Research Of Massive Access Techniques For Low-Power Wide-Area Networks

With the rapid popularization of smart devices,network needs to carry more and more end devices.In order to meet this situation,many Internet of Things technology solutions are gradually emerging.Among all of these solutions,Low-Power Wide-Area Networks(LPWAN),whose end devices can achieve ultra long transmission distance with ultra-low power consumption and low network organizing costs,has received extensive attention.As a member of LPWAN solutions,LoRaTM(Long Range)has become the most commonly deployed IoT private network solution and the development situation is in full swing.Therefore,this paper will focus on the LoRa WAN,which is the MAC layer protocol for LoRa,and analyzes the existing pain points of it.In order to ensure the success of device accessing network under massive access situation,this paper will propose a more effective resource allocation strategy due to unreasonable allocation of spreading factors and channel selection strategies in current protocol.The reset of the article will carry out specific research from the following two aspects.Firstly,the problem of spreading factors allocation in LoRa WAN networks is studied.Based on the ACK acknowledgement mechanism in LoRa WAN protocol,the collision probability of data packets on each spreading factor is deduced.Then,the problem of minimizing the maximum collision probability is established for the fairness to optimize the collision probability on different spreading factors.In this thesis,a heuristic algorithm based on greedy algorithm is designed to obtain the optimal devices allocation number for each spreading factor,and the spreading factor is allocated according to the RSSI(Received Signal Strength Indicator)of each device.All of the effort is to reduce the collision probability of the entire network to allow more devices to communicate and increase the throughput.Secondly,the channel selection for multi users in unknown channel condition will be studied.With the Multi-Armed Bandit model build on the LoRa WAN specification,end device can learn to choose best channel to transmit packets without any information of the channel.The paper focuses on the strategies based on ?-Greedy algorithm and Upper Confidence Bound and Thompson Sampling and will proposed an improved algorithm based on Thompson Sampling.All of the channel selection strategies will be continuously iterated and updated with balancing the exploration and exploitation.As a result,end devices work with these strategies can find the best channel during the process of learning.Finally,the simulation shows that the channel access scheme provided can effectively select the data transmission channel in the case of massive access and the data transmission delay can be close to the optimal solution.