Provably Asymptotically Near-Optimal Motion Planning with Sparse Data Structures
Loading...
Authors
Dobson, Andrew J.
Issue Date
2012
Type
Thesis
Language
Keywords
Asymptotic , Asymptotically optimal , Motion Planning , Optimal , PRM , resource constrained
Alternative Title
Abstract
Asymptotically optimal planners, such as PRM*, guarantee thatsolutions approach optimal as iterations increase. Roadmaps with this property, however, may grow too large. If optimality is relaxed,asymptotically near-optimal solutions produce sparser graphs by notincluding all edges. The idea stems from graph spanner algorithms,which produce sparse subgraphs that guarantee near-optimal paths.Existing asymptotically optimal and near-optimal planners, however,include all sampled configurations as roadmap nodes. Consequently, only infinite graphs have the desired properties. This work proposes an approach that provides the following asymptotic properties: (a) completeness, (b) near-optimality and (c) the probability of adding nodes to the spanner roadmap converges to zero as iterations increase. Thus, the method suggests that finite-size data structures might have near-optimality properties. The method brings together ideas from various planners but deviates from existing integrations of PRM* with graph spanners. Simulations for rigid bodies show that the method indeed provides small roadmaps and results in faster query resolution. The rate of node addition is shown to decrease over time and the quality of solutions satisfies the theoretical bounds. Smoothing provides a more favorable comparison against alternatives with regards to path length.
Description
Citation
Publisher
License
In Copyright(All Rights Reserved)