Date of Degree
M. Umit Uyar
In this dissertation, we present a bio-inspired decentralized topology control mechanism, called force-based genetic algorithm (FGA), where a genetic algorithm (GA) is run by each autonomous mobile node to achieve a uniform spread of mobile nodes and to provide a fully connected network over an unknown area. We present a formal analysis of FGA in terms of convergence speed, uniformity at area coverage, and Lyapunov stability theorem.
This dissertation emphasizes the use of mobile nodes to achieve a uniform distribution over an unknown terrain without a priori information and a central control unit. In contrast, each mobile node running our FGA has to make its own movement direction and speed decisions based on local neighborhood information, such as obstacles and the number of neighbors, without a centralized control unit or global knowledge.
We have implemented simulation software in Java and developed four different testbeds to study the effectiveness of different GA-based topology control frameworks for network performance metrics including node density, speed, and the number of generations that GAs run.
The stochastic behavior of FGA, like all GA-based approaches, makes it difficult to analyze its convergence speed. We built metrically transitive homogeneous and inhomogeneous Markov chain models to analyze the convergence of our FGA with respect to the communication ranges of mobile nodes and the total number of nodes in the system. The Dobrushin contraction coefficient of ergodicity is used for measuring convergence speed for homogeneous and inhomogeneous Markov chain models of our FGA. Furthermore, convergence characteristic analysis helps us to choose the nearoptimal values for communication range, the number of mobile nodes, and the mean node degree before sending autonomous mobile nodes to any mission.
Our analytical and experimental results show that our FGA delivers promising results for uniform mobile node distribution over unknown terrains. Since our FGA adapts to local environment rapidly and does not require global network knowledge, it can be used as a real-time topology controller for commercial and military applications.
Sahin, Cem Safak, "Design and Performance Analysis of Genetic Algorithms for Topology Control Problems" (2010). CUNY Academic Works.