A subway shield door automatic linkage opening and closing control adjusting system

Abstract

This invention discloses an automatic linkage opening and closing control and adjustment system for subway platform screen doors, specifically relating to the field of rail transit automation control technology. It includes a central linkage control platform that communicates with the train automatic monitoring system to generate optimized strategy parameters for platform screen door opening and closing; and multiple edge intelligent gate control units that communicate with the central linkage control platform to receive the optimized strategy parameters. This invention solves the problems of slow response and single-point failure in traditional centralized control by combining global optimization with millisecond-level local response. It constructs a multi-layered active safety protection system combining prediction, identification, and emergency braking, which can accurately distinguish obstacle types and implement differentiated strategies, improving safety accuracy. It also possesses the ability to continuously evolve through federated learning and achieve high availability through redundant communication and neighbor collaboration, realizing a synergistic leap in efficiency, safety, and reliability for subway platform screen door systems.

Description

Technical Field

[0001] This invention relates to the field of rail transit automation control technology, and more specifically, to an automatic linkage opening and closing control and adjustment system for subway platform screen doors. Background Technology

[0002] The subway platform screen door system is a key piece of equipment for ensuring passenger safety, improving platform order, and achieving energy-saving operation. Its core function is to achieve precise, reliable, and safe linkage between the subway platform screen door and the train door. With the continuous growth of urban rail transit capacity and the continuous improvement of the level of intelligence, the core requirement for the linkage control between subway platform screen doors and train doors lies in the coordination of safety, efficiency, and reliability.

[0003] However, the current subway platform screen doors, which are widely used in systems based on centralized logic control or simple local feedback, struggle to balance real-time control with global optimization capabilities. Centralized control relies on a central processor to aggregate all information and issue commands, resulting in significant signal transmission and processing delays that fail to meet millisecond-level safety response requirements. Furthermore, the fixed control strategies cannot be dynamically adjusted based on real-time passenger flow and train stopping deviations, leading to low throughput efficiency. Existing technologies primarily rely on contact sensors or safety strips at the door gaps, triggering protection only after physical compression, which is a reactive measure. They cannot identify obstacle types such as people or luggage, nor can they provide pre-contact warnings, easily leading to frequent emergency stops or missed detections. The level of intelligence is low, and safety protection is passive and simplistic. Once deployed, the parameters and logic of traditional systems become fixed, unable to learn and evolve from operational data to continuously optimize performance. At the same time, the system has poor fault tolerance; failure of the central node or a single gate control unit often leads to partial or complete loss of functionality, lacking autonomous degradation and rapid recovery capabilities. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides an automatic linkage opening and closing control and adjustment system for subway platform screen doors to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an automatic linkage opening and closing control and adjustment system for subway platform screen doors, comprising: The central linkage control platform communicates with the train automatic monitoring system to generate optimized strategy parameters for the opening and closing of the platform screen doors; Multiple edge intelligent gating units are communicatively connected to the central linkage control platform to receive the optimization strategy parameters and autonomously generate and execute control commands based on local real-time sensor data. The digital twin synchronization optimization module is connected to the central linkage control platform and is used to generate the optimization strategy parameters through simulation and deduction.

[0006] Preferably, it further includes a multi-source heterogeneous sensor network, and the edge intelligent gating unit is communicatively connected to the multi-source heterogeneous sensor network; The multi-source heterogeneous sensor network includes at least a positioning base station for detecting the position of the train door, an environmental perception sensor for collecting passenger flow data, and a contact detection sensor embedded in the door. The edge intelligent gating unit is equipped with an AI inference module, which is used to fuse and process the data of the multi-source heterogeneous sensor network and execute differentiated security policies.

[0007] Preferably, the differentiated security strategy includes: If the identified object is a human body or human torso, the door will be controlled to retract urgently and remain open. If the object is identified as luggage, the door will be controlled to pause its movement and attempt to close slowly after a preset delay; If the identified object is a foreign object smaller than a preset size, the door will be controlled to complete the closing action normally, and an alarm record will be generated.

[0008] Preferably, the system is connected via a heterogeneous redundant communication network; The heterogeneous redundant communication network includes at least a wired link for transmitting real-time control commands, a wireless private network link for transmitting sensor data, and a cooperative link for direct communication between edge units. The edge intelligent gating unit can switch to the backup link within 50ms when a primary link failure is detected.

[0009] Preferably, the digital twin synchronization optimization module is used for: The digital twin synchronization optimization module is used to integrate the passenger flow simulation model and the train stop prediction model, and generate and optimize the combination of strategy parameters for different operating scenarios through offline simulation and deduction. The central linkage control platform is used to select matching strategy parameters from the strategy parameter combination according to the train operation plan, and send them to the corresponding edge intelligent gating unit in advance. The edge intelligent gating unit is used to take the received strategy parameters as a baseline control framework when the train arrives at the station, and dynamically adjust them in combination with real-time sensing data to generate the final execution command.

[0010] Preferably, the AI ​​inference module pre-installed in the edge intelligent gating unit is configured as a lightweight reinforcement learning agent; The reinforcement learning agent is configured to: use the optimization strategy parameters issued from the central linkage control platform as the baseline strategy, and combine them with real-time acquired local sensor data to output a dynamic adjustment amount for the platform screen door control command.

[0011] Preferably, the optimization strategy parameters are generated collaboratively by the central linkage control platform and the digital twin synchronous optimization module, and the generation method includes: Collect all historical operational data of the entire site, and train and solve the optimization strategy parameters in the virtual environment constructed by the digital twin synchronization optimization module through a multi-agent reinforcement learning algorithm.

[0012] Preferably, the system further includes a door servo drive module, which adopts a dual closed-loop control strategy. The inner loop is a closed loop based on the position, speed and current of the encoder, and the outer loop is a torque adaptive closed loop based on the feedback of the contact detection sensor. When the closing resistance is detected to exceed the safety threshold, the outer loop control takes priority, causing the door to retract.

[0013] Preferably, it also includes a self-learning optimization engine deployed on the central linkage control platform, used to periodically update the AI ​​model parameters distributed to each edge intelligent gating unit through a federated learning framework.

[0014] Preferably, the edge intelligent gating unit has neighbor collaboration and fault takeover functions; When a certain edge intelligent gate control unit fails, adjacent units automatically form a temporary control group to take over the basic linkage functions of the gate controlled by the failed unit.

[0015] The technical effects and advantages of this invention are as follows: By constructing a new architecture that integrates a central linkage control platform, edge intelligent gate control units, and a digital twin synchronous optimization module, control responsibilities are layered. The central platform and digital twin module focus on offline global strategy generation and optimization based on historical and simulation data to form optimized strategy parameters, while the edge units use lightweight AI models to make autonomous decisions and execute within milliseconds. This mode of central offline global optimization and edge online autonomous decision-making not only utilizes the powerful computing power of the center for complex calculations and long-term strategy planning, but also fully leverages the low latency and high reliability advantages of edge computing. This enables the system to dynamically adjust the action of each door in a complex and ever-changing operating environment, significantly improving overall passage efficiency and system response speed, while avoiding the single point of failure risk of centralized control. By integrating multi-source sensor data, obstacle types can be identified or differentiated in real time before and during the closing action, and differentiated safety strategies can be implemented. The strictest emergency retraction and opening is implemented for human bodies; a cautious approach of slow closing after pausing is adopted for luggage; and alarms are recorded for small foreign objects, completing the cycle. This balances safety and efficiency. At the final physical execution layer, torque adaptive outer loop control based on contact detection sensor feedback is adopted. When abnormal resistance is detected, this outer loop control has the highest priority and can immediately override the position command of the inner loop, forcing the door to retract. This provides the most direct and fastest physical safety redundancy. From predictive warning to accurate identification to physical emergency stop, the in-depth defense system realizes the leap from post-event response to pre-event prevention and intelligent handling in the event of an incident in safety protection. Through the multi-agent reinforcement learning algorithm of the digital twin module, the strategy is continuously simulated and optimized in a virtual environment. The federated learning framework enables the AI ​​models distributed across all edge units to evolve collaboratively without sharing original sensitive data. The system continuously optimizes performance using actual operating data from the entire line, allowing the system control strategy to adapt to long-term changes in passenger flow patterns and equipment aging. In addition, through a heterogeneous redundant communication network and a neighbor collaboration and fault takeover mechanism, the system can automatically switch or reorganize in a very short time when a communication link or a single edge unit fails, with the adjacent unit taking over control. This ensures that basic linkage functions are not interrupted, meeting the stringent reliability requirements of rail transit scenarios. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] This invention provides an automatic linkage opening and closing control and adjustment system for subway platform screen doors, which is used to realize safe, efficient and adaptive linkage control between subway platform screen doors and train doors.

[0019] like Figure 1 As shown, this system mainly includes a central linkage control platform, multiple edge intelligent gate control units, a digital twin synchronization optimization module, a multi-source heterogeneous sensor network, a heterogeneous redundant communication network, a gate servo drive module, and related software function modules.

[0020] 1. Centralized linkage control platform The central control platform is deployed in the station control room, and its hardware is based on high-performance industrial servers or server clusters. The platform communicates in real time with the metro's existing Automatic Train Supervision (ATS) system through standard industrial communication protocols such as OPCUA or customized data interfaces to obtain train operation plans, real-time locations, train formation information, and dispatching instructions.

[0021] One of the core functions of the platform is to generate "optimized strategy parameters" for the opening and closing of platform screen doors. These parameters are not direct motion commands, but rather include advanced control settings such as the baseline value for the door opening width, suggested opening and closing speed curves, safe dwell time, and response modes for different passenger flow densities. The platform generates these parameters based on a comprehensive decision made from ATS information, historical operational data, and optimization suggestions from the digital twin module.

[0022] 2. Digital Twin Synchronization Optimization Module The digital twin synchronization optimization module is logically tightly coupled with the central platform and can be deployed on the same server or an independent simulation computing node. It builds and maintains a high-fidelity virtual environment that is consistent with the physical station's geometric dimensions and equipment layout.

[0023] The core of this module lies in integrating and running two key analytical models: a passenger flow simulation model and a train stop prediction model. The passenger flow simulation model can simulate passenger gathering, movement, and boarding / alighting behavior on the platform based on historical passenger flow data, such as the OD matrix. The train stop prediction model can predict the deviation between the actual stopping position and the standard position of the train after it arrives at the station, based on track, signal, and historical stopping data.

[0024] The workflow of this module is as follows: it calls the above model in the form of offline or low-priority background tasks, and conducts a large number of simulations for different operational scenarios, such as morning peak, evening peak and emergencies. By simulating the system performance under different strategy parameters, it evaluates its traffic efficiency and safety indicators, and then generates and optimizes a set of "strategy parameter combinations" for different scenarios as an expert database for the central platform's decision-making.

[0025] 3. Multiple edge intelligent gating units Each edge intelligent gate control unit controls a movable shielded door. Its main hardware is an industrial controller that integrates AI computing units such as GPUs, NPUs, or high-performance MCUs. Each unit is connected to the central linkage control platform and a multi-source heterogeneous sensor network through a heterogeneous redundant communication network.

[0026] Its core task is to receive the "optimization strategy parameters" issued by the central platform and make autonomous decisions within the real-time control cycle of train arrival. Specifically, the pre-installed AI inference module in the unit is configured as a lightweight reinforcement learning agent, for example, using the PPO or DQN algorithm framework and undergoing model pruning and quantization. The agent's policy network uses the aforementioned optimized strategy parameters as its initial weights or benchmark value function.

[0027] During operation, the reinforcement learning agent works in millisecond cycles: it receives and fuses multi-source sensor data in real time to form a state vector describing the current local environment; then, using this state vector and optimization policy parameters as input, it outputs a "dynamic adjustment amount" to the baseline control command through forward inference of the neural network, such as increasing or decreasing the baseline door opening width by a certain amount, or fine-tuning the parameters of the speed curve. Finally, the unit converts the adjusted result into an execution command that can directly drive the door servo drive module.

[0028] 4. Multi-source heterogeneous sensor networks Multi-source heterogeneous sensor networks provide real-time, multi-dimensional environmental perception data for edge intelligent decision-making, which includes at least: Positioning base stations used to detect the location of train doors: Specifically, UWB ultra-wideband positioning systems can be used. The base stations are deployed along the platform, communicate with onboard tags, and calculate and provide the precise position information of the train doors relative to the corresponding platform screen doors in real time, with an accuracy of up to centimeters.

[0029] Environmental sensing sensors used to collect passenger flow data include at least a millimeter-wave radar array or a stereo vision camera. The millimeter-wave radar is used to stably generate a 3D point cloud of the area in front of the door under complex lighting conditions to detect the number, location, and movement trajectory of passengers. The stereo vision camera is used to assist in the contour recognition and classification of passengers and luggage, as well as the detection of abnormal behavior.

[0030] Contact detection sensors embedded in the door: Specifically, a distributed fiber optic grating sensor array can be used, which is deployed in the sealing strip of the movable door. When an object is clamped in the door gap, the wavelength of the fiber optic grating will shift. Through demodulation, the location and force of the force can be sensed with high sensitivity.

[0031] The AI ​​inference module of the edge intelligent gating unit performs spatiotemporal alignment and feature-level fusion on the aforementioned multi-source data to form a unified environmental perception. Based on this, the system executes differentiated security policies: If the identified object is a human body or human torso, the highest priority security response is immediately triggered, controlling the door to retract urgently and remain open for at least 3 seconds.

[0032] If the object is identified as luggage, the door movement is paused, a holding torque is applied to prevent the door from sliding, and after a 2-second delay, the door is slowly closed at a lower speed, such as 50% of the normal speed.

[0033] If the identified object is smaller than a preset size, such as a foreign object with a diameter of less than 2 cm, the event log is recorded locally, and the door is controlled to complete the current closing cycle normally. At the same time, the alarm information including time and location information is uploaded to the central platform.

[0034] 5. Heterogeneous Redundant Communication Networks The system connects each module through a heterogeneous redundant communication network to ensure the reliability and real-time performance of data transmission. This network includes at least three independent physical or logical links: Wired links for transmitting real-time control commands: Time-Sensitive Network (TSN) Ethernet is preferred to provide deterministic low latency for door servo drive commands, typically requiring a transmission channel of <10ms.

[0035] Wireless private network links for transmitting sensor data: 5G private network 5G-U or industrial Wi-Fi 6 can be used to carry high-bandwidth data streams from visual sensors, radar, etc.

[0036] Collaborative links for direct communication between edge units: CANFD bus, i.e., Controller Area Network Flexible Data Rate, can be used to achieve low-latency state synchronization and direct transmission of emergency commands between adjacent edge intelligent gating units.

[0037] Each edge intelligent gating unit integrates a multi-mode communication gateway and connects to the aforementioned networks. The system pre-sets primary and backup links for different types of data, such as control commands, video streams, and heartbeat signals. The edge intelligent gating unit continuously monitors the link status, and when it detects that the communication quality of the primary link is below the threshold or is completely interrupted, it can automatically switch the data stream to the backup link within 50ms to ensure uninterrupted communication.

[0038] 6. Door Servo Drive Module The door servo drive module consists of a servo motor, a driver, a high-precision absolute encoder, and a transmission mechanism, and is controlled by the corresponding edge intelligent door control unit.

[0039] This module employs a dual closed-loop control strategy. The inner loop is a position-speed-current closed loop based on motor encoder feedback, ensuring the door accurately tracks its movement trajectory. The outer loop is a torque adaptive closed loop based on contact detection sensors, such as a distributed fiber optic grating array, providing real-time feedback signals. The outer loop continuously monitors the resistance torque during the closing process. When the detected resistance exceeds a preset safety threshold, which can be set based on human biomechanical data, such as 150N, the outer loop control will receive the highest priority, forcibly overriding the position command of the inner loop and immediately driving the door to perform a retraction action. This is the last direct physical guarantee for achieving safety and preventing pinching.

[0040] 7. System's self-learning and fault-tolerance functions To improve the long-term performance and reliability of the system, the system also includes the following advanced features: The self-learning optimization engine, deployed on a central control platform, periodically initiates a federated learning process, such as every night. Specifically, the central platform distributes the current global AI model parameters to each edge unit. Each unit uses newly generated local data to perform a round of local training, calculating the updated gradients of the model parameters (not the original data), and uploads the encrypted gradients. The central platform aggregates all gradients, updates the global model, and then distributes the new parameters. In this way, the system can continuously evolve while protecting data privacy.

[0041] Neighbor collaboration and fault takeover functions: Each edge intelligent gate control unit maintains periodic heartbeat communication with its left and right neighboring units through a collaborative link, such as CANFD. When a unit is detected to have a fault, such as heartbeat loss or self-test abnormality, its left and right neighboring units automatically form a temporary control group through the collaborative link. They share the necessary sensor status information and, through a preset simple collaborative algorithm, jointly decide to generate basic door opening and closing control commands for the faulty door, thereby taking over its basic linkage function. At this time, the entire system enters a degraded "local collaborative mode" and alarms the central platform.

[0042] 8. Brief Description of System Workflow Within a typical control cycle: Before the train enters the station, the central platform matches and pre-downloads parameters from the optimization strategy parameter library provided by the digital twin module to the corresponding edge units according to the ATS plan. During the train's entry into the station and after it stops, each edge unit autonomously generates the final control command based on the pre-loaded parameters and real-time fused sensor data, driving the door to complete precise and safe linkage with the train door. After the entire process is completed, the performance data is recorded and used for subsequent federated learning updates.

[0043] Through the above-mentioned hardware and software collaborative architecture design, this invention realizes the transformation of platform screen door control from traditional fixed procedures or simple responses to intelligent collaborative control based on prediction, real-time optimization, autonomous decision-making and continuous learning capabilities, which significantly improves the safety, efficiency and adaptability of subway operation.

[0044] Finally, it should be noted that the accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other. In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An automatic linkage opening and closing control and adjustment system for subway platform screen doors, characterized in that, include: The central linkage control platform communicates with the train automatic monitoring system to generate optimized strategy parameters for opening and closing the platform screen doors; Multiple edge intelligent gating units are communicatively connected to the central linkage control platform to receive the optimization strategy parameters and autonomously generate and execute control commands based on local real-time sensor data. The digital twin synchronization optimization module is connected to the central linkage control platform and is used to generate the optimization strategy parameters through simulation and deduction.

2. The automatic linkage opening and closing control and adjustment system for subway platform screen doors according to claim 1, characterized in that, It also includes a multi-source heterogeneous sensor network, and the edge intelligent gating unit is communicatively connected to the multi-source heterogeneous sensor network; The multi-source heterogeneous sensor network includes at least a positioning base station for detecting the position of the train door, an environmental perception sensor for collecting passenger flow data, and a contact detection sensor embedded in the door. The edge intelligent gating unit is equipped with an AI inference module, which is used to fuse and process the data of the multi-source heterogeneous sensor network and execute differentiated security policies.

3. The automatic linkage opening and closing control and adjustment system for subway platform screen doors according to claim 2, characterized in that, The differentiated security strategy includes: If the identified object is a human body or human torso, the door will be controlled to retract urgently and remain open. If the object is identified as luggage, the door will be controlled to pause its movement and attempt to close slowly after a preset delay; If the identified object is a foreign object smaller than a preset size, the door will be controlled to complete the closing action normally, and an alarm record will be generated.

4. The automatic linkage opening and closing control and adjustment system for subway platform screen doors according to claim 1, characterized in that, The system is connected via a heterogeneous redundant communication network; The heterogeneous redundant communication network includes at least a wired link for transmitting real-time control commands, a wireless private network link for transmitting sensor data, and a cooperative link for direct communication between edge units. The edge intelligent gating unit can switch to the backup link within 50ms when a primary link failure is detected.

5. The automatic linkage opening and closing control and adjustment system for subway platform screen doors according to claim 1, characterized in that, The digital twin synchronization optimization module is used for: The digital twin synchronization optimization module is used to integrate the passenger flow simulation model and the train stop prediction model, and generate and optimize the combination of strategy parameters for different operating scenarios through offline simulation and deduction. The central linkage control platform is used to select matching strategy parameters from the strategy parameter combination according to the train operation plan, and send them to the corresponding edge intelligent gating unit in advance. The edge intelligent gating unit is used to take the received strategy parameters as a baseline control framework when the train arrives at the station, and dynamically adjust them in combination with real-time sensing data to generate the final execution command.

6. The automatic linkage opening and closing control and adjustment system for subway platform screen doors according to claim 1, characterized in that, The AI ​​inference module pre-installed within the edge intelligent gating unit is configured as a lightweight reinforcement learning agent. The reinforcement learning agent is configured to: use the optimization strategy parameters issued from the central linkage control platform as the baseline strategy, and combine them with real-time acquired local sensor data to output a dynamic adjustment amount for the platform screen door control command.

7. The automatic linkage opening and closing control and adjustment system for subway platform screen doors according to claim 5, characterized in that, The optimization strategy parameters are generated collaboratively by the central linkage control platform and the digital twin synchronous optimization module, and the generation methods include: Collect all historical operational data of the entire site, and train and solve the optimization strategy parameters in the virtual environment constructed by the digital twin synchronization optimization module through a multi-agent reinforcement learning algorithm.

8. The automatic linkage opening and closing control and adjustment system for subway platform screen doors according to claim 2, characterized in that, The system also includes a door servo drive module, which adopts a dual closed-loop control strategy. The inner loop is a closed loop based on the position, speed and current of the encoder, and the outer loop is a torque adaptive closed loop based on the feedback of the contact detection sensor. When the closing resistance is detected to exceed the safety threshold, the outer loop control takes priority, causing the door to retract.

9. The automatic linkage opening and closing control and adjustment system for subway platform screen doors according to claim 1, characterized in that, It also includes a self-learning optimization engine deployed on the central linkage control platform, which is used to periodically update the AI ​​model parameters distributed to each edge intelligent gating unit through a federated learning framework.

10. The automatic linkage opening and closing control and adjustment system for subway platform screen doors according to claim 1, characterized in that, The edge intelligent gating unit has neighbor collaboration and fault takeover functions; When a certain edge intelligent gate control unit fails, adjacent units automatically form a temporary control group to take over the basic linkage functions of the gate controlled by the failed unit.

Images