Full-automatic control device and control method for boarding bridge

Abstract

The invention discloses a boarding bridge full-automatic control device and a control method, and relates to the technical field of airport ground equipment automation. The device comprises a physical isolation special network system, a bimodal sensing system, a PLC control system, an encryption storage system, a man-machine interaction system and a security policy execution system. Data isolation is achieved through an independent private network, cabin door recognition and obstacle detection are achieved through the dual-mode sensing technology of a camera and a laser radar, full-automatic approaching and bridge withdrawing operation is achieved in combination with a PLC control system, and operation safety of the system is guaranteed through multiple encryption and a safety strategy. The method can improve the operation efficiency of the boarding bridge, reduces manual intervention, enhances the operation safety, and is suitable for automatic upgrading of various airport boarding bridges.

Description

Technical Field

[0001] This invention belongs to the field of airport ground equipment automation technology, specifically relating to a fully automatic control device and method for boarding bridges. Background Technology

[0002] Traditional boarding bridge operations have long relied on manual on-site control, which presents several technical bottlenecks: In terms of operational efficiency: the manual docking process requires repeated position calibration, which limits the timeliness of flight support and makes it difficult to meet the needs of high-density flight turnaround. From a safety management perspective: Operators are limited by their field of vision, which can easily create blind spots in their perception of critical parts of the aircraft (such as engines and wings), and there is an inherent delay in emergency response, resulting in a persistently high risk of collision. From an ergonomics perspective: Operators need to maintain a high level of concentration in a complex environment with multiple variables, resulting in a heavy workload and fatigue accumulation that can easily lead to misoperation. At the data governance level: the entire operation process lacks a systematic data recording and traceability mechanism, fault tracing and performance optimization rely too much on experience-based judgment, and the decision-making is not scientific enough.

[0003] While some semi-automatic boarding bridges currently incorporate basic automation technologies, they still have significant technical shortcomings: Insufficient reliability of the perception system: The positioning accuracy of the visual positioning module is significantly reduced in complex lighting environments (such as strong light, backlight, and low illumination), and the lidar has weak ability to suppress dynamic environmental interference (such as the movement of people and equipment on the ground), resulting in poor stability of environmental perception. Weak cybersecurity protection: The control network lacks a physical isolation and defense-in-depth system, lacks proactive defense mechanisms, and faces significant risks of unauthorized access and data tampering, making it difficult to meet industrial control network security standards; Limited control precision performance: The motion control module lacks millimeter-level precision braking capability for the end effector, and is prone to potential damage to the aircraft door due to control deviations in the docking margin. The security mechanism is not well designed: the identity authentication mechanism for switching operating modes is not strong enough, the permission boundaries are vague, and there is a significant delay in the emergency response link, making it difficult to achieve rapid closed-loop handling of risks.

[0004] The aforementioned technical shortcomings mean that existing solutions cannot meet the high reliability and high security requirements of modern airports for fully automated operation of boarding bridges, thus hindering the intelligent upgrading of airport ground support processes.

[0005] The information disclosed in this background section is intended only to enhance the understanding of the overall background of the invention and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a fully automatic control device and method for boarding bridges to overcome the problems in the prior art.

[0007] To achieve the above objectives, the present invention provides an automatic boarding bridge docking and evacuation system, comprising the following hierarchical modules: The perception layer is used to receive environmental data and includes: an aircraft positioning module, an anchor marker positioning module, a ground equipment status recognition module, a path obstacle and anti-collision detection module, a multi-sensor fusion algorithm, and a multi-modal collaborative perception system composed of a camera-based visual door recognition module and a lidar-based multi-modal collaborative perception system. The multi-modal collaborative perception system is configured as a spatiotemporal synchronous fusion processing unit. The camera processes images using OpenCV and employs an adaptive distortion correction algorithm, while the lidar processes point clouds using the PCL library and integrates a dynamic background modeling module. The decision layer is used to receive data from the perception layer and generate path planning instructions. It includes: a multi-threaded system architecture, a take-off and withdrawal path control algorithm, system anomaly monitoring, and a safety-priority PLC control system based on the PROFINET architecture (including a CPU1512SP F-1 PN controller, IO module and motion control unit, supporting three operation mode switching, with local operation having the highest priority and remote switching requiring two-factor authentication). The execution layer is used to drive mechanical actions according to decision instructions, and includes: a human-machine interaction module and a drive execution module; The storage and network layer is used for data storage and information exchange, including: an attack-resistant storage system using MySQL 8.0 database, dynamically rotating AES-GCM encryption for alarm information and status data, storage frequency once per second, and a physical isolation-dynamic defense network system constructed from fiber optic cables and network cables; Each module interacts through an independent, closed network, and a publish-subscribe model based on delegates and event queues is used to achieve cross-thread asynchronous communication in order to decouple control.

[0008] Further preferred configurations of the perception layer include: Aircraft and anchorage marker positioning module: Based on a deep vision learning model, it achieves anti-occlusion recognition and positions the aircraft by analyzing images of the nose, landing gear, doors, and anchorage markers. Ground equipment status recognition module: determines the connection status between ground equipment and aircraft through visual recognition, and sends occupancy signals to constrain operation permissions; Visual door recognition module: after the camera acquires images, it performs online calibration to remove distortion, activates multi-frame super-resolution reconstruction under low light, inputs the YOLOv8m model to convert it into TensorRT format to output the door bounding box, improves the positioning accuracy to ±3px through perspective transformation and sub-pixel edge fitting, and sends the coordinate data to the decision system via MQTT after CRC verification; The lidar collision avoidance module receives spatiotemporally asynchronous point cloud data from lidar and blind spot radar, aligns the timestamps using a motion compensation algorithm, and transforms them to the bridge coordinate system. It then uses a combination of plane fitting and deep learning to filter out ground and self-point clouds. The remaining point clouds are clustered using DBSCAN and a three-dimensional Euclidean distance field is constructed. When an obstacle is detected encroaching on the dynamic safety boundary, the power is cut off; otherwise, a pre-alarm is issued. The dynamic safety boundary is ≥50cm for the engine area and ≥30cm for the wing area.

[0009] Further preferred configurations for the decision-making level include: Multi-threaded system architecture: divided into execution mechanism, decision control, data communication, and human-computer interaction threads, with modular isolation to prevent information blockage; The approach and withdrawal path control algorithm includes a hierarchical finite state machine and approach and withdrawal path planning. The hierarchical finite state machine manages the approach and withdrawal mode switching and integrates safety strategies to monitor risks. The approach and withdrawal path planning includes an initial pre-approach point and a camera fine-tuning stage. The fine-tuning adopts the pinhole imaging principle and the formula θ=arctan(λ×ΔP). The withdrawal path includes an initial pre-approach point and anchor marker positioning. When the marker fails, a multi-sensor fusion algorithm is used. The motion control module of the safety-type PLC control system: When automatically approaching, the contact prediction model is activated when the data from both distance sensors are ≤60cm and the difference is <10cm. The speed is reduced by 10% in the intervals of 60-50cm, 50-40cm, etc., and the speed is idled to 0cm and braked when the distance is 10cm. When removing the bridge, a dynamic obstacle avoidance topology map is constructed based on the LiDAR data. Active safety strategy: integrates minimum risk strategy (strategy triggered according to risk level when operating conditions deviate) and emergency takeover link (local emergency stop button directly connected to PLC power relay, response delay <10ms).

[0010] Further preferred, the human-computer interaction module of the execution layer includes: a one-click operation interface: providing encrypted user account login, operation mode switching buttons, and emergency stop function; encryption adopts a combination of salt-valued MD5 encryption and behavior auditing, and forces passwords to contain letters, numbers, and special characters; 3D situation display unit: Real-time rendering of motion trajectory, obstacle distribution, and aircraft position; Parameter configuration interface: Supports dynamic adjustment of extrinsic parameter matrix, geometric model, and safety threshold; Status visualization unit: Displays operating status via HMI and LED display screen; the bridge body is equipped with warning language and warning light modules. It supports dual control via HMI touchscreen and physical buttons.

[0011] More preferably, the driver execution module of the execution layer includes: Core control command execution: Receives path planning commands, converts them into parameters such as walking speed and angle, and controls the walking mechanism, docking compartment rotation, canopy extension and retraction, etc. through the IO module; Multi-system collaborative communication: Real-time interaction with image recognition and lidar systems via PROFINET protocol, coordinating the synchronization of sensing data and control commands; acting as a communication hub to process and forward camera video signals and lidar data. Safety and anomaly response: When a high-risk value is triggered, the PLC will stop the walking mechanism and send an alarm; local / remote / fully automatic mode access interlocking is implemented to ensure the highest control authority of the local operator console; Hardware interface management: The PLC connects to the sensor interface to collect data such as bridge angle and hatch position, and uses a first-order low-pass filter combined with an integral compensation algorithm to suppress high-frequency noise.

[0012] Further preferred configurations for the storage and network layers include: Anti-attack storage system: The stored content includes more than 50 types of alarm information (including VDGS interlock signals, safety shoe fault signals, etc.) and status data (including bridge length, infrared distance, etc.). Blockchain timestamps are added to the encrypted data, and local and back-end use sharded redundant storage. Physical Isolation-Dynamic Defense Network System: Deployed in DCR, SCR, and bridge computer rooms, it achieves three-layer segmentation of management network, control network, and sensor network through logical isolation gateways; the topology includes S5560X-30F-EI switches (with built-in deep packet inspection engine to intercept non-PROFINET protocol data) in DCR computer rooms, fiber optic patch devices in SCR computer rooms, and S5130S-PWR-EI switches (PoE power supply module with overload fuse protection) in bridge computer rooms.

[0013] Further preferably, the path obstacle and collision avoidance detection module of the perception layer includes: Point cloud preprocessing module: processes radar point clouds, accurately identifies and distinguishes ground and obstacles, and removes point clouds from the bridge body; Multi-radar synchronization processing module: Enables simultaneous multi-radar, multi-angle obstacle identification through a caching mechanism; Collision risk assessment unit: It analyzes the point cloud type and distance through distance clustering algorithm, and obtains the risk value through collision risk calculation formula. When the risk value is too high, it sends a signal to the decision layer to suspend operation.

[0014] Further preferred, system anomaly monitoring includes: data communication anomaly detection: including timeout determination and content verification anomaly, with content verification anomalies including data integrity and logical rationality; Sensor failure detection: Sends a risk value when sensor data is lost; Risk detection and assessment: including LiDAR obstacle risk assessment and integrated sensor data risk assessment.

[0015] The system control method includes the following steps: S1. Spatiotemporal synchronization phase: Aligning the lidar timestamps using a dedicated protocol; S2. Collaborative perception stage: Map the visually recognized door coordinates and the lidar point cloud to a unified coordinate system to generate a multi-source confidence fusion trajectory; S3. Path Planning and Execution Phase: The decision-making layer executes state machine decisions, generates path tracking instructions and sends them to the drive unit; it updates the state and receives manual instructions through the interaction thread; S4. Adaptive docking phase: The bridge pitch angle is dynamically adjusted according to the aircraft door height, and the contact prediction model is activated at a distance of 50cm from the door; S5. Safety Handling Phase: When an anomaly or collision risk is detected, an emergency strategy is triggered and an alarm is sent. The emergency strategy is to cut off power or stop the vehicle abruptly.

[0016] Compared with the prior art, the present invention has the following beneficial effects: Improved operational efficiency: The fully automated docking and bridge removal process requires no manual intervention, reducing docking time to within 2 minutes and minimizing the risk of flight delays; Enhanced safety: The dual-modal sensing system enables high-precision environmental detection, and collision risk threshold control and safety no-go zone setting can effectively prevent collision accidents; Data security and reliability: Independent dedicated network and multiple encryption mechanisms ensure the security of data transmission and storage, meeting airport information security requirements; Flexible and controllable operation: Supports switching between local, remote and fully automatic modes, and a priority mechanism ensures manual takeover capability in emergencies; High traceability: Comprehensive data recording and storage functions facilitate operation process traceability, fault analysis, and performance optimization. Attached Figure Description

[0017] Figure 1 This is a hardware diagram of the boarding bridge docking and evacuation system according to the present invention; Figure 2 A flowchart illustrating the functional composition of the boarding bridge docking and evacuation system according to the present invention; Figure 3 A schematic diagram of the sensing layer structure of the boarding bridge docking and evacuation system according to the present invention; Figure 4 This is a schematic diagram of the decision-making layer structure of the boarding bridge docking and evacuation system according to the present invention; Figure 5 A schematic diagram of the execution layer boarding bridge structure according to the boarding bridge docking and evacuation system of the present invention; Figure 6 This is a schematic diagram of the protection strategy described in the boarding bridge docking and evacuation system according to the present invention. Detailed Implementation

[0018] The specific embodiments of the present invention will be described in detail below, but it should be understood that the scope of protection of the present invention is not limited to the specific embodiments.

[0019] Example 1

[0020] Reference Figure 1 The system for automatic passenger boarding bridge docking and disembarking is shown in block diagram format. The main hardware of the system is divided into two parts: local and remote. The local part includes (1) the PLC controller 3210 and boarding bridge drive mechanism 3212 of the execution layer 300. (2) the vision sensor 110, lidar 130 and bridge body sensor 140 installed in the perception layer 100. The remote part includes (1) the PLC controller 3210 of the execution layer 300, (2) the deep vision learning model 110 of the perception layer 100, (3) the decision algorithm computer 220 of the decision layer 200, and (4) the human-computer interaction console 311 and the monitoring screen 312. The execution layer 300 and the decision layer 200 communicate using the Profinet communication protocol, and the decision layer 200 and the perception layer 100 communicate using the MQTT communication protocol. Arrows are used to indicate the direction of information interaction between the block diagrams.

[0021] Reference Figure 2 The flowchart illustrates the perception layer 100, the decision-making layer 200, and the execution layer 300.

[0022] Furthermore, the perception layer 100 is used to receive environmental data. It includes an aircraft and anchor marker location module 110, a ground equipment usage status detection module 120, a path obstacle and collision avoidance detection module 130, and a multi-sensor fusion algorithm 140. Among them, the aircraft and anchorage marker positioning and detection module 110 is built based on a deep visual learning model. Through the feature extraction layer, it performs multi-dimensional analysis on the aircraft nose contour features, nose landing gear structural features, door morphology features, and anchorage marker visual features to achieve target recognition and spatial positioning under anti-obstruction conditions. The ground equipment operation status monitoring module 120 uses visual recognition technology to detect the physical connection status between ground equipment and aircraft in real time, and sends equipment occupancy status signals to the decision-making layer to dynamically adjust operating permissions. The path obstacle and anti-collision detection module 130 includes a closed-loop detection system consisting of a point cloud preprocessing unit, a multi-radar synchronous processing unit, and a collision risk assessment unit. The multi-sensor fusion positioning module 140 uses multiple types of sensors distributed at key nodes of the bridge to calculate the three-dimensional spatial coordinates of the bridge body in real time and upload them to the decision-making layer.

[0023] Specifically, the deep vision learning module includes three core components: (1) a training dataset, which integrates multi-weather condition calibration images and dynamic illumination compensation samples, and improves environmental adaptability through data augmentation technology; (2) an edge distortion processing algorithm, which receives the original image stream from the ultra-wide-angle camera and outputs geometrically corrected images using a polynomial distortion correction model; and (3) a video matrix processor, which integrates video signals from multiple cameras, generates a standardized image frame sequence that can be processed by the recognition engine, and outputs the monitoring screen synchronously.

[0024] The workflow of the path obstacle and collision avoidance detection module 130 includes: a point cloud preprocessing unit performs noise reduction filtering on the original radar point cloud, and separates the ground reflection points, aircraft feature points and obstacle point clouds based on a feature classification algorithm, while filtering out interference from the bridge's own point cloud; a multi-radar synchronous processing unit realizes spatial data fusion of multiple radar devices through a timestamp alignment mechanism and cache queue management; and a collision risk assessment unit applies a clustering algorithm to perform spatial density analysis on the obstacle point cloud, and triggers an emergency stop command when the risk value exceeds a preset threshold, based on a preset collision risk formula.

[0025] The sensor network of the multi-sensor fusion positioning module 140 includes three types of key sensors: bridge abutment angular displacement sensors, located at the hub of the receiving platform, used to detect the horizontal deflection angle of the bridge body; bridge strain sensors, equidistantly distributed along the main body of the covered bridge, used to measure the deflection deformation of the bridge body; and wheel frame tilt sensors, set at the fulcrum of the traveling mechanism, used to monitor the attitude changes of the supporting structure. The data from each sensor are calibrated using a filtering algorithm, and the final output is the positioning data of the bridge end effector in the global coordinate system.

[0026] Reference Figure 3 This diagram illustrates the connection between the visual sensors and LiDAR in the perception layer 100 of this system. Figure 2 As shown, it includes: Visual sensor group 110, used to acquire visual images LiDAR sensor group 130 is used to acquire surrounding point cloud data. Linux device 150 is used to process the acquired LiDAR sensor group data, store the deep learning visual recognition model, and send the calculation results to the Windows device of decision layer 200.

[0027] Windows device 200 is used by the decision-making layer 200 to calculate the running path and monitor anomalies.

[0028] PLC device 321 is used by execution layer 300 to perform data processing and communication, and to perform operations.

[0029] The images acquired by the vision sensor group 110 include the aircraft nose, aircraft doors, anchor markers, and ground bridge-mounted equipment.

[0030] The point cloud data acquired by the lidar sensor group 130 includes bridge point cloud, ground point cloud, and aircraft point cloud.

[0031] Correspondingly, the Linux devices in the perception layer 100 are specifically used to identify and determine the positions of the aircraft nose, nose landing gear, aircraft doors, anchor markers, and the activation status of ground bridge equipment through deep learning visual recognition models; by filtering out irrelevant point clouds and clustering point clouds, they can realize obstacle detection on the wheel frame travel path and realize the function of preventing collisions between the boarding bridge and the aircraft, aviation ground support equipment, and personnel.

[0032] In this embodiment of the invention, after the aircraft is in position, the visual sensor captures an image of the aircraft and sends the image data to a Linux device. A deep learning visual model identifies the aircraft's position coordinates and continuously sends this result to the Windows device of the decision layer 200. When the decision layer 200 triggers a docking command, the boarding bridge moves, and the lidar sensor group continuously scans the surrounding point cloud data. When an obstacle is in the path and enters a danger threshold range, a risk threshold is triggered, and the boarding bridge stops moving. When the system is in an abnormal state, and the boarding bridge deviates from its normal path, and there is a risk of collision with the aircraft at the bridgehead, the decision layer 200 triggers a risk threshold, and the boarding bridge stops moving. When the boarding bridge is in an evacuation state, the visual sensor captures evacuation markers. When the deep learning visual model identifies that the boarding bridge wheel frame has entered the coordinate range specified by the evacuation markers, the decision layer 200 determines that the evacuation is successful. The obstacle detection function for the evacuation path is the same as the obstacle detection function for the docking path.

[0033] The decision layer 200 receives environmental data transmitted from the perception layer and generates path planning instructions. It includes a multi-threaded system architecture 210, a path control algorithm 220, and system anomaly monitoring 230. The multi-threaded control system architecture 210 adopts a four-layer logical division: an actuator drive layer, a decision control core layer, a data communication management layer, and a human-machine interface layer. Each functional thread achieves physical isolation of data channels through an inter-process communication isolation mechanism, effectively avoiding the risk of data blocking caused by thread competition. In particular, the data communication layer has a priority system and implements a bandwidth guarantee strategy for critical control commands.

[0034] The retrieval path control algorithm 220 includes two major functional components: (1) a hierarchical finite state machine, which is equipped with a state transition logic with a safety interlock mechanism to coordinate the switching of working states between the retrieval mode and the withdrawal mode, and integrates an environmental risk perception strategy; (2) an intelligent path planning engine, which constructs retrieval path optimization functions and withdrawal path optimization functions based on kinematic models.

[0035] The working logic of the hierarchical finite state machine includes: activating the three-dimensional obstacle avoidance strategy during the approach and evacuation phases, initiating the emergency braking response protocol, and triggering the state machine to enter the protection mode and cut off the power output when the lidar detects that the collision risk value exceeds the safety threshold.

[0036] The intelligent path planning engine's computational process includes: In the docking path calculation stage, a two-stage precise positioning strategy is adopted. First, the initial pre-docking point coordinates are determined through visual recognition. After reaching the pre-docking point, the door position is then located using pinhole imaging based on the visual servo control principle. The pinhole imaging positioning formula is: θ = arctan(λ×ΔP), where θ represents the pose adjustment angle, λ is the camera focal length calibration parameter, and ΔP is the feature point pixel coordinate offset. In the evacuation path calculation stage, the anchor marker visual navigation scheme is used first. When the anchor marker recognition fails, it automatically switches to the multi-sensor fusion navigation mode. By comparing the sensor parameters of the current and initial positions, the boarding bridge is guided back to the anchor position.

[0037] The system anomaly monitoring module 230 includes a triple protection mechanism: (1) a data communication diagnosis unit, which implements a communication link timeout judgment and a dual verification mechanism for transmitted data. The dual verification includes CRC cyclic redundancy check and data logic compliance verification; (2) a sensor health monitoring unit, which starts a redundancy fault tolerance mechanism and sends a device failure alarm code and outputs it on the human-machine interface when the sensor data stream is interrupted or the value exceeds the limit; (3) a comprehensive risk assessment unit, which integrates the lidar point cloud risk analysis model and the multi-sensor data cross-verification algorithm, and outputs the system comprehensive risk level through a dynamic weight allocation strategy.

[0038] Reference Figure 4 This displays the logic diagram of the 200 decision-making algorithm in the decision-making layer of this system. For example... Figure 4 As shown, it includes: The multi-threaded architecture 210 is used to isolate and decouple various types of information interactions.

[0039] Path tracking algorithm 221 is used to convert pixel deviations into physical angles through geometric optical relationships.

[0040] Anomaly monitoring 230 is used to monitor abnormal system states and can trigger emergency stop commands.

[0041] In this embodiment of the invention, a multi-threaded architecture is used to isolate and decouple the signals of each module. When the operator issues a docking or evacuation command through the human-machine interface 311, the decision control thread 211 drives the boarding bridge to move. During the movement, it receives data from the perception layer 100 and system abnormal information. When it receives data from the visual sensor of the perception layer, it uses the path tracking algorithm 221 to locate the specific coordinates of the cabin door and anchor markers and plans the driving path. When abnormal information is detected, such as the presence of obstacles or collision risks, or abnormal verification of transmitted information content, a safety threshold is triggered, and the boarding bridge is driven to an emergency stop.

[0042] The execution layer 300 is used to drive the mechanical mechanism to complete actions according to decision instructions. It includes a human-machine interaction module 310 and a drive execution module 320; The human-machine interaction module includes: (i) a one-click operation unit, which integrates: an encrypted identity authentication module for user account login verification; a mode switching control panel for setting up fully automatic / semi-automatic / manual operation mode selection controls; and an emergency stop trigger device, which is configured with a physical emergency stop button and a software emergency stop command interface; (ii) a three-dimensional situation display unit, which is configured to present real-time visualization: obstacle spatial distribution point cloud map; and aircraft real-time positioning coordinates; (iii) a dynamic parameter configuration interface, which supports: external parameter calibration matrix input and correction; adjustment of motion mechanism geometric model parameters; and dynamic setting of collision safety threshold; (iv) a status visualization unit, which includes: an HMI human-machine interface device for real-time display of equipment operating status codes; an LED status indicator screen for displaying feedback on the current operating status of the system through LED display; and an audio-visual warning component that integrates a voice alarm and a rotating warning light; and (v) a video monitoring matrix unit that receives and displays operation scene images captured by multiple cameras.

[0043] The drive execution module includes: (i) a core control instruction execution unit, configured to receive path planning instructions from the decision system, convert them into walking speed and rotation angle control parameters, and control the movement of the boarding bridge's walking mechanism, the connecting cabin's rotating mechanism, and the canopy's telescopic mechanism through the IO module; (ii) a multi-system collaborative communication unit, which uses the Profinet protocol to achieve real-time data interaction with the image recognition system and the lidar system. The communication unit is equipped with a network interface management module for processing the distribution and transmission of camera video signals and lidar obstacle data. (III) Safety assurance and abnormal response unit, including: emergency stop control module, which triggers emergency stop command of walking mechanism through PLC and starts human-machine interaction alarm simultaneously when a high-risk value signal is received; operation permission management module, which configures priority interlock logic between local operating console, remote control terminal and fully automatic mode to ensure that local operating console has the highest control authority; (IV) hardware interface management unit, which connects bridge body angle sensor and hatch position sensor through PLC to perform real-time filtering processing on the collected sensor data.

[0044] The core control command execution unit includes: a walking speed control submodule, which converts path planning commands into walking motor speed parameters; a rotation angle control submodule, which generates rotation angle control quantities according to the docking cabin's position and posture adjustment requirements; and a telescopic stroke control submodule, which drives the canopy telescopic mechanism according to preset stroke parameters.

[0045] The multi-system collaborative communication unit includes: a visual data channel, which transmits the video stream captured by the camera to the image recognition system; a point cloud data channel, which forwards the obstacle point cloud collected by the lidar to the processing module; and a control command channel, which transmits the console commands to the PLC controller.

[0046] The emergency stop control module's response process includes: receiving a high-risk signal output by the collision risk assessment module; cutting off the power supply to the walking mechanism; and sending an audible and visual alarm command to the human-machine interface.

[0047] Reference Figure 5 This shows the specific drive mechanism structure diagram of the execution layer 300 of this system. For example... Figure 5 As shown, the boarding bridge and its ancillary functional structures in the execution layer 300 include: Anchor marker 111 is used for visual sensor identification during evacuation.

[0048] The boarding bridge drive mechanism 321 is used for boarding bridge travel, connection angle adjustment, height adjustment, and docking mechanism retraction and extension.

[0049] 140 multi-type sensors are installed on the boarding bridge to sense the status of the boarding bridge, surrounding objects, and aircraft position.

[0050] In this embodiment of the invention, the boarding bridge drive mechanism includes a wheel frame 3211, left and right drive wheels 3212, a connecting compartment drive motor 3213, a bridge height drive motor 3214, a canopy 3215, a rain door 3216, a leveling wheel 3217, an LED display screen, and a warning voice 3218.

[0051] In this embodiment of the invention, when the execution layer 300 initiates the docking command, the LED display shows the docking status of the boarding bridge, the warning voice broadcast system issues a docking voice prompt, the boarding bridge enters the initialization state, and the PLC drives the boarding bridge to a preset state. After reaching the initialization state, the execution layer detects the operating status of each module and each thread architecture. If there are no abnormalities and the operation is within the safety threshold range, the execution layer drives the boarding bridge to move. The PLC controls the left and right wheels of the boarding bridge to move forward to a distance of 2 meters from the aircraft. The visual sensor of the perception layer locates the coordinates, and the angle of the docking cabin is adjusted through precise coordinates. The boarding bridge continues to be driven to approach the aircraft door. When the boarding bridge is 0.5 meters away from the aircraft, the docking attitude is adjusted, and the boarding bridge is slowly and accurately driven to dock with the aircraft door, the canopy is deployed, the automatic leveling wheels are extended, the safety boots are deployed, and the boarding bridge rain door is opened. When the execution layer initiates the evacuation command, the LED screen displays that the boarding bridge is in the evacuation state, the warning voice broadcast system issues a evacuation voice prompt, the PLC controls the boarding bridge to retract the automatic leveling wheels, retract the canopy to cancel the safety boot engagement, and close the boarding bridge's rain and wind doors. The execution layer checks the operating status of each module and thread architecture. If there are no abnormalities and the operation is within the safety threshold range, the execution layer drives the boarding bridge to slowly evacuate the aircraft and head towards the return point.

[0052] Reference Figure 6 This demonstrates the specific process by which the system detects an anomaly and triggers security measures. For example... Figure 6 As shown, the safety protection measures include the following components: Perception Layer 100: Sends visual learning model and lidar point cloud recognition data to the decision layer via the MQTT communication protocol.

[0053] Bridge-mounted device 330: Used to send constraint signals to the decision-making level when bridge-mounted device use is detected.

[0054] Boarding bridge hardware 321: When a hardware device malfunctions, the decision-making level will identify the abnormal state and trigger protection strategies.

[0055] The human-computer interaction interface 311 allows operators to actively press the stop button to send an emergency stop signal when the boarding bridge environment is found to be abnormal but the decision-making level does not respond.

[0056] The decision-making level 200 is responsible for environmental safety monitoring and action decisions.

[0057] In this embodiment of the invention, when the system triggers the protection strategy, the boarding bridge will cut off the power output and send an alarm signal to the human-machine interface. If the system is in a static ready state, and the perception layer 100, the bridge-mounted device 330, or the boarding bridge hardware 321 experiences an abnormal fault, and the decision layer does not receive information or the received data verification is abnormal, then the system is determined to be abnormal and an alarm signal is sent to the human-machine interface. When the system is in operation, if the visual sensor and lidar in the perception layer malfunction, or if the boarding bridge hardware malfunctions and environmental data cannot be transmitted, the decision layer will trigger the protection strategy. When the lidar identifies an obstacle or the distance between the aircraft and the boarding bridge exceeds a threshold, the decision layer will also trigger the protection strategy. If the system detects that the bridge-mounted device is running during the pre-operation environmental safety detection state, the decision layer will trigger the protection strategy. When the system is in operation and an environmental abnormality is manually detected but the decision layer does not trigger the protection strategy, intervention can be made in the following ways: pressing the emergency stop button to trigger the braking system, using the isolation button to block the execution layer communication, or clicking the stop button on the software UI interface to terminate the boarding bridge autopilot.

Claims

1. An automatic docking and evacuation system for boarding bridges, characterized in that, Includes the following hierarchical modules: The perception layer is used to receive environmental data and includes: an aircraft positioning module, an anchor marker positioning module, a ground equipment status recognition module, a path obstacle and anti-collision detection module, a multi-sensor fusion algorithm, and a multi-modal collaborative perception system composed of a camera-based visual door recognition module and a lidar-based multi-modal collaborative perception system. The multi-modal collaborative perception system is configured as a spatiotemporal synchronous fusion processing unit. The camera processes images using OpenCV and employs an adaptive distortion correction algorithm, while the lidar processes point clouds using the PCL library and integrates a dynamic background modeling module. The decision layer is used to receive data from the perception layer and generate path planning instructions. It includes: a multi-threaded system architecture, a take-off path control algorithm, system anomaly monitoring, and a safety-priority PLC control system based on the PROFINET architecture. The execution layer is used to drive mechanical actions according to decision instructions, and includes: a human-machine interaction module and a drive execution module; The storage and network layer is used for data storage and information exchange, including: an attack-resistant storage system using MySQL 8.0 database, dynamically rotating AES-GCM encryption for alarm information and status data, storage frequency once per second, and a physical isolation-dynamic defense network system constructed from fiber optic cables and network cables; Each module interacts through an independent, closed network, and a publish-subscribe model based on delegates and event queues is used to achieve cross-thread asynchronous communication in order to decouple control.

2. The system according to claim 1, characterized in that, The specific configuration of the perception layer includes: Aircraft and anchorage marker positioning module: Based on a deep vision learning model, it achieves anti-occlusion recognition and positions the aircraft by analyzing images of the nose, landing gear, doors, and anchorage markers. Ground equipment status recognition module: determines the connection status between ground equipment and aircraft through visual recognition, and sends occupancy signals to constrain operation permissions; Visual door recognition module: after the camera acquires images, it performs online calibration to remove distortion, activates multi-frame super-resolution reconstruction under low light, inputs the YOLOv8m model to convert it into TensorRT format to output the door bounding box, improves the positioning accuracy to ±3px through perspective transformation and sub-pixel edge fitting, and sends the coordinate data to the decision system via MQTT after CRC verification; The lidar collision avoidance module receives spatiotemporally asynchronous point cloud data from lidar and blind spot radar, aligns the timestamps using a motion compensation algorithm, and transforms them to the bridge coordinate system. It then uses a combination of plane fitting and deep learning to filter out ground and self-point clouds. The remaining point clouds are clustered using DBSCAN and a three-dimensional Euclidean distance field is constructed. When an obstacle is detected encroaching on the dynamic safety boundary, the power is cut off; otherwise, a pre-alarm is issued. The dynamic safety boundary is ≥50cm for the engine area and ≥30cm for the wing area.

3. The system according to claim 1, characterized in that, The specific configuration of the decision-making layer includes: Multi-threaded system architecture: divided into execution mechanism, decision control, data communication, and human-computer interaction threads, with modular isolation to prevent information blockage; The approach and withdrawal path control algorithm includes a hierarchical finite state machine and approach and withdrawal path planning. The hierarchical finite state machine manages the approach and withdrawal mode switching and integrates safety strategies to monitor risks. The approach and withdrawal path planning includes an initial pre-approach point and a camera fine-tuning stage. The fine-tuning adopts the pinhole imaging principle and the formula θ=arctan(λ×ΔP). The withdrawal path includes an initial pre-approach point and anchor marker positioning. When the marker fails, a multi-sensor fusion algorithm is used. The motion control module of the safety-type PLC control system: When automatically approaching, the contact prediction model is activated when the data from both distance sensors are ≤60cm and the difference is <10cm. The speed is reduced by 10% in the intervals of 60-50cm, 50-40cm, etc., and the speed is idled to 0cm and braked when the distance is 10cm. When removing the bridge, a dynamic obstacle avoidance topology map is constructed based on the LiDAR data. Active safety strategy: Integrates minimum risk strategy and emergency takeover link. The minimum risk strategy is a strategy triggered according to risk level when the operating conditions deviate. The emergency takeover link is a local emergency stop button directly connected to the PLC power relay, with a response delay of <10ms.

4. The system according to claim 1, characterized in that, The human-computer interaction module of the execution layer includes: a one-click operation interface: providing encrypted user account login, operation mode switching buttons, and emergency stop function; encryption adopts a combination of salt value encryption MD5 and behavior auditing, and forces passwords to contain letters, numbers and special characters; 3D situation display unit: Real-time rendering of motion trajectory, obstacle distribution, and aircraft position; Parameter configuration interface: Supports dynamic adjustment of extrinsic parameter matrix, geometric model, and safety threshold; Status visualization unit: Displays operating status via HMI and LED display screen; the bridge body is equipped with warning language and warning light modules. It supports dual control via HMI touchscreen and physical buttons.

5. The system according to claim 1, characterized in that, The execution layer's driver execution module includes: Core control command execution: Receives path planning commands, converts them into parameters such as walking speed and angle, and controls the walking mechanism, docking compartment rotation, canopy extension and retraction, etc. through the IO module; Multi-system collaborative communication: Real-time interaction with image recognition and lidar systems via PROFINET protocol, coordinating the synchronization of sensing data and control commands; acting as a communication hub to process and forward camera video signals and lidar data. Safety and anomaly response: When a high-risk value is triggered, the PLC will stop the walking mechanism and send an alarm; local / remote / fully automatic mode access interlocking is implemented to ensure the highest control authority of the local operator console; Hardware interface management: The PLC connects to the sensor interface to collect data such as bridge angle and hatch position, and uses a first-order low-pass filter combined with an integral compensation algorithm to suppress high-frequency noise.

6. The system according to claim 1, characterized in that, The specific configuration of the storage and network layer includes: Anti-attack storage system: The stored content includes more than 50 types of alarm information and status data. Blockchain timestamps are added to the encrypted data, and the local and backend use sharded redundant storage. Physical Isolation-Dynamic Defense Network System: Deployed in DCR, SCR and bridge computer rooms, it achieves three-layer segmentation of management network, control network and sensor network through logical isolation gateways; the topology includes S5560X-30F-EI switches in DCR computer rooms, fiber optic patch devices in SCR computer rooms and S5130S-PWR-EI switches in bridge computer rooms.

7. The system according to claim 1, characterized in that, The path obstacle and collision avoidance detection module of the sensing layer includes: Point cloud preprocessing module: processes radar point clouds, accurately identifies and distinguishes ground and obstacles, and removes point clouds from the bridge body; Multi-radar synchronization processing module: Enables simultaneous multi-radar, multi-angle obstacle identification through a caching mechanism; Collision risk assessment unit: It analyzes the point cloud type and distance through distance clustering algorithm, and obtains the risk value through collision risk calculation formula. When the risk value is too high, it sends a signal to the decision layer to suspend operation.

8. The system according to claim 1, characterized in that, The system anomaly monitoring includes: data communication anomaly detection: including timeout determination and content verification anomaly, the content verification anomaly including data integrity and logical rationality; Sensor failure detection: Sends a risk value when sensor data is lost; Risk detection and assessment: including LiDAR obstacle risk assessment and integrated sensor data risk assessment.

9. The system according to any one of claims 1-8, characterized in that, The control method of the system includes the following steps: S1. Spatiotemporal synchronization phase: Aligning the lidar timestamps using a dedicated protocol; S2. Collaborative perception stage: Map the visually recognized door coordinates and the lidar point cloud to a unified coordinate system to generate a multi-source confidence fusion trajectory; S3. Path Planning and Execution Phase: The decision-making layer executes state machine decisions, generates path tracking instructions and sends them to the drive unit; it updates the state through an interactive thread and can receive manual intervention instructions; S4. Adaptive docking phase: The bridge pitch angle is dynamically adjusted according to the aircraft door height, and the contact prediction model is activated at a distance of 50cm from the door; S5. Safety Handling Phase: When an anomaly or collision risk is detected, an emergency strategy is triggered and an alarm is sent. The emergency strategy is to cut off power or stop the vehicle abruptly.

Images