改进的PSO-BP算法在工业机器人末端位姿误差补偿中的应用

Application of Improved PSO-BP Algorithm in the Compensation of End-pose Error of Industrial Robot

提出了一种全梯度标准粒子群优化反馈(FGSPSO-BP)神经网络的工业机器人末端位姿补偿模型.首先,提出一种运动学逆变换算法,通过机器人末端位姿对机器人各关节角度值进行计算,并采用Matlab验证了运动学逆变换算法的准确性.然后,提出一种基于全梯度下降法的FGSPSO-BP算法,将机器人实际末端位姿参数作为输入样本,实际位姿与理想位姿的各关节角度值之差作为输出样本,对网络进行训练,以得到机器人实际末端位姿参数与各关节角度值差的关系,采用测试样本对网络模型算法进行了验证.最后,利用新松机器人所采集的实际位姿和理想位姿数据,通过神经网络的方法,实现了对机器人各实际关节角度值的补偿,使机器人达到了理想的末端位置与姿态.实验结果表明,相对于传统的PSO-BP与YSPSO-BP神经网络,FGSPSO-BP神经网络对于机器人末端位姿误差的补偿精度更高,稳定性更好.

We propose an end-pose compensation model for the industrial robot based on a full-gradient standard particle swarm optimization-back propagation(FGSPSO-BP)neural network.First,we propose an inverse kinematics transformation algorithm that calculates the angle of each joint of the robot based on its end-pose,the accuracy of which is verified by Matlab.Then,we propose an FGSPSO-BP algorithm based on the full-gradient descent method.The actual end-pose parameters of the robot are used as input samples,and the difference between the actual and ideal poses of each joint angle is used as an output sample to train the network.To obtain the relationship between the actual end-pose parameters of the robot and the angle difference of each joint,the performance of the network model algorithm was verified based on test samples.Finally,using actual-and ideal-pose data collected by the Siasun robot,we use the neural network method to realize compensation of the actual joint angle values of the robot,to enable the robot to reach the ideal end position and posture.The experimental results show that compared with the traditional PSO-BP and YSPSO-BP neural networks,the proposed FGSPSO-BP neural network obtains higher compensation accuracy and better stability for robot end-pose errors.