基于BP神经网络的机器人容错控制

Fault Tolerance Control Robot Manipulators Based on BP Neural Network

提出一种基于神经网络的机器人故障检测系统结构,给出在线学习逼近机器人标称模型的故障检测方法。控制系统可运行于两种工作模式:系统正常时,采用常规计算力矩控制器;有故障发生时,引入补偿控制来削弱故障影响,保障系统的正常运行。对机器人的仿真控制结果表明,设计的控制策略在机器人发生故障时,不仅能够继续完成预定工作任务而且还保证了跟踪精度,表明了该方案的有效性。

A fault detection and fault tolerant control algorithm for robot is presented. The control system can run in two models： common computed torque controller is in use when system is normal, otherwise, compensative control is introduced to reduce the effects of faults and guarantee the performance of system, when faults occur. The control strategy presented above can not only complete the scheduled missions but also keep a highly tracking accuracy.