Probabilistic Guidance Technology Research For Swarms Of Autonomous Agents

Compared with the method of early formation flight and muti-spacecraft cooperation to complete the task,the main challenge of the spacecraft swarm research progress is the large number of agent in swarm,when the number of the agent in formation is 10 or 20,corresponding to the number of swarms desired spacecraft contains can reach hundreds or even thousands.In order to solve the problem of large number of agents,this paper studies the probability guidance algorithm,breaking the commen way of original formation,in the view of probability we take advantage of the law of large numbers to swarm as a unit which can be counted,make the large number of agents in the research transfer into configuration advantage,eventually realize the swarm formation.Before application of probabilistic guidance algorithm for satellite swarms on the distribution of space density,we should study the probabilistic guidance algorithm at the two-dimensional space.according to the actual research called “AARe ST”plan,we used the probabilistic guidance algorithm space to configuration satellite swarms which were on orbit,each agents in the swarm make decisions according to the probability of their transfer,these decisions only based on their own current state,so that the whole swarm eventually achieve the desired density distribution in the space.This paper takes two different approaches to design the probabilistic guidance method in the transition matrix,the one is based on linear matrix inequality(LMIs),and the other is based on the Inhomogeneous Markov Chain(IMC)algorithm.The LMI method can guarantee the desired convergence rate,furthermore it also can be used minimize the cost function to reflect the consumption of fuel,while the IMC method is utilized to obtain the cluster density feedback gain planning the next distribution of swarm.In addition,the two methods have the ability of self repair,when the current density distribution and the desired density distribution are different,each agent in the swarm can move itself automaticily to achieve the desired space configuration.