Base pose estimation using shared reachability maps for manipulation tasks

Manipulation tasks for robots with a robotic arm and a mobile base are one of the most error-prone operations in service robotics. Especially for low-cost robots with a small reachable workspace of the arms, the selection of base poses for reaching tasks is a challenging problem. In this paper, we focus on the estimation of base poses for robots such that they are able to reach a given end effector target pose using space-efficient discretizations of the robots workspace, so-called reachability maps. We further present a method to share these reachability maps among robots by storing them on a cloud-database, annotated with a semantic description. These are then integrated in a knowledge-based task execution system for service robots. Last, we conclude with the results of an experiment showing the cooperation of robots of different hardware making use of each other's reachability maps.