Adaptive collision detection algorithm design for six-axis industrial manipulator with accurate payload estimation

Abstract

Collision detection capabilities of industrial manipulator is essential for ensuring the human safety during human machine collaboration. The payload variations will significantly decrease the detection accuracy and sensitivity. However, difficulty in parameter tuning of time-variant threshold and time-consuming training procedure make existed collision detection methods not easily generalized to industrial manipulator systems with payload variation. This paper presents an adaptive collision detection algorithm for six-axis industrial manipulator with accurate payload estimation. Specifically, the unknown payload is estimated online through a generalized momentum-based indirect adaptation law. Compared to existed adaptive collision detection algorithms, the joint acceleration is not needed in the adaptation law, which makes the algorithm easy to be applied to industrial manipulator systems. Based on the estimated payload, a generalized momentum external force observer is developed to estimate the external collision force. Finally, the collision detection can be realized by comparing the estimated collision force with constant threshold values. Comparative simulations and experiments indicate that the proposed adaptive collision detection method can achieve accurate and robust detection under different loads.