With the booming development of high-speed railways (HSRs), key technologies of wireless communications need to be constantly innovated. In particular, the frontier issue of low delay of the handover for the fifth generation (5G) in fast-moving scenarios has attracted attention from both industry and academia. Based on an analysis of a large number of measured data and the location of the user equipment (UE), a fast handover algorithm is proposed to solve the problem of long delay for a train moving at high speed in a 5G-railway (5G-R). By calculating the speed of a train and its direction of movement, a reasonable handover mode is selected and the handover chain of neighboring cells is identified. The location of the train can be calculated to determine whether UE enters the defined identification zone of pre-handover. Depending on the values collected in the measurement report, the command of the handover is triggered when the weight of the target cell is greater than that of the source cell. Our experimental results show that the delay of the fast handover algorithm is reduced by 2.03%, and the success rate of the handover is increased by 0.42%. Research directions for smart railways are discussed based on these findings.

When the wireless communication network is optimized for the high-speed railway scenario (HSRS), GPS connections are prone to frequent interruptions. The causes of this phenomenon are analyzed after collecting a large quantity of measured data. A positioning algorithm based on the fingerprint database by twice-fuzzy clustering is proposed to obtain the locations of the terminal inside the carriage of high-speed train (HST) in real time. After collecting more than 300,000 sampled data of both network characteristics and location information, the database of original fingerprints has been constructed. The identification and elimination of abnormal fingerprints are helpful to improve the quality of the fingerprint database. The longitude and latitude of the terminal, which is losing the signals of GPS, can be calculated by setting the fingerprint integral counter, threshold value and weighted measurement eigenvalues and by constructing a matrix of dissimilarity. The experimental results show that the proportion of similar fingerprints between 15 and 200 by once-fuzzy clustering is as high as 90.81%; Additionally, the number of over 95.57% of the similar fingerprints is between 1-20 by twice-fuzzy clustering. The proportion of samples with positioning accuracy less than 10 m is 63.33%, and those less than 5 m account for 41.67%. The average positioning accuracy of the proposed algorithm is 9.02 m, which is suitable for acquiring location information when the signals of GPS are losing in HSRS.