Abstract:

Aiming at the problem of consensus collaborative control of unmanned underwater vehicle (UUV) clusters under adverse communication conditions, considering the adverse communication characteristics of high latency, low bandwidth, and the need for stealth in underwater inter-cluster communication, a reinforcement learning based event-triggered intelligent consensus collaborative control architecture is designed to achieve effective collaboration of UUV clusters under adverse communication conditions. Firstly, an event-triggered distributed observer is devised which utilizes dynamic interaction information between leaders and neighbors to estimate the tracking reference signal required for UUVs under adverse communication conditions. Subsequently, reinforcement learning methods are used to directly learn the optimal control strategy from system interactions. Finally, the effectiveness of the proposed method is validated through simulation results.

Key words: adverse communication condition, event-triggered, consensus cooperative control, reinforce-ment learning