Abstract
This article presents a novel fault-tolerant position-force optimal control method for constrained reconfigurable manipulators with uncertain actuator failures. On the basis of the radial basis function neural network (RBFNN)-estimated manipulators dynamics, the proposed force-position error fusion function and the estimated actuator failure are utilized to construct an improved optimal performance index function, which reflects the faults and optimizes system comprehensive performance as well as the energy consumption simultaneously. Based on the policy iteration (PI) scheme and the adaptive dynamic programming (ADP) algorithm, the Hamiltonian-Jacobi-Bellman (HJB) equation is solved by constructing the critic neural network (NN), and then the approximated fault-tolerant position-force optimal control policy can be derived correspondingly. The closed-loop manipulator system is proved to be asymptotically stable by using the Lyapunov theory. Finally, simulations are provided to demonstrate the effectiveness of the proposed method.
Highlights
The replacement of manual production by manipulators is a significant development trend of digitalization, industrial automation, and intelligence
For the unknown dynamics of reconfigurable manipulators, Zhou et al [20] proposed a force/position fault-tolerant control method of constrained reconfigurable manipulators, which consisted of a modified sliding mode controller to ensure the force/position tracking performance and a radial basis function neural network (RBFNN)-based compensation controller to increase the robustness of the manipulator system
Dong et al [37] concentrated on studying the decentralized optimal control method for the reconfigurable manipulators, designing a model-based compensation controller and an adaptive dynamic programming (ADP)-based optimal controller to deal with the influence of the unknown internal dynamics and the interconnected dynamic coupling, respectively
Summary
The replacement of manual production by manipulators is a significant development trend of digitalization, industrial automation, and intelligence. For the unknown dynamics of reconfigurable manipulators, Zhou et al [20] proposed a force/position fault-tolerant control method of constrained reconfigurable manipulators, which consisted of a modified sliding mode controller to ensure the force/position tracking performance and a RBFNN-based compensation controller to increase the robustness of the manipulator system All these methods mentioned above ignored the problem of the comprehensive optimization of the control performance and power consumption in the event of actuator failures. Dong et al [37] concentrated on studying the decentralized optimal control method for the reconfigurable manipulators, designing a model-based compensation controller and an ADP-based optimal controller to deal with the influence of the unknown internal dynamics and the interconnected dynamic coupling, respectively These ADP-based optimal control methods focused on solving the position tracking problem of the complex nonlinear systems, while there were only a few investigations addressing the problems of actuator failures with reinforcement learning theory of the constrained manipulator systems.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have