Data-Driven Model of Off-Line Collision Safety Based on Neural Network for Cooperative Robot

In the application of industrial production, it is crucial for cooperative robots to avoid collision and retain safety. This paper presents an off-line model to assess the level of collision risk based on BP neural network. A detection device was designed to collect the required data composed of force and displacement, while the risk level was assessed according to the relevant standard with collision force and the calculated collision power. The collected data was used for training of neural network and the off-line model is built. The preliminary experiment is carried out, while the performance and the feasible are verified by the results.