Research On Coordinated Algorithm Of Multiple Readers In Large RFID System

Radio Frequency Ident ificat ion(RFID)is an effect ive non-contact auto mat ic ident ification techno logy,which can effec t ively sense and co llect informat ion and is generally used in many scenarios such as warehouse management.In RFID systems,the RFID reader ident ifies and co llects relevant informat ion o f RFID tags by the radio frequency signal,and the read throughput is an important indicator to measure the efficiency of the system.To fully cover the monitoring area,people must deplo y mult iple readers to work cooperatively in large RFID systems,but the problem o f reader caused co llisio ns great ly limits the read efficie ncy o f the system.How to efficiently coordinate mult iple readers to work together is crit ical for high throughput in RFID systems.Exist ing researchers focus on designing efficient reader scheduling strategies that arrange adjacent readers to work at diff erent t imes to avo id signal co llisio ns.However,the improvement o f the throughput is st ill limited by tag distribution and algorithm’s strategy.In order to further improve the read throughput of mult i-reader RFID systems,this paper carries out related r esearch,and the main research results are as fo llows:(1)Two load balancing algorithms for readers are proposed.Due to the different tag distribut ion,the number o f tags covered by each reader(i.e.the load o f it)is unbalanced.This imbalance leads to different execut ion t imes for readers: the heavy load readers take a longer t ime to co llect all tags,while the other readers that finish execut ion earlier have to wait for nothing,causing a waste o f t ime.Aiming at the problem o f the unbalanced load of readers,we adjust the interrogat ion radius o f readers by adjust ing the readers’ transmit power,thereby coordinat ing the readers’ loads.First ly,the restricted load balancing algorithm is proposed to coordinate the load o f the readers that works simultaneously under the condit io n o f satisfying the limitat ions,and the interrogat ion radius of each reader is adjusted to meet the target load.We also propose an est imat ion method to obtain the radius of the reader when it satisfies the target load.Then the l imitat ion of the algorithm is analyzed,and we propose the co mplete load balancing algorithm so that the load of the readers working at the same t ime can co mpletely be balanced.Simulat ion experiments prove that our algorithms can be co mbined with the exist ing scheduling algorithms to perform tag protoco ls such as tag ident ificat ion,which can well coordinate the load of the readers and further improve the read throughput.(2)A reader scheduling algorithm co mbined with deep reinforcement learning algorithm is proposed.Since the reader scheduling problem is usually so lved by the heurist ic algorithm,it is difficult to find the opt imal so lut ion.This paper co mbined with the DQN algorithm,mult iple readers are regarded as mult iple agents,and the act ivat ion o f the reader is regarded as different actions.The scheduling problem o f mult iple readers is modeled as the act ion decision problem o f mult iple agents.The DQN network is trained for each agent to learn the best scheduling strategy,with the goal o f minimizing the overall tag ident ificat ion t ime,a reward mechanism and input state informat ion are designed,and the most time-efficient scheduling strategy is obtained through training and learning.Experimental results show that the algorithm can effectively learn scheduling strategies,reduce the t ime to read all tags and improve the read throughput.