An Architecture for Intelligent Task Automation

This report discusses the Martin Marietta Intelligent Task Automation Project (ITA). The purpose of the ITA project is to integrate Artificial Intelligence (AI) task planning, path planning, vision, and robotics technologies into a system designed to autonomously perform manufacturing tasks in dynamic or unstructured environments. The application domain chosen for primary demonstrations is dimensional measurement of an F-l 5 bulkhead. The overall goal is to be able to perform the inspection an order of magnitude faster than the current manual method, which takes about 24 hours for about 1000 inspection points. The project was conducted in two phases. Phase I, completed in December 1984, demonstrated the readiness of the technologies in each of the areas making up the ITA system. Phase II, which was mostly complete in June 1987, demonstrated that the technologies can be integrated into a working system and that the system can be transferred to other applications. The architecture of the ITA system is discussed with an emphasis on the AI components making up the system. The strengths and weaknesses of the architecture and AI techniques applied are discussed.