Iretiayo Akinola, Jingxi Xu, and Peter Allen

Columbia Robotics Lab


Grasping in dynamic environments presents a unique set of challenges. A stable and feasible grasp can become infeasible as the object moves; this problem becomes pronounced when there are obstacles in the scene. One common approach is to switch between a set of pre-planned grasps but it usually results in large swinging arm trajectory motions as the closest grasp in the list may require a complete change in arm motion. This can result in failure especially if the objects is moving relatively fast. In this work, we tackle some of these challenges by developing a workspace-aware online grasping system that is able to plan and update grasps in near real-time as the object moves. Our approach jointly optimizes for stability and reachability of the grasp. It takes a grasp from a previous time-step as a seed and optimizes quickly to find a stable and reachable grasp in the neighboorhood. Using this formulation, we show that our method is able to find stable grasps for fast moving objects and also in cases where the target moves within static obstacles.