Adaptive Fixed-Time Optimal Formation Control for Uncertain Nonlinear Multiagent Systems Using Reinforcement Learning
Ping Wang, Chengpu Yu, Maolong Lv, Jinde Cao
Abstract
This article explores the application of reinforcement learning (RL) strategy to achieve an adaptive fixed-time (FxT) optimized formation control of uncertain nonlinear multiagent systems. The primary obstacle in this process is the difficulty in attaining FxT stability under the actor-critic setting due to intermediate estimation errors and generic system uncertainties. To overcome these challenges, the RL control algorithm is implemented using an identifier-actor-critic structure, where the identifier is utilized to address the system uncertainty involving unknown nonlinear dynamics and external disturbances. Furthermore, a novel quadratic function is introduced to establish the boundedness of the estimation error of the actor-critic learning law, which plays a pivotal role in the FxT stability analysis. Finally, a unified FxT optimized formation control strategy is developed, which guarantees the realization of the predetermined formation at a fixed time while optimizing the given performance measure. The effectiveness of the proposed control algorithm is verified through simulation of a team of marine surface vessels.