Research On Strategy Evolution Method For Swarm Robots Under Local Communication

Driven by the growing demand for intelligence,swarm robots are asked to have the ability of autonomous task execution in order to adapt to the more and more complex task environment and produce a marked effect.Making swarm robots evolve their behavior strategy independently is the key way to realize intelligence.Limited by the harsh external environment and the ability of swarm robots,the traditional behavior strategy evolution method is not enough to support the development of swarm robots’ intelligence.Therefore,this paper draws on the methods of communication interaction and intelligent model learning and training,faces special challenges of the local communication,aims at the core goal of improving the evolutionary efficiency of swarm robots’ behavior strategies,and carries out the following four key technical researches:(1)Facing the structural requirements of autonomous task execution of swarm robots,aiming at the challenges of multi-task,weak communication,and low computing power faced by swarm robots in the behavior strategy evolution,and considering the local communication task environment,a three-layer structure of swarm robots system is proposed.Based on the architecture,the behavior strategy evolution process and evolution architecture of individual robots in the swarm are designed.The method realizes the evolution of an individual’s behavior strategy,drives the evolution of behavior strategy of the whole swarm robots and provides stable and feasible architecture support for the evolution of behavior strategy of swarm robots under the condition of local communication.(2)Facing the communication interaction requirements in the process of task execution and behavior strategy evolution of swarm robots,aiming at the challenges of consensus reached by members within the swarm under the communication limited environment,a consensus initiative cooperation architecture is designed.On this basis,a communication object priority selection method based on Ad-hoc is proposed,and the corresponding algorithm is designed.The method realizes the situation consensus of swarm robots under communication constraints.Experiments show that the proposed method makes the swarm robots adapt to different communication conditions,and realizes the consensus initiative cooperation of swarm robots through low communication cost.(3)Facing the needs of behavior strategy evolution of homogeneous swarm robots,aiming at the challenges of low efficiency and small search space of behavior strategy evolution of swarm robots under local communication conditions,a heterogeneous-homogeneous swarm coevolution method,Torch,is proposed to improve the evolution ability of swarm robots.In this method,a behavior strategy expression method called behavior expression tree is proposed to improve the task execution ability of swarm robots and realize the evolution of behavior strategies of swarm robots under local information conditions.Experiments verify the effectiveness of the proposed method and verify the improvement of the method in task execution ability and behavior strategy evolution performance.(4)Facing the needs of swarm robots to perform area search tasks under local communication conditions,aiming at the challenges of task execution by harsh environment(unpredictable environment and lack of GPS signal),a swarm cooperative area search method,C-search,is proposed.In this method,behavior strategy evolution modules corresponding to different environmental conditions are designed to realize global accurate planning and autonomous evolution of behavior strategies in a harsh environment.The experimental results show that the design of C-search increases the adaptability of swarm robots to a variety of environments and the robustness to harsh environments.