This grant is for continued development of Dr. Kuipers' approach to exploratory map construction and qualitative robotic navigation. Current simulation studies will be extended to control of a momentum-dominated robot in an unknown and dynamically changing 2-D environment. The perceptual approach is to model topological features of the robot's world before extracting error-prone estimates of geometric features. A major theme will be recognition and resolution of topological ambiguities, with modeling of both parametric and qualitative changes in world states. Another theme is smooth switching between anticipatory control strategies in a momentum-dominated world.
