A method for optimal control of approach lights based on reinforcement learning

Abstract

The application provides an optimal control method for navigation lights based on reinforcement learning, and relates to the field of data processing. In the method, a navigation light relationship graph and an airport environment perception result are obtained, an incorrect guiding path is identified, and a path judgment state is generated; a path interference action combination is generated based on a reinforcement learning model, and after safety verification, the path interference action combination is converted into a navigation light node control instruction to change the visual structure of the incorrect guiding path; and through elimination effect evaluation and simulation interference environment training, the safe and continuous elimination of multiple incorrect guiding paths is realized. The technical solution provided by the application can simultaneously consider the elimination of incorrect guiding paths, the maintenance of real runway paths, and the limitation of light fixture heat attenuation in the reinforcement learning control process, so as to ensure the stability of real runway path identification while avoiding the navigation light nodes from being in a long-term high-heat operation state.

Description

Technical Field

[0001] This application relates to the field of data processing technology, specifically to an optimal control method for navigation lights based on reinforcement learning. Background Technology

[0002] With the development of intelligent airport control, navigation light control methods are increasingly being applied to runway guidance and lighting adjustments in complex weather conditions. Existing technologies typically adjust the brightness of navigation lights dynamically based on visibility, ambient light, and visual recognition performance to improve the ability to identify actual runway paths.

[0003] However, most existing reinforcement learning models only focus on the light output effect and lack constraints on the thermal decay state of LED lights. When some navigation light nodes operate at high brightness for extended periods, the corresponding LED lights experience a significant increase in temperature rise, leading to a decrease in luminous efficiency and a reduction in actual light output. Upon detecting this decrease in light output, the reinforcement learning model further increases the driving power for brightness compensation, creating a vicious cycle of "high brightness, high heat, low efficiency, and then increased brightness again." This vicious cycle not only easily leads to a decrease in the lifespan of navigation light nodes and an imbalance in brightness distribution, but also causes the reinforcement learning model to continuously rely on high-brightness area control, resulting in concentrated heat load and decreased system stability.

[0004] Therefore, there is an urgent need for an optimal control method for navigation lights that can simultaneously consider error path elimination, real runway path maintenance, and lamp thermal attenuation limitations during reinforcement learning control, so as to ensure the stability of real runway path recognition while avoiding navigation light nodes from falling into a long-term high-heat operation state. Summary of the Invention

[0005] This application provides an optimal control method for navigation lights based on reinforcement learning, which can simultaneously consider error path elimination, real runway path maintenance, and lamp thermal attenuation limitation during the reinforcement learning control process, so as to ensure the stability of real runway path recognition while avoiding navigation light nodes from falling into a long-term high-heat operation state.

[0006] The first aspect of this application provides a reinforcement learning-based optimal control method for navigation lights. The method includes: acquiring a light relationship diagram of navigation lights within a target airport, the light relationship diagram describing the light relationships between navigation light nodes in a real runway path; collecting airport environment perception results of the target airport, identifying erroneous guidance paths based on the airport environment perception results, and comparing the erroneous guidance paths with the light relationship diagram to generate a path judgment state describing the completeness of the erroneous guidance paths; constructing a set of path interference actions based on the path judgment state, and generating a combination of path interference actions from the set of path interference actions based on the path judgment state using a reinforcement learning model, wherein the combination of path interference actions includes centerline disruption actions, boundary misalignment actions, entrance desymmetry actions, etc. The system includes: depth truncation and true path maintenance actions; performing safety checks on the path interference action combinations and restricting the execution of path interference action combinations that disrupt the true runway path lighting relationship based on the safety check results; converting path interference action combinations that pass the safety check into navigation light node control commands; changing the visual structure of the misguided path based on the navigation light node control commands; and re-collecting airport environmental perception results after executing the navigation light node control commands to evaluate the elimination effect of the path interference action combinations on the misguided path; constructing a simulated interference environment to train the reinforcement learning model to identify and eliminate the misguided path based on the simulated interference environment; and adjusting the strategy of the reinforcement learning model based on safety checks during the online operation phase to achieve continuous elimination of multiple misguided paths.

[0007] A second aspect of this application provides an optimal control device for navigation lights based on reinforcement learning. The device includes an acquisition module and a processing module. The acquisition module is used to acquire a light relationship diagram of navigation lights within a target airport, the light relationship diagram describing the light relationships between navigation light nodes in a real runway path. The processing module is used to collect airport environment perception results of the target airport, identify erroneous guidance paths based on the airport environment perception results, and compare the erroneous guidance paths with the light relationship diagram to generate a path judgment state describing the completeness of the erroneous guidance paths. The processing module is further used to construct a set of path interference actions based on the path judgment state, and generate a combination of path interference actions from the set of path interference actions based on the path judgment state using a reinforcement learning model. The combination of path interference actions includes centerline disruption actions, boundary misalignment actions, and entrance deviance actions. The processing module includes symmetrical maneuvers, depth truncation maneuvers, and true path-keeping maneuvers. It further performs safety checks on the combinations of path interference maneuvers and restricts the execution of combinations that disrupt the true runway path lighting relationships based on the safety check results. Simultaneously, it converts the path interference maneuvers that pass the safety check into navigation light node control commands. The processing module also modifies the visual structure of the misguided path based on the navigation light node control commands and re-acquires airport environmental perception results after executing the navigation light node control commands to evaluate the effectiveness of the path interference maneuvers in eliminating the misguided path. Furthermore, the processing module constructs a simulated interference environment to train the reinforcement learning model to identify and eliminate the misguided path. During online operation, it restricts the reinforcement learning model's strategy based on safety checks to achieve continuous elimination of multiple misguided paths.

[0008] A third aspect of this application provides an electronic device comprising a processor and a memory; the memory having a stored computer program, wherein the computer program, when executed by the processor, implements the reinforcement learning-based optimal control method for navigation lights as described above.

[0009] In a fourth aspect of this application, a non-transitory computer-readable storage medium is provided, the non-transitory computer-readable storage medium storing instructions that, when executed, perform the reinforcement learning-based optimal control method for navigation lights as described above.

[0010] In summary, one or more technical solutions provided in this application have at least the following technical effects or advantages: By constructing a light relationship diagram, identifying erroneous guidance paths, generating path interference action combinations, and performing safety checks, the reinforcement learning model can proactively weaken erroneous guidance paths while maintaining the stability of the real runway path, thereby improving runway guidance safety and the accuracy of real path recognition under complex weather conditions. Simultaneously, by converting path interference action combinations into navigation light node control commands and performing closed-loop evaluation based on re-acquired airport environmental perception results, the reinforcement learning model can continuously optimize control strategies, improving its adaptability to complex interference environments such as water reflection, wet snow scattering, background light interference, and refraction shift. Furthermore, by constructing a simulated interference environment and introducing safety check constraints during online operation, the reinforcement learning model can be limited to generating control actions that disrupt the real runway path light relationships, reducing the risk of thermal decay caused by long-term high-brightness operation of navigation light nodes, and improving the operational stability and lifespan of the navigation light system. Therefore, erroneous guidance path elimination, real runway path maintenance, and lamp thermal decay limitation can be considered simultaneously during the reinforcement learning control process, ensuring the stability of real runway path recognition while preventing navigation light nodes from entering a long-term high-heat operation state. Attached Figure Description

[0011] Figure 1 A flowchart illustrating an optimal control method for navigation lights based on reinforcement learning, provided in an embodiment of this application; Figure 2 A schematic diagram of a module for an optimal control device for navigation lights based on reinforcement learning, provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0012] Explanation of reference numerals in the attached figures: 21. Acquisition module; 22. Processing module; 31. Processor; 32. Communication bus; 33. User interface; 34. Network interface; 35. Memory. Detailed Implementation

[0013] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0014] In the description of the embodiments of this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a specific manner.

[0015] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. In addition, the terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0016] To address the aforementioned technical problems, this application provides a reinforcement learning-based optimal control method for navigation lights, referring to... Figure 1 , Figure 1 A flowchart illustrating an optimal control method for navigation lights based on reinforcement learning, provided in this application embodiment, is shown. This method is applied to a server and includes steps S110 to S160, as follows: S110. Obtain the lighting relationship diagram of the navigation lights within the target airport. The lighting relationship diagram is used to describe the lighting relationship between each navigation light node in the actual runway path.

[0017] Specifically, the server first obtains basic information about each navigation light node from the target airport's navigation light maintenance database, construction completion data, airport lighting control system, on-site inspection terminals, and navigation light control cabinets. A navigation light node refers to a navigation light entity that can be individually located, identified, or controlled, including runway centerline lights, runway edge lights, entrance lights, touchdown area lights, stop row lights, taxiway centerline lights, and taxiway edge lights. The installation location in the basic information describes the position of the navigation light node in the target airport coordinate system; the lamp type distinguishes the lighting function undertaken by the navigation light node; the emission color distinguishes the visual guidance meaning of the navigation light node; the illumination direction describes the main direction of the navigation light node's beam; the rated brightness indicates the allowable output capacity of the navigation light node under standard operating conditions; the lamp group to which the navigation light node belongs indicates the functional area to which the navigation light node belongs; the power supply circuit indicates the power source of the navigation light node; the control circuit indicates the control channel of the navigation light node; the communication address indicates the address at which the server locates the navigation light node when issuing control commands; and the operating status indicates whether the navigation light node is currently in a normal, fault, maintenance, degraded, or off state. After obtaining the above basic information, the server compares the information from different data sources for the same navigation light node for consistency. If there is a conflict in the installation location, communication address, or light group to which the same navigation light node belongs, the data most recently confirmed by the on-site inspection terminal and the navigation light control system shall prevail, and the confirmed basic information shall be written into the navigation light node data table so that subsequent spatial registration, establishment of light relationship, and generation of light relationship diagram are all based on the same set of navigation light node data.

[0018] When the server performs spatial registration of each navigation light node based on basic information, it uniformly transforms the installation position of each navigation light node to the airport coordinate system of the target airport and calculates the relative position of the navigation light node with respect to the actual runway path. The airport coordinate system refers to a spatial coordinate system established with a preset reference point within the target airport as the origin and the runway centerline and lateral directions as the coordinate axes. The actual runway path refers to the real guideable path defined by the runway threshold, runway centerline, runway boundaries, touchdown area, stop position, and taxiway exit. For each navigation light node, the server determines its projection position, lateral offset, longitudinal sequence, and functional area on the actual runway path based on its installation location, thereby determining which spatial function the navigation light node performs on the actual runway path: centerline guidance, boundary definition, threshold identification, touchdown indication, stop blocking, or taxiway connection.

[0019] in, Indicates the first The position of each navigation light node in the airport coordinate system is obtained by the server after performing coordinate transformation on the original installation position. This represents the navigation light node number, and its value range is the set of numbers of all navigation light nodes within the target airport; It represents the coordinate transformation relationship from the original data coordinate system to the airport coordinate system, which is determined by the correspondence between the target airport reference point, the runway centerline direction, the runway lateral direction, and the original data coordinate system; Indicates the first The installation positions of each navigation light node in the original data coordinate system are obtained from navigation light design documents, as-built documents, or on-site inspection data. This formula unifies the installation positions from different sources into the same airport coordinate system through coordinate transformation, enabling the server to determine the relative positional relationship between each navigation light node and the actual runway path under a unified spatial reference.

[0020] After determining the relative positional relationship between each navigation light node and the actual runway path, the server establishes lighting relationships based on the light group to which each navigation light node belongs and the spatial registration results. Lighting relationships refer to the visual guidance association jointly formed by two or more navigation light nodes on the actual runway path, rather than ordinary spatial adjacency relationships. For navigation light nodes arranged sequentially along the runway centerline and belonging to the runway centerline lights, the server establishes a centerline continuity relationship to represent the longitudinal extension direction of the actual runway path. For navigation light nodes located on the left and right sides of the runway and belonging to the runway edge lights, the server establishes a boundary parallel relationship to represent the lateral range of the actual runway path. For navigation light nodes located on both sides of the runway threshold and having threshold recognition function, the server establishes a threshold symmetry relationship to represent the starting position of the actual runway path. For navigation light nodes located in the touchdown area and providing progressive indications along the approach direction, the server establishes a touchdown area progression relationship to represent the visual transition of the aircraft from approach to the landing area. For navigation light nodes located in the stop position and providing a no-pass indication, the server establishes a stop-block relationship to indicate the control meaning of not being allowed to continue taxiing or enter the runway. For navigation light nodes located in the area connecting the runway path and the taxiway path, the server establishes a path connection relationship to represent the connection between the actual runway path and the legal taxiway path.

[0021] in, Indicates the first The navigation light node and the first The lighting relationships between navigation light nodes can be categorized into centerline continuity, boundary parallelism, entrance symmetry, grounding area progression, stop / blockage, path connection, or no relationship. Indicates the relationship with the first The number of the other navigation light node used for relationship determination among the navigation light nodes, with a value range excluding... The set of navigation light node numbers other than those mentioned above; The function for determining the relationship between lights is determined by the consistency of the light groups, spatial adjacency, relative direction, left-right correspondence, and functional areas. and They represent the first The navigation light node and the first The light group to which a navigation light node belongs is obtained from the light group field in the basic information; and They represent the first The navigation light node and the first The position of each navigation light node in the airport coordinate system is obtained by spatial registration; and They represent the first The navigation light node and the first The functional area of ​​each navigation light node is determined by its relative position to the actual runway path. This formula, by simultaneously determining the light group attributes, spatial location, and functional area, enables the server to transform the spatial arrangement of physical lights into lighting relationships with runway guidance significance.

[0022] When the server performs semantic annotation on light relationships, it writes the actual runway path guidance meaning for each light relationship. This enables it to determine whether a suspected light spot arrangement has a structure similar to the actual runway path when identifying incorrectly guided paths. Semantic annotation refers to writing the functional meaning, display order, safety level, and controllable boundaries corresponding to the light relationships into the relationship attributes. For example, the actual runway path guidance meaning of a continuous centerline relationship is to instruct the aircraft to align along the longitudinal direction of the runway; the actual runway path guidance meaning of a parallel boundary relationship is to define the left and right boundaries of the runway and maintain lateral recognition stability; the actual runway path guidance meaning of a symmetrical entrance relationship is to indicate the location and direction of the runway entrance; the actual runway path guidance meaning of a progressive touchdown area relationship is to indicate the visual transition of the touchdown area from far to near; the actual runway path guidance meaning of a stop / block relationship is to indicate stop, wait, or prohibit crossing; and the actual runway path guidance meaning of a path connection relationship is to indicate the legal transition position between the runway path and the taxiway path. Through semantic annotation, the server not only knows which navigation light nodes are adjacent, but also knows what role these adjacent relationships play when pilots or airborne vision equipment identify the actual runway path.

[0023] in, Indicates the first The navigation light node and the first The semantic annotation results between each navigation light node include the actual runway path guidance meaning, safety level, permissible control range, and display order requirements; The semantic annotation function is determined by the light relationship type, node functional attributes, and relationship quality. Indicates the first The navigation light node and the first The lighting relationships between each navigation light node are obtained by the lighting relationship determination function. and They represent the first The navigation light node and the first The node attributes of each navigation light node, including lamp type, emission color, illumination direction, rated brightness, operating status, and safety constraint level, are obtained from the basic information and operating status confirmation results. Indicates the first The navigation light node and the first The quality of the relationship between navigation light nodes is determined by a combination of their spatial spacing, directional consistency, left-right correspondence stability, and display order consistency. This formula combines light relationships with node attributes, giving each light relationship a clear meaning for guiding the actual runway path, thus providing semantic constraints for subsequent safety checks and path interference actions.

[0024] When the server generates a light relationship graph by treating each navigation light node as a graph node and each light relationship as a graph edge, it first encapsulates each navigation light node as a graph node, writing its installation location, light type, emission color, illumination direction, rated brightness, light group, power supply circuit, control circuit, communication address, and operating status into the corresponding graph node. Then, it encapsulates each light relationship as a graph edge, writing the relationship type, actual runway path guidance meaning, adjacent order, spatial direction, display order, safety constraint level, and allowable control range into the corresponding graph edge. The light relationship graph simultaneously describes the physical layout, control mapping, and actual runway path guidance relationships of the navigation light nodes. This allows the server to compare suspected light spot arrangements in the environment with the light relationship graph after collecting airport environmental perception results, determine whether they constitute incorrect guidance paths, and identify which navigation light nodes can participate in control and which must maintain the actual runway path light relationships when generating path interference action combinations.

[0025] in, The diagram represents the relationship between lights, which is generated by the server based on the navigation light nodes and the relationships between lights. This represents a set of graph nodes, with each graph node corresponding to a navigation light node. The node attributes are written by the basic information and spatial registration results. This represents the set of graph edges, where each edge corresponds to a light relationship, and its edge attributes are written by the light relationship and semantic annotation results; It represents the set of attributes of the graph node, including installation location, lamp type, light color, illumination direction, rated brightness, lamp group, power supply circuit, control circuit, communication address and operating status; This represents the set of graph edge attributes, including relationship type, actual runway path guidance meaning, adjacent order, spatial direction, display order, safety constraint level, and permissible control range. This formula organizes navigation light nodes and their relationships into structured graph data, enabling the server to directly call the light relationship graph in subsequent steps to complete error guidance path identification, path judgment state generation, path interference action combination generation, and safety verification.

[0026] S120. Collect the airport environment perception results of the target airport, identify the wrong guidance path based on the airport environment perception results, and compare the wrong guidance path with the light relationship diagram to generate a path judgment state to describe the completeness of the wrong guidance path.

[0027] Specifically, when the server collects the airport environment perception results of the target airport, it first takes the actual runway path of the target airport as the center and simultaneously acquires continuous image frames, low-altitude visibility, precipitation status, ground water status, wet snow coverage status, low-altitude refraction status, background light distribution, aircraft attitude changes, and navigation light node operation status. Continuous image frames refer to runway area images obtained continuously over a period of time by airborne vision equipment, tower monitoring equipment, or runway ground monitoring equipment; low-altitude visibility refers to the visible distance near the aircraft's approach or taxiing altitude; precipitation status refers to the impact of rain, snow, freezing rain, or mixed precipitation on light propagation; ground water accumulation status refers to the formation of reflective surfaces on the runway or taxiway surfaces; wet snow coverage status refers to the state where wet snow covers the runway surface or surrounding area, causing light spot diffusion; low-altitude refraction status refers to the state where changes in low-altitude temperature and humidity cause light path deviation; background light distribution refers to the distribution of city lights, vehicle lights, or other non-navigational lights around the target airport in the field of view; aircraft attitude changes refer to the changes in the aircraft's heading angle, pitch angle, roll angle, and position relative to the actual runway path; navigation light node operation status refers to the brightness, on / off status, fault status, communication status, and control loop status of each navigation light node. The server aligns the above data according to the time of collection and organizes the time-aligned data into airport environmental perception results, so that the airport environmental perception results can simultaneously represent the status of lighting display, weather interference, ground reflection, background interference, and aircraft observation.

[0028] in, Indicates time The corresponding airport environment perception results are obtained by the server performing time alignment on multi-source data within the same time window. Indicates from time At that time consecutive image frames, This indicates the length of the time window, which is preset by the system based on the image acquisition frequency and path recognition stability requirements. Indicates time Low-altitude visibility is obtained by meteorological monitoring equipment or runway low-altitude visibility sensors; Indicates time The precipitation status was obtained from meteorological monitoring equipment; Indicates time The water level on the runway surface is determined by runway surface monitoring equipment, image reflection analysis results, or surface slippage sensors. Indicates time The state of wet snow cover is determined by image texture recognition results, ground temperature data, and precipitation status. Indicates time The low-altitude refraction state is determined by the low-altitude temperature and humidity gradient, image spot drift results, and historical refraction offset records. Indicates time The background light distribution was extracted from the brightness, color, and spatial position of the light spots in the non-navigation light areas of consecutive image frames. Indicates time The changes in aircraft attitude are obtained from airborne navigation data, surveillance radar data, or airport surface surveillance data. Indicates time The operational status of the navigation light nodes is obtained by the navigation light control system and navigation light operation monitoring equipment. This formula organizes multi-source environmental data into a unified state at the same time, enabling subsequent error guidance path identification to simultaneously consider the combined effects of weather, reflection, background light, aircraft perspective, and navigation light operational status.

[0029] When the server maps the airport environment perception results to the spatial coordinates corresponding to the lighting relationship map, it first reads the graph node coordinates of each navigation light node and the spatial direction of each lighting relationship in the lighting relationship map. Then, it converts the pixel positions, ground water accumulation positions, wet snow coverage positions, background light distribution positions, and low-altitude refraction influence areas in consecutive image frames to the spatial coordinates corresponding to the lighting relationship map. Spatial coordinates refer to the coordinates relative to the lighting relationship. Figure 1 The airport coordinate system is defined; the mapping result refers to the spatial correspondence between the light spots, reflection areas, scattering areas, and background light areas in the airport environmental perception results and the navigation light nodes in the lighting relationship diagram. The server extracts a candidate light spot set based on the mapping result. The candidate light spot set refers to the set of light spots with significant brightness, duration, and spatial arrangement characteristics in consecutive image frames. Among them, candidate light spots that can stably match the navigation light nodes in the lighting relationship diagram are marked as real light spots, while candidate light spots that cannot stably match the navigation light nodes but are similar to the navigation light spots in brightness, color, shape, or arrangement are marked as suspected interference spots.

[0030] in, Indicates the first The position of each candidate light spot in spatial coordinates is obtained by back-projecting the pixel positions in consecutive image frames by the server; This represents the candidate spot number, and its value range is the set of all candidate spot numbers extracted within the current time window; The back projection function represents the projection from the image pixel position to the spatial coordinate position, which is determined by the camera calibration parameters, the aircraft attitude changes, and the runway plane constraints. Indicates the first The pixel positions of candidate light spots in consecutive image frames are obtained by image brightness threshold segmentation, color filtering and light spot connected component extraction; Indicates time The aircraft's attitude changes are used to determine the direction of the image's line of sight; Indicates time The runway planar constraints are determined by the target airport runway geometry data and the coordinates of the navigation light nodes in the lighting relationship diagram. This formula transforms candidate light spots in the image to those corresponding to the lighting relationship diagram through back projection. Figure 1 In the spatial coordinates, the server can determine whether the candidate light spot originates from the real navigation light node.

[0031] in, Indicates the first The spatial deviation between the nearest navigation light node in the relationship diagram between each candidate light spot and the light source is obtained by the server calculating the minimum distance between the candidate light spot position and the position of each node in the diagram. The diagram shows the corresponding number of lights in the relationship diagram. A graph node for each navigation light node; This represents the set of nodes in the lighting relationship diagram. Indicates the first The position of each candidate light spot in spatial coordinates; Indicates the first The position of each navigation light node in spatial coordinates is obtained from the light relationship diagram. If If the candidate light spot's color characteristics and brightness changes are consistent with the corresponding navigation light node's operating status, the server will mark the candidate light spot as a real light spot; if... If the matching conditions are not met, but the candidate spot has visual characteristics similar to the navigation light spot, the server will mark the candidate spot as a suspected interference spot.

[0032] When constructing a suspected path set based on the candidate spot set, the server first filters suspected interference spots from the candidate spot set, and then associates them according to the spatial distance, arrangement direction, brightness progression, left-right distribution, continuous occurrence time, and motion consistency between the suspected interference spots. A suspected path set refers to a set composed of multiple suspected interference spots that can be visually interpreted as a path structure. Centerline continuity refers to multiple suspected interference spots arranged continuously along a longitudinal direction, forming a visual feature similar to the runway centerline. Parallel boundary features refer to two groups of suspected interference spots located on opposite sides of the path, forming approximately parallel boundaries. Entrance symmetry features refer to the visual feature where suspected interference spots on both sides of the suspected path end form a left-right correspondence. Depth extension features refer to the visual feature where suspected interference spots form depth cues along the aircraft's observation direction. The server performs a structural comparison between the suspected path set and the light relationship diagram. If a suspected path structurally approximates the centerline continuity, boundary parallelism, entrance symmetry, or grounding area progression in a real runway path, but its main spot cannot stably match the real navigation light node, then the suspected path is identified as an incorrect guidance path.

[0033] in, Indicates the first The continuity of the centerline of a suspected path is calculated by the server based on the stability of the spacing and the consistency of the orientation of adjacent suspected interfering spots in the suspected path; This represents the suspected path number, and its value range is all path numbers in the suspected path set. Indicates the first The number of suspected interference spots contained in each suspected path is obtained from the correlation results of suspected interference spots; Indicates the first The sequence number of adjacent suspected interfering spot pairs in the suspected path; Indicates the first The spatial distance between adjacent suspected interference spots is calculated from the positions of the two suspected interference spots in spatial coordinates. Indicates the first The average distance between all adjacent suspected interfering spots in the suspected path is calculated from all... We get the average. Indicates the first The angle between the direction of the line connecting adjacent suspected interference spots and the main direction of the suspected path is calculated from the fitting result of the suspected path direction and the direction of the line connecting adjacent suspected interference spots; This represents a stability parameter to prevent the denominator from being zero, and is preset to a positive number by the system. The formula evaluates whether suspected paths have a continuous arrangement similar to runway centerlines by assessing spacing stability and directional consistency; a higher value indicates that the suspected path is more easily identified as a continuous centerline.

[0034] in, Indicates the first The parallelism of the boundaries of the suspected paths is calculated by the server based on the consistency of the direction and the uniformity of the width of the suspected interfering light spots on both sides; Indicates the first The angle difference between the left and right boundary directions of a suspected path is calculated by fitting the boundary lines to the suspected interference spots on the left and right sides respectively. Indicates the first The average distance from the left boundary of a suspected path to the centerline of the suspected path is calculated from the position of the suspected interference spot on the left and the centerline of the suspected path. Indicates the first The average distance from the right boundary of a suspected path to the centerline of the suspected path is calculated from the position of the suspected interference spot on the right and the centerline of the suspected path. This represents a stability parameter to prevent the denominator from being zero, and is preset to a positive number by the system. The formula evaluates whether a suspected path has a parallel structure similar to the runway boundary by assessing the similarity in the left and right boundary directions and the similarity in the left and right widths. Higher values ​​indicate that the suspected path is more easily identified as a runway boundary structure.

[0035] in, Indicates the first The degree of symmetry at the entrance of a suspected path is calculated by the server based on the similarity of the number and brightness distribution of light spots on the left and right sides of the suspected path ends; Indicates the first The number of suspected interfering spots on the left side of the suspected path end is obtained by counting the candidate spots on the left side of the end region. Indicates the first The number of suspected interference spots on the right side of the suspected path end is obtained by counting the candidate spots on the right side of the end region. Indicates the first The average brightness of the suspected interference spot on the left side of the suspected path end is obtained by averaging the brightness of the suspected interference spot on the left side of the end. Indicates the first The average brightness of the suspected interference spot on the right side of the suspected path end is obtained by averaging the brightness of the suspected interference spot on the right side of the end. This represents a stable parameter to prevent the denominator from being zero, and is preset to a positive number by the system. The formula evaluates whether a suspected path has a symmetrical structure similar to a runway entrance by checking if the number and brightness distribution of light spots on the left and right sides of the end are similar.

[0036] in, Indicates the first The depth of a suspected path is calculated by the server based on the proximity of the main direction of the suspected path to the observation direction of the aircraft and the proportion of the forward light spot. Indicates the first The angle between the main direction of the suspected path and the observation direction of the aircraft is determined by the fitting result of the suspected path direction and the change in the aircraft's attitude. Indicates the first The number of suspected interference spots located within the forward observation area of ​​the aircraft in the suspected path is determined by the spatial location of the suspected interference spots and the observation direction of the aircraft. Indicates the first The number of suspected interfering light spots contained in each suspected path; This represents a stability parameter to prevent the denominator from being zero, and is preset to a positive number by the system. The formula evaluates whether a suspected path can form a traceable depth clue along the aircraft's observation direction using directional proximity and forward distribution.

[0037] When determining the cause of a misguided path based on the correspondence between the misguided path and the airport environmental perception results, the server correlates the location, shape, brightness changes, and stability of the misguided path with the ground water accumulation, wet snow cover, background light distribution, taxiway intersection area, and low-altitude refraction state. If the misguided path is located in the area corresponding to the ground water accumulation state, and its suspected interference light spot is mirrored with the actual light spot, the cause is determined to be a water reflection path; if the misguided path is located in the area corresponding to the wet snow cover state, and its suspected interference light spot has diffused edges and blurred brightness boundaries, the cause is determined to be a wet snow scattering path; if the misguided path highly overlaps with the background light distribution, and its color is close to the color of the navigation lights, the cause is determined to be a background light interference path; if the misguided path is located in the taxiway intersection area, and its extension direction deviates from the actual runway path but has a deep extension characteristic, the cause is determined to be an intersection light path; if the misguided path undergoes periodic spatial shifts with changes in the low-altitude refraction state, the cause is determined to be a refraction shift path.

[0038] in, Indicates the first The causes of incorrect guidance paths include water reflection paths, wet snow scattering paths, background light interference paths, intersecting light paths, or refraction offset paths. The function representing the cause determination is jointly determined by the spatial overlap, visual morphology, and temporal changes between the erroneous guidance path and the environmental interference state. Indicates the first The incorrect guidance path is determined by comparing the set of suspected paths with the structure of the light relationship diagram. It indicates the state of water accumulation on the ground and is used to determine whether there is a source of mirror reflection; This indicates the state of wet snow cover and is used to determine whether there is a source of scattering and diffusion. This indicates the background light distribution, used to determine whether there is a source of background light other than navigation lights; This indicates the low-altitude refraction state and is used to determine whether there is a source of light spot shift. This diagram, representing the lighting relationship, is used to determine whether an incorrect guidance path is located in a taxiway intersection area or whether it resembles the actual runway path structure. By integrating environmental disturbance conditions and the lighting relationship diagram, this formula enables the server not only to identify the existence of incorrect guidance paths but also to determine their source, providing a basis for selecting subsequent path disturbance action combinations.

[0039] When the server generates the path judgment state based on the relationship diagram between the misguided path and the lighting, it organizes the following parameters for each misguided path into a structured state: centerline continuity, boundary parallelism, entrance symmetry, depth extension, directional proximity, spatial proximity, salience, stability, formation cause, and the current confirmed stability of the actual runway path. Directional proximity describes the proximity between the extension direction of the misguided path and the aircraft's observation direction or the direction of the actual runway path; spatial proximity describes the spatial proximity between the misguided path and the actual runway path; salience describes the visibility intensity of the misguided path in the aircraft's field of view; stability describes the stability of the misguided path in consecutive image frames; and the current confirmed stability of the actual runway path describes whether the actual runway path can be stably identified in the current environment. The path judgment state does not simply indicate the presence or absence of a misguided path, but rather describes why the misguided path has complete path characteristics, why it is easily misidentified, and which path characteristics should be prioritized for destruction.

[0040] in, Indicates the first The path determination status corresponding to each error guidance path is generated by the server based on the relationship diagram between the error guidance path and the light. Indicates the first The continuity of the centerline of the erroneous guidance path is calculated by the spacing stability and directional consistency of adjacent suspected interference spots; Indicates the first The parallelism of the boundaries of the erroneous guidance path is calculated by the proximity of the left and right boundaries and the balance of their widths; Indicates the first The degree of symmetry at the entrance of the erroneous guidance path is calculated by the similarity of the number and brightness distribution of the left and right light spots at the ends; Indicates the first The depth of the erroneous guidance path is calculated by the proximity of the path direction to the aircraft's observation direction and the proportion of the forward light spot. Indicates the first The degree of directional similarity of the incorrect guidance path is calculated from the angle between the main direction of the incorrect guidance path and the direction of the actual runway path or the direction of observation of the aircraft. Indicates the first The spatial proximity of the erroneous guidance paths is calculated from the minimum distance and overlap between the erroneous guidance paths and the actual runway paths; Indicates the first The significance of a faulty guidance path is determined by the average brightness, color similarity, and image area occupied by the suspected interfering spots in the faulty guidance path. Indicates the first The stability of a faulty path is determined by the duration, positional fluctuations, and morphological fluctuations of the faulty path in consecutive image frames; Indicates the first The cause of each erroneous boot path is determined by the cause determination function; Indicates time The stability of the current confirmed real runway path is calculated by the matching stability of the correspondence between the real light spot and the real runway path in the light relationship diagram. This formula unifies the structural integrity, misidentification probability, formation cause, and real runway path recognition state of the misguided path into a path judgment state, enabling the reinforcement learning model to select centerline destruction action, boundary misalignment action, entrance desymmetry action, depth truncation action, or real path preservation action based on the path judgment state.

[0041] S130. Construct a set of path interference actions based on the path judgment state, and generate a combination of path interference actions from the set of path interference actions based on the path judgment state using a reinforcement learning model. The combination of path interference actions includes centerline destruction actions, boundary misalignment actions, entrance desymmetry actions, depth truncation actions, and true path preservation actions.

[0042] Specifically, when the server constructs the path interference action set based on the continuity of the centerline, the parallelism of the boundaries, the symmetry of the entrance, the depth extension, and the current confirmed stability of the actual runway path, it first reads the previously generated path judgment state and maps each state indicator in the path judgment state to the navigation light nodes and light relationships in the light relationship diagram. The path interference action set refers to the set of candidate control actions that can weaken the visual structure of the misguided path without disrupting the light relationships of the actual runway path; the centerline disruption action is used to disrupt the continuous arrangement of runway centerlines in the misguided path; the boundary misalignment action is used to disrupt the parallel arrangement of runway boundaries in the misguided path; the entrance desymmetry action is used to disrupt the left-right symmetry arrangement of runway entrances in the misguided path; the depth truncation action is used to disrupt the followable extension formed along the aircraft's viewing direction in the misguided path; and the actual path maintenance action is used to maintain the continuity of the centerline, the parallelism of the boundaries, the symmetry of the entrance, the progressive relationship of the touchdown area, the stopping and blocking relationship, and the path connection relationship of the actual runway path. When constructing a set of path interference actions, the server does not directly use highlighting or debrightening as the only action. Instead, it generates corresponding structural weakening actions based on the most prominent path features of the erroneous guidance path and ensures that subsequent action combinations do not weaken the real runway path by maintaining the real path.

[0043] in, Indicates the first The set of path interference actions corresponding to each erroneous boot path is determined by the server based on the first... The path determination status of each incorrect guidance path is constructed; This represents the error guidance path number, and its value range is the set of all error guidance path numbers identified at the current moment. This indicates a centerline disruption action, triggered by the continuity of the centerline. This indicates a boundary misalignment action, triggered by the degree of boundary parallelism. This indicates a desymmetry action at the entrance, triggered by the degree of symmetry at the entrance. This indicates a depth cutoff action, triggered by the degree of depth extension. This represents the true path-maintaining action, triggered by the current confirmed stability level of the true runway path. This formula organizes actions corresponding to different path characteristics into a unified set of path interference actions, enabling the server to select appropriate path interference actions based on the structural integrity of the erroneous guided path.

[0044] When the server constructs a centerline disruption action based on the centerline continuity, it first determines whether suspected interference spots in the misguided path form a continuous arrangement along a certain direction. Then, it identifies navigation light nodes that are spatially adjacent to this continuous arrangement but do not belong to the critical maintenance area of ​​the actual runway path from the light relationship diagram. The critical maintenance area of ​​the actual runway path refers to the area that must remain stable to ensure the pilot or airborne vision equipment can confirm the actual runway path, including the runway centerline, runway boundaries, runway threshold, touchdown area, and stop row lights area. If the centerline continuity is high, the server sets navigation light nodes that can affect the visual continuity of the misguided path as candidate control objects for the centerline disruption action. By changing the brightness change sequence, brightness change amplitude, or display duration of these navigation light nodes, the continuous arrangement in the misguided path is interrupted, while maintaining the continuous display relationship of the runway centerline lights in the actual runway path.

[0045] in, This indicates a centerline disruption action, generated by the server when the continuity of the centerline meets the action triggering conditions. The function representing the centerline failure action is determined by the continuity of the centerline, the light relationship diagram, the error guidance path, and the current confirmed stability of the actual runway path. Indicates the first The continuity of the centerline of the erroneous guidance path is calculated by the spacing stability and directional consistency of adjacent suspected interference spots; The diagram shows the relationship between the navigation lights, which is used to determine which navigation light nodes belong to the critical maintenance area of ​​the actual runway path and which navigation light nodes can participate in centerline breaking actions. Indicates the first The incorrect guidance path is determined by comparing the set of previous suspected paths with the light relationship diagram; Indicates time The current confirmed stability of the actual runway path is calculated from the matching stability of the actual runway path in the actual light spot and light relationship diagram. This formula generates a centerline disruption action under the condition that the centerline continuity is high and the current confirmed stability of the actual runway path allows it, so that the action target is limited to weakening the centerline continuity features of the misguided path, rather than weakening the actual runway path.

[0046] When the server constructs boundary misalignment actions based on the degree of boundary parallelism, it first identifies whether there are approximately parallel suspected interference light spots arranged on both sides of the erroneous guidance path. Then, based on the light relationship diagram, it excludes the runway edge lights and their associated light relationships that are responsible for runway width identification in the real runway path. If the degree of boundary parallelism is high, the server selects the navigation light nodes located on both sides of the erroneous guidance path that can affect the stability of the false boundary as candidate control objects for the boundary misalignment action. It then controls these navigation light nodes to have subtle differences in the timing, magnitude, or duration of brightness changes, so that the suspected boundaries on both sides of the erroneous guidance path no longer maintain a synchronized, equidistant, and parallel visual state. The purpose of the boundary misalignment action is not to produce obvious light flickering, but to make the erroneous guidance path no longer meet the stable parallel structure that the runway boundary should have.

[0047] in, This indicates a boundary misalignment action, which is generated by the server when the parallelism of the boundary meets the action triggering conditions. The boundary misalignment action generation function is determined by the degree of boundary parallelism, the light relationship diagram, the error guidance path, and the current confirmed stability of the actual runway path. Indicates the first The parallelism of the boundary of the erroneous guidance path is calculated by the directional similarity and width uniformity of the suspected interfering light spots on the left and right sides; A light relationship diagram is used to determine the boundary parallel relationship of runway edge lights and the navigation light nodes that can participate in boundary misalignment actions. Indicates the first A series of error-guided paths are used to determine the locations of false boundaries that need to be weakened; This indicates the current confirmed stability level of the actual runway path, used to limit boundary misalignment actions from weakening the boundary parallelism of the actual runway path. The formula determines whether to disrupt the false boundary structure based on the boundary parallelism and uses a lighting diagram to prevent boundary misalignment actions from mistakenly affecting the actual runway path boundaries.

[0048] When the server constructs the desymmetry action based on the symmetry of the entrance, it first determines whether there are suspected interfering light spots at the end of the erroneous guidance path that are similar in number, brightness, or position on the left and right sides. Then, it uses the light relationship diagram to determine whether the end area has spatial overlap or functional association with the entrance lights of the actual runway path. If the entrance symmetry is high and the end of the erroneous guidance path does not belong to the entrance area of ​​the actual runway path, the server selects the navigation light node adjacent to the end of the erroneous guidance path as the candidate control object for the entrance desymmetry action. By adjusting the left and right brightness ratio, display start and end times, or brightness recovery order, the end of the erroneous guidance path no longer presents the left and right symmetry relationship required for the runway entrance. If the end of the erroneous guidance path is close to the entrance area of ​​the actual runway path, the server increases the priority of the actual path maintenance action to avoid the entrance desymmetry action weakening the recognition stability of the actual entrance lights.

[0049] in, This indicates the entry point desymmetry action, which is generated by the server when the symmetry of the entry point meets the action triggering conditions. The function for generating desymmetric actions at the entrance is determined by the degree of symmetry at the entrance, the lighting relationship diagram, the error guidance path, and the current confirmed stability of the actual runway path. Indicates the first The degree of symmetry at the entrance of the erroneous guidance path is calculated by the similarity of the number and brightness distribution of the suspected interfering light spots on the left and right sides of the end. A diagram showing the lighting relationships is used to determine the entrance lights and their symmetrical relationships on the actual runway path. Indicates the first A false entry path is used to determine the location of the false entry point. This indicates the current confirmed stability level of the actual runway path, used to limit de-symmetry actions at the entrance to prevent abnormal left-right displays of the actual entrance lights. The formula determines whether a false entrance structure needs to be weakened based on the entrance symmetry and maintains the symmetrical recognition relationship of the actual entrance lights through a light relationship diagram.

[0050] When the server constructs a depth truncation action based on the depth of extension, it first determines whether the misguided path forms a followable extension clue along the aircraft's observation direction. Then, it combines directional proximity, spatial proximity, and salience to determine the most visually appealing distant misidentification area along the misguided path. The distant misidentification area refers to the area where the misguided path extends along the aircraft's observation direction, causing the pilot or onboard vision equipment to have a continuous tendency to follow it. If the depth of extension is high, the server selects navigation light nodes related to the distant misidentification area from the light relationship diagram, but which do not affect the progression relationship and path connection relationship of the actual runway path grounding area, as candidate control objects for the depth truncation action. By adjusting the brightness progression relationship, display continuity, or display duration of these navigation light nodes, the distant extension of the misguided path is weakened, thereby reducing its ability to mislead the aircraft's observation direction.

[0051] in, This indicates a depth cutoff action, which is generated by the server when the depth extension meets the action triggering conditions. The depth cutoff action generation function is determined by the depth extension, directional proximity, spatial proximity, salience, light relationship diagram, error guidance path, and the current confirmed stability of the actual runway path. Indicates the first The depth of the erroneous guidance path is calculated by the proximity between the main direction of the path and the direction of observation of the aircraft, as well as the proportion of the forward suspected interference spot. Indicates the first The degree of directional similarity of the incorrect guidance path is calculated from the angle between the main direction of the incorrect guidance path and the actual runway path or the direction observed by the aircraft. Indicates the first The spatial proximity of the erroneous guidance paths is calculated from the minimum distance and overlap between the erroneous guidance paths and the actual runway paths; Indicates the first The significance of a faulty guidance path is determined by the average brightness of the suspected interfering spot, the similarity of its colors, and the area occupied by the image. This diagram illustrates the relationship between lights, used to determine the progressive relationship of grounding areas and the connection relationship of paths; Indicates the first A series of error guidance paths are used to locate remote misidentified areas; This indicates the current confirmed stability of the actual runway path and is used to limit the depth truncation action from disrupting the depth guidance of the actual runway path. By simultaneously considering extensibility, directionality, spatial proximity, and salience, this formula prioritizes depth truncation actions on the far-end misidentification areas most likely to cause persistent misguidance.

[0052] When the server constructs the real path holding action based on the current confirmed stability of the actual runway path, it first determines whether the matching between the actual light spot and the actual runway path in the light relationship diagram is stable, and whether the combination of path interference actions may indirectly affect the actual runway path. If the current confirmed stability of the actual runway path is low, or the directional and spatial similarity between the erroneous guidance path and the actual runway path is high, the server sets the real path holding action as a mandatory action. The real path holding action maintains the brightness ratio, display order, and safety display relationship among the runway centerline lights, runway edge lights, entrance lights, landing area lights, stop row lights, and taxiing connection lights, so that the combination of path interference actions weakens the erroneous guidance path without reducing the confirmability of the actual runway path.

[0053] in, This indicates that the actual path is maintained, and is generated by the server based on the current confirmed stability of the actual runway path. The function representing the real path-keeping action generation function is jointly determined by the current confirmed stability of the real runway path, the light relationship diagram, and the operational status of the navigation light nodes. Indicates time The stability of the actual runway path is currently confirmed by calculation based on the matching stability of the correspondence between the actual light spot and the actual runway path in the light relationship diagram; A lighting relationship diagram is used to determine the necessary continuous centerline relationships, parallel boundary relationships, symmetrical entrance relationships, progressive grounding area relationships, stopping and blocking relationships, and path connection relationships; Indicates time The operational status of the navigation light nodes is collected by the navigation light control system and navigation light operation monitoring equipment. This formula generates hold-up actions based on the stability of the actual runway path identification and the operational status of the navigation light nodes, providing a safe hold-up basis for subsequent path interference actions.

[0054] When the server marks the applicable conditions for each path interference action in the path interference action set, it binds each path interference action to its applicable error guidance path type, path judgment state range, allowed navigation light nodes, prohibited navigation light nodes, allowed brightness change method, allowed duration, allowed recovery method, and action combination restrictions. The applicable conditions refer to the boundary conditions under which path interference actions can be selected and executed by the reinforcement learning model; the types of erroneous paths correspond to their causes, including water reflection paths, wet snow scattering paths, background light interference paths, intersecting light paths, and refraction offset paths; the path judgment state range is used to limit the conditions under which a certain action can be selected based on the continuity of the centerline, the parallelism of the boundary, the symmetry of the entrance, the depth extension, and the current confirmed stability of the real runway path; the allowed navigation light nodes are those that can participate in weakening erroneous paths without disrupting the lighting relationships of the real runway path; the prohibited navigation light nodes are the key holding nodes in the real runway path; the allowed brightness change methods include gradual dimming, short-term phase reversal, sequential delay, and smooth recovery; the allowed duration is used to limit the duration of the action; the allowed recovery method is used to limit the way to return to the safe display state after the action ends; the action combination restriction is used to constrain the superposition relationship between multiple path interference actions.

[0055] in, Indicates path interference action The applicable conditions labeling results are generated by the server based on the light relationship diagram, path judgment status, and airport safety constraints; This represents any path interference action, and its values ​​include centerline destruction action, boundary misalignment action, entrance desymmetry action, depth truncation action, or true path preservation action; Indicates path interference action The applicable error boot path type is determined by the cause of the error. Indicates path interference action The applicable range of path judgment status is determined by the range of values ​​for centerline continuity, boundary parallelism, entrance symmetry, depth extension, and the current confirmed stability of the actual runway path. Indicates path interference action The set of navigation light nodes that are allowed to participate is determined by the non-critical hold-up nodes in the light relationship diagram and the navigation light nodes adjacent to the misguided path; Indicates path interference action The set of navigation light nodes that are prohibited from participating is determined by the critical maintenance nodes of the actual runway path in the light relationship diagram; Indicates path interference action The permitted brightness variation method is determined by the navigation light node control capabilities and safety display rules; Indicates path interference action The permissible duration is determined by the navigation light node response delay, the aircraft observation period, and the safety constraint level. Indicates path interference action The permitted recovery method is determined by the requirement to retain the actual path after the action ends; Indicates path interference action The action combination constraint is determined by whether the superposition of multiple path interference actions disrupts the real runway path lighting relationship. This formula structures the execution boundary of each path interference action, ensuring that the reinforcement learning model can only select and combine path interference actions within the allowed range.

[0056] After the server inputs the path judgment state into the reinforcement learning model, the model selects the corresponding path interference action from the set of path interference actions based on the path judgment state to generate a path interference action combination. The reinforcement learning model is a model that continuously adjusts the control strategy through states, actions, and feedback. In this scheme, it does not directly output the brightness value of a single navigation light node, but first outputs the selection result of the path interference action, and then determines the specific navigation light nodes involved in the control based on the light relationship diagram. A path interference action combination refers to the combined result of multiple path interference actions executed simultaneously or sequentially for the same misguided path. When selecting actions, the reinforcement learning model uses centerline continuity, boundary parallelism, entrance symmetry, depth extension, directional proximity, spatial proximity, salience, stability, cause of formation, and the currently confirmed stability of the actual runway path as state information. It also uses historical experience data to determine which path interference actions can weaken the misguided path under similar states while maintaining the stability of the actual runway path.

[0057] in, Indicates the first The combination of path interference actions corresponding to each erroneous guidance path is selected by the reinforcement learning model from the set of path interference actions. This indicates selecting the action combination that maximizes the evaluation value within the parentheses; This represents a candidate path interference action combination, which consists of one or more path interference actions from the path interference action set. Indicates the first The set of path interference actions corresponding to each erroneous guidance path; This represents the state of the reinforcement learning model's path decision. Next, execute the candidate path interference action combination. The value assessment is based on the model parameters. It was trained using historical experience data; The policy parameters of the reinforcement learning model are obtained by training in an offline simulated disturbance environment and adjusting the online constrained policy. Indicates the first Determine the status of the path corresponding to each incorrect guidance path; The weight of the security risk penalty is pre-set by the target airport's security level and navigation light control rules. Indicates the combination of interference actions for candidate paths Relative to the lighting relationship diagram The safety risk assessment is determined by whether the candidate path interference action combination may disrupt the actual runway path lighting relationship. This formula selects path interference action combinations through a combination of reinforcement learning value evaluation and safety risk penalty, enabling the reinforcement learning model to prioritize action combinations that can weaken misguided paths without disrupting the actual runway path.

[0058] When determining the primary and secondary path interference actions in the path interference action combination based on the cause of the error, the server first reads the cause of the erroneous path and then selects the priority action according to the misleading mechanism corresponding to different causes. The primary path interference action is the path interference action that decisively weakens the main misidentification features of the current erroneous path; the secondary path interference action is the path interference action used to cooperate with the primary path interference action, enhance the elimination effect, or maintain the stability of the real runway path. If the cause is a water reflection path, the misleading effect mainly comes from the continuous extension of the mirror image. The server prioritizes centerline disruption or depth truncation as the primary path interference action. If the cause is a wet snow scattering path, the misleading effect mainly comes from the false boundary formed by light spot diffusion. The server prioritizes boundary misalignment as the primary path interference action and uses the real path preservation action as an auxiliary path interference action. If the cause is a background light interference path, the misleading effect mainly comes from the similarity between the background light arrangement and the entrance or depth structure. The server prioritizes entrance desymmetry or depth truncation as the primary path interference action. If the cause is a cross-light path, the server prioritizes boundary misalignment or depth truncation as the primary path interference action. If the cause is a refraction offset path, the server prioritizes centerline disruption as the primary path interference action and uses the real path preservation action as an auxiliary path interference action.

[0059] in, This indicates the main path interference action, which is determined by the server from the combination of path interference actions based on the cause of the interference and the path judgment status. This indicates auxiliary path interference actions, which can be one or more path interference actions, determined by the server based on the main path interference actions and the current confirmed stability of the actual runway path. This represents the function that determines the primary and secondary actions, which is jointly determined by the cause of formation, the combination of path interference actions, and the path judgment state. Indicates the first The cause of this error path is determined by the cause of the preceding error. Indicates the first The path interference action combination corresponding to each erroneous guidance path; Indicates the first The path judgment state corresponding to each error guidance path is used to determine the most prominent misidentification feature. This formula determines the primary and secondary action relationship through the cause and path judgment state, enabling different types of error guidance paths to adopt different action combination logics.

[0060] When the server determines the navigation light nodes participating in control and their corresponding action execution parameters based on the light relationship diagram, it first maps each path interference action in the path interference action combination to the light relationship diagram. It then identifies navigation light nodes that are spatially adjacent to the erroneous guidance path, related to the corresponding light relationship, and not belonging to the set of prohibited navigation light nodes as participating navigation light nodes. Action execution parameters refer to the control parameters required to apply the path interference action to a specific navigation light node, including the direction of brightness change, the magnitude of brightness change, the order of brightness change, the duration of the action, and the action recovery method. The direction of brightness change indicates whether the navigation light node brightens, dims, or remains bright during action execution; the magnitude of brightness change indicates the size of the brightness change; the order of brightness change indicates the execution sequence among multiple participating navigation light nodes; the duration of the action indicates the duration the path interference action is maintained; and the action recovery method indicates how the navigation light node returns to its true path-maintained state or normal display state after the action ends. Based on the control loop, communication address, response latency, security constraint level, and allowable control range in the light relationship diagram, the server generates specific action execution parameters that can be sent to the navigation light control system.

[0061] in, Indicates the first The set of action execution parameters corresponding to each error guidance path is generated by the server based on the combination of path interference actions and the light relationship diagram. The action execution parameter generation function is determined by the combination of path interference actions, the light relationship diagram, the erroneous guidance path, the applicable condition label, the current confirmed stability of the actual runway path, and the operating status of the navigation light nodes. Indicates the first The path interference action combination corresponding to each erroneous guidance path; This diagram illustrates the relationship between navigation lights, used to identify the navigation light nodes involved in control, control loops, communication addresses, and safety constraint levels. Indicates the first An error guidance path is used to determine the spatial area where the path interference action needs to take effect; This indicates the labeling results of the applicable conditions for each path interference action in the path interference action set, which is used to limit the navigation light nodes that are allowed to participate, the navigation light nodes that are prohibited from participating, the allowed brightness change methods, the allowed duration, the allowed recovery methods, and the action combination restrictions; This indicates the current confirmed stability level of the actual runway path, which is used to determine the control strength of the actual path-holding action; This indicates the operational status of navigation light nodes, used for troubleshooting, maintenance, downgrading, or communication anomalies. The formula combines path interference actions with light relationship diagrams, erroneous guidance paths, and applicable condition annotations to generate action execution parameters specific to each navigation light node, providing a clear control basis for subsequent safety checks and navigation light node control command conversions.

[0062] S140. Perform a safety check on the path interference action combination, and restrict the execution of path interference action combinations that disrupt the actual runway path lighting relationship based on the safety check result. At the same time, convert the path interference action combination that passes the safety check into navigation light node control commands.

[0063] Specifically, when the server maps path interference action combinations to navigation light nodes in the lighting relationship diagram, it first reads each path interference action in the combination and its corresponding execution parameters. Then, it reads the node position, light group, lighting relationship, control loop, communication address, and safety constraint level of each navigation light node in the lighting relationship diagram. Node position refers to the spatial location of the navigation light node in the airport coordinate system; light group refers to the functional category undertaken by the navigation light node; lighting relationship refers to the centerline continuity, boundary parallelism, entrance symmetry, grounding area progression, stop / blocking, or path connection relationships formed by the navigation light node and other navigation light nodes; the safety constraint level indicates the importance of the navigation light node to the safety of real runway path identification. Based on the proximity between the effective area of ​​the path interference action combination and the node position of the navigation light node, the importance of the navigation light node in the real runway path lighting relationship, and the safety constraint level of the navigation light node, the server classifies the participating navigation light nodes into safety-maintaining nodes and interference control nodes. Safety maintenance nodes are navigation light nodes that must maintain the stability of the actual runway path lighting relationship; interference control nodes are navigation light nodes that are allowed to perform path interference actions within the range that does not disrupt the actual runway path lighting relationship.

[0064] in, Indicates the first The safety maintenance score of each navigation light node is a higher value, indicating that the navigation light node should be classified as a safety maintenance node. This represents the navigation light node number, and its value range is the set of all navigation light node numbers in the light relationship diagram. Indicates the first The safety constraint level of each navigation light node is obtained from the node attributes in the light relationship diagram; Indicates the first The importance of each navigation light node in the actual runway path lighting relationship is calculated by the number and safety level of the centerline continuity, boundary parallelism, entrance symmetry, touchdown zone progression, stop and blockage relationship, and path connection relationship connected by the navigation light node. Indicates the first The actual path maintenance requirement for each navigation light node is determined by the current confirmed stability of the actual runway path and the light group to which the navigation light node belongs. Indicates the first The spatial proximity between a navigation light node and the area of ​​effect of a misguided path is calculated from the minimum distance from the location of the navigation light node to the misguided path; the closer the distance, the higher the value. , , and The weights of the corresponding factors are pre-set by the target airport's safety rules and navigation light control strategies. This formula, by comprehensively considering safety constraints, the importance of the actual runway path relationship, the actual path maintenance requirements, and the spatial proximity of the misguided path, determines whether a navigation light node is more suitable for maintaining the stability of the actual runway path or for participating in mitigating misguided paths.

[0065] When the server performs a centerline safety check on a centerline disruption action, it compares the corresponding navigation light node with the centerline continuity in the light relationship diagram to determine whether the centerline disruption action will cause the runway centerline lights in the actual runway path to experience continuous display interruptions, abnormal brightness progression, or centerline direction deviation. The centerline safety check prioritizes protecting the longitudinal recognition capability of the actual runway path, ensuring that pilots or airborne vision equipment can continuously confirm the runway centerline direction. If the centerline disruption action only affects the interference control node and does not change the runway centerline light display relationship in the safety holding node, the centerline safety check passes. If the centerline disruption action weakens any critical centerline continuity in the actual runway path, the centerline safety check fails.

[0066] in, This represents the centerline safety check value; a higher value indicates a more stable continuity of the centerline. The set of graph edges in the lighting relationship diagram that belong to the continuous relationship of the center line is obtained from the lighting relationship diagram; The first line representing the continuous relationship of the centerline The navigation light node and the first One navigation light node; Indicates the effect of the centerline breaking action on the first The brightness change caused by each navigation light node is obtained from the action execution parameters in the path interference action combination; Indicates the effect of the centerline breaking action on the first The amount of brightness change caused by each navigation light node; Indicates the first The navigation light node and the first The maximum allowable brightness difference between each navigation light node is determined by the target airport navigation light display rules, the node's rated brightness, and the safety constraint level. This represents a stable parameter to prevent the denominator from being zero, and is preset to a positive number by the system. The formula determines whether a centerline disruption action will damage the longitudinal continuity of the actual runway path by judging whether the difference in brightness between the navigation light nodes at both ends of the centerline continuity exceeds the allowable range.

[0067] When the server performs boundary safety checks on boundary misalignment actions, it compares the corresponding navigation light node with the boundary parallelism in the light relationship diagram to determine whether the boundary misalignment action will cause imbalance in the recognition of the left and right boundaries of the actual runway path, abnormal runway width perception, or misleading boundary direction. Boundary safety checks ensure that the lateral range of the actual runway path defined by the runway edge lights is not disrupted by path interference actions. If the boundary misalignment action only weakens the false boundary parallelism features in the erroneous guidance path, and the runway edge lights in the actual runway path still maintain left-right correspondence and consistent boundary direction, then the boundary safety check passes. If the boundary misalignment action causes the brightness difference, timing difference, or directional relationship between the left and right edge lights of the actual runway path to exceed the allowable range, then the boundary safety check fails.

[0068] in, This represents the boundary security check value; a higher value indicates a more stable boundary parallel relationship. This represents the difference in the change between the left and right boundary directions of the actual runway path after the boundary misalignment action is performed. It is calculated from the boundary fitting direction corresponding to the runway edge light. The average brightness change of the left runway edge light under boundary misalignment action represents the actual runway path, which is obtained from the statistics of the control parameters of the left runway edge light. The average brightness change of the right-side runway edge light under boundary misalignment action represents the actual runway path, which is obtained from the statistics of the right-side runway edge light control parameters. The current average brightness of the runway edge lights on the left side of the actual runway path is obtained from the operating status of the navigation light nodes; This indicates the current average brightness of the runway edge lights on the right side of the actual runway path; This represents a stable parameter to prevent the denominator from being zero, and is preset to a positive number by the system. The formula uses the differences in left and right boundary direction and brightness to determine whether boundary misalignment disrupts the lateral boundary recognition of the actual runway path.

[0069] When the server performs entrance safety checks on entrance desymmetry actions, it compares the corresponding navigation light node with the entrance symmetry relationship in the light relationship diagram to determine whether the entrance desymmetry action will cause inconsistencies in the left and right display of the entrance lights on the actual runway path, delays in entrance position recognition, or misjudgments of the entrance direction. Entrance safety checks ensure that the starting position and entry direction of the actual runway path are stable and identifiable. If the entrance desymmetry action only affects interference control nodes around the end of the erroneous guidance path, and the left-right brightness ratio, display order, and color meaning of the entrance lights on the actual runway path remain unchanged, then the entrance safety check passes. If the entrance desymmetry action causes a mismatch between the left and right entrance lights on the actual runway path, then the entrance safety check fails.

[0070] in, This represents the entry security check value; a higher value indicates a more stable symmetry relationship at the entry point. The brightness of the left entrance light at the actual runway path entrance after control is obtained by superimposing the current brightness of the entrance light and the brightness change corresponding to the entrance desymmetry action. This indicates the brightness of the right-hand entrance light at the actual runway path entrance after control. The start time of the display of the left entrance light at the actual runway path entrance is determined by the navigation light node control command; Indicates the start time of the display of the right entrance light at the actual runway path entrance; This represents a stable parameter to prevent the denominator from being zero, and is preset to a positive number by the system. The formula determines whether the desymmetry action at the entrance disrupts the identification of the actual runway path entrance by checking the consistency of brightness on the left and right sides of the entrance lights and the consistency of display timing.

[0071] When the server performs depth safety verification on a depth truncation maneuver, it compares the corresponding navigation light node with the grounding area progression and path connection relationships in the light relationship diagram to determine whether the depth truncation maneuver will cause a visual interruption in the actual runway path grounding area, unclear legal connection positions between the runway path and taxiway path, or distortion of the depth relationship. Depth safety verification is used to ensure the continuous guidance capability of the actual runway path in the aircraft's observation direction. If the depth truncation maneuver only truncates the false depth extension features of the erroneous guidance path and does not affect the grounding area progression and path connection relationships in the actual runway path, then the depth safety verification passes; if the depth truncation maneuver weakens the continuous identification of the actual grounding area or the legal taxiway connection area, then the depth safety verification fails.

[0072] in, This represents the depth security check value; a higher value indicates a more stable relationship between the grounding area progression and the path connection. The set of graph edges representing the progressive relationship and path connection relationship of the grounding area in the lighting relationship diagram is obtained from the lighting relationship diagram; This represents the number of edges in the graph edge set; This indicates two navigation light nodes that constitute a progressive relationship or path connection relationship in the grounding area; Indicates the first The depth display intensity of each navigation light node after control is determined by the brightness, display duration, and spatial order of the navigation light node after control. Indicates the first The depth indication intensity of each navigation light node after control; This represents a stable parameter to prevent the denominator from being zero, and is preset to a positive number by the system. The formula determines whether the depth cutoff action affects the near-far guidance relationship of the actual runway path by judging whether the display intensity of adjacent navigation light nodes in the depth direction of the actual runway path remains progressively consistent.

[0073] When the server performs a safety check on the actual path holding action, it comprehensively compares the centerline continuity, boundary parallelism, entrance symmetry, grounding zone progression, stop / blocking relationships, and path connection relationships of the corresponding navigation light nodes in the light relationship diagram. This comparison determines whether the actual path holding action can offset the indirect impact of other path interference actions on the actual runway path. The safety check does not determine whether a particular mitigation action is safe, but rather whether the actual path holding action is sufficient to maintain the current confirmed stability of the actual runway path. If the actual path holding action can maintain the brightness ratio, display order, and safe display relationships of key navigation light nodes, it increases the likelihood of safe passage of the path interference action combination. If the actual path holding action is insufficient to stabilize the actual runway path, it requires regeneration or enhancement of the actual path holding action.

[0074] in, This indicates the value for maintaining safety; the higher the value, the stronger the ability of the real path holding action to maintain the relationship between the actual runway path and the lights. This indicates the centerline safety check value; This represents the boundary security check value; This represents the entry security check value; This represents the depth security check value; The stop blocking safety check value is calculated based on whether the stop row of lights maintains the "no passage" display after control. , , , and These represent the weights of the corresponding verification items, determined by the safety requirements of the target airport at different operational phases. This formula, by comprehensively considering the safety status of various lighting relationships along the actual runway path, determines whether the actual path-holding actions can support the execution of a combination of path-interference actions.

[0075] When the server performs an overall safety check on a combination of path interference actions, it not only determines whether each individual path interference action passes its corresponding safety check, but also whether the superposition of multiple path interference actions creates new lighting conflicts. The overall safety check is used to identify situations where individual actions appear safe, but the combination disrupts the actual runway path lighting relationship. The server determines whether the combination of path interference actions causes directional conflicts between the actual runway path and taxiway path, whether it weakens the stop-and-block relationship, whether it causes multiple critical navigation light nodes to dim simultaneously, whether it causes adjacent navigation light nodes to flash abnormally, and whether it prevents the pilot or onboard vision equipment from distinguishing between the actual runway path and taxiway path. If none of the above issues exist, a safety check pass result is generated; if any of the issues exist, a safety check fail result is generated.

[0076] in, This represents the overall security check value; the higher the value, the safer the overall combination of path interference actions. This represents the minimum value among all individual safety check values, used to reflect the limitation of the weakest lighting relationship on overall safety; The penalty for directional conflict is calculated based on whether the actual runway path and the taxiing path cause confusion in directional prompts after control is initiated. The indication of stopping the weakening of the penalty is calculated based on whether the prohibition of passage still exists after the control of the traffic lights is stopped. The penalty for abnormal flickering is calculated based on whether the frequency, amplitude, and timing of brightness changes in adjacent navigation light nodes exceed the safety display rules. , and The weight of the corresponding penalty item is set by the target airport's security policy. This formula evaluates whether a combination of path interference actions is suitable for execution by combining the minimum value of individual checks and the combined conflict penalty, avoiding the formation of new visual risks by superimposing multiple actions with small security boundaries.

[0077] When the server generates a security check failure result, it performs restriction processing on the corresponding path interference actions based on the reason for failure. The reason for failure refers to the specific type of light relationship disruption that caused the path interference action combination to fail the security check, including weakened centerline continuity, disrupted boundary parallelism, abnormal entrance symmetry, interrupted grounding area progression, weakened stop / block relationship, conflicting path connections, or increased risk of abnormal flickering. Restriction processing includes deleting the corresponding path interference action, reducing the brightness change amplitude of the corresponding path interference action, shortening the action duration of the corresponding path interference action, changing the brightness change order of the corresponding path interference action, or enhancing the real path maintenance action. After completing the restriction processing, the server resubmits the restricted path interference action combination for security check until a security check success result is generated, or it is determined that there is no executable path interference action combination in the current path judgment state.

[0078] in, This represents the combination of path interference actions after the restriction processing; The constraint processing function is determined by the original path interference action combination, the reason for failure, and the light relationship diagram. This indicates a combination of path interference actions prior to security verification. This represents the set of reasons for failure, obtained from the overall security check and the output of each individual security check. This diagram illustrates the light relationships on the actual runway path, used to determine the light relationships that should be protected and the navigation light nodes that should be restricted. The formula, by applying reverse constraints to the path interference action combinations without considering the cause, makes the restricted path interference action combinations closer to the actual runway path safety boundaries.

[0079] When generating a security verification result, the server reads the control loop, communication address, current brightness, allowed brightness change method, response delay, and security constraint level of the participating navigation light nodes according to the light relationship diagram. It then generates corresponding navigation light node control instructions based on the path interference action combination. Navigation light node control instructions refer to control data that can be directly executed by the navigation light control system, including the navigation light node's communication address, control loop, target brightness, brightness change sequence, action duration, action recovery method, and execution time. When generating navigation light node control instructions, the server transforms the structured actions in the path interference action combination into specific lighting control content, ensuring that centerline disruption actions, boundary misalignment actions, entrance desymmetry actions, depth truncation actions, and true path maintenance actions are implemented at their respective navigation light nodes.

[0080] in, Indicates the first The set of control instructions for the navigation light nodes corresponding to each incorrect guidance path; The control command generation function is determined by the combination of path interference actions that have passed safety verification, the light relationship diagram, and the operating status of the navigation light nodes. This indicates a combination of path interference actions that have passed security checks; It represents a lighting relationship diagram, used to read the control loop, communication address, allowed brightness change method, response delay, and safety constraint level of the navigation light nodes involved in the control; Indicates time The system monitors the operational status of navigation light nodes, including current brightness, fault status, communication status, and control loop status. This formula combines safe path interference actions with executable navigation light node information to generate control commands that can be directly issued to the navigation light control system.

[0081] When adjusting the control commands for navigation light nodes based on the control requirements corresponding to safety maintenance nodes and interference control nodes, the server employs different control boundaries for different types of navigation light nodes. For safety maintenance nodes, navigation light node control commands are only allowed to maintain the current display state or perform actions that maintain the true path; actions that would weaken the brightness changes in the lighting relationship of the true runway path are not permitted. For interference control nodes, navigation light node control commands are allowed to perform centerline disruption actions, boundary misalignment actions, entrance desymmetry actions, or depth truncation actions within the safety verification allowable range. For navigation light nodes that are adjacent to both the true runway path and the misguided path, the server prioritizes meeting the requirements for maintaining the lighting relationship of the true runway path, and then performs path interference actions within the remaining allowable control range. Through this adjustment method, the final output navigation light node control commands can change the visual structure of the misguided path without compromising the recognition stability of the true runway path.

[0082] in, Indicates the first The final output of each navigation light node is the navigation light node control command; The diagram shows the corresponding number of lights in the relationship diagram. A graph node for each navigation light node; This represents the set of nodes that maintain security, determined by the security maintenance score and the security constraint level. This represents the set of interference control nodes, determined by the area of ​​influence of the error guidance path, node locations, and security constraint levels. Indicates the first The hold control command for each navigation light node is determined by the actual path hold action and the current display status; Indicates the first The interference control command for each navigation light node is obtained by combining and converting path interference actions. This represents a control range clipping function, used to restrict interference control commands to the allowable control range; Indicates the first The minimum control boundary allowed for each navigation light node is determined by the safety constraint level, the allowed brightness variation method, and the current operating status. Indicates the first The maximum allowable control boundary for each navigation light node is determined by rated brightness, control loop capability, and safety display rules. This formula, by employing different control rules for safety holding nodes and interference control nodes, enables the final navigation light node control commands to mitigate erroneous guidance paths while protecting the actual runway path.

[0083] S150. Change the visual structure of the erroneous guidance path according to the navigation light node control command, and re-acquire the airport environment perception results after executing the navigation light node control command to evaluate the effect of the path interference action combination on eliminating the erroneous guidance path.

[0084] Specifically, when determining the navigation light node, brightness change direction, brightness change amplitude, brightness change sequence, action duration, and action recovery method in the navigation light node control command, the server first parses the navigation light node control command obtained after the prior security check, and binds each control content with the path interference action in the path interference action combination. The navigation light node refers to the light node that actually receives the control command; the brightness change direction refers to whether the navigation light node increases, decreases, or maintains brightness within the current control cycle; the brightness change amplitude refers to the magnitude of the brightness change of the navigation light node relative to the current brightness; the brightness change sequence refers to the control order among multiple navigation light nodes; the action duration refers to the duration of the corresponding path interference action; and the action recovery method refers to the way the navigation light node returns to the true path maintenance state or normal display state after the path interference action ends. The server uses this correspondence to determine which navigation light nodes are affected by centerline disruption actions, boundary misalignment actions, entrance desymmetry actions, depth truncation actions, and true path preservation actions. This ensures that the navigation light node control commands are not aimlessly adjusting brightness, but rather are used to change the specific visual structure of the erroneous guidance path.

[0085] in, Indicates the first The correspondence between the instructions and actions corresponding to each incorrect guidance path is obtained by the server by matching the navigation light node control instructions with the path interference actions. This represents the error path number, and its value range is the set of all error path numbers currently identified. The instruction action mapping function is determined by the combination of navigation light node control instructions, path interference actions that have passed safety checks, and the light relationship diagram. Indicates the first The set of control commands for the navigation light nodes corresponding to each erroneous guidance path is generated after the safety verification is passed. This indicates a combination of path interference actions that have passed the safety check, including centerline disruption actions, boundary misalignment actions, entrance desymmetry actions, depth truncation actions, and true path preservation actions. This represents a lighting relationship diagram used to determine the correspondence between navigation light nodes and the actual runway path lighting. This formula, by associating control commands, path interference actions, and the lighting relationship diagram, enables the server to clearly define the action attribution and visual structure change target of each navigation light node's control command.

[0086] After distinguishing between safety maintenance nodes and interference control nodes based on the safety verification results, the server controls the safety maintenance nodes to maintain the lighting relationship of the actual runway path, and controls the interference control nodes to execute corresponding path interference actions. Safety maintenance nodes are key navigation light nodes in the actual runway path that play a role in ensuring centerline continuity, boundary parallelism, entrance symmetry, grounding area progression, stopping / blocking, or path connection. Their control objective is to maintain the stability of the actual runway path identification. Interference control nodes are navigation light nodes located around the misguided path and within the allowable range of the safety verification that can participate in path interference actions. Their control objective is to weaken the visual structure of the misguided path. Specifically, if an interference control node corresponds to a centerline disruption action, the server controls that interference control node to change the brightness continuity between it and adjacent suspected interference spots; if it corresponds to a boundary misalignment action, the server controls that interference control node to change the brightness synchronization relationship on both sides of the misguided path; if it corresponds to an entrance desymmetry action, the server controls that interference control node to change the left-right correspondence at the end of the misguided path; if it corresponds to a depth truncation action, the server controls that interference control node to change the brightness progression relationship at the far end of the misguided path. Through the above controls, the erroneous guidance path no longer stably exhibits characteristics such as continuous centerline, parallel boundary, symmetrical entrance, or extended depth.

[0087] in, Indicates the first The target brightness of each navigation light node after executing the navigation light node control command is determined by the server based on the node type and the corresponding path interference action. The diagram shows the corresponding number of lights in the relationship diagram. A graph node for each navigation light node; This represents the set of nodes that maintain safety, determined by the safety verification results and the light relationship diagram. This represents the set of interference control nodes, determined by the safety verification results, the area affected by the erroneous guidance path, and the light relationship diagram. Indicates the first The brightness of each safety-maintaining node is determined by the actual path maintenance action and the current operating state; Indicates the first The current brightness of each interference control node before it executes control is obtained from the operating status of the navigation light node. Indicates the first The brightness change of each interference control node under corresponding path interference actions is obtained from the action execution parameters in the navigation light node control command. This formula maintains the stability of the actual runway path lighting relationship by applying different control rules to the safety maintenance node and the interference control node, while simultaneously weakening the visual structure of the erroneous guidance path.

[0088] The server continuously collects execution feedback during the execution of navigation light node control commands and adjusts the participation weight of the corresponding navigation light node in subsequent path interference action combinations based on the execution feedback. Execution feedback refers to the actual response of the navigation light node after executing the control command, including actual brightness, response delay, control loop status, communication status, and operational status. Actual brightness is used to determine whether the navigation light node has reached the target brightness; response delay is used to determine whether the navigation light node has completed the brightness change within the expected time; control loop status is used to determine whether the corresponding control channel is normal; communication status is used to determine whether the communication between the server and the navigation light node is reliable; and operational status is used to determine whether the navigation light node is in a normal, faulty, maintenance, degraded, or closed state. If the execution result of a certain interference control node deviates significantly from the control command of the navigation light node, the participation weight of that interference control node in subsequent path interference action combinations is reduced; if a certain interference control node executes stably and can effectively change the visual structure of the erroneous guided path, the participation weight of that interference control node is increased; if a certain safety holding node exhibits an abnormal response, the real path holding action is immediately strengthened, and the control changes involving that safety holding node are reduced.

[0089] in, Indicates the first The participation weight of each navigation light node in the next control cycle is updated by the server based on the feedback from this execution. Indicates the first The participation weight of each navigation light node in the current control cycle is obtained from the previous round of learning data or the default node weight; , and These represent the influence coefficients of brightness response, response delay, and abnormal state on the weighting of the participants, which are preset by the target airport control strategy. Indicates the first The actual brightness of each navigation light node is collected by the navigation light operation monitoring equipment. Indicates the first The target brightness of each navigation light node is determined by the navigation light node control command. Indicates the first The actual response delay of each navigation light node is calculated from the time the control command is issued and the time when the brightness reaches the target range. Indicates the first The maximum allowable response delay for each navigation light node is determined by the type of light, control loop capability, and safety constraint level. Indicates the first The abnormal status flag of each navigation light node is set to a larger value when the control loop status, communication status or operation status is abnormal, and a smaller value when all are normal. This represents a stable parameter to prevent the denominator from being zero, and is preset to a positive number by the system. The formula updates the participation weights of navigation light nodes based on brightness following capability, response timeliness, and abnormal states, ensuring that subsequent path interference actions prioritize reliable navigation light nodes.

[0090] After completing the navigation light node control commands, the server re-acquires airport environmental perception results and re-extracts candidate light spot sets and reconstructs suspected path sets based on these re-acquired results. The re-acquired airport environmental perception results maintain the same data source and spatial coordinates as the previous results, including continuous image frames, low-altitude visibility, precipitation status, ground water accumulation status, wet snow cover status, low-altitude refraction status, background light distribution, aircraft attitude changes, and navigation light node operating status. The server extracts candidate light spot sets in the same manner as before, marking light spots that can stably match the navigation light nodes in the light relationship diagram as real light spots, and marking light spots that cannot stably match but have similar characteristics to navigation light spots as suspected interference spots. Subsequently, based on the spatial distance, arrangement direction, brightness progression, left-right distribution, and duration of the suspected interference spots, the server reconstructs the suspected path set and determines the changes in the erroneous guided path after executing the navigation light node control commands based on the light relationship diagram.

[0091] in, Indicates the first The result of the path judgment status change before and after the execution of the navigation light node control command for the erroneous guided path is obtained by the server calculating the difference between the path judgment status before and after execution. This indicates the path determination status regenerated after executing the navigation light node control command. It is generated from the newly collected airport environment perception results, candidate light spot set, suspected path set, and light relationship diagram. This indicates the path determination state before executing the navigation light node control command, obtained from the preceding sequence identification. This formula, by comparing the path determination states before and after execution, enables the server to determine whether the visual structure of the erroneous path has been weakened, rather than simply determining whether the navigation light node has completed a brightness change.

[0092] The server evaluates the effectiveness of path interference action combinations by considering changes in the continuity of the centerline, the parallelism of the boundaries, the symmetry of the entrance, and the depth of the misguided path, combined with the current confirmed stability of the actual runway path. If the continuity of the centerline decreases after execution, it indicates that the visual structure of the continuous centerline of the misguided path has been weakened; if the parallelism of the boundaries decreases, it indicates that the false boundary structure of the misguided path has been weakened; if the symmetry of the entrance decreases, it indicates that the false entrance structure at the end of the misguided path has been weakened; if the depth of the path decreases, it indicates that the following ability of the misguided path along the aircraft's observation direction has been weakened. Simultaneously, the server also determines whether the current confirmed stability of the actual runway path remains stable or improves. If the main misidentification features of the misguided path decrease and the current confirmed stability of the actual runway path does not decrease, an effective action combination is generated; if the misguided path is weakened but the current confirmed stability of the actual runway path decreases, a risky action combination is generated; if the misguided path is not effectively weakened, an ineffective action combination is generated.

[0093] in, Indicates the first The elimination effect score corresponding to each error guidance path is calculated by the server based on the changes in the error guidance path structure and the changes in the stability of the actual runway path before and after execution. , , , and The evaluation weights for centerline continuity, boundary parallelism, entrance symmetry, depth extension, and current confirmed stability of the actual runway path are respectively set by the safety requirements of the target airport at different operational stages. Indicates the first before execution The continuity of the centerline of a faulty path is obtained from the path judgment status before execution; Indicates the number of executions after execution. The continuity of the centerline of the erroneous guidance path is calculated from the re-collected airport environmental perception results; and These represent the first and second execution times, respectively. The degree of parallelism of the boundaries of the incorrect guidance paths; and These represent the periods before and after execution, respectively. The degree of symmetry of the entry points of each error-guided path; and These represent the periods before and after execution, respectively. The extent to which a wrong path extends; This indicates the current confirmed stability level of the actual runway path before execution, calculated from the airport environment perception results before execution. This indicates the current confirmed stability of the actual runway path after execution, calculated from the re-collected airport environmental perception results. This formula rewards both the weakening of erroneous guidance path structures and the stability of the actual runway path, enabling the assessment of the elimination effect to reflect whether the combination of path interference actions truly improves guidance safety.

[0094] in, Indicates the first The evaluation type of the combination of interference actions corresponding to each erroneous guidance path; The effective elimination threshold is pre-set by the target airport based on the error guidance path elimination requirements and safety redundancy. Indicates the first Scores for the elimination effect corresponding to each erroneous guidance path; This indicates the current confirmed stability level of the actual runway path after execution; This indicates the current confirmed stability of the actual runway path before execution. The formula categorizes path interference action combinations into effective, risky, or ineffective combinations by simultaneously assessing the elimination effect score and the stability of the actual runway path.

[0095] The server writes effective action combinations, risky action combinations, and ineffective action combinations, along with their corresponding path judgment states, safety verification results, navigation light node control commands, execution feedback, and re-collected airport environmental perception results, into the reinforcement learning model's learning data. The learning data refers to the dataset used to update the reinforcement learning model's action selection strategy. It includes the path judgment state before execution, the actual path interference action combinations executed, whether the safety verification passed, the specific navigation light node control commands, execution feedback, the airport environmental perception results after execution, and the elimination effect evaluation type. For effective action combinations, the reinforcement learning model increases their selection probability under similar path judgment states; for risky action combinations, the reinforcement learning model decreases their selection probability in scenarios where the current confirmed stability of the real runway path is low; for ineffective action combinations, the reinforcement learning model decreases their selection probability under the corresponding misguided path type. Through this learning data write-back process, the reinforcement learning model can gradually form a control strategy oriented towards misguided path elimination and constrained by the safety of the real runway path.

[0096] in, Indicates the first The learning data corresponding to each error guidance path is written into the reinforcement learning model by the server after the elimination effect evaluation is completed; This indicates the pre-execution path judgment status, used to describe the completeness of the erroneous guidance path before the combination of interference actions in the execution path; This indicates a combination of path interference actions that have passed security checks; It indicates the security verification result, including the result of passing the security verification, the result of failing the security verification, and the content to be processed that is restricted; Indicates the control command for the navigation light node; This indicates the execution feedback, including actual brightness, response delay, control loop status, communication status, and operating status; This indicates the airport environment perception results that were re-collected after execution. This indicates the type of elimination effect evaluation, including effective action combinations, risky action combinations, or ineffective action combinations. This formula organizes the states, actions, safety constraints, execution feedback, and evaluation results of a complete control loop into learning data, enabling the reinforcement learning model to adjust the generation of subsequent path interference action combinations based on the actual execution results.

[0097] in, This indicates the updated reinforcement learning model parameters, which are obtained by the server based on the learning data. This represents the parameters of the reinforcement learning model before the update; The policy update step size is determined by the strength of the safety constraints and the model update frequency during the online operation phase. This indicates the direction of gradient adjustment for the parameters of the reinforcement learning model, which is calculated during the training process of the reinforcement learning model; This represents the value evaluation of the combination of path-interference actions by the reinforcement learning model in the pre-execution path judgment state; The feedback evaluation function is determined by the elimination effect assessment type, the elimination effect score, and the current confirmed stability of the actual runway path after execution. Indicates the evaluation type of the path interference action combination; Indicates the elimination effect score; This indicates the current confirmed stability of the actual runway path after execution. This formula, by increasing the feedback evaluation for effective action combinations and decreasing the feedback evaluation for risky and ineffective action combinations, makes the reinforcement learning model more inclined to generate safe and effective path interference action combinations in subsequent similar path judgment states.

[0098] S160. Construct a simulated interference environment to train a reinforcement learning model based on the simulated interference environment to identify and eliminate erroneous guidance paths. During the online operation phase, adjust the strategy of the reinforcement learning model according to the security verification constraints to achieve the continuous elimination of multiple erroneous guidance paths.

[0099] Specifically, when constructing a simulated interference environment based on the light relationship diagram, the server first reads the actual runway path, taxiway path, navigation light nodes, light relationships, control loops, communication addresses, and safety constraint levels from the light relationship diagram, and writes these contents into the simulation space. The simulated interference environment refers to the training environment used to reproduce the target airport's light displays, environmental interference, and path misidentification phenomena; the actual runway path refers to the actual guidance path defined by the runway centerline lights, runway edge lights, entrance lights, touchdown area lights, and stop row lights; the taxiway path refers to the ground taxiing guidance path defined by the taxiway centerline lights and taxiway edge lights; the control loop represents the electrical channel through which the navigation light nodes receive control signals; the communication address represents the address where the server locates and issues control commands to the navigation light nodes; and the safety constraint level represents the importance of the navigation light nodes for the safe identification of the actual runway path. In the simulated interference environment, the server maintains consistency between the navigation light nodes, light relationships, and control mappings with the target airport, ensuring that the path interference action combinations generated by the reinforcement learning model in the simulated interference environment correspond to executable navigation light node control commands in the target airport.

[0100] in, This represents a simulated interference environment, constructed by the server based on a light relationship diagram of the target airport. This diagram represents the lighting relationships and is used to provide information on navigation light nodes and their relationships. The actual runway path is represented by the corresponding nodes and edges of the runway centerline lights, runway edge lights, entrance lights, landing area lights, and stop row lights in the lighting relationship diagram; The taxiing path is represented by the graph nodes and edges corresponding to the taxiway centerline lights and taxiway side lights in the lighting relationship diagram; This represents the set of navigation light nodes, obtained from the graph nodes in the light relationship diagram; The set of lighting relationships is represented by the edges in the lighting relationship diagram. This represents the set of control loops, obtained from the basic information of each navigation light node; This represents the set of communication addresses, provided by the navigation light control system. This represents the set of safety constraint levels, determined by the target airport's safety rules and node attributes in the lighting relationship diagram. This formula incorporates the lighting structure, control structure, and safety constraints of a real airport into the same simulation object, enabling the simulated interference environment to reproduce the control and guidance relationships between navigation light nodes in a real airport.

[0101] When constructing erroneous guidance paths in a simulated interference environment based on their causes, the server generates water reflection paths, wet snow scattering paths, background light interference paths, intersecting light paths, and refraction offset paths. Water reflection paths are false paths generated when real light spots are mirrored in water-filled areas of the runway or taxiway; wet snow scattering paths are false paths generated when wet snow covers areas that cause navigational aid light spots to diffuse and form boundary-like structures; background light interference paths are false paths generated when non-navigational aid lights around the target airport form a continuous arrangement in the field of view; intersecting light paths are false paths generated when the arrangement of lights in taxiway intersection areas is mistaken for the runway's depth; and refraction offset paths are false paths generated when low-altitude refraction causes a shift in the position of the real light spot. The server generates different types of erroneous guidance paths by adjusting the interference intensity and combination method. The interference intensity represents the degree of influence of the interference source on the brightness, shape, position, and stability of the light spot, while the combination method represents the simultaneous or sequential action of multiple interference sources within the same time window.

[0102] in, Represents the first generated in the simulated interference environment Error boot path; This represents the error boot path number, and its value range is the set of all error boot path numbers generated in the simulated interference environment; The function representing the error guidance path generation function is determined by the cause of the error, the intensity of the interference, the combination method, the simulated interference environment, and the light relationship diagram. Indicates the first The causes of incorrect guidance paths include water reflection paths, wet snow scattering paths, background light interference paths, intersecting light paths, or refraction offset paths. Indicates the first The interference intensity of an incorrectly guided path is determined by the degree of reflection, the degree of scattering, the brightness of the background light, the significance of the intersecting lights, or the degree of refraction offset; Indicates the first The combination of error guidance paths is determined by whether multiple causes occur simultaneously, consecutively, or overlap. This represents a simulated interference environment; This diagram represents the lighting relationship and is used to constrain the structural similarity between erroneous guidance paths and actual runway paths. By changing the cause, intensity, and combination of these factors, the formula enables the simulated interference environment to generate various erroneous guidance paths that are similar to actual runway paths but originate from incorrect sources.

[0103] When generating simulated airport environment perception results based on simulated interference environments, the server simulates continuous image frames, low-altitude visibility, precipitation status, ground water accumulation status, wet snow coverage status, low-altitude refraction status, background light distribution, aircraft attitude changes, and navigation light node operation status within the simulated interference environment. This simulated data is then organized into simulated airport environment perception results according to time windows. The simulated airport environment perception results have the same data structure as the airport environment perception results during online operation, allowing the path recognition methods learned by the reinforcement learning model during offline training to be directly transferred to the online operation phase. The server then extracts a candidate light spot set based on the simulated airport environment perception results. This candidate light spot set includes real light spots and suspected interference light spots. Next, a suspected path set is constructed based on the spatial position, arrangement direction, brightness progression, left-right distribution, and duration of the suspected interference light spots. Finally, the suspected path set is compared with the light relationship diagram to generate a path judgment status.

[0104] in, Indicates time The simulated airport environment perception results are generated by the server in a simulated interference environment; This represents a simulated sensing function used to generate a data structure consistent with the actual data acquisition process. This represents a simulated interference environment; This represents the error boot path in a simulated interference environment; Indicates time The simulated aircraft attitude changes are generated by preset approach trajectories, taxiing trajectories, or abnormal yaw trajectories; Indicates time The simulated operational status of navigation light nodes is generated from the node attributes, control loop capabilities, and simulated fault states in the lighting relationship diagram. This formula generates a perception state consistent with the actual operational phase in a simulated interference environment, enabling the use of the same path decision logic in both offline training and online operation phases.

[0105] When training a reinforcement learning model based on path judgment states, the server uses these states as the model's state data and enables the model to select combinations of path interference actions from a set of path interference actions. Path judgment states include centerline continuity, boundary parallelism, entrance symmetry, depth extension, directional proximity, spatial proximity, salience, stability, cause of formation, and the current confirmed stability of the actual runway path. The set of path interference actions includes centerline disruption actions, boundary misalignment actions, entrance desymmetry actions, depth truncation actions, and actions that maintain the actual path. A combination of path interference actions refers to multiple path interference actions selected by the reinforcement learning model for the same erroneous path and their execution order. The server determines the navigation light nodes involved in control based on the lighting relationship diagram and generates action execution parameters for these nodes. These parameters include the direction of brightness change, the magnitude of brightness change, the order of brightness change, the duration of the action, and the action recovery method, enabling the path interference action combinations to be transformed from model output into executable control actions.

[0106] in, Indicates the simulation training phase for the first The path interference action combination generated by the incorrect guidance path; The policy function of a reinforcement learning model is represented by the model parameters. control; The policy parameters of the reinforcement learning model are updated jointly by the simulated training process and the constrained policy adjustment during the online operation phase. Indicates the first The simulated path judgment status corresponding to each erroneous guidance path is generated from the simulated airport environment perception results, candidate light spot set, suspected path set, and light relationship diagram. Indicates the first The set of path interference actions corresponding to each erroneous guidance path; This represents a light relationship diagram used to determine the navigation light nodes that can be affected by path interference actions and the actual runway path light relationships that must not be disrupted. The formula inputs the path judgment state, the set of path interference actions, and the light relationship diagram into a reinforcement learning model, enabling the model to learn to generate combinations of path interference actions under safety structure constraints.

[0107] When a path interference action combination is submitted for safety verification, the server first determines whether the combination disrupts the lighting relationships of the actual runway path. Safety verification involves performing centerline safety checks, boundary safety checks, entrance safety checks, depth safety checks, maintenance safety checks, and overall safety checks on the path interference action combination to determine whether it weakens the continuity of the centerline, the parallelism of the boundaries, the symmetry of the entrance, the progression of the grounding zone, the stopping and blocking relationship, or the path connection relationship of the actual runway path. If the path interference action combination fails the safety verification, the server restricts its execution and feeds the restriction result back to the reinforcement learning model; if the path interference action combination passes the safety verification, the server applies it to the simulated interference environment and modifies the visual structure of the erroneous path in the simulated interference environment based on the path interference action combination.

[0108] in, Indicates the simulation training phase for the first The security verification result of the error boot path, which can be either a passing security verification result or a failing security verification result; The safety verification function is determined by the combination of path interference actions, the light relationship diagram, and the path judgment state. This represents the combination of path interference actions generated during the simulation training phase; This diagram illustrates the lighting relationships and safety constraint levels along the actual runway path. This indicates the simulated path judgment state, used to determine the proximity and risk of misidentification between the current erroneous path and the actual runway path. This formula, by performing safety checks in advance during the simulation training phase, enables the reinforcement learning model to learn that it should not select path interference action combinations that disrupt the lighting relationships of the actual runway path.

[0109] After simulating a disruptive environment, the server re-collects the simulated airport environmental perception results and evaluates the effectiveness of the path interference action combination in eliminating misguided paths. The elimination effect is used to indicate whether the continuity of the centerline, the parallelism of the boundaries, the symmetry of the entrance, and the depth extension of the misguided path decrease, while also indicating whether the current confirmed stability of the real runway path remains stable or improves. If the misguided path is weakened and the current confirmed stability of the real runway path remains stable or improves, the server generates an effective action combination; if the misguided path is weakened but the current confirmed stability of the real runway path decreases, the server generates a risky action combination; if the misguided path is not weakened, the server generates an ineffective action combination. Through this evaluation method, the reinforcement learning model does not only learn how to change lighting, but also learns how to eliminate misguided paths without disrupting the real runway path.

[0110] in, Indicates the simulation training phase. The elimination effect score corresponding to each error guidance path is calculated by the server based on the changes in the path judgment state before and after execution. , , , and The evaluation weights, representing the continuity of the centerline, the parallelism of the boundaries, the symmetry of the entrance, the depth of the extension, and the current confirmed stability of the actual runway path, are pre-set by the safety requirements of the target airport. and These represent the order before and after the execution of the path interference action combination. The continuity of the centerline of the incorrect guidance path; and These represent the periods before and after execution, respectively. The degree of parallelism of the boundaries of the incorrect guidance paths; and These represent the first and second execution times, respectively. The degree of symmetry of the entry points of each error-guided path; and These represent the first and second execution times, respectively. The extent to which a wrong path extends; and These represent the current confirmed stability levels of the actual runway path before and after execution. This formula rewards the weakening of erroneous path structures while simultaneously rewarding the stability of the actual runway path, enabling effective action combinations to demonstrate a safety elimination effect.

[0111] When updating the reinforcement learning model based on effective action combinations, risky action combinations, and ineffective action combinations, the server writes the simulated path judgment state, path interference action combinations, safety verification results, simulated airport environment perception results after execution, and elimination effect evaluation results into the learning data, and adjusts the policy parameters of the reinforcement learning model accordingly. For effective action combinations, the server increases the probability of action selection in similar path judgment states; for risky action combinations, the server decreases the probability of action selection in scenarios where the current confirmed stability of the real runway path is low; for ineffective action combinations, the server decreases the probability of action selection in the corresponding cause and corresponding path judgment state. After completing multiple rounds of training, the server deploys the updated reinforcement learning model to the online runtime phase, enabling it to generate candidate path interference action combinations based on real airport environment perception results.

[0112] in, This represents the parameters of the reinforcement learning model after offline training is completed. The initial parameters of the reinforcement learning model are obtained from a pre-trained model or random initialization. This represents the offline training step size, which is set by the training stability requirements and the number of simulated samples. This indicates the direction of gradient adjustment relative to the model parameters, calculated during the reinforcement learning training process; This represents the value evaluation of the combination of path interference actions by the reinforcement learning model in the simulated path judgment state; The simulated feedback evaluation function is determined by the action combination type, elimination effect score, and safety verification result. This indicates the type of action combination during the simulation training phase, including effective action combinations, risky action combinations, or ineffective action combinations. Indicates the elimination effect score; This indicates the security verification result. The formula adjusts the reinforcement learning model parameters by simulating training samples, ensuring the model learns a safe and effective error-direction elimination strategy before deployment.

[0113] During the online operation phase, when identifying multiple misguided paths based on airport environmental perception results, the server follows the same identification logic as before, extracting candidate spot sets, constructing a set of suspected paths, identifying misguided paths, and generating path judgment statuses from the airport environmental perception results. When multiple misguided paths exist simultaneously, the server determines the processing priority based on salience, stability, directional proximity, spatial proximity, cause of formation, and the current confirmed stability of the real runway path. Processing priority indicates the order in which multiple misguided paths are processed; salience indicates the visibility intensity of the misguided path in the field of view; stability indicates the persistence of the misguided path in consecutive image frames; directional proximity indicates the proximity between the misguided path and the direction of the real runway path or the aircraft's observation direction; spatial proximity indicates the spatial proximity between the misguided path and the real runway path; cause of formation indicates the origin of the misguided path; and the current confirmed stability of the real runway path indicates whether the real runway path is easily and stably confirmed in the current environment.

[0114] in, Indicates the first The processing priority score for each error path is given, with higher scores indicating that the error should be processed first. Indicates the first The significance of each erroneous guidance path is calculated based on the average brightness, color similarity, and image area occupied by the suspected interfering light spots; Indicates the first The stability of a faulty path is calculated from its duration, positional fluctuations, and morphological fluctuations in consecutive image frames; Indicates the first The degree of similarity between the directions of the incorrect guidance paths is calculated by converting the angle between the main direction of the incorrect guidance path and the direction of the actual runway path or the direction observed by the aircraft. Indicates the first The spatial proximity of the erroneous guidance paths is calculated from the minimum distance and overlap between the erroneous guidance paths and the actual runway paths; Indicates the first The risk weights corresponding to the causes of each misguided path are pre-set by the target airport based on the misidentification risk of water reflection paths, wet snow scattering paths, background light interference paths, intersecting light paths, and refraction offset paths. This indicates the current confirmed stability level of the actual runway path; , , , , and These represent the priority weights of the corresponding factors, set by the target airport's operational phase and safety strategy. This formula determines the sequential elimination order of multiple misguided paths by comprehensively considering the misidentification intensity of the misguided path itself and the stability of the actual runway path.

[0115] When generating corresponding path interference action combinations based on processing priority, the server prioritizes the erroneous guidance path with the highest processing priority and inputs its path judgment status into the reinforcement learning model, which then generates the corresponding path interference action combination. This path interference action combination still needs to undergo security verification. Only after passing the security verification can it be converted into a navigation light node control command and applied to the target airport navigation light node. After execution, the server re-collects the airport environmental perception results to determine whether the current erroneous guidance path has been eliminated and whether other erroneous guidance paths have become new priority targets due to changes in lighting, environment, or visual structure. In this way, multiple erroneous guidance paths are not arbitrarily controlled simultaneously, but are eliminated one by one or in groups according to processing priority, avoiding the superposition of multiple path interference action combinations that would disrupt the actual runway path lighting relationship.

[0116] in, This indicates the error path number that is prioritized for handling during the current online operation phase; This indicates that the error guidance path with the highest processing priority score has been selected; Indicates time The identified set of error redirection paths; Indicates the first The processing priority score for each misguided path is determined by this formula. This formula prioritizes the misguided paths with the highest processing priority, ensuring that the reinforcement learning model processes the most likely misguided paths to cause false positives.

[0117] During the online operation phase, when the reinforcement learning model adjusts action selection probabilities, navigation light node participation weights, action execution parameters, and cause identification weights based on security checks, the server only allows the reinforcement learning model to adjust its strategy within the range of path interference action combinations that have passed security checks. Action selection probability refers to the probability that the reinforcement learning model will select a certain type of path interference action given the same path judgment state; navigation light node participation weight refers to the priority of a navigation light node participating in subsequent path interference action combinations; action execution parameters refer to the brightness change direction, brightness change amplitude, brightness change order, action duration, and action recovery method corresponding to the path interference action combination; cause identification weight refers to the weight the server uses when determining whether an incorrectly guided path belongs to a water reflection path, wet snow scattering path, background light interference path, intersecting light path, or refraction offset path. If a path interference action combination passes security checks and is evaluated as a valid action combination, the server allows the reinforcement learning model to increase the relevant action selection probability and navigation light node participation weight; if a path interference action combination fails security checks or is evaluated as a risky action combination, the server restricts the reinforcement learning model from further enhancing similar strategies and reduces the relevant action selection probability, action execution parameter range, or cause identification weight.

[0118] in, This represents the updated reinforcement learning model parameters during the online execution phase; This represents the parameters of the reinforcement learning model before it is deployed to the online running stage; This indicates that the online strategy adjustment step size is set by the security constraint strength during the online operation phase and is smaller than the offline training step size; the results are verified through security checks. This represents the security check constraint factor. It takes a value of one when the path interference action combination passes the security check, and a value of zero when it fails the security check. This indicates the direction of gradient adjustment relative to the parameters of the reinforcement learning model; This represents the value evaluation of the reinforcement learning model for the combination of safe path interference actions in the online path judgment state; Indicates the online operation phase. The path determination status of the incorrect guidance path; This indicates a combination of path interference actions that have passed security checks; The online feedback evaluation is jointly determined by the elimination effect assessment results, execution feedback, and the current confirmed stability of the actual runway path. This formula controls the online policy adjustment through a safety check constraint factor, ensuring that path interference action combinations that fail the safety check are not absorbed as positive policies by the reinforcement learning model. This guarantees that the lighting relationship of the actual runway path remains intact during the continuous elimination of multiple misguided paths.

[0119] This application also provides an optimal control device for navigation lights based on reinforcement learning, referring to... Figure 2 , Figure 2 This application provides a schematic diagram of a reinforcement learning-based optimal control device for navigation lights. The device is a server, comprising an acquisition module 21 and a processing module 22. The acquisition module 21 acquires a light relationship diagram of navigation lights within the target airport, describing the light relationships between navigation light nodes on the actual runway path. The processing module 22 collects airport environment perception results from the target airport and identifies erroneous guidance paths based on these results. It also compares the erroneous guidance paths with the light relationship diagram to generate a path judgment state describing the completeness of the erroneous guidance paths. Furthermore, the processing module 22 constructs a set of path interference actions based on the path judgment state and generates a combination of path interference actions from the set based on the path judgment state using a reinforcement learning model. This combination of path interference actions includes centerline disruption actions and boundary... The processing module 22 is used to perform safety checks on combinations of path interference actions, and restrict the execution of path interference action combinations that disrupt the real runway path lighting relationship based on the safety check results. At the same time, it converts the path interference action combinations that pass the safety check into navigation light node control commands. The processing module 22 is also used to change the visual structure of the erroneous guidance path according to the navigation light node control commands, and re-collect the airport environmental perception results after executing the navigation light node control commands to evaluate the elimination effect of the path interference action combinations on the erroneous guidance path. The processing module 22 is also used to construct a simulated interference environment to train a reinforcement learning model to identify and eliminate erroneous guidance paths based on the simulated interference environment, and adjust the strategy of the reinforcement learning model according to the safety check restrictions during the online operation phase to achieve continuous elimination of multiple erroneous guidance paths.

[0120] This application also provides an electronic device, with reference to... Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: at least one processor 31, at least one network interface 34, a user interface 33, a memory 35, and at least one communication bus 32.

[0121] The communication bus 32 is used to enable communication between these components.

[0122] The user interface 33 may include a display screen and a camera. Optionally, the user interface 33 may also include a standard wired interface and a wireless interface.

[0123] The network interface 34 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0124] The processor 31 may include one or more processing cores. The processor 31 connects to various parts of the server via various interfaces and lines, executing instructions, programs, code sets, or instruction sets stored in the memory 35, and calling data stored in the memory 35 to perform various server functions and process data. Optionally, the processor 31 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 31 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content to be displayed on the screen; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 31 and may be implemented as a separate chip.

[0125] The memory 35 may include random access memory (RAM) or read-only memory. Optionally, the memory 35 may include a non-transitory computer-readable storage medium. The memory 35 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 35 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 35 may also be at least one storage device located remotely from the aforementioned processor 31. Figure 3 As shown, the memory 35, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program for an optimal control method of navigation lights based on reinforcement learning.

[0126] exist Figure 3 In the electronic device shown, the user interface 33 is mainly used to provide an input interface for the user and to obtain the user input data; while the processor 31 can be used to call an application program stored in the memory 35 that is a reinforcement learning-based optimal control method for navigation lights. When executed by one or more processors, the electronic device executes one or more methods as described in the above embodiments.

[0127] This application also provides a non-transitory computer-readable storage medium storing instructions. When executed by one or more processors, these instructions cause an electronic device to perform one or more of the methods described in the above embodiments.

[0128] The foregoing description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Those skilled in the art will readily conceive of other embodiments of this disclosure upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The specification and embodiments are considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

Claims

1. A method for optimal control of navigation lights based on reinforcement learning, characterized in that, The method includes: Obtain a light relationship diagram of the navigation lights within the target airport. The light relationship diagram is used to describe the light relationship between each navigation light node in the actual runway path. The airport environment perception results of the target airport are collected, and the wrong guidance path is identified based on the airport environment perception results. At the same time, the wrong guidance path is compared with the light relationship diagram to generate a path judgment state to describe the completeness of the wrong guidance path. Based on the path judgment state, a set of path interference actions is constructed, and a combination of path interference actions is generated from the set of path interference actions based on the path judgment state using a reinforcement learning model. The combination of path interference actions includes centerline destruction actions, boundary misalignment actions, entrance desymmetry actions, depth truncation actions, and true path preservation actions. The path interference action combination is subjected to a safety check, and the execution of path interference action combinations that disrupt the actual runway path lighting relationship is restricted based on the safety check result. At the same time, the path interference action combination that passes the safety check is converted into navigation light node control commands. The visual structure of the misguided path is changed according to the navigation light node control command, and the airport environment perception results are re-acquired after the navigation light node control command is executed, so as to evaluate the effect of the path interference action combination on the elimination of the misguided path; A simulated interference environment is constructed to train the reinforcement learning model to identify and eliminate the misguided paths. During the online operation phase, the reinforcement learning model is adjusted according to security checks to achieve the continuous elimination of multiple misguided paths.

2. The optimal control method for navigation lights based on reinforcement learning according to claim 1, characterized in that, The acquisition of the lighting relationship diagram of navigation lights within the target airport specifically includes: Obtain basic information about each navigation light node within the target airport, including installation location, light type, light color, illumination direction, rated brightness, light group, power supply circuit, control circuit, communication address, and operating status. Based on the aforementioned basic information, spatial registration is performed on each of the navigation light nodes to determine the relative positional relationship between each of the navigation light nodes and the actual runway path; The lighting relationships are established based on the light group and spatial registration results of each navigation light node. The lighting relationships include centerline continuity, boundary parallelism, entrance symmetry, grounding area progression, stop and blockage, and path connection. The lighting relationships are semantically annotated to determine the actual runway path guidance meaning corresponding to each lighting relationship; The light relationship graph is generated by treating each of the navigation light nodes as graph nodes and each of the light relationships as graph edges.

3. The optimal control method for navigation lights based on reinforcement learning according to claim 1, characterized in that, The process involves collecting airport environment perception results of the target airport, identifying erroneous guidance paths based on these results, and comparing these erroneous guidance paths with the light relationship diagram to generate a path judgment state describing the completeness of the erroneous guidance paths. Specifically, this includes: The system collects continuous image frames, low-altitude visibility, precipitation status, ground water accumulation status, wet snow coverage status, low-altitude refraction status, background light distribution, aircraft attitude changes, and navigation light node operation status of the target airport. Based on these data, the system generates the airport environmental perception results. The airport environment perception results are mapped to the spatial coordinates corresponding to the light relationship diagram, and a candidate light spot set is extracted based on the mapping results. The candidate light spot set includes real light spots and suspected interference light spots. A set of suspected paths is constructed based on the set of candidate light spots, and the set of suspected paths is structurally compared with the light relationship diagram to identify erroneous guiding paths with features such as continuous centerline, parallel boundary, symmetrical entrance, or extended depth. The cause of the error guidance path is determined based on the correspondence between the error guidance path and the airport environmental perception result. The cause of the error guidance path includes water reflection path, wet snow scattering path, background light interference path, intersecting light path, and refraction offset path. The path judgment state is generated based on the erroneous guidance path and the light relationship diagram. The path judgment state includes the continuity of the centerline, the parallelism of the boundary, the symmetry of the entrance, the depth extension, the directional proximity, the spatial proximity, the significance, the stability, the cause of formation, and the current confirmed stability of the actual runway path.

4. The optimal control method for navigation lights based on reinforcement learning according to claim 3, characterized in that, The step of constructing a set of path interference actions based on the path judgment state, and generating a combination of path interference actions from the set of path interference actions based on the path judgment state using a reinforcement learning model, specifically includes: Based on the continuity of the centerline, a centerline disruption action is constructed; based on the parallelism of the boundary, a boundary misalignment action is constructed; based on the symmetry of the entrance, an entrance desymmetry action is constructed; based on the depth extension, a depth truncation action is constructed; and based on the current confirmed stability of the actual runway path, an actual path maintenance action is constructed, so as to generate the set of path interference actions. Each path interference action in the path interference action set is labeled with applicable conditions, wherein the applicable conditions include the type of erroneous guidance path, the range of path judgment status, the navigation light nodes that are allowed to participate, the navigation light nodes that are prohibited from participating, the allowed brightness change method, the allowed duration, the allowed recovery method, and the action combination restrictions. The path judgment state is input into the reinforcement learning model, and a corresponding path interference action is selected from the path interference action set according to the path judgment state to generate the path interference action combination; Based on the causes of the interference, the main path interference action and the auxiliary path interference action in the path interference action combination are determined, and the navigation light nodes involved in the control and the corresponding action execution parameters are determined based on the light relationship diagram. The action execution parameters include the direction of brightness change, the amplitude of brightness change, the order of brightness change, the duration of the action, and the method of action recovery.

5. The optimal control method for navigation lights based on reinforcement learning according to claim 2, characterized in that, The process involves performing a safety check on the path interference action combination, and restricting the execution of path interference action combinations that disrupt the actual runway path lighting relationship based on the safety check result. Simultaneously, the path interference action combinations that pass the safety check are converted into navigation light node control commands, specifically including: The path interference action combination is mapped to the navigation light nodes in the light relationship diagram to determine the node position, light group, light relationship, control loop, communication address and safety constraint level of the navigation light node participating in the control. Based on the node position and the safety constraint level, the navigation light node is divided into safety maintenance node and interference control node. Perform centerline safety checks on the centerline disruption actions, perform boundary safety checks on the boundary misalignment actions, perform entrance safety checks on the entrance desymmetry actions, perform depth safety checks on the depth truncation actions, and perform maintenance safety checks on the real path maintenance actions to determine whether each path interference action disrupts the real runway path lighting relationship. An overall safety check is performed on the combination of path interference actions to determine whether the superposition of multiple path interference actions leads to directional conflict between the real runway path and the taxiway path, weakening of the stop and blockade relationship, simultaneous dimming of key navigation light nodes, or abnormal flashing of adjacent navigation light nodes. When generating a security check failure result, the corresponding path interference actions are restricted according to the reason for failure, and the restricted path interference action combination is resubmitted to the security check. When generating a security verification result, the control loop, communication address, current brightness, allowed brightness change method, response delay and security constraint level of the navigation light node involved in the control are read according to the light relationship diagram, and the corresponding navigation light node control command is generated according to the path interference action combination. The control commands for the navigation light node are adjusted according to the control requirements corresponding to the safety maintenance node and the interference control node.

6. The optimal control method for navigation lights based on reinforcement learning according to claim 1, characterized in that, The process of altering the visual structure of the misguided path according to the navigation light node control command, and re-acquiring airport environmental perception results after executing the navigation light node control command, to evaluate the effect of the path interference action combination on eliminating the misguided path, specifically includes: The navigation light node, brightness change direction, brightness change amplitude, brightness change sequence, action duration, and action recovery method in the navigation light node control command are determined, and the navigation light node control command is matched with the centerline destruction action, boundary misalignment action, entrance desymmetry action, depth truncation action, and true path preservation action in the path interference action combination. Based on the safety verification results, distinguish between safety maintenance nodes and interference control nodes, control the safety maintenance nodes to maintain the actual runway path lighting relationship, and control the interference control nodes to execute corresponding path interference actions to change the visual structure of the erroneous guidance path; During the execution of the navigation light node control command, execution feedback is continuously collected, including actual brightness, response delay, control loop status, communication status and operating status. The participation weight of the corresponding navigation light node in the subsequent path interference action combination is adjusted according to the execution feedback. After completing the control command for the navigation light node, the airport environment perception results are re-acquired, and the candidate light spot set and the suspected path set are reconstructed based on the re-acquired airport environment perception results. Then, the change status of the erroneous guidance path after executing the control command for the navigation light node is determined based on the light relationship diagram. Based on the changes in the continuity of the centerline, the parallelism of the boundaries, the symmetry of the entrance, and the depth of the erroneous guidance path, and in conjunction with the current confirmed stability of the actual runway path, the elimination effect of the path interference action combination is evaluated to generate effective action combinations, risky action combinations, or ineffective action combinations. The effective action combination, the risky action combination, and the invalid action combination, along with the corresponding path judgment status, safety verification results, navigation light node control commands, execution feedback, and the re-collected airport environmental perception results, are written into the learning data of the reinforcement learning model, so as to adjust the generation results of subsequent path interference action combinations based on the learning data.

7. The optimal control method for navigation lights based on reinforcement learning according to claim 1, characterized in that, The construction of a simulated interference environment, used to train the reinforcement learning model to identify and eliminate the misguided paths, and the adjustment of the reinforcement learning model's strategy based on security checks during online operation to achieve continuous elimination of multiple misguided paths, specifically includes: A simulated interference environment is constructed based on the aforementioned light relationship diagram, and the actual runway path, taxiway path, navigation light nodes, light relationships, control loops, communication addresses, and safety constraint levels are written into the simulated interference environment. Based on the causes, water reflection paths, wet snow scattering paths, background light interference paths, cross light paths, and refraction offset paths are constructed in the simulated interference environment, and different types of error guidance paths are generated by adjusting the interference intensity and combination method. Based on the simulated interference environment, a simulated airport environment perception result is generated, and based on the simulated airport environment perception result, a candidate spot set is extracted, a suspected path set is constructed, and a path judgment state is generated; The reinforcement learning model is trained based on the path judgment state to generate a combination of path interference actions from the set of path interference actions, and the navigation light nodes involved in the control and the corresponding action execution parameters are determined based on the light relationship diagram. The path interference action combination is sent to the security verification, and after passing the security verification, it is applied to the simulated interference environment. Then, the effect of the path interference action combination on eliminating the erroneous guidance path is evaluated based on the re-collected simulated airport environment perception results, so as to generate effective action combinations, risky action combinations and ineffective action combinations. The reinforcement learning model is updated based on the effective action combination, the risky action combination, and the ineffective action combination, and the updated reinforcement learning model is deployed to the online running phase. During the online operation phase, multiple erroneous guidance paths are identified based on airport environmental perception results, and processing priorities are determined based on significance, stability, directional proximity, spatial proximity, cause of formation, and the current confirmed stability of the actual runway path. Based on the processing priority, a corresponding combination of path interference actions is generated, and the corresponding combination of path interference actions is executed after passing the security check, so as to achieve the continuous elimination of multiple erroneous guidance paths; During the online operation phase, the reinforcement learning model is adjusted based on the security verification restrictions, including the action selection probability, the weight of the navigation light node participation, the action execution parameters, and the weight of the cause identification.

8. A navigation light optimal control device based on reinforcement learning, characterized in that, The apparatus is used to execute the reinforcement learning-based optimal control method for navigation lights as described in any one of claims 1 to 7, the apparatus comprising an acquisition module and a processing module, wherein... The acquisition module is used to acquire a light relationship diagram of navigation lights within the target airport. The light relationship diagram is used to describe the light relationship between each navigation light node in the actual runway path. The processing module is used to collect the airport environment perception results of the target airport, identify the wrong guidance path based on the airport environment perception results, and compare the wrong guidance path with the light relationship diagram to generate a path judgment state to describe the completeness of the wrong guidance path. The processing module is further configured to construct a set of path interference actions based on the path judgment state, and generate a combination of path interference actions from the set of path interference actions based on the path judgment state using a reinforcement learning model. The combination of path interference actions includes centerline destruction actions, boundary misalignment actions, entrance desymmetry actions, depth truncation actions, and true path preservation actions. The processing module is also used to perform safety verification on the path interference action combination, and restrict the execution of path interference action combinations that disrupt the real runway path light relationship based on the safety verification result, while converting the path interference action combination that passes the safety verification into navigation light node control commands. The processing module is also used to change the visual structure of the misguided path according to the navigation light node control command, and to re-acquire the airport environment perception results after executing the navigation light node control command, so as to evaluate the effect of the path interference action combination on the elimination of the misguided path. The processing module is further configured to construct a simulated interference environment, train the reinforcement learning model to identify and eliminate the misguided paths based on the simulated interference environment, and adjust the strategy of the reinforcement learning model according to security verification during the online operation phase, so as to achieve the continuous elimination of multiple misguided paths.

9. An electronic device, characterized in that, The electronic device includes a processor and a memory; the memory stores a computer program, wherein the computer program, when executed by the processor, implements the reinforcement learning-based optimal control method for navigation lights as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores instructions that, when executed, perform the reinforcement learning-based optimal control method for navigation lights as described in any one of claims 1 to 7.

Images