Textureless object recognition and arm planning for a mobile manipulator

Manipulator robots require of comprehensible perception systems that permit them to interact with the environment, localize target objects and perform complex manipulation tasks. To do that, a full-step manipulation process has to be performed involving recognition of supporting planes, object clustering, its recognition and finally smooth arm trajectory planning with obstacle avoidance. The goal of this paper is to propose a general and practical strategy for designing a 3D perception system architecture that can robustly plan for the location of texture and untextured objects for the purpose of grasping. Our proposed system includes RGB-D calibrated system to acquire 3D color scenario data, a 7-DOF arm controlled with a PMAC 2-PCI and a precise arm-end grasp. A validation of the overall proposed idea with some experimental results and a comparison of the benefits with similar initiatives is presented.