Joint Exploration of Local Metrics and Geometry in Sampling-based Planning

Loading...
Thumbnail Image
Degree type
Graduate group
Discipline
Subject
GRASP
Kodlab
Electrical and Computer Engineering
Engineering
Systems Engineering
Funder
Grant number
License
Copyright date
Distributor
Related resources
Author
Pacelli, Vincent
Contributor
Abstract

This thesis addresses how the local geometry of the workspace around a system state can be combined with local metrics describing system dynamics to improve the connectivity of the graph produced by a sampling-based planner over a fixed number of configurations. This development is achieved through generalization of the concept of the local free space to inner products other than the Euclidean inner product. This new structure allows for naturally combining the local free space construction with a locally applicable metric. The combination of the local free space with two specific metrics is explored in this work. The first metric is the quadratic cost-to-go function defined by a linear quadratic regulator, which captures the local behavior of the dynamical system. The second metric is the Mahalanobis distance for a belief state in a belief space planner. Belief space planners reason over distributions of states, called belief states, to include modeled uncertainty in the planning process. The Mahalanobis distances metric for a given belief state can be exploited to include notions of risk in local free space construction. Numerical simulations are provided to demonstrate the improved connectivity of the graph generated by a sampling-based planner using these concepts.

Advisor
Date of degree
2017-08-01
Date Range for Data Collection (Start Date)
Date Range for Data Collection (End Date)
Digital Object Identifier
Series name and number
Volume number
Issue number
Publisher
Publisher DOI
Journal Issue
Comments
Recommended citation
BibTeX: @MastersThesis{pacelli_MScThesis2017, author = {Vincent Pacelli}, title = {Joint Exploration of Local Metrics and Geometry in Sampling-based Planning}, school = {University of Pennsylvania}, year = {2017}, }