Collision detection method for industrial robots based on multi-channel information fusion

Collision detection is an important part of ensuring its safety and reliability, and it is a low-cost and low-requirement collision detection method to detect and identify collision events based on the data of sensors or monitoring systems inside the robot. In view of the shortcomings of the existing deep learning methods in processing the robot motion state data, this paper constructs a CA-Transformer network with multi-channel information fusion, takes the real-time load as the judgment index, realizes the fast and effective prediction of the robot load, and integrates a variety of sensor data in turn to improve the analysis ability of the robot state. The research results can provide ideas for the safe operation detection of robots.