Many of today's robotic work cells are unable to detect when an action failure has occurred. This results in faulty products being sent down the line, and/or downtime for the cell as failures are detected and corrected. This article examines a novel knowledge-driven system that provides added agility by detecting and correcting action failures. The system also provides for late binding of action parameters, thus providing flexibility by allowing plans to adapt to changing environmental conditions. The key feature of this system is its knowledge base that contains the necessary relationships and representations to allow for failure detection and correction. This article presents the ontology that stores this knowledge as well as the overall system architecture. The manufacturing domain of kit construction is examined as a sample test environment.
Proceedings Title: Proceedings of the 2nd International Coonference on Robot Intelligence Technology and Application 2013
Conference Dates: December 18-20, 2013
Conference Location: Denver, CO
Pub Type: Conferences
failure detection, manufacturing, ontology, robotics, PDDL