Research On Handover Method In GSM-R

China High-speed railway had been developing rapidly. In the early railway development, people mainly focused on the construction of railway infrastructure. However, with the increasement of the speed of train, the traditional schedule communication could no longer meet the demands of train's safe operation. GSM-R (GSM for Railway) was the special railway mobile communication system which was applied in CTCS-3 (China Train Control System Level Three). GSM-R supported the speed which was between 300 km/h and 500 km/h of train. Schedule communication was the important guarantee for the train's safe operation in high-speed environment. The count of dropped calls caused by handover occupies the rate up to 40% of the total number. So it was significant to do research on handover.The analysis on handover showed that the failure of handover has always occurred in the case of which the network parameters did not meet the communication requirements. The value range of network parameters had statistical regularity. This paper researched the handover mining association rules based on Apriori and Hotspot algorithm, mining the value range of network parameters which was strongly associated with handover event. This paper adopted the theory of electromagnetic wave propagation to analyse and explain the strong association between the handover event and the value range of network parameters. This paper proposed the probability model of satisfactory communication and established the probability satisfactory communication model of the rxlev, the distance between BS and MS, the speed of train. In order to find the handover decision output which had the optimal probability of satisfactory communication, this paper proposed an advanced handover algorithm based on BP neural network. The inputs of BP neural network were the rxlev, the distance between BS and MS, the speed of train. The output of BP neural network was handover decision which had the optimal probability of satisfactory communication. Because handover occurred on the point which had the optimal probability of satisfactory communication, the probability of satisfactory communication and the rate of success handover had been improved effectively.In the results of the handover event mining association rules, the strong association value range of the network parameters were consistent with the electromagnetic wave theory. This paper obtained the handover diagnostic information after analysing the results. In the model of probability satisfactory communication, this paper simulated the traditional handover algorithm, the location predict handover algorithm and the relay assisted handover scheme and the BP neural network handover algorithm. The experimental results showed that the unsatisfied communication probability of BP neural network handover algorithm can be reduced the rate of 41.1% averagely comparing with the traditional handover algorithm. In contrast with the location predict handover algorithm and the relay assisted handover scheme, the algorithm which proposed by this paper gained lower probability of unsatisfactory communication and lower cost.