Online optimal consensus control based on PI algorithm for nonlinear multi-agent system with completely unknown dynamics

摘要

In this paper, the optimal control of continuous-time nonlinear multi-agent systems with completely unknown dynamic is addressed via adaptive dynamic programming (ADP) technique. A NN identifier with filter variables is used to estimate the unknown dynamic system, and policy iteration (PI) algorithm is employed to get the optimal policies. In the implement of the algorithm, a NN based on data sampling and least-square method is adopted to approximate the value function weights, which can avoid solving the value function. Moreover, the two NNs work simultaneously, that is to say the optimal control policies can be calculated online and the actor NN can be removed. In particular, compared with the existing literatures, the state derivative is not necessary in the sampling process, and the implementation of the algorithm has enhanced. Finally, a simulation is given to illustrate the effectiveness and the whole process of the algorithm.