Abstract:

Aimed to make the robot manipulator tracking desired trajectory with high precision, a hybrid controller composed of a FLNN controller and PD controler was proposed. Firstly, a dynamics model of the manipulator was given and the characteristics of the model were presented. Then, the property of the FLNN to approximate nonlinear function unformly was investigated. After that, a control system with two closed control loop was constructed for the robot manipulator, and the error dynamics equation for the system was obtained, and an uncertainty dynamics function for the manipulator was derived. Then, a FLNN was intrduced into the system to approximate the uncertainty dynamics function. Finally, a FLNN-PD controller based on the FLNN's approximation was designed for the system, and a weight learning algorithm was planned for the FLNN. Simulations were performed on that system so as to validate the proposed controller. The results show that the position tracking error and speed tracking error of the robot joints are controlled within ±0.002rad and ±0.1rad/s respectively which means that the a robot based on the FLNN-PD controller is able to tracking desired trajectory with high precision. 

Key words: robot, functional link neural network(FLNN), weight learning algorithm , PD controller, tracking control

摘要：

为了实现直接驱动机器人精密控制,研究了直接驱动机器人的函数链神经网络-PD（FLNN-PD）控制问题。首先,对机器人动力学模型及FLNN的函数逼近特性进行分析;然后为机器人构建双闭环控制系统,推导出系统的误差动力学方程及不确定性动力学函数;引入FLNN逼近不确定性函数并为系统规划出FLNN-PD控制律;最后进行了仿真控制实验。结果显示,该控制器可使系统转角误差和角速度误差控制在±0.002rad和±0.1rad/s之内,且其对系统的参数变化及外部扰动具有较强的自适应性和鲁棒性。

关键词: 机器人, 函数链神经网络, 权值学习算法, PD控制器, 轨迹跟踪控制