Research On Algorithm Of Multi-robot Pattern Formation Based On Swarm Intelligence

With the development of robot technology and the needs of practical applications, research on the applications of multi-robot systems has been a trend. Multi-robot pattern formation is one of the typical problems of multi-robot cooperation. Aggregation and pattern formation by initially randomly distributed multi-robot system can be used as basis of other more complicated applications, such as self-assembly, connected movement and self-organized construction.Traditional multi-robot coordination algorithms are mainly based on central coordination or negotiation and task allocation mechanism. When the scale of multi-robot system increases, with the increase of the nodes number, these methods will encounter bottleneck, the traffic of inter-node communication and the algorithm complexity will be significantly increase.This study designed a multi-robot patten formation algorithm based on swarm intelligence, using a distributed self-organizing mechanism to enable multi-robot system to form a predefined pattern ont the grid map. Through the inter-node communication and the stigmergy between the swarm and the environment, swarm intelligence makes the individuals response to local information according to certain rules and complete complex tasks on the swarm level. Comparing with traditional methods, this swarm intelligence coming through the emergence which is a fully distributed mechanism is more efficient, well scalable and robust.In this algorithm, the diffusible pheromone is used to describe and indicate the predefined pattern and to helpe the multi-robot system evenly distribute to the grids constituting the pattern; the virtual pheromone is used as stigmergy mechanism to realize coordination and experience sharing between individuals and to help the multi-robot system search ang move to the targeted grids effectively; particle swarm optimization (PSO) is used as the movement decision making mechanism, on the basis of diffusible pheromone and virtual pheromone information.In this study, simulation platform is designed and used to validate the algorithm. Experiment results confirm the effectivity, efficienceny and scalability of the algorithm. Based on experiments on the simulation platform, the affects of key parameters on the multi-robot pattern formation algorithm are discussed. The paper also describes design principles of the robot platform, the e-mouse and the initial implementation of algorithm on this hardware platform.