Research On RFID Multi-Reader Anti-Collision Algorithm Based On Reinforcement Learning

With the popularity of Internet of Things technology,Radio Frequency Identification(RFID)technology is more and more widely applied to various lines,such as warehousing logistics,retailing,manufacturing,service industry,identification.RFID is one of the most rapid new technologies in the world.It has the advantages of non-contact,long-distance recognition,anti-interference ability,environmental adaptability and high confidentiality.In the actual application of the RFID system,it is often necessary to place a large number of readers in an area.If they communicated with the tags at the same frequency and at the same time,the readers can not recognize the response signal of the tag,which will cause collision between the readers.In real life,there are many applications of intensive readers,collisions between readers must exist.With the rise of reinforcement learning,it has a mechanism of self-correction and feedback,which is suitable for the model of multi-reader selection channel.Therefore,the research focus on the multi-reader anti-collision algorithm in RFID system based on reinforcement learning.The main research contents of th:is paper and the innovations are as follows:1.Propose a multi-reader anti-collision algorithm based on Q-Learning.The algorithm effectively improves the HiQ algorithm.The complex hierarchical structure in HiQ algorithm was eliminated.The algorithm not only apply the ?-greedy strategy,but also improved the reward function.By adaptively interacting and learning with the surrounding environment through intelligent algorithms,a Q-valued function is generated to obtain an optimal channel resources allocation.The simulation results show that when the number of readers is 9,the frequency collision rate of this algorithm is 28.60%lower than HiQ and 36.03%lower than the EHiQ;when the number is 16,the frequency collision rate of this algorithm is 31.08%lower than HiQ and 47.74%lower than the EHiQ;when the number is 25,the frequency collision rate of this algorithm is 54.94%lower than HiQ and 67.03%lower than the EHiQ.2.Propose a multi-reader anti-collision algorithm based on Deep Q Network.The algorithm combines the idea of the Q-Learning with a neural network.The entire algorithm process includes defining states and actions,establishing replay memory,updating weights of neural network,and finding optimal strategy to allocate channel resources.The simulation results show that when the number of readers is 49,the frequency collision rate of this algorithm is 10.77%lower than HiQ and 35.52%lower than the EHiQ;when the number of readers is 64,the frequency collision rate of this algorithm is 35.78%lower than HiQ algorithm and 46.27%lower than the EHiQ algorithm.