Research On M2M Load Prediction For Random Access

With the rapid development of the Internet of Things and mobile communication networks,more and more machine-type communication devices have the ability to connect and need to communicate.It is becoming a reality that all terminal devices are connected to the network.However,with the explosive growth of machine-type communication devices,the problem of preamble collision of access terminal devices has become more and more prominent.This not only seriously affects the number of terminal devices successfully accessed in the random access process,but also causes a large number of terminal devices to seriously affect human life because they cannot access the network on time.In addition,many studies on random access mechanisms are based on known load,such as uplink resource allocation studies,ACB load control mechanisms,and so on.Therefore,real-time load prediction for the number of M2M terminals in the random access process has important practical significance.Firstly,the random access process is analyzed and researched,and a corresponding load forecasting model is established based on actual access scenario.According to the preambles status information detected by the base station,the number of access terminal devices is predicted by using nonlinear regression and back propagation neural networks respectively.Meanwhile,the Kalman filter is used to correct the previous load prediction based on the current preambles status in the real-time load prediction phase.Compared with the simulation of existing load prediction algorithm,the simulation results show that the proposed load algorithm can improve the absolute error,access success rate and average number of retransmissions effectively.Secondly,this paper proposes an improved random access scheme based on load prediction.In the new random access scheme,when multiple terminal devices select the same preamble,the collision probability of the preamble can be effectively reduced by the access state contention resolution mechanism.Since the current random access system is based on the slot access protocol,the device that failed to access in the previous time slot needs to re-access in the next time slot,which may cause severe load and unnecessary retransmission of the network.Therefore,this paper proposes a grouping backoff algorithm based on load prediction,which reduces the number of retransmission times of the device by distributing the backoff device to multiple time slots in the future.The simulation results show that when the number of activated devices increases from 40,000 to 70,000 and the maximum retransmission times equals ten,the M2M device access success rate increases by two percent on average,and the preamble retransmission times decrease by an average of two times.In this paper,a real-time load prediction is performed on the number of access devices in the random access process,which effectively improves the access success rate.By improving the existing random access mechanism based on the load prediction algorithm,the network performance is further improved.