Javascript must be enabled for the correct page display

Multi-agent survivor search at an earthquake disaster area

Schokker, H. J. (2015) Multi-agent survivor search at an earthquake disaster area. Master's Thesis / Essay, Industrial Engineering and Management.

[img] Text
toestemming.pdf - Other
Restricted to Backend only

Download (21kB)
[img] Text
Master_IEM_2015_HendrikSchokker.pdf - Published Version
Restricted to RUG campus only

Download (6MB)

Abstract

An assignment control for a group of robots (agents) to multiple collapsed buildings (targets) is proposed for search and rescue purposes in earthquake scenarios. The research focuses on two parts, routing and agent-target assignment. The routing and assignment takes place on the mission space which contains agent locations, target locations and obstacles. For the routing part, two algorithms are considered. “ConnectGraph” is proposed for creating movable edges on the mission space and “Dijkstra’s algorithm” is then incorporated for finding the shortest routes from agents to targets. Agent performance is measured with a payoff function that represents the time that can be spent looking for survivors. The global performance is formulated as the sum of all payoffs. A congestion game is formed which allows for the introduction of a potential game. With cardinal potential games the change in payoff for any player switching targets can be expressed with a single function, the potential function. Various algorithms including the BRU (Best-Response-Update) and JSFP (Joint Strategy Fictitious Play) are compared to find the optimum agent-target assignments. For larger scenarios where the number of agents and targets exceed 10 by 10 the JSFP algorithm is most suitable due to computational time and near optimal assignments.

Item Type: Thesis (Master's Thesis / Essay)
Degree programme: Industrial Engineering and Management
Thesis type: Master's Thesis / Essay
Language: English
Date Deposited: 15 Feb 2018 08:03
Last Modified: 15 Feb 2018 08:03
URI: http://fse.studenttheses.ub.rug.nl/id/eprint/12560

Actions (login required)

View Item View Item