New approaches for self-organizing aggregation behavior in swarm robotics

In this thesis, an algorithm and two different methods have developed for self-organizing aggregation behavior in swarm robotics. The algorithm developed enables swarm robots to exhibit a flexible and scalable aggregation behavior. To evaluate the flexibility and scalability of the aggregation algorithm, systematic simulations have carried out in different arena sizes, different numbers of robots and different sensing radii. The obtained results have statistically analyzed according to the number of robots, arena size and sensing radii. One of the developed methods is fuzzy logic-based self-organizing aggregation behavior. The method allows swarm robots to evaluate a small number of sensor inputs with fuzzy logic rules and to exhibit aggregation behavior. Noisy and noiseless systematic simulations have applied to evaluate the performance of the fuzzy logic-based method. These simulations have applied with different arena sizes, different numbers of robots and different sensing area sizes. The second method is dynamically interactive self-organizing aggregation method. This method aims to exhibit the aggregation behavior with the help of a state selector by using the obstacle and robot detection sensors that swarm robots have. The status selector decides which of the three different controllers to operate for the swarm robots to exhibit the aggregation behavior. These controllers allow swarm robots to approach and alignment, avoid obstacle, and move randomly. The control units used in the method offer a dynamic interaction structure between swarm robots and neighboring robots. Systematic simulations have applied according to arena size, number of robots and sensing radius to evaluate the performance of dynamic interactive aggregation behavior. Three different aggregation metrics used in the literature have used to measure the performance of the aggregation method. The aggregation metrics of a developed method have compared with the other two methods in the literature. As a result of the comparisons, it has observed that the performance of the dynamic interacting aggregation behavior method is more successful than the other methods.