Uncertainty-Aware Arm-Base Coordinated Grasping Strategies for Mobile Manipulation

Dong Chen, Ziyuan Liu, Georg von Wichert

Abstract: The ability to reliably perform grasping operations is key for successful applications of mobile manipulation robots. While robots robustly perform in controlled environment like factories, humans still significantly outperform robots in unconstrained environments. This is particularly true when it comes grasping. Human grasping is faster and tremendously more robust, especially in the presence of significant uncertainty. We aim at improving this situation and propose two major building blocks. Firstly, we consider how to effectively use the mobility of the robot base. Secondly, we show an approach to effectively handle grasping uncertainty. In this paper, we introduce a general system architecture for mobile manipulators to execute grasping tasks. This architecture allows a mobile manipulator to employ arm-base coordinated motions during grasping. The architecture also supports the active handling of uncertainty by means of adaptive grasp strategies. To purposefully handle uncertainty we propose two versatile grasping strategies. Small uncertainty can be directly handled by a rapid grabbing strategy, while large uncertainty can be handled by means of pre-grasp manipulation. The targeted selection of the strategy for a specific case takes both grasp success probability and execution time into consideration. We evaluate our approach on a real robot to show that our approach is feasible in real applications, and that it outperforms a traditional grasping procedure.