Abstract: Robot milling is an important method of large complex curved surface processing, however, because of the industrial robot structure and components manufacturing, installation error, the absolute accuracy of its trajectory is low in the process of large stroke movement, significantly restrict the application conditions range of milling. The current sensing and control methods of robot precision mainly focus on the prediction of positioning error based on vision and the measurement of trajectory error based on laser tracker, the former is difficult to account for milling track errors, while the latter is operation complex and extremely expensive. Thus, a method of tool end position error calculation and machining error compensation in robotic milling is proposed, and the precision prediction and compensation of robot milling errors are realized. By training the BP neural network optimized by particle swarm optimization (PSO), the position error prediction model of robot machining tool TCP is established, the position error iteration model is established based on the truncation method, and the comprehensive compensation strategy of trajectory error is formulated. The experimental results show that the maximum cutting error of robotic milling is reduced from 1.354 mm to 0.244 mm, which provides a theoretical and technical basis for the extension of robotic milling conditions.

Key words: robot milling, vision tracking measurement, machining error, comprehensive compensation, BP neural network