The positioning accuracy of the cucumber picking robot manipulator might be influenced seriously due to the slight deviation of the stucture parameters, therefore, the high precision three-coordinate measuring machine Platinum Faro Arm was used to calibrate the stucture parameters of the robot manipulator and the forward kinematic model was constructed based on the modified structure parameters,and the error analysis of the ideal inverse kinematics indicated that the angle error of the waist joint far exceeded the accuracy of the position encoder. So the Levenberg-Marquardt Back Propagation ( LMBP) neural network is proposed to solve the modified inverse kinematics,and the network output combined with the ideal inverse kinematics is used to compensate for the position accuracy. Experimental simulation was carried out for verifying the feasibility of the presented algorithm,and the results showed that the maximum value of angle error was about 0.006rad, which greatly improved the positioning accuracy of the cucumber picking robot manipulator from 10.57 millimeters to 3.77 millimeters.%考虑到黄瓜采摘机械手结构参数的微小偏差可能会对末端定位精度造成较大的影响,因此,利用高精度三坐标测量仪Platinum FaroArm对机械手的结构参数进行了标定,建立了基于修正参数的正运动学模型,在此基础上对理想逆运动学进行误差分析,发现腰关节的角度误差远远大于位置编码器的精度.因此,提出采用LMBP神经网络算法求解修正后的关节角度,并将网络输出与理想逆运动学结合起来,达到补偿机械手定位精度的目的.为了验证算法的可行性,进行了仿真试验,结果表明:LMBP神经网络输出角度误差的最大值约为0.006rad,能将末端位置误差从10.57mm补偿到3.77mm,大大提高了黄瓜采摘机械手的定位精度.
展开▼