Adaptive Dynamic Programming-Based Decentralized Guaranteed Cost Control for Reconfigurable Manipulators with Uncertain
- PDF / 3,052,322 Bytes
- 16 Pages / 595.276 x 790.866 pts Page_size
- 111 Downloads / 258 Views
ORIGINAL ARTICLE
Adaptive Dynamic Programming‑Based Decentralized Guaranteed Cost Control for Reconfigurable Manipulators with Uncertain Environments Yuanchun Li1 · Weinign Jin1 · Bing Ma1 · Bo Dong1 Received: 1 April 2020 / Revised: 1 June 2020 / Accepted: 28 July 2020 © The Korean Institute of Electrical Engineers 2020
Abstract In this paper, a novel decentralized guaranteed cost control method is designed for reconfigurable manipulators with uncertain environments based on the adaptive dynamic programming (ADP) approach. Each joint module, which is the basic unit for constructing the reconfigurable manipulators, is regarded as a subsystem with model uncertainties that include the error of frictional modeling and the interconnection dynamic coupling (IDC) effect. Then, by employing a robust controller and a neural network (NN) identifier-based compensation controller, the decentralized guaranteed cost control issue with uncertain environments can be changed into the optimal control issue of reconfigurable manipulators. Based on ADP algorithm, the critic neural network is introduced to approximate the modified cost function, and then the Hamilton–Jacobi–Bellman equation is addressed by the policy iterative algorithm, thus making the obtention of approximate optimal control policy doable. The stability of the robotic system under the proposed control policy is demonstrated by employing the Lyapunov theory. Finally, the effectiveness of the proposed control policy for reconfigurable manipulators with different configurations is verified by simulation experiments. Keywords Reconfigurable manipulators · Adaptive dynamic programming (ADP) · Decentralized guaranteed cost control · Policy iterative · Critic neural network
1 Introduction Compared with the traditional industrial manipulators, which are designed with the characteristic of large size, single task and fixed degree of freedom, reconfigurable manipulators [1] are composed of different number of joint modules and link modules, and these modules have standard electrical interface and communication interface. Reconfigurable manipulators can change the configuration and degree of * Bo Dong [email protected] Yuanchun Li [email protected] Weinign Jin [email protected] Bing Ma [email protected] 1
Department of Control Science and Engineering, Changchun Univerity of Technology, Changchun, China
freedom according to the requirements of different tasks. At present, reconfigurable manipulators have become a research focus and have a wide application prospects [2–5], such as deep space exploration, emergency relief, medical service, etc. In these dangerous and complex uncertain environments, the reconfigurable manipulators may contact or collide with unknown object or obstacle. Consequently, it is essential to design proper control systems to ensure the stability and accuracy of the robot when operating special tasks. The reconfigurable manipulator systems with uncertain environments, which is essentially a class of complex nonlinear system, can introduce
Data Loading...
