Decentralized asynchronous particle swarm optimization with dynamic neighborhood topology and its implementation on multi robot search task

In this thesis decentralized asynchronous particle swarm optimization (PSO) with dynamic neighborhood topology and its implementation to a searchtask of a multi-agent system were studied. Particle swarm optimization with dynamic neighborhood topology studied in this thesis allows the neighborsof particles or basically the neighborhood topology to change dynamically with time. Such a view of the algorithm is advantageous for its paralleland distributed implementations. The algorithm with dynamic neighborhood topology was tested on various benchmark functions under differentneighborhood dynamics. Decentralized asynchronous realization of particle swarm optimization algorithm is suitable for parallel and distributedimplementations. Such a version of the algorithm allows particles to exchange information and update their estimates at totally independent timeinstants and dynamically change neighborhood topology of particles with respect to time. Simulations were performed using a single processor andmultiple processors in a computer network. The proposed algorithm is used for a search task of a multi-agent system in an unknown environment whichconsist of small robots with limited sensing capability. The method adopts asynchronous mechanism for information exchange and position updates ofthe agents and dynamic neighborhood topology of communicating agents. Simulations with a realistic simulator and implementation with real robots wereperformed to show the effectiveness of the proposed algorithm.