ACTOR POSITIONING IN WIRELESS SENSOR AND ACTOR NETWORKS USING MATCHING THEORY

Author

Ismail Guneydas, Southern Illinois University CarbondaleFollow

Date of Award

1-1-2008

Degree Name

Master of Science

Department

Computer Science

First Advisor

Akkaya, Kemal

Abstract

AN ABSTRACT OF THE THESIS OF ISMAIL GUNEYDAS, for the Master of Science degree in Computer Science, presented on 5th November 2008, at Southern Illinois University Carbondale. TITLE: ACTOR POSITIONING IN WIRELESS SENSOR AND ACTOR NETWORKS USING MATCHING THEORY. MAJOR PROFESSOR: Dr. KEMAL AKKAYA In most of the Wireless sensor and actor network (WSAN) applications, the locations for the actors are determined autonomously by the collaboration of actors and/or sensors in order to eliminate human intervention as much as possible. Particularly, sensors can collaborate in a distributed manner and elect cluster-heads (CHs) among them which will be taking into account the distribution of the sensors within the region. In such cases, the actors can then move to such sensor locations (i.e., replace them as cluster-heads) as they have the ability to move by talking to nearby sensors/actors. Such movement, however, should be done wisely in order to minimize the total distance that will be traveled by the actors so that their lifetimes can be extended. Nevertheless, this may not be possible since not all the actor and CH locations will be known to each actor. In addition, the actors may not be reachable to each other and thus conflicts in assignments can easily occur. In this thesis, we propose an actor-CH location matching algorithm which will detect the CH locations and assign the actors to such locations in a distributed manner with the minimized travel distance. We adapt the Gale-Shapley (G-S) stable matching algorithm from Matching Theory in order to prevent conflicts and minimize the travel distance. In this matching algorithm, actors are regarded as men and CHs are regarded as women. First, we detect the CH locations through running a quorum-based search within the sensor network. Later, G-S is run on actor and CH locations. Once the locations are determined, each actor moves to that location. We evaluated the performance of our approach through simulation and have shown that our approach can produce results very close to the brute force approach.