With the acceleration of urbanization and increase in population mobility, the passenger volume handled by urban rail transit systems is continuously increasing. Subways are one of the primary methods by which to effectively alleviate traffic pressure in large cities. Thus, accurate train position information is crucial for ensuring the safe operation of subways. The communication-based train control (CBTC) system, which is the primary subway train operation control system, relies on train position information to ensure the safe and efficient operation of trains on the line. Existing CBTC systems employ train positioning technologies, such as axle counters, wireless local area network (WLAN), long-term evolution (LTE), and cross-induction loops. However, these positioning technologies have issues, such as a large number of trackside devices, susceptibility to electromagnetic interference, scarcity of spectrum resources, and low positioning accuracy, which limit the development of CBTC systems. In recent years, visible light communication technology has developed rapidly, exhibiting features such as concurrent lighting and communication, simple equipment, strong anti-interference capabilities, a license-free spectrum, and high positioning accuracy. Such technology is widely applied in fields such as medical care, visual signal and data transmission, and underwater communication. Therefore, there are broad prospects for applications of visible light communication in CBTC systems for subway, including high-precision train positioning.

To address the issue of high-reliability positioning for trains operating in subway tunnels, a train positioning method using visible light communication with target tracking and particle filtering is proposed. First, LED lamps installed on the tunnel walls at fixed intervals were modulated. A camera mounted on top of the head of the train captures images of the LED lamps on the tunnel walls, demodulates them to obtain the corresponding LED identity information, and then queries the line database to acquire the world coordinates of the LED lamps. Second, using Kalman filtering and the Camshift algorithm, the target was tracked and its trajectory was predicted to quickly locate the positions of the LED lamps in the imaging plane. Thus, the center coordinates of the light spot were obtained. The attitude angle information obtained from the inertial measurement unit (IMU) and conversion relationship between the coordinate systems were then utilized in combination with geometric features to calculate the world coordinates of the train's position. Next, a particle filter fusion algorithm was employed, wherein the train positioning results were used as observations and the information obtained from the speed sensor was used as the state transition equation. This fusion process combines the training positioning results with the position information from the speed sensor to optimize the positioning accuracy of the train. Finally, a visible light communication positioning experimental platform was set up, and MATLAB software was used to conduct simulation experiments on the positioning algorithm. We conducted tests on the static and dynamic positioning performance of the proposed positioning method and compared it with other positioning methods.

A train positioning experimental platform with dimensions of 20 m×1.5 m×1.0 m is established to validate the effectiveness of the proposed train positioning algorithm. Twenty test points, spaced at 0.5 m intervals, are set up at a horizontal distance of 1.5 m from the LED lamps. The experiments are conducted at speeds of 0, 40, 60, 80, 100, and 120 km/h. To reduce random errors, each test is repeated five times, and the average of the five tests is considered as the final positioning result for performance comparison and analysis. The proposed method of train positioning is compared with the methods of perspective arcs and Bayesian forecasting using visible-light imaging. The experimental results show that the maximum positioning error in the static state is 24.53 cm, and 99% of the static positioning errors are within 25 cm, indicating good overall positioning performance. Under moving conditions, the closer the receiver is to the transmitter and the slower the speed of the train, the more accurate the positioning is. At a speed of 120 km/h, the maximum positioning error is 38.25 cm, and 84% of the dynamic positioning errors are within 25 cm. Hence, the overall positioning accuracy is stable and good, demonstrating that the proposed method achieves good positioning accuracy in most cases. In comparison, the positioning accuracy of the proposed algorithm is higher, and its overall positioning stability is better than those of visible light imaging positioning methods and methods based on machine learning and visible light imaging. These results suggests that the proposed method improves the accuracy and robustness of the train positioning.

To address the issues of insufficient positioning accuracy and the interruption of train positioning in subway tunnel environments, a visible light communication train positioning method based on target tracking and a particle filter is proposed. This method effectively reduces the number of trackside devices, provides high positioning accuracy, and possesses a strong anti-interference ability. By simulating the operating environment of subway trains, the proposed method uses LED lamps as transmitters and cameras as receivers to capture the identity information of LED lamps, and thereby convey position information. Through target prediction and tracking, this method quickly finds the position of the LED lamps in an image and obtains the center coordinates of the lamp spot, thus solving the problems of spectrum resource scarcity and susceptibility to electromagnetic interference in existing train positioning technologies. A particle filter algorithm is utilized to optimize the train positioning results, resolve the issue of positioning interruption when the number of LED lamps is insufficient during the train positioning process, and effectively improve the accuracy and robustness of positioning under train operating conditions. Hence, this method provides a new option for train positioning in CBTC systems oriented toward vehicle-to-vehicle communication.