A next generation expert system for flexible assembly

An attempt is made to demonstrate the feasibility of an engineering approach to machine learning through the development of a robotic system that learns to perform assembly tasks from clearly specified training instances. A major technological impediment to increasing productivity in the manufacturing sector is the relative absence of automation at most stages of the product design process. One such stage involves the translation of nonprocedural goal files-describing the desired result of some aspect of the manufacturing process-into procedural task files. The task files detail the exact sequence of required steps as typified by an assembly operation. Flexibility is introduced with an expert translator. The technique provides for coordination among the assembly robots and minimizes costly down-time. The system has been implemented in C and is supported by the Sun 3/50 platform. An advanced prototype has been implemented at the Naval Command, Control, and Ocean Surveillance Center on a DAP-610 supercomputer