Subways are necessary to alleviate the pressure of public transportation in metropolises. Currently, communications-based train control (CBTC) systems are the mainstream train operation control systems for subways. And the train positioning information obtained by positioning technology is an important parameter for ensuring safe train operation. Traditional train-positioning methods include balises, inductive loops, axle counters, wireless local area networks (WLANs), and long-term evolution (LTE) technology, but the above methods are facing certain inevitable shortcomings and hard to completely meet the needs of continuous positioning. The main drawbacks to these methods cover high cost, difficulty of maintenance, low precision, susceptibility to electromagnetic interference, limited spectrum resources and so on. Therefore, it is of great practical significance to conduct research on a train positioning method with high positioning accuracy, strong real-time performance, and low cost. With the recent advances in visible light communication (VLC), the technology is widely used in many fields owing to its high transmission rate, considering lighting and communication, environmental protection, and free spectrum resources. In this study, optical camera communication technology was applied to CBTC systems, using light-emitting diode (LED) lamps in the tunnel as the transmitter of an optical camera communication system. A train positioning method is proposed based on optical camera communication with double LED lamps, which has advantages such as high positioning accuracy, excellent real-time performance, low cost, and strong anti-interference ability. It can be used as a supplement or alternative to the existing train positioning methods in CBTC systems and has broad application prospects.
To improve the real-time performance and accuracy of train positioning, we propose a train positioning algorithm based on backpropagation (BP) neural network and optical camera communication. The proposed algorithm is divided into two parts: the detection and recognition of LED lamp information and a positioning algorithm of optical camera communication with double LED lamps. First, in the detection and recognition stage, the feature images of different LED lamps, such as frequency, area, and duty cycle, are obtained using an optical camera communication system. After training by machine learning based on the BP neural network, a training model is developed. Furthermore, in the train positioning stage, the imaging principle and geometric principle are adopted. The characteristics of the LED lamp images are extracted when the camera captures the images. After decoding, the identity (ID) information of the LED lamps is collected for the location coordinates of the lamps. Due to the distance between two adjacent LED lamps in the subway tunnel is 10 m (taking Chengdu Metro Line 1 as an example), the distance between the two adjacent LED lamps in the imaging plane can be determined using image processing technology. The focal length of the camera is provided, and the image coordinates of the LED lamps are calculated. Using geometric principles and coordinate conversion, the world coordinates of the train position can be determined. Finally, an experimental platform of optical camera communication for train positioning is established, and the proposed train positioning algorithm is verified by MATLAB. The static and dynamic performance of train positioning in the proposed algorithm is tested.
To verify the effectiveness of the algorithm for train positioning, we set up an experimental platform for train positioning with a size of 4 m×1 m×1.2 m and arranged 20 test points for testing 400 times at four different vertical heights. The average of five tests at each position was taken as the final train positioning results for comparison. The experimental results show that the average positioning error in the proposed algorithm did not fluctuate significantly when the vertical height changed. Considering the positioning results with a vertical height of 0 for error statistics, 90.1% of the error was less than or equal to 2.650 cm. To verify the effectiveness of the proposed train positioning algorithm during operation, when a train was running at a speed of 4 m/s in the same scenario, the algorithm was used to determine the real-time train position. The experimental results show that the position coordinates predicted by the proposed algorithm were basically the same as the actual trajectory coordinates, and the positioning error was within 5 cm. The positioning results could be obtained 20 times in 1 s, and the average positioning time was 51.34 ms. The positioning accuracy reached the centimeter level, and the positioning time was near the millisecond level using the proposed algorithm.
Based on the fixed arrangement of LED lamps in subway tunnels, we attempted to apply optical camera communication technology to CBTC systems. We classified the information characteristics of the collected LED lamps through a BP neural network to identify the ID information of LED lamps, which can solve certain problems such as few addresses, more recognition conditions, and susceptibility to electromagnetic interference in signal modulation. In addition, when the location coordinates of a single LED lamp are found, the proposed algorithm can achieve positioning results and reduce the positioning time without affecting accuracy. In summary, the proposed algorithm can improve the real-time performance and accuracy of train positioning as a supplement or alternative to the existing train positioning methods in CBTC systems.
