Three-dimensional multi-finger caging for capturing convex objects in the space

Abstract

Multi-finger caging offers an adaptive and robust object grasping approach, but caging objects in the space is complicated and difficult due to the complex relative motion between the object and the multi-fingered mechanism, and the high degrees of freedom of fingers. In this paper, an integrative three-dimensional multi-finger caging algorithm for tumbling targets in the space is presented. The main core of the proposed algorithm is the “leader-follower” mode: (1) in order to realize the synchronous motion of the tumbling target, the leading finger of the multi-fingered mechanism is selected and its base joint is controlled to track the caging point of the target using the model predictive control strategy; (2) to simplify the degrees of freedom, the leading finger adjusts its joint angles to resemble its shape as closely as possible with the caging edge based on One Way Distance, then possible configurations of other following fingers are determined according to structural constraints of the multi-fingered mechanism; and (3) the effective caging configuration is determined according to the derived multi-finger caging conditions. The main advantages of this work are (1) the simple task mode – “leader-follower” mode can significantly simplify the calculation burden of caging configuration design; (2) the adaptivity and robustness for tumbling targets with different shapes benefited from the flexibility of the caging configuration. The proposed algorithm is applied in the capture of a dysfunctional satellite with solar panels and a discarded upper stage rocket, respectively. The simulation results show that the proposed algorithm can find the effective caging configuration efficiently.