Sensorless Collision Detection for Robots Based on Load Torque Observer

This paper proposes a sensorless collision detection strategy for robots based on load torque online observation of drive motors in joints. This proposed strategy utilizes a model reference adaptive system to identify the inertia and viscosity coefficient of the drive motor. With these identified motor parameters, a state observer can be designed to observe the load torque of drive motor. The differential of observed load torque is utilized to detect collision and it will be processed to avoid misjudgements. Once the processed differential value crosses the predetermined threshold value, it is considered a collision happens. Experimental results demonstrate that this strategy can ensure expected performance of collision detection and protect the safety of operators effectively.