure (i.e., ex-situ pre-training and in-situ re-training) and several training techniques, the number of weight updates can be significantly reduced and thus it will be suitable for efficient in-situ learning implementations. As a case study, we consider the task of balancing an inverted pendulum, a classical problem in both RL and control theory. We believe that this study shows the promise of using memristor-based hardware neural networks for handling complex tasks through in-situ reinforcement learning.

@article{tao_2020,
title = {Modeling and Emulating a Physiotherapist’s Role in Robot‐Assisted Rehabilitation},
author = {Tao, R.; Ocampo, R.; Fong, J.; Soleymani, A.; Tavakoli, M.},
journal = {Advanced Intelligent Systems},
year = {2020},
institution = {University of Alberta, Canada},
abstract = {In home‐based rehabilitation, one possible approach is haptic teleoperation in which a hospital‐based therapist is haptically linked and tele‐presented to a home‐based patient to effectively simulate traditional in‐hospital therapies over a distance. In this context, this paper proposes a learn‐and‐replay (LAR) paradigm that consists of two phases: A therapist‐in‐loop (interactive) phase where the therapist interacts through the haptic teleoperation loop with the patient to perform one or more repetitions of the cooperative therapy task, and a therapist‐out‐of‐loop (standalone) phase where the therapist’s cooperative role of the task is played by the patient‐side robot in future repetitions. During the interactive phase, the therapist demonstrates impedance during cooperating with the patient in performing the task. During the standalone phase, the patient‐side robot is automatically controlled to mimic the therapist’s demonstrated impedance which is learned in the interactive phase. The Direct Force Reflection (DFR) architecture is utilized as the control method for the bilateral telerehabilitation system. Case studies involving one degree‐of‐freedom and two‐degree‐of‐freedom cooperative manipulation tasks are tested for proof of concept. In the experiment, the therapists and patients were simulated by using healthy participants (the authors of the paper) and were located in the same room to show a proof‐of‐concept idea of the work. The results show that the impedance of the therapist’s arm can be replicated by the patient‐side robot for both tasks and proposed LAR telerehabilitation paradigm assists the therapist in the rehabilitation procedure to takecare of other tasks or attend to other patients. },
keywords = {rehabilitation robotics, haptic teleoperation, impedance estimation},
language = {English},
publisher = {John Wiley & Sons, Inc.}
}

In home‐based rehabilitation, one possible approach is haptic teleoperation in which a hospital‐based therapist is haptically linked and tele‐presented to a home‐based patient to effectively simulate traditional in‐hospital therapies over a distance. In this context, this paper proposes a learn‐and‐replay (LAR) paradigm that consists of two phases: A therapist‐in‐loop (interactive) phase where the therapist interacts through the haptic teleoperation loop with the patient to perform one or more repetitions of the cooperative therapy task, and a therapist‐out‐of‐loop (standalone) phase where the therapist’s cooperative role of the task is played by the patient‐side robot in future repetitions. During the interactive phase, the therapist demonstrates impedance during cooperating with the patient in performing the task. During the standalone phase, the patient‐side robot is automatically controlled to mimic the therapist’s demonstrated impedance which is learned in the interactive phase. The Direct Force Reflection (DFR) architecture is utilized as the control method for the bilateral telerehabilitation system. Case studies involving one degree‐of‐freedom and two‐degree‐of‐freedom cooperative manipulation tasks are tested for proof of concept. In the experiment, the therapists and patients were simulated by using healthy participants (the authors of the paper) and were located in the same room to show a proof‐of‐concept idea of the work. The results show that the impedance of the therapist’s arm can be replicated by the patient‐side robot for both tasks and proposed LAR telerehabilitation paradigm assists the therapist in the rehabilitation procedure to takecare of other tasks or attend to other patients.

@article{razmi_2020,
title = {Near-Optimal Neural-Network Robot Control with Adaptive Gravity Compensation},
author = {Razmi, M.; Macnab, C.J.B.},
journal = {Neurocomputing},
year = {2020},
institution = {University of Calgary, Canada},
abstract = {Adaptive nonlinear optimal control methods, as proposed in the literature, give rise to some questions around practical implementation in robotics, especially how to find a solution in a reasonable time and how to deal with gravity. This paper proposes a method to solve these problems by using a neural network with local basis-function domains, specifically the Cerebellar Model Articulation Controller (CMAC). The algorithm uses the local domains in order to keep track of the value of local cost-functionals. In this way, it can freeze the learning of the network’s weights in a feedforward-like component in the CMAC when the bias has been overcome identified by using an error-based cost-functional e.g. automatic gravity compensation in a robot. After the feedforward component has been established, the algorithm then starts to learn another set of weights which contribute to feedback-like terms in the CMAC and these weights get frozen when they no longer reduce a cost-functional that includes additional control effort e.g. in a robot the control effort beyond that needed to compensate for gravity is penalized. Lyapunov methods guarantee uniformly ultimately bounded signals and ensure weight drift and bursting do not occur. One advantage is that the training time for finding a near-optimal control does not increase over previous neural-adaptive methods. Another advantage is that penalizing the control effort in a search for optimization does result in any steady-state error due to gravity. Simulations show that the proposed method significantly outperforms a standard adaptive-CMAC control using e-modification, without increasing control effort or training time. An experimental flexible-joint robot verifies that the adaptive method significantly outperforms a linear quadratic regulator. },
keywords = {nonlinear optimal control direct adaptive control neural-adaptive control overlearning weight drift bursting Cerebellar Model Articulation Controller elastic-joint robot},
language = {English},
publisher = {Elsevier B.V.}
}

Adaptive nonlinear optimal control methods, as proposed in the literature, give rise to some questions around practical implementation in robotics, especially how to find a solution in a reasonable time and how to deal with gravity. This paper proposes a method to solve these problems by using a neural network with local basis-function domains, specifically the Cerebellar Model Articulation Controller (CMAC). The algorithm uses the local domains in order to keep track of the value of local cost-functionals. In this way, it can freeze the learning of the network’s weights in a feedforward-like component in the CMAC when the bias has been overcome identified by using an error-based cost-functional e.g. automatic gravity compensation in a robot. After the feedforward component has been established, the algorithm then starts to learn another set of weights which contribute to feedback-like terms in the CMAC and these weights get frozen when they no longer reduce a cost-functional that includes additional control effort e.g. in a robot the control effort beyond that needed to compensate for gravity is penalized. Lyapunov methods guarantee uniformly ultimately bounded signals and ensure weight drift and bursting do not occur. One advantage is that the training time for finding a near-optimal control does not increase over previous neural-adaptive methods. Another advantage is that penalizing the control effort in a search for optimization does result in any steady-state error due to gravity. Simulations show that the proposed method significantly outperforms a standard adaptive-CMAC control using e-modification, without increasing control effort or training time. An experimental flexible-joint robot verifies that the adaptive method significantly outperforms a linear quadratic regulator.

@article{ma_2020,
title = {Observer integrated backstepping control for a ball and plate system},
author = {Ma, J.; Tao, H.; Huang, J.},
journal = {International Journal of Dynamics and Control},
year = {2020},
institution = {Beijing University of Chemical Technology, China},
abstract = {An observer integrated backstepping control is designed for a ball and plate system with the characters of cascade and model uncertainties. The model of the system is written in the cascaded strict feedback form. The uncertainties of the model and derivatives of the virtual controls are estimated by a linear extended state observer and a tracking differentiator, respectively. The convergence of the derivative estimation is studied. The actual control and virtual controls are designed based on the cascaded model with the Lyapunov theory. The stability of the closed-loop system is consequently proven. Simulations and experimental results demonstrate the excellent tracking performance of the control design. },
keywords = {Ball and plate system, backstepping, Linear extended state observer (LESO), Tracking differentiator (TD)},
language = {English},
publisher = {Springer Nature Switzerland}
}

An observer integrated backstepping control is designed for a ball and plate system with the characters of cascade and model uncertainties. The model of the system is written in the cascaded strict feedback form. The uncertainties of the model and derivatives of the virtual controls are estimated by a linear extended state observer and a tracking differentiator, respectively. The convergence of the derivative estimation is studied. The actual control and virtual controls are designed based on the cascaded model with the Lyapunov theory. The stability of the closed-loop system is consequently proven. Simulations and experimental results demonstrate the excellent tracking performance of the control design.

@article{ariba_2020,
title = {Performance analysis for time-delay systems and application to the control of an active mass damper},
author = { Ariba, Y.; Gouaisbaut, F.},
year = {2020},
institution = {LAAS-MAC, France},
abstract = {This paper proposes a method to analyze, beyond stability, the performances of linear time-delay systems. Using robust analysis techniques, a sufficient condition that analyzes the location of eigenvalues in the complex plane is presented. More precisely, a set of quadratic inequality constraints are designed to define an admissible region for the infinitely many eigenvalues of a time-delay system and the quadratic separation theorem is applied to assess that the eigenvalues are effectively belonging to that stability region. This method is then used for the control of an active mass damper. A standard state feedback control is replaced with a static output feedback plus a static delayed output feedback. This strategy avoids the full measurement of the state and shows that delays in the dynamic may be helpful for stabilization. The closed-loop system is then expressed as a time-delay system and the performance criterion is exploited to analyze the stability and the damping properties. Simulations and experimental tests support the approach. }
}

This paper proposes a method to analyze, beyond stability, the performances of linear time-delay systems. Using robust analysis techniques, a sufficient condition that analyzes the location of eigenvalues in the complex plane is presented. More precisely, a set of quadratic inequality constraints are designed to define an admissible region for the infinitely many eigenvalues of a time-delay system and the quadratic separation theorem is applied to assess that the eigenvalues are effectively belonging to that stability region. This method is then used for the control of an active mass damper. A standard state feedback control is replaced with a static output feedback plus a static delayed output feedback. This strategy avoids the full measurement of the state and shows that delays in the dynamic may be helpful for stabilization. The closed-loop system is then expressed as a time-delay system and the performance criterion is exploited to analyze the stability and the damping properties. Simulations and experimental tests support the approach.