A Framework for Guided Motion Planning

Randomized sampling based algorithms are widely used in robot motion planningdue to the problem's intractability, and are experimentally effective on a widerange of problem instances. Most variants bias their sampling using variousheuristics related to the known underlying structure of the search space. Inthis work, we formalize the intuitive notion of guided search by defining theconcept of a guiding space. This new language encapsulates many seeminglydistinct prior methods under the same framework, and allows us to reason aboutguidance, a previously obscured core contribution of different algorithms. Wesuggest an information theoretic method to evaluate guidance, whichexperimentally matches intuition when tested on known algorithms in a varietyof environments. The language and evaluation of guidance suggests improvementsto existing methods, and allows for simple hybrid algorithms that combineguidance from multiple sources.