A 3D perception-based robotic manipulation system for automated truck unloading

Abstract

This paper presents an autonomous robotic manipulation system that uses 3D vision processing algorithms to locate and unload cardboard boxes from floor-loaded shipping containers or semi-truck trailers. A key novel aspect of this system is the proposed 3D box detection algorithm as well as multiple box manipulation (or gripping) strategies that can be intelligently determined on the fly right after the box has been detected. With the approach, the robotic manipulator can unload product with no-prior knowledge of a loading pattern and can avoid collisions even in loads with piled-up boxes and no clear geometric structure of the loading environment.