Abstract
This paper presents a method of manipulating contact states for robotic assembly based on a multi-agent system. Each of agents has a model of local contact states between corners of assembly objects. The model is represented by 3-DOF contact state transitions network and has motion constraints in each contact states. Based on this model, individually, each of agents senses contact state and suggests robot motion for next contact state and motion constraints. Suggestions are integrated and the robot motion is determined to satisfy motion constrains. This method is applied in simulations about 4 kinds of insertion task; single-peg insertion, dual-peg insertion, L-shaped-peg insertion, and T-shaped-peg insertion. The results are presented to show the effectiveness of this method.
