The communications-based train control (CBTC) systems play a vital role in ensuring the safe and effective operation of urban mass transit, which can further improve headway and reduce the number of wayside equipment. The emergence of vehicle-to-vehicle(V2V) communications in railway signaling industry has made it crucial that metros ensure safe train separation with a moving block, where a train determines its location, direction, and speed by itself. Therefore, identifying the accurate location of trains becomes a tremendous challenge for CBTC using V2V communications. To determine its location on the rails, many devices including tachometers, accelerometers, transponders (or tags), radar, wireless local area network (WLAN), and long-term evolution (LTE) are utilized by a train. There are still some intrinsic drawbacks to train positioning in existing CBTC systems, which are characterized by the low standard of precision, discontinuity, and vulnerability to jamming attacks in wireless networks. In recent years, however, the technology of visible light communication (VLC) has been gaining increasing attention as it has a wide range of application scenarios such as indoor localization, traffic lights management, and sensing, which can provide both illumination and data communications to address the urgent problems of spectrum crunch, wireless jamming and so on. Moreover, the VLC technology has great potential and can also be applied to determine the train location for new CBTC systems via light-emitting diode (LED) lamps, which are usually installed on the tunnel walls of metros. In the present study, the needs of train localization through the combination of VLC and binocular stereo vision are satisfied to achieve autonomous train positioning, particularly in tunnels. Hopefully, the basic strategies and findings obtained can be conducive to autonomous train positioning for CBTC systems adopting V2V communications.

In the present study, LED lamps installed on the tunnel walls of metros are used as the transmitter of VLC, while the binocular stereo vision system fixed on the top of a cab serves as the receiver of VLC. In this way, the autonomous train positioning in CBTC systems is realized. Firstly, the receiver captures the images of LED lamps and transmits them to the on-board equipment. In light of differences in frequencies from highly flickering LED lamps, a running train can acquire the unique identification (UID) of the corresponding LED lamp in real time and the location of the lamp can be precisely pinpointed in world coordinate system through on-board database. Then, the center coordinates of LED bright spots on the image of LED lamps can be extracted based on the gray weighted centroid algorithm and ordinary least squares as a single feature point for stereo matching to calculate the relative distance between the train and the LED lamp. Next, in terms of the principle of binocular stereo vision, the initial train location can be determined through coordinate conversions. Finally, to obtain the actual conditions of train running, the Wiener filter and inertial measurement unit (IMU) have been adopted to compensate for the train positioning error caused by motion blur of images and mechanical vibration from the receiver respectively, and realize autonomous train positioning at different speeds. In addition, an experimental platform of VLC and binocular stereo vision for autonomous train positioning are established and the experimental results of static and dynamic train positioning are analyzed by MATLAB. The results are combined with the real line date and equipment information in Chengdu Metro Line 1 to demonstrate the feasibility and effectiveness of the proposed method.

A series of experiments on static and dynamic train positioning within a range of 20 test points are carried out as well as train running direction. For static train positioning, 90% of train location errors can be controlled within 20 cm and the maximum error is 29.73 cm (Fig. 11). The proposed method shows less deviation from the actual train location in terms of the train positioning results compared with the train positioning method based on VLC and monocular vision, when a train is running at the edge of a positioning unit and far from the LED lamp (Fig. 12). As to dynamic train positioning, the course angle error of the binocular stereo vision system is reduced from 7.5° to 0.5° after the compensation of motion blur using IMU when a train is running at the speed of 20 km/h, and the maximum train location error is decreased from 35.24 cm to 32.09 cm (Fig. 13). Moreover, the maximum train location errors are 32.09 cm, 33.05 cm, 34.25 cm, 34.75 cm and 36.11 cm at the speed of 20 km/h, 40 km/h, 60 km/h, 80 km/h and 100 km/h, respectively (Fig. 14). In addition, 75% of the dynamic train positioning errors are less than 18 cm, 19 cm, 22 cm, 23 cm and 23.5 cm, respectively (Fig. 15), and the maximum time of train positioning is 51.32 ms (Table 2). Overall, the results of static and dynamic train positioning can meet the requirements of the IEEE 1474.1—2004 standard for train positioning.

In the present study, a novel kind of train positioning method, combining VLC and binocular stereo vision, is specifically provided to achieve the autonomous train positioning of CBTC systems in metro tunnels, which can be taken as a supplement to traditional train positioning methods. According to the above empirical study, the maximum errors of static train positioning and dynamic train positioning are 29.73 cm and 36.11 cm and the maximum time of train positioning is 51.32 ms, demonstrating the real train location ultimately. Meanwhile, the standard of train positioning precision has dropped slowly when a train is gradually picking up speed. Additionally, the precision of train positioning can reach the centimeter level in the proposed method, and the maximum error and the maximum time of train positioning are much smaller than ±10 m and 2 s, respectively, which are in line with the strict IEEE 1474.1—2004 standard. This study shows that the proposed method can satisfy the needs of autonomous train positioning in tunnels and provide some alternative approaches to train positioning of CBTC systems with V2V communications.