Studies On The Flocking-algorithm Of Cluster-control In Multiagents Systems

Cluster-control is an up-to-date research field that very popular recently.It is inspired by biological phenomena and will gradually transplant to human social behavior and artificial intelligence development.With the help of mathematical language and machine language,by means of rigorous analysis,it will make kinds of cluster more efficient and stable.After canvassing the backgrounds and application value of cluster-control,this dissertation investigates its present research state.First we dissect the development history of cluster-control and summarize its application in different fields.Second we list a number of research methods that frequently-used and describe their applicable scopes.After that,we introduce several representative studies,and discuss their advantages and disadvantages.The main result of this dissertation serveys the problems of Flocking-algorithm,which is an important branch in Cluster-control.After summarizing and pondering the previous research results,we consider those imperfect environments in reality.Take the multi-agents system,for example,we propose two creative improved algorithms.After describing their characteristics and advantages,we provide the stimulation experiment proves in software MATLAB,and also gave theoretical analysis proves in mathematics.In the most research findings,they use virtual leader,which is not suited every situation.So in the first algorithm,different from the past,we turn the virtual leader into real leader,that is to pick some individuals of the colony and let them become the real leaders.And considered the split phenomena that may appear,we propose the thought of interval and rendezvous point as the connection policy of the colony.In the mean time,we use an obstacle avoidance strategy with anticipation which can reduce the cost of obstacle avoidance.In the second algorithm,we consider the limitations that individuals in reality may not be able to obtain the speed informations of the units around.In this case,realization of flocking can only be based on the location informations.In order to compensate for the lack of speed informations,we propose the thought of path priority,which make the individuals do the screening works.This will improve the utilization efficiency of the information they've got.Further,this algorithm needs not real-time feedback information from the leaders so that the algorithm can be more applicable and can better conform to the reality.This dissertation provide the detailed introduction of two algorithms above,and also give the stimulation experiment proves and theoretical analysis proves in mathematics.In the end,we made a summary of the whole dissertation and look tothe future of cluster-control research.