Retrieving objects from clutter using a mobile robotic manipulator

This paper presents a task and motion planning method for retrieving a target object from clutter using a mobile robotic manipulator. We consider a configuration of objects where the density of the space is high. Thus, the robot may have to relocate some obstacles to retrieve the target without collisions. Our goal is to reduce the time for accomplishing the retrieval task by computing a relocation plan, which is a sequence of obstacles to be relocated, and determining the pose of the mobile base where the plan can be executed successfully. We propose a planning method that aims to minimize the number of obstacles to be relocated. Depending on the poses of the obstacles to be manipulated, the base pose of the robot is determined such that the travel distance of the end-effector manipulating the obstacles is minimized. We test the proposed method in a high-fidelity simulated experiment. Our method is compared to a method finding a relocation plan within a limited range of search space.