Cornell University
Challenges in Continuous Path Planning: Rarefactions, Uncertainty and Reinforcement Learning
Abstract
dc:description.abstractWe consider three optimal control problems, which focus on continuouspath-planning applications, and each problem deals with a specific challenge. First, we introduce a new local factoring technique which can remove numerical artifacts arising from Eikonal equations with non-smooth conditions. Next, we deal with path-planning under an initial uncertainty, which can be removed later at some certainty time, and further discuss methods suitable for different notions of optimality. The third section considers an online-learning path-planning problem with unknown surveillance intensity and develops a Bayesian reinforcement learning method. In addition to the three path-planning problems, a new method of parameterizing neural networks is introduced. It follows the continuous optimal control interpretation of deep learning and uses B-spline basis functions to parameterize. For each problem we use numerical experiments to show the advantages of our proposed methods.
Degree
thesis:*- Name thesis:degree_name
- Ph. D., Applied Mathematics
- Level thesis:degree_level
- Doctor of Philosophy
- Discipline thesis:degree_discipline
- Applied Mathematics
- Grantor
- Cornell University
- Year dc:date.issued
- 2023
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Qi, Dongping
- Chair dc:contributor.chair
-
- Vladimirsky, Alexander
- Committee members dc:contributor.committeemember
-
- Bindel, David
- Damle, Anil
Subjects
dc:subject × 6Rights
- Language dc:language.iso
- en
Identifiers
dc:identifier.*- Dc Identifier Bibid
- 16219366
- Dc Identifier Other
-
ProQuest Submission ID: 13778
ProQuest Publication ID: 30569263