Files in this item



application/pdfECE499-Sp2013-chang.pdf (2MB)Restricted to U of Illinois
(no description provided)PDF


Title:Development and Analysis of Adversarial Agent Control Algorithms in Mobile Sensor Networks
Author(s):Chang, Jerry
Contributor(s):Hutchinson, Seth
Subject(s):mobile sensor networks (MSNs)
adversarial agent control algorithms
adversarial network sensors
Abstract:Mobile sensor networks (MSNs) can be described as a network of sensing units, each of which has the ability to move. Because the sensors are able to move, they can move to improve the coverage of the network dynamically. Much of today's research on MSNs assume optimal conditions such that all of the sensors are functioning properly. There is little research that takes into account the possibility of sensors that function improperly. In the worst case, these adversarial sensors may move in a manner that could reduce the coverage of the network. The research presented in this thesis proposes three algorithms that model adversarial sensor motion. The first proposed algorithm generates a rapidly-exploring random tree (RRT) for each adversarial agent. Each node in an RRT represents the steady-state position of its associated agent. After generating an RRT for each adversarial agent, each sensor is assigned a path from its associated RRT. If each adversarial agent traversed its path, the resulting steady-state configuration would have the maximal cost out of all configurations resulting from all combinations of paths from the RRTs. The second method involves letting the adversarial agents operate under Lloyd's algorithm with respect to each other while the functioning sensors operate normally under Lloyd's algorithm with respect to every sensor. Based on simulations, the steady-state cost is only locally maximal for some of the simulations but not all. While generally an increase in cost will result, if there are a few adversarial agents, then a decrease in cost will actually occur. The third approach seeks to herd the functioning sensors into a minimal area. This is done by gathering the adversarial agents on one side of the area, then synchronously moving them towards the function sensors. Generally, this algorithm generates a steady-state configuration that has a final cost that is larger than costs generated in previous algorithms. However, this algorithm has the least amount of research and still requires more work before it is formalized. Ultimately, these proposed methods are just the early algorithms for adversarial sensor control in MSNs. Further research would involve the development and analysis of other algorithms.
Issue Date:2013-05
Publication Status:unpublished
Peer Reviewed:not peer reviewed
Date Available in IDEALS:2014-03-19

This item appears in the following Collection(s)

Item Statistics