A robot collision detection method based on VAE-LSTM

Abstract

The application provides a robot collision detection method based on VAE-LSTM, which does not need additional external sensors, and only uses the runtime sequence data collected by the robot itself for training and detection, effectively reducing the system cost. The method collects joint operation data during the movement of the robot in real time, and performs data preprocessing, learns the complex mode of joint torque change by using a VAE-LSTM hybrid model, predicts the reconstruction window features of the next sample by using the original window features of the current sample, and finally determines whether the robot collides by comparing the model prediction result score with a threshold value. The method can accurately identify the collision event and output the robot stop motion signal in time, and can be widely applied to industrial production environments, and effectively improves the safety of the robot work.

Images