Potential Gap
Potential Gap: A Gap-Informed Reactive Policy for Safe Hierarchical Navigation
This project considers the integration of gap-based local navigation methods with artificial potential field (APF) methods to derive a local planning module, called potential gap, for hierarchical navigation systems. Central to the construction of the local planner is the use of sensory-derived local free-space models that detect gaps and use them for the synthesis of the APF. Trajectories derived from the APF are provably collision-free for idealized robot models. The provable property is lost when applied to more realistic models. A set of algorithm modifications correct for these errors and enhance robustness to non-ideal models, in particular a nonholonomic robot model. Integration of the potential gap local planner into a hierarchical navigation system provides the local goals and trajectories needed for collision-free navigation through unknown environments. Monte Carlo experiments in benchmark worlds confirm the asserted safety and robustness properties.
A presentation video is here.