Robot independent assembly sequence plannin

This thesis presents an extensive blueprint for the architecture of a system which solves the robot independent assembly sequence planning problem. The system uses a decompositional approach to methodically convert CAD information about a set of parts into a robust assembly sequence when no a priori information is known about the assembling robot. The refining process converts the part interrelationships into a parts graph (an analog of the exploded-view diagram). The parts graph data is then reduced into a set of candidate assembly sequences for which it is geometrically possible to perform all the operations. These candidate sequences are arranged in a graph form so that common-sense metrics can be used to weight each sequence for later selection by a graph searching algorithm. The three main metrics proposed for use are the stability of subassemblies, the robustness of grasp sites, and a new metric, the size of the gripper envelope. The gripper envelope is analyzed by sweeping a plane along the direction of part mating to determine the effects of obstacles in the assembly environment. Experimental results indicate the robustness and applicability of the gripper envelope metric to the rating of assembly sequences when no a priori information is known about the assembling robot.