基于粒子群优化的多智能体强化学习算法在水下机器人定位系统中的应用  

The Application of Multi-agent Reinforcement Learning Algorithm Based on Particle Swarm Optimization in Positioning System of Underwater Vehicle

作　　者：高延增[1] 叶家玮[1] 龚幼[1] 

机构地区：[1]华南理工大学船舶与海洋工程系,广东广州510641

出　　处：《机床与液压》2010年第19期64-67,71,共5页Machine Tool & Hydraulics

摘　　要：提出一种融合多传感器的超小型远程操纵型水下机器人(ROV)快速定位系统。利用基于粒子群优化的多智能体强化学习算法融合水下机器人超短基线定位声纳及位姿传感器信号,可提高ROV的定位速度。给出ROV定位系统试验平台硬件结构以及多线程、模块化的软件系统搭建方案。在船模实验池中进行ROV的定位验证,结果表明该算法满足ROV对定位速度的要求。An intelligent underwater positioning system for remote operated vehicle（ROV） was presented.Multi-agent reinforcement learning algorithm based on particle swarm optimization was used to fuse signals from ultra-short baseline（USBL） position sonar and pose sensors,so that positioning velocity of ROV was quickened.The hardware frame of the intelligent navigation system and the multithreading and modularizing software system were proposed.The results of navigation experiment made in ship model pool indicate the proposed intelligent navigation system is feasible.

关 键 词：水下机器人 定位系统 Q学习 粒子群优化 

分 类 号：TP24[自动化与计算机技术—检测技术与自动化装置]