Smart placement of a two-arm assembly for an everyday object manipulation humanoid robot based on capability maps

Abstract

A robot that will execute everyday object manip- ulation tasks needs a competent body that can handle as many different objects as possible in as many ways as possible. To accomplish this, we must design the robot body in such a way that will allow it to achieve many different ways to manipulate the objects. In this work, we present a design method to mount a two-arm assembly in a torso and a mobile base (to complete a humanoid robot) taking into account a voxelized structure of points to reach and a variable orientation and relative position of both arms. We make use of the concept of capability maps to calculate a score for different possible two-arm assembly placements. An object to manipulate is selected. In particular, we are using the object size as the most important manipulation feature. We test many possible arm mounting configurations with many different object positions and we run an object reaching simulation for each of these combinations. The successfulness of each of these trails is used to derive a capability map. A score for each map (that corresponds to a different arm mounting configuration) is derived and used to select the final placement of the robot. With this work, we present an effective process to determine the best arms bases placement accord to the collaborative reaching results for a two-arm humanoid robot.