Planning, Training and Learning in Supervision of Flexible Assembly Systems

Abstract

In the context of balanced automation systems, a generic architecture for evolutive supervision of robotized assembly tasks is presented. This architecture provides, at various abstraction levels, functions for dispatching actions, execution monitoring, and diagnosing and recovering from failures. A planning strategy and domain knowledge for nominal plan execution and error recovery is described. Through the use of machine learning techniques, the supervision architecture will be given capabilities to improve its performance over time. The participation of humans in the training and supervision activities is considered essential. The combination of human interactivity with automatic aspects (planning, learning,..) is discussed.