Problem decomposition for assembly planning

Automatic planning of an assembly task suitable for execution by a robot system requires analysis of constraints arising from task and object geometry as well as constraints arising from the specific configuration of the robot system. This paper describes our current efforts toward building a high-level planner that plans robot assembly tasks while confining its attention to constraints arising from task and object characteristics. The robot itself is modelled in terms of its basic capabilities such as, grasp, free motion and certain fine-motion commands. The planner works from a specification of the assembly operations, the objects involved in the operations, and an incomplete geometric model of the objects. The assembly planning problem is viewed as comprising of the determination of positions and orientations of component objects relative to each other and the determination of the robot motion primitives required to achieve these relations. The paper discusses how we decompose the assembly planning problem to reflect this view, the intermediate representations used, and our current implementations by use of an example assembly task.