Georgia Institute of Technology
Distributed Filtering and Smoothing with High-Dimensional Data
Abstract
dc:description.abstractDistributed state estimation is a difficult and fundamental problem that is commonly a requirement for realizing higher-level autonomous functionality, such as navigation and control. Differing from single-agent state estimation, distributed state estimation strategies benefit from communication and inter-robot perception in order to share or reduce uncertainty amongst team members. Central to state estimation strategies are the use of probabilistic representations describing both the best state estimate available and the uncertainty associated with the estimate as a probability density function (PDF). Such strategies are often split into two estimation objectives, filtering and smoothing, which aim to find the best estimate for the state given either all information up to the current time-step or all information up to a future time-step. This dissertation focuses on developing tools for distributed state estimation under the context of particle-based methods, which represent uncertainty as a set of weighted candidate states. While many distributed extensions of the particle filter and particle smoother have been proposed, these strategies must circumvent the use of a pointwise product between particle-based PDFs as products between particle distributions are ill-defined. This lack of a well-defined product is one of the major drawbacks to distributed particle-based methods as the pointwise product operation is commonly used to fuse information from multiple robots. In this Dissertation, we propose a product strategy for particle distributions in order to facilitate distributed filtering and smoothing by leveraging common distributed methods that require a pointwise product. Our strategy is coupled with nonlinear measurements in the form of camera images in order to estimate the hidden state of the agent.
Degree
thesis:*- Level thesis:degree_level
- Doctoral
- Department dc:contributor.department
- Electrical and Computer Engineering
- Grantor dc:publisher
- Georgia Institute of Technology
- Year dc:date.issued
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Lin, Tony Xiaotong
- Advisors dc:contributor.advisor
-
- Zhang, Fumin
- Coogan, Samuel
Subjects
dc:subject × 4Rights
- Language dc:language.iso
- en_US
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/1853/75199