神经网络与计算力矩复合的机器人运动轨迹跟踪控制

Tracking Control of Robot Using Hybrid Controller Based on Neural Network and Computed Torque

为了实现机器人精密运动控制,在其关节系统引入计算力矩法(CTC)与神经网络复合的控制器,旨在通过CTC实现系统的初步控制并利用神经网络补偿机器人的不确定动力学特性所带来的运动误差。首先,建立了机器人的动力学模型并对其不确定性动力学量进行了描述;然后,为机器人构建了双闭环控制系统,并依据机器人标称模型规划出CTC控制律;进而,引入函数链神经网络(FLNN)对不确定性动力学量进行估值,并推导出FLNN的学习律;最后,对系统进行了仿真,结果显示,该复合控制器可将关节位置和速度跟踪误差控制在±0.001 rad和±0.001 rad/s之内,且其对机器人的参数变化及外部扰动具有较强的自适应性与鲁棒性。

In order to improve robot manipulator＇s tracking accuracy,a hybrid controller consisting of a functional link neural network sub-controller（FLNNC） and a computed torque sub-controller（CTC） was introduced into the manipulator,which made use of CTC to drive the manipulator reaching its desired position roughly while employed the FLNNC to compensate the tracking error caused by the dynamic uncertainty and disturbance of the robot.To accomplish this,firstly,a nominal dynamic model of the manipulator was established,and the dynamic uncertainty of the robot manipulator was modeled and formulized.And then,a control system with two close loops was built for the manipulator,and the computed-torque control law based on the nominal manipulator model was planned for the system.Moreover,a functional link neural network（FLNN） being capable of approximating the dynamic uncertainty term of the robot was designed in the system,and the weight learning algorithm for the FLNN was derived.Finally,simulations were made on that system so as to validate the hybrid controller.The results showed that both the position error and speed tracking error of the robot joints could be controlled within ±0.001 rad and±0.001 rad/s,which meant that the proposed hybrid controller was able to make the robots tracking desired trajectory with high precision.