Automated assembly in the presence of significant system errors

A primary source of difficulty in automated assembly is the uncertainty in the relative position of the parts being assembled. A logic branching approach to solving this problem is discussed. Force sensor information, responses to recent moves, and results from previous assemblies are used to generate the branching decisions. Several heuristic assembly algorithms are presented. The proposed approach generates efficient compliant motion strategies for any set of hard, smooth parts that can be modeled as a peg and hole. Two of the algorithms converge to acceptable performance levels in less than 100 assembly trials. This implies that a real assembly cell using these algorithms would converge quickly enough for the learning to be done online. This would eliminate the modeling errors introduced by learning with an assembly simulator. Logic branching is compared with other machine learning and expert system techniques