| At present,with the rapid development of computer network technology,the computers are changing fast to an opening,network,cooperation working mode.The theory and technique based on agent,especially multi-agent systems,bring us a brand-new mode to design and implement software system which runs in a distributied and opening environment.RoboCup(The RobotWorld Cup) is a typical multi-agent system,which has currently chosen soccer as its standard task.In the RoboCup,the purpose of path planning is to find a path without collision which meets some estimate criterions.The path planning is applied in the lower policy which is used as the foundation of the robots' basic actions,the planning result directly has an effect on the action's real-time and veracity.So,everyone who studies on robot treats path planning as an emphases.This thesis chooses robot soccer as research background,and pay more attention to path planning problem of the system to get effective methods to resolve path planning problem in real world.Through analysis and research on traditional path planning methods,we propose a path planning algorithm based on adaptive isolation niche genetic algorithm(AINGA).The main work and the innovation of this paper list as follows:(1)On the basis of the analysis of theory of genetic algorithm,niche genetic algorithm(NGA) and isolation niche in detail,we make an improvement on isolation niche genetic algorithm:introduce the exchange pool of optimal individuals,which can avoid local optimal solution,and accelerate convergence.(2)Through the analysis of evolution and the research on traditional adaptive genetic algorithm,we design new crossover operator and mutation operator which are associated with generation.It can accelerate convergence,too.(3) According to the two improvements above,we propose an adaptive isolation niche genetic algorithm.Then it is applied to RoboCup2D,compared to Q-learning,it turns out to be more effective and efficient in planning efficiency. |
PDF Full Text Request