The Library
Enhancing vehicle destination prediction using latent trajectory information
Tools
Tools
Tools
Van Hinsbergh, James (2022) Enhancing vehicle destination prediction using latent trajectory information. PhD thesis, University of Warwick.
|
PDF
WRAP_THESIS_VanHinsbergh_2022.pdf - Submitted Version - Requires a PDF viewer. Download (25Mb) | Preview |
Official URL: http://webcat.warwick.ac.uk/record=b3894860
Abstract
Intelligent transportation systems have the potential to provide road users with a range of useful applications, including vehicle preconditioning, traffic flow management and intelligent parking recommendations. The majority of these applications can benefit from knowledge of vehicle activities (common situations that a vehicle encounters e.g. traffic), along with the upcoming destinations that a vehicle will visit. We focus on the trajectories that vehicles provide, and the data contained within them, in order to ascertain information about the patterns in individuals' mobility data.
Machine learning has been used in many different vehicle applications, and we focus on using these techniques to predict the activity of a vehicle and its future destinations. Clustering methods can be applied at the level of trajectories or the individual instances within them, and we explore both of these alternatives in this thesis. Additionally, we explore several classification approaches to predict activities and destinations. In developing our methods, we make use of a combination of both geospatial and temporal data along with on-board vehicle sensor data.
This thesis presents novel methods for filtering stay points to identify points of interest and applying destination prediction to vehicle trajectories. Existing methods for stay point detection are not specific to vehicles, and therefore any region of low mobility is potentially considered to be of interest. We propose a novel method for filtering the extracted stay points to identify points of interest, using vehicle data to predict vehicle activities. The predicted activities are further used to represent trajectories as sequences of annotated locations, to inform the detection of similarities between journeys. Finally, this thesis presents a novel method for using additional properties of a trajectory to cluster trajectories into groupings of similar trajectories with the aim of improving the accuracy of destination prediction. We evaluate our proposed methods on a set of vehicle datasets, varying in purpose and the data available.
| Item Type: | Thesis (PhD) | ||||
|---|---|---|---|---|---|
| Subjects: | G Geography. Anthropology. Recreation > GA Mathematical geography. Cartography Q Science > QA Mathematics > QA76 Electronic computers. Computer science. Computer software T Technology > TA Engineering (General). Civil engineering (General) T Technology > TE Highway engineering. Roads and pavements T Technology > TL Motor vehicles. Aeronautics. Astronautics |
||||
| Library of Congress Subject Headings (LCSH): | Intelligent transportation systems, Machine learning, Automated vehicles -- Data processing, Trajectories (Mechanics), Trip generation, Geospatial data, Geographic information systems, Location-based services | ||||
| Official Date: | April 2022 | ||||
| Dates: |
|
||||
| Institution: | University of Warwick | ||||
| Theses Department: | Department of Computer Science | ||||
| Thesis Type: | PhD | ||||
| Publication Status: | Unpublished | ||||
| Supervisor(s)/Advisor: | Griffiths, Nathan | ||||
| Sponsors: | Jaguar Land Rover (Firm) ; Engineering and Physical Sciences Research Council | ||||
| Format of File: | |||||
| Extent: | vi, 176 pages : illustrations, maps | ||||
| Language: | eng |
Request changes or add full text files to a record
Repository staff actions (login required)
 |
View Item |
