论文部分内容阅读
针对含有建模误差和不确定干扰的关节机器人轨迹快速跟踪控制,提出了一种改进神经网络PID控制器的设计方法。该方法采用了双控制器鲁棒控制,神经网络通过学习PID的输入输出特性,快速补偿关节机器人系统的建模误差和不确定干扰,而利用最小二乘法和收敛后的神经网络输入输出特性优化PID控制参数,能够削弱建模误差对控制效果的干扰。控制器在李雅普诺夫意义下是稳定的。以两关节机械臂为被控对象进行了仿真实验,实验结果表明改进控制器的优越性。
Aiming at the fast tracking control of joint robot with modeling errors and uncertainties, a design method of improved neural network PID controller is proposed. This method uses dual controller robust control. Neural network can quickly compensate for the modeling errors and uncertainties of the joint robot system by learning the input and output characteristics of the PID. By using the least square method and the optimized input and output characteristics of the neural network, PID control parameters, can weaken the modeling error on the control effect of interference. The controller is stable in the Lyapunov sense. The simulation experiment was carried out using the two articulated robots as the controlled object. The experimental results show that the superiority of the controller is improved.