Research Group Based On Swarm Intelligence Robot Formation

Swarm robots can carry out and eventually complete those complex tasks because of swarm intelligence emerged from interactions between individuals, in case the required capability being far beyond that possessed by any single member. It is significant that swarm robots move in formation for implementing such complex tasks, such as investigation and security patrols as well as space exploration. Besides, formation is one of standard academic problems in swarm robotics. Therefore, it makes sense that the author presents the topic by combining swarm intelligence theory with the typical application of swarm robotics.In this thesis, the search environment is represented as a two-dimensional plain workspace, where robots are distributed randomly. To control swarm robots in specific formation, two approaches and corresponding strategies, i.e., centralized and distributed, are presented. The former implements formation tasks by making use of the traditional particle swarm optimization algorithm. Concretely, a fitness function is constructed to evaluate position information about all robots in search space. It is thus clear that the optimal value stands for the desired formation. Therefore, the PSO algorithm can further be used to optimize the above object function. The direction step by step approaches the optimum can be viewed as the direction of robot moving velocity in the process of update evolution when iteration occurs. The corresponding simulations indicate that triangle, linear, circular and hexagon shapes can be formed in this way. Aiming at the distributed computation in reality, the author investigates and solves the same problem of formation in a full distributed fashion. The difference between them is that each individual robot computes its own fitness and cognitive position, catches the best position of swarm through interactions, and generates its own moving velocity following the swarm intelligent principles. The thesis also researches the transformation among four formations. According to the position information of the neighbor robots, a closed-loop control strategy is used to transform each other. Simulation results show the effectiveness.