基于鲁棒学习算法的水下机器人神经网络控制  

Neural network control of underwater vehicles based on robust learning algorithm

作　　者：梁霄 甘永 李晔[1] 孙玉山[1] 方少吉[1] 

机构地区：[1]哈尔滨工程大学船舶工程学院,黑龙江哈尔滨150001

出　　处：《哈尔滨工程大学学报》2006年第B07期74-78,共5页Journal of Harbin Engineering University

基　　金：国家“863”计划基金资助项目（2002AA420090）.

摘　　要：针对水下机器人神经网络控制系统响应速度慢及对噪声较敏感的问题，依据变结构控制理论，结合误差反向传播学习算法，推导出一种新颖的强鲁棒性学习算法，并详细讨论其全局稳定性条件，最后在水下综合探测机器人仿真平台上进行了试验研究．结果表明，控制器对学习率的改变和外界扰动有很强的鲁棒性，大大降低了机器人机械传动系统的磨损，且能够保证神经网络快速、稳定地学习，从而满足实时性控制的要求，具有较高的理论和实用价值．Aiming at low response speed and sensitization to noises in control system of underwater vehicles by adopting neural network, a novel robust learning algorithm is deduced based on variable structure control theory and error back propagation algorithm, and the global stability conditions are discussed in details. Finally, simulation experiments are carried out on General Detection Remotely Operated Vehicle. The results show that it has good robustness to external disturbance and changing of learning-ratio, which reduces the abrasion of the mechanically-driven system greatly, and it keeps learning of neural network fast and stably, which meets the requirement of real-time control and has theoretical and practical value.

关 键 词：水下机器人 神经网络 鲁棒学习算法 变结构控制理论 全局稳定性 

分 类 号：TP24[自动化与计算机技术—检测技术与自动化装置]