Research On Flocking Algorithm For Multi-agents Oriented To Obstacle And Guiders

Agent is a significant concept in the field of artificial intelligence, and flocking is one of those examples where a coordinated task is performed by multi-agents. Research on flocking provides theoretical basis for the design of practical engineering group model, and it also has a great inspiration in the optimal design of engineering systems, laying the foundation for the development of distributed coordination control as well as multi-agent systems. As a result, all algorithms for multi-agent flocking have great theoretical value and engineering guidance.Firstly, the existing multi-agent avoidance algorithms consider nothing but only the speed of agents. And due to the thoughtlessness of the direction of the velocity it is difficult to confirm the direction while in the process of avoiding obstacles. In this case, an improved algorithm called MAAO is proposed. Before analyzing and simulating, a theoretical analysis of the MAAO algorithm is performed. In spite of O(n2) complexity of MAAO, the higher speed of MAAO algorithm proves its higher efficiency of avoiding obstacles.Secondly, the existing multi-agent flocking algorithm is kind of redundant in engineering application, for the fact that all agents have to be well aware of the exact location of the target when the existing algorithms are accomplished. Worse still, it is unrealistic under some specific circumstances. To solve this problem, guiders are introduced, which are selected among all the agents and have information about target information to guide others complete flocking finally. This article gives theoretical analysis of the feasibility of the guiders as well as the simulation to support the view of point. The convergence time on different proportion of guiders is calculated and summarized, which is "the more guiders involved, the less time taken".Thirdly, the condition that target information is not available for all agents is also well considered. Research of flocking is performed under the condition of not knowing the target information and communication distance. Also, this article covers the method of calculating flocking when multi-agent flocking becomes a probabilistic event. Through comparison of actual values and the theoretical calculated values, the probability value calculation is proved to be correct.