Abstract: 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.