Study On RFID Network Planning Based On Evolutionary Computation

In recent years,due to the advantages of low requirements of working environment,low cost and fast response,RFID network has attracted lots of attention in academia,and has been widely used in business,industry,military and so on.RFID network planning is the most important research.RFID network is divided into static network and dynamic network because of the difference of tags and readers' state.Because the readers' reading range is limited,designing reasonable reader deployment strategy has become hotspot.The purpose of RFID network planning is to find the appropriate deployment strategy through the algorithm,so that RFID network can achieve great performance in terms of coverage,total cost,interference and load balance.Many effective RFID network planning algorithms have been proposed,but most of the research are just applying the existing optimization algorithm.They do not combine evolutionary algorithms with the characteristics of RFID network.On the one hand,this thesis designs a static RFID network planning algorithm based on multi-objective optimization,and parallelizes it to solve large-scale problems.On the other hand,combined with the characteristics of dynamic RFID network,a dynamic RFID network planning algorithm based on particle swarm optimization algorithm is designed.The main work is summarized as follows:A Decomposition-based Multi-objective Self-adaptive Differential Evolution Algorithm for RFID Network Planning: Evolutionary algorithm has been widely used in optimization problems because of strong robustness,self-adaptive and self-learning,and multi-objective evolutionary algorithms based on decomposition have low computational complexity,better diversity and high accuracy.In this thesis,the multi-objective evolutionary algorithm based on decomposition is applied to static RFID network planning,and combined with the characteristics of static RFID network,a decomposition-based multi-objective self-adaptive differential evolution algorithm(Decomposition-based Multi-objective Self-adaptive Differential Evolution Algorithm for RFID Network Planning,MOSDE/D)is proposed.A unique coding method,an improved population initialization strategy,an improved adaptive differential evolution and an improved crowding distance ranking strategy are designed.A large number of experiments have been carried out on different types of static RFID network examples.The results show that MOSDE/D can design great deployment strategies.Compared with the existing two algorithms,it shows better planning performance.A Parallel MOSDE/D based on coarse-grained model: We combine the parallel design based on the coarse-grained model with the MOSDE/D and apply it to the large-scale static RFID network planning(PMOSDE/D).The sub population division algorithm,unique topological structure and migration strategy are designed.In the experiment,PMOSDE/D shows desirable planning performance.Compared with MOSDE/D,PMOSDE/D has obvious acceleration performance.Dynamic RFID network planning based on multi-population PSO: Compared with static RFID network,dynamic RFID network planning is a relatively new research topic.Most of the research focuses on the design of reading protocol to solve the "collision problem" caused by tag movement.There is little research on the design of dynamic RFID network planning algorithm based on the characteristics of dynamic RFID network.Therefore,we combine the dynamic evolutionary optimization method based on multi-populations with the characteristics of dynamic RFID network,and propose a dynamic RFID network planning algorithm based on multi-populations PSO(PSO-MP).The dynamic adaptive objective function,sensing and response strategy,dynamic PSO velocity operator and rejection operation are designed.Experimental results show that PSO-MP has a better dynamic planning performance.