Q-learning based on particle swarm optimization for positioning system of underwater vehicles

The paper presents an intelligent underwater positioning system for remotely operated vehicle (ROV). We used multi-agents reinforcement learning algorithms based on particle swarm optimization fusing signals from ultra-short baseline (USBL) position sonar and pose sensors, so that the USBL can be accelerated and be in-phase with pose sensors. We proposed the frame work of the hardware of the intelligent navigation system, and the multithreading and modularizing software system. Navigation experiment taken in ship model tank indicated the feasibility of the proposed intelligent navigation system.