Abstract: As the external interference factor in the belt grinding environment, it is one of the key problems to study the anti-jamming ability of hand-eye calibration algorithm in the robot belt grinding system. By analyzing the mathematical model of hand-eye calibration in the robot grinding system and using the calibration tool, the measurement data are obtained quickly. Based on the original singular value decomposition algorithm, an improved scheme based on weight is proposed. After calculating the initial hand-eye relationship and the calibration error for each measurement data and according to the size of the measurement data error to reallocate its corresponding weight, and the hand-eye relationship would be recalculated finally. Through the simulation test, the improved algorithm is less sensitive than the original algorithm to the measurement data, and the calibration accuracy is good. The experimental results show that the improved algorithm is 45.9% less than the original algorithm, and the maximum error is reduced by 24.4% compared with the original algorithm. The improved algorithm is proved by the actual grinding test, and is more suitable for the robot grinding system.

Key words: hand-eye calibration, robot grinding, singular value decomposition, weighted