Research On Optimization Of Switching Algorithm In FRMCS

In recent years,high-speed railways and mobile communication systems have developed rapidly,and the narrowband communication technology GSM-R has been unable to meet such high mobility scenarios.LTE-R in the Next Generation Railway Mobile Communication System(FRMCS)has the following advantages over GSM-R: First,it has a flat network structure,which makes the delay shorter;Second,OFDM is the key technology.Increased spectrum efficiency,reduced network deployment and maintenance costs,and greater performance improvement;Third,MIMO technology makes full use of space resources,multi-input and multi-output transmission mode increases the channel capacity of the system;Fourth,LTE-R It supports a high-speed mobile terminal environment with a maximum speed of 500km/h,which is more suitable for high-speed mobile scenes.The main steps of handover in the LTE-R system include handover measurement phase,handover decision,and handover execution phase.Factors affecting the occurrence of handover are handover hysteresis parameter,handover trigger delay,and handover trigger threshold.The A3 event-based handover algorithm is the most widely used handover algorithm in the railway mobile communication handover algorithm.The A3 event-based algorithm mainly involves two parameters: switching hysteresis threshold parameters and switching trigger delay.At present,most scholars study that the LTE-R system switching algorithm is mostly based on the A3 event,but the A3 event-based handoff algorithm has inherent problems such as inaccurate decision,high ping-pong effect rate and high dropped call rate.The fault tolerance of zone switching is not high.With the increase of the railway running speed,the Doppler effect becomes more obvious and the switching time is shorter and shorter.The traditional A3 algorithm can not meet the communication needs of high-speed railway,and an excellent handover algorithm is urgently needed.In recent years,machine learning(including deep learning)algorithms have achieved great success in image classification,speech recognition,and the like.The algorithm can learn the deep and subtle features of the desired model by means of feature learning.It has the advantages of high fault tolerance and wide adaptability.The neural network algorithm can achieve the success rate of traditional algorithms and reduce the computational complexity and has a stronger advantage.For the LTE-R handover algorithm,the A3 event-based handoff algorithm is prone to ping-pong effect(PPHO)and radio link connection failure(RLF)when the train is running at high speed.This paper proposes a radial basis.Radial Basis Function(RBF)neural network handoff optimization algorithm.The algorithm firstly collects a large number of Hys and TTT parameter sets with good switching effects when the trains are running at different speeds(0-100m/s)in a specific environment,and sends them to the RBF neural network for training.The RBF neural network contains one input layer and one.The layer hidden layer and one layer of output layer,the number of neurons in the hidden layer is determined by the trial and error method,and the RBF neural network parameter update uses the clustering algorithm.After the network training is completed,the nonlinear models of Hys and TTT at different speeds are obtained.Then,according to the received signal received quality(RARQ)of thetrain,the self-correction term is added to adjust and optimize the Hys and TTT.Finally,experiments on the MATLAB simulation platform show that the algorithm reduces the call drop rate and the ping-pong effect rate compared with the A3 algorithm,which can effectively improve the switching success rate of the train in high-speed operation environment and increase the robustness.The RBF neural network handoff optimization algorithm has the problem that the weight of the hidden layer to the output layer is slower in emergency convergence.To improve this problem,this paper uses the gray wolf algorithm to optimize the RBF neural network.Firstly,the optimized RBF neural network weight matrix is ??the wolf individual in the grey wolf algorithm.The fitness function of each individual is given by the designed fitness function,and the wolf is continuously updated until the model converges or reaches the specified stopping condition.In order to overcome the problem that the gray wolf algorithm has slow convergence rate,easy to fall into local maxima or minimum value and does not consider its own experience,this paper introduces the memory function in particle swarm optimization(PSO)to improve it.The weight of the wolf individual update speeds up the convergence rate of the algorithm in the middle and later stages,and increases the global search ability,reducing the probability of falling into local maxima or minima.The simulation results show that the optimized algorithm can better search the global optimal solution,further improve the switching success rate and achieve better switching effect.The analysis of the complexity of the algorithm shows that the improved algorithm has a shorter delay and better real-time performance.