Lessons Learned from the Animate Agent Project (so far)

This paper gives a brief overview of the Animate Agent Project at the University of Chicago, describing skills, RAPs, and a recently added spatial planning module. The RAP system takes in symbolic goals and refines them into sequences of continuous skills that move the robot in real-time. The RAP system uses the spatial module as an assistant when spatial problems arise during goal refinement. As an example, the skills, RAPs, and spatial planning actions for a simple trash cleanup task are also briefly described. The paper finishes with answers to the seven symposium questions. The Animate Agent Project The Animate Agent Project at the University of Chicago is aimed at building robotic and software systems for carrying out a variety of everyday goals in realistic, everyday environments (Firby &: Swain 1991). A major goal of the project is to create a software architecture that allows existing plans and behaviors to be used as building blocks for additional tasks. The architecture currently being used consists of three major components: the RAP reactive plan execution system, a modular, skill-based control system, and a spatial planning subsystem incorporating a simple path planner. Both the control system and the spatial planning system are essentially under the control of the RAP system.