Safety-Enhanced Formation Maneuver Control for Electric Vehicle With Edge-Weighted Topology and Reinforcement Learning Strategy
Bing Huang, Yuzhou Song, Hongde Qin, Jianming Miao, Cheng Zhu
Abstract
Electric Vehicle Fleet (EVF) provides a promising solution for complex missions, while excessive fleet members and obstacle-rich mission environments increase collision probability, thereby threatening system security. Within this context, this paper investigates a collision-free formation maneuver control strategy for EVF. Specifically, an Edge-Weighted Laplacian Matrix (EWLM) is developed to evaluate collision risk and ensure inter-vehicle safety distances. Through obstacle detection, virtual nodes are identified and incorporated into the EWLM. This modification endows each member with obstacle-escaping capabilities in a similar manner. Meanwhile, for inherent hydrodynamic effects and external disturbances, a reinforcement learning echo state network (RL-ESN) is proposed to match uncertainties. Compared with ESN trained by conventional method, RL-ESN provides improved weight convergence rates and more precise uncertainty matching. For convincing, simulation results are presented to demonstrate the superiority of the collision-free control scheme.