Event-triggered receding horizon control via actor-critic design

Abstract

In this paper, we propose a novel event-triggered near-optimal control for nonlinear continuoustime systems. The receding horizon principle is utilized to improve the system robustness and obtain better dynamic control performance. In the proposed structure, we first decompose the infinite horizon optimal control into a series of finite horizon optimal problems. Then a learning strategy is adopted, in which an actor network is employed to approximate the cost function and an critic network is used to learn the optimal control law in each finite horizon. Furthermore, in order to reduce the computational cost and transmission cost, an event-triggered strategy is applied. We design an adaptive trigger condition, so that the signal transmissions and controller updates are conducted in an aperiodic way. Detailed stability analysis shows that the nonlinear system with the developed event-triggered optimal control policy is asymptotically stable. Simulation results on a single-link robot arm with different noise types have demonstrated the effectiveness of the proposed method.

Performance‐guaranteed containment control for pure‐feedback multi agent systems via reinforcement learning algorithm

“…Remark Different from the existing optimal methods, such as References 15‐23,44‐47, the updating laws of actor and critic NN are obtained with complex calculation through the stability analysis and the gradient descent algorithm of Bellman error square. However, this article utilizes a simplified RL algorithm, where the actor and critic updating laws can be constructed based on the uniqueness of the optimal solution.…”

“…For the past few years, as one of the most effective and popular online learning methods, actor-critic RL was extensively applied in the research field of optimal control. [15][16][17][18][19][20][21][22][23] By utilizing a modified cost function, an optimal robust control strategy was proposed for a class of uncertain nonlinear systems. 19 In Reference 22, based on RL algorithm and improved Hamilton-Jacobi-Bellman (HJB) function, the finite-time optimal control problem was addressed for uncertain nonlinear systems with dead-zone input.…”

“…Based on actor‐critic structure and neural networks (NNs) or fuzzy logic systems, an approximate optimal cost function and an optimal controller are obtained through RL algorithm. For the past few years, as one of the most effective and popular online learning methods, actor‐critic RL was extensively applied in the research field of optimal control 15‐23 . By utilizing a modified cost function, an optimal robust control strategy was proposed for a class of uncertain nonlinear systems 19 .…”

“…Notably, in most existing containment researches, such as References 1‐11, only the realization of containment goal is considered, and the optimal control problem is ignored. In addition, in contrast with References 15‐23,41‐47, the optimal control design in this article is predigested via a simpler learning mechanism to construct the actor and critic updating laws instead of gradient descent algorithm. (2) By constructing observer and filter, the assumption that the states are measurable can be removed and the algebraic loop problem in pure‐feedback MASs is resolved.…”

Simplified prescribed performance tracking control of uncertain nonlinear systems

“…Based on the method given in [4][5][6], the neural network approximation technology and the disturbance observer method can be used to estimate f (x(t)) − fr(xr(t)) and D(t) with bounded estimation errors. The estimation of unknown nonlinear function is defined as Ŵaσ(V x).…”