Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.771732
Title: Agent-based lost person movement modelling, prediction and search in wilderness
Author: Mohibullah, W.
ISNI:       0000 0004 7659 6066
Awarding Body: UCL (University College London)
Current Institution: University College London (University of London)
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Abstract:
In this research we investigate the problem of searching for a Lost Person (LP) in wilderness using an autonomous Unmanned Aerial Vehicle (UAV). The problem of search with a UAV is often treated as gridded environment search where the state of each grid (cell) is examined individually for the presence or absence of the target. However, this idealised way of search fails to exploit many potentially valuable dependencies and secondary cues - such as material deposited or left by the LP or topographical features such as natural tracks (trails) - which could significantly aid the search process. We discuss the need for such a system and review the current state-of-the-art work. Since key to a quick and successful search is a well defined initial distributions. We further argue the need to generate the initial distribution over the trajectory of the LP, not merely the end location, usually done in literature. We propose a search framework consisting of three key phases: information gathering, initial distribution generation and search. In the information gathering phase, we collect detailed information related to both the LP and the search environment. Then in the initial distribution generation phase, using the information gathered, we generate distribution over the LP's trajectory using particles. Each particle represented by an agent model of LP movement with sampled parameters, navigating and interacting with the environment represented using data-sets in the form of terrain elevation, topography and vegetation. To ensure, the agent model is a good representation of the LP behaviour, we calibrate its parameters using the method called SMC2 . Finally in the Search phase, a UAV is deployed to explore the search area and detect the LP, any evidence features or changes in the environment. All information detected are localised and used to update the distribution over the LP trail until either the LP is located or the search is terminated.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.771732  DOI: Not available
Share: