School of Computer Science, Carnegie Mellon University
To appear in the Ninth International Workshop on the Principles of Diagnosis (DX98)
We present an implementation of our situation-dependent learning approach in a real robotic system, ROGUE. ROGUE learns situation-dependent costs for arcs in a topological map of the environment; these costs are then used by the path planner to predict and avoid failures. In this article, we present the representation of the path planner and the navigation modules, and describe the execution trace. We show how training data is extracted from the execution trace. We present experimental results from a simulated, controlled environment as well as from data collected from the actual robot. Our approach effectively refines models of dynamic systems and improves the efficiency of generated plans.