Low-altitude unmanned aerial vehicle dynamic landing point selection method and system

Abstract

The application discloses a low-altitude unmanned aerial vehicle dynamic landing point selection method and system, relates to the unmanned aerial vehicle data processing technical field, and comprises the following steps: acquiring candidate landing points, unmanned aerial vehicle states and environment data, determining basic features, emergency landing point attributes and basic risk values; determining a current emergency level, generating a dynamic evaluation weight under a non-emergency level and calculating a comprehensive priority; verifying reachability according to energy consumption, flight time, residual energy and residual flight time, grading and screening reachable candidate results, forming an emergency level distance path result, an automatic candidate result or a manual selection result; uniformly collecting all path results to determine a selected landing point and a formed path, executing target maintenance or update determination according to a current target state, and executing hysteresis type change, maneuvering feasibility and endurance margin verification when applicable, determining a final target landing point and outputting navigation guidance information. The application makes the landing point evaluation adapt to state changes and improves the executability of candidate results.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) data processing technology, specifically to a method and system for dynamic landing point selection for low-altitude UAVs. Background Technology

[0002] As the application scenarios for low-altitude drones continue to increase, drones may be affected by changes in energy status, environmental conditions, and mission status during mission execution, requiring the selection of suitable landing sites based on the current conditions. Existing methods for selecting landing sites for low-altitude drones typically determine candidate landing sites based on preset rules, distance information, or fixed evaluation methods.

[0003] However, during actual flight, the status of the UAV and environmental conditions may change dynamically, and relying on a single evaluation factor is insufficient to comprehensively reflect the availability, risk, and actual reachability of candidate landing sites. Furthermore, some existing methods do not adequately consider energy constraints, time constraints, and processing requirements under different emergency conditions, which may lead to selection results that do not meet actual flight conditions.

[0004] Therefore, how to dynamically evaluate and determine the reachability of candidate landing sites based on the real-time status of the UAV and the environmental conditions, and how to form a corresponding target landing site selection and update process under different emergency conditions, is a technical problem that needs to be solved in this field. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for dynamic landing point selection of low-altitude unmanned aerial vehicles (UAVs) to solve the technical problem mentioned in the background art of difficulty in dynamically evaluating, determining reachability, and updating targets by combining the real-time status of the UAV, the environmental status, and the comprehensive conditions of candidate landing points.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for dynamic landing point selection of a low-altitude unmanned aerial vehicle, comprising the following steps:

[0008] Acquire candidate landing point data, UAV real-time status data, and environmental data for the current decision-making cycle, and acquire the latest status data of the current target landing point if a current target landing point exists.

[0009] Based on the candidate landing point data, the UAV real-time status data, and the environmental data, the basic characteristic data, emergency landing point attributes, and basic risk values ​​of each candidate landing point are determined.

[0010] The current emergency level is determined based on the real-time status data of the UAV and the environmental data, and the corresponding candidate landing point processing strategy is determined based on the current emergency level; when the current emergency level is non-urgent, the dynamic evaluation weight of the current decision cycle is generated, and the comprehensive priority of each candidate landing point is determined based on the basic feature data and the dynamic evaluation weight.

[0011] Based on the energy consumption and flight time required for the UAV to reach each candidate landing point, as well as the UAV's remaining energy and remaining flight time, the accessibility of each candidate landing point is verified to form an accessible candidate result.

[0012] The reachable candidate results are classified and screened according to the current emergency level, and the candidate landing point processing strategy is used to form emergency-level distance path results, automatic candidate results, or manually selected results.

[0013] The results of the emergency-level distance path, the automatic candidate results, or the manual selection results are uniformly summarized, and the proposed landing point and the formation path corresponding to the proposed landing point are determined based on the unified summary results.

[0014] Based on the formation path corresponding to the proposed landing point and the status of the current target landing point, a target retention or target update determination is performed, and a hysteresis change determination, maneuver feasibility verification, and endurance margin verification are performed where applicable, in order to determine the final target landing point or the status of the current decision cycle in which no valid final target landing point has been formed.

[0015] When determining the final target landing point, target landing navigation guidance information is generated and output; while the dynamic landing point selection function remains enabled, the next decision cycle is initiated.

[0016] Furthermore, the basic feature data includes feature data that respectively characterize the basic qualifications, distance, runway length, wind direction matching, availability, occupancy status, and data confidence of the candidate landing point; the basic risk value is determined based on the availability, occupancy status, wind direction matching, runway length, and data confidence of the candidate landing point, so that the basic conditions and risk factors of the candidate landing point can participate in the subsequent evaluation.

[0017] Furthermore, the current emergency level includes normal, early warning, emergency, and extremely urgent levels; when the current emergency level is normal, early warning, or emergency, the dynamic evaluation weight is generated based on the baseline weight of the corresponding emergency level and the UAV status and environmental status of the current decision cycle; when the current emergency level is extremely urgent, an extremely urgent fast path identifier is formed and the determination of the comprehensive priority is skipped, so that different emergency levels adopt corresponding candidate landing point processing paths.

[0018] Furthermore, the reachability verification includes performing hard energy consumption and hard time constraints verification on the total energy consumption required for the UAV to reach the candidate landing point and the estimated total flight time, respectively. Only candidate landing points whose reachability calculation is valid and which pass both the hard energy consumption and hard time constraints verification are determined as the reachable candidate results, so as to exclude candidate landing points that do not meet the energy consumption or time constraints.

[0019] Furthermore, when the current emergency level is extremely urgent, the candidate landing points selected through hierarchical screening are used to form the extremely urgent distance path result; when the current emergency level is normal, early warning, or emergency, a candidate solution set is formed with arrival time, flight energy consumption, and landing risk as multiple optimization objectives, and the automatic candidate result is formed based on the multi-objective evaluation results of the candidate solution set in the automatic decision-making mode, or the manual selection result is formed based on the manual selection received for the candidate solution set in the manual decision-making mode, so as to adapt to different emergency levels and decision-making modes.

[0020] Furthermore, the unified aggregation includes uniformly receiving the urgent distance path results, the automatic candidate results, or the manually selected results. When a unique valid result is received, the candidate landing point corresponding to the unique valid result is determined as the proposed landing point, while retaining the formation path of the proposed landing point. The proposed landing point does not directly replace the current target landing point. Instead, based on the formation path and the status of the current target landing point, the final target landing point is determined after target retention or target update determination, maneuver feasibility verification, and endurance margin verification, so that candidate results formed by different paths enter a unified target determination process.

[0021] Furthermore, when the formation path of the proposed landing point is an automatically decided path and the current target landing point is valid, a hysteresis change decision is performed based on the combined priority of the proposed landing point and the current target landing point; when the formation path of the proposed landing point is an urgent distance path or a manually decided path, or when there is no valid current target landing point, a bypass access decision is performed; for proposed landing points that have been admitted through the hysteresis change decision or the bypass access decision and are different from the current target landing point, horizontal maneuver feasibility verification, vertical maneuver feasibility verification, and endurance margin verification are performed, and when all the verifications pass, the proposed landing point is determined as the final target landing point, so as to reduce the situation where the current target landing point is directly updated simply because of changes in candidate results.

[0022] Secondly, based on the same inventive concept, the present invention also provides a dynamic landing point selection system, including a data acquisition module, a feature determination module, an emergency classification and dynamic weight generation module, a comprehensive evaluation and accessibility verification module, a classification screening module, a multi-target candidate result formation module, a proposed landing point aggregation module, a target switching determination module, and an output and closed-loop module; each module is respectively used to perform the aforementioned data acquisition, basic feature and basic risk determination, emergency classification and dynamic weight generation, comprehensive evaluation and accessibility verification, classification screening, multi-target candidate result formation, proposed landing point aggregation, target maintenance or target update determination, target landing navigation guidance information output, and next decision cycle initiation process.

[0023] Thirdly, the present invention also provides an electronic device, including a processor and a memory, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the low-altitude unmanned aerial vehicle dynamic landing point selection method described in the first aspect.

[0024] Fourthly, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the low-altitude unmanned aerial vehicle dynamic landing point selection method described in the first aspect.

[0025] Compared with the prior art, the beneficial effects of the present invention are:

[0026] This invention acquires real-time status data, environmental data, and candidate landing point data of UAVs, performs dynamic evaluation, accessibility verification, and multi-path result aggregation on candidate landing points, and determines the final target landing point by combining the target switching judgment process, thereby forming a landing point selection and processing flow for dynamic environments.

[0027] This invention determines the emergency level based on real-time UAV status data and environmental data, and generates dynamic evaluation weights under non-emergency levels. This allows the evaluation strategy for candidate landing sites to be adjusted according to the current emergency level, UAV status, and environmental status, thereby improving the problem that fixed evaluation methods are insufficient to adapt to dynamic status changes.

[0028] This invention performs reachability verification based on the energy consumption and flight time required for the UAV to reach the candidate landing point, as well as the UAV's remaining energy and remaining flight time. Only candidate landing points that pass both the energy consumption hard constraint verification and the time hard constraint verification are retained as reachable candidates, thereby reducing the possibility of candidate landing points that do not meet the actual energy consumption or time conditions entering the subsequent selection process.

[0029] This invention unifies and summarizes the distance path results for emergency response, automatic candidate results, and manual selection results. It determines the proposed landing point and retains its formation path only when a unique valid result is obtained, so that candidate results formed under different emergency levels and decision-making modes can enter the subsequent target determination process according to a unified data boundary.

[0030] This invention performs target hold or target update determination based on the formation path of the proposed landing point and the status of the current target landing point, and performs hysteresis change determination, maneuver feasibility verification and endurance margin verification when applicable; when the dynamic landing point selection function is enabled, the next decision cycle is started, thereby reducing the situation where the proposed landing point directly replaces the current target landing point, and enabling the target landing point selection to continue with the periodic data update. Attached Figure Description

[0031] Figure 1 This is a flowchart of the low-altitude unmanned aerial vehicle (UAV) dynamic landing point selection method of the present invention;

[0032] Figure 2 This is a diagram illustrating the functions and data interaction of the dynamic landing point selection system of this invention. Detailed Implementation

[0033] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0034] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this invention should have the ordinary meaning understood by those skilled in the art. The terms "first," "second," and similar terms used in the embodiments of this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word covers the element or object listed after the word and its equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are only used to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0035] like Figure 1 and Figure 2 As shown:

[0036] First embodiment:

[0037] This embodiment provides a dynamic landing point selection method for low-altitude unmanned aerial vehicles (UAVs). This method is applicable to UAVs during flight, dynamically determining a target landing point from multiple candidate landing points based on the UAV's own status, the external environment status, and the real-time status of multiple candidate landing points. In the following embodiment, an industrial-grade fixed-wing UAV or compound-wing UAV with runway takeoff and landing capabilities is used as an example. This UAV has a flight control system, a navigation module, a positioning module, a status monitoring module, and a communication interface for data interaction with the landing point selection system.

[0038] The landing point selection system is configured to acquire candidate landing point data, real-time UAV status data, and environmental data. Based on the acquired data, it performs candidate landing point evaluation, reachability verification, emergency screening, multi-target selection, and target landing point switching judgment. The landing point selection system can be installed in the UAV's onboard computing equipment and communicate with the flight control system and navigation module to receive UAV status data and output target landing point information and navigation guidance information.

[0039] In this embodiment, the landing site selection method is repeatedly executed according to a decision cycle. Each decision cycle is recalculated based on the candidate landing site data, UAV real-time status data, and environmental data acquired at the current moment. As the UAV's position, remaining range, engine status, weather conditions, and the openness of candidate landing sites change, the evaluation results of candidate landing sites obtained in each decision cycle can change accordingly.

[0040] In one specific implementation, the dynamic weight model performs online inference according to a preset update interval, for example, generating a weight adjustment every 10 seconds based on the updated UAV state and environmental state. This update interval describes a specific update method for the dynamic weights and does not imply that the execution cycle of the entire landing point selection process must be fixed at 10 seconds. The overall execution cycle of the landing point selection method can be set according to the UAV's flight stage, data update frequency, and the computing power of the processing device.

[0041] To ensure clarity regarding the objects and results in subsequent processing, this implementation method distinguishes the relevant terms as follows.

[0042] "Candidate landing sites" refer to potential landing sites for drones, obtained at the start of the current decision-making cycle based on ground station data, airborne identification results, shared data from other drones, or pre-stored landing site data. Candidate landing sites can include regular airports, standard take-off and landing sites, temporary take-off and landing sites, field emergency landing sites, and emergency landing sites with emergency support capabilities.

[0043] The "reachability candidate set" refers to the set of candidate landing points that meet the reachability conditions after undergoing availability, energy consumption, and flight time verification. A candidate landing point has a higher overall priority, but this does not mean that the candidate landing point will definitely be included in the reachability candidate set; if the energy consumption or flight time required for the UAV to reach the candidate landing point exceeds the corresponding safety limit, the candidate landing point is excluded.

[0044] The "candidate set for optimization" refers to the set of candidate landing sites formed after further processing of the reachable candidate set according to the current emergency level using corresponding screening rules. Different emergency levels may correspond to different runway lengths, wind direction deviations, distance ranges, and emergency landing site attribute requirements.

[0045] The "Pareto candidate solution set" refers to the set of non-dominated candidate landing points obtained by non-dominated sorting of candidate landing points in the candidate set to be optimized, with arrival time, flight energy consumption and landing risk as multiple optimization objectives.

[0046] "Proposed landing point" refers to the recommended landing point obtained in the current decision-making cycle through multi-objective optimization, automatic decision-making, manual selection, or simplified ranking methods in urgent situations. The proposed landing point only represents the recommendation result obtained in this decision-making cycle and has not directly replaced the landing point currently being used by the navigation module.

[0047] "Current target landing point" refers to the landing point that has been determined before the start of the current decision cycle and is currently used by the navigation module for trajectory planning and flight guidance. When this method is executed for the first time and no current target landing point exists, the proposed landing point obtained in this round can be used as the initial target landing point to be verified.

[0048] "Final target landing point" refers to the landing point that is maintained or updated after a hysteresis-based change judgment between the proposed landing point and the current target landing point, and after verifying the maneuverability of the turns, altitude changes, and range margins required for target switching. The final target landing point is output to the navigation module for subsequent trajectory planning and flight guidance.

[0049] This implementation also involves two different evaluation metrics: overall priority and overall relevance.

[0050] The overall priority is determined by multiple characteristic values ​​of candidate landing points and dynamic weights for the current decision cycle, reflecting the overall priority of candidate landing points under the current UAV and environmental conditions. In addition to evaluating candidate landing points, the overall priority is also used to compare the priority difference between the proposed landing point and the current target landing point during subsequent target switching.

[0051] The comprehensive proximity score is used in automated decision-making mode to determine the proposed landing point from the Pareto candidate solution set. The comprehensive proximity score is determined based on the candidate landing point's performance on three objectives: arrival time, flight energy consumption, and landing risk, as well as its distance relative to the positive and negative ideal solutions. The comprehensive proximity score is only used for further selection within the Pareto candidate solution set and does not replace the role of comprehensive priority in hysteresis-based change decision-making.

[0052] S100: Obtain landing point decision data for the current decision cycle.

[0053] At the start of each decision cycle, the landing point selection system acquires candidate landing point data, UAV real-time status data, and environmental data corresponding to the current moment. If the navigation module already has a target landing point in progress, the system reads the identifier of the current target landing point from the navigation module and matches the latest status data of the current target landing point from the candidate landing point data of the current decision cycle based on the identifier. All of the above data together constitute the landing point decision data for the current decision cycle.

[0054] The candidate landing site data describes the various potential landing locations and their statuses that can be acquired by the landing site selection system. The candidate landing sites obtained in the current decision-making cycle are grouped into a candidate landing site set, and the [number of] candidate landing sites is selected from this set. The candidate landing sites are denoted as candidate landing sites. .

[0055] Candidate landing sites can include regular airports, standard takeoff and landing sites, temporary takeoff and landing sites, field emergency landing sites, and emergency landing sites with emergency support capabilities. Regarding candidate landing sites... The acquired candidate landing site data includes at least:

[0056] Candidate landing site identifier, candidate landing site type, longitude, latitude and altitude of candidate landing site, runway length, runway orientation;

[0057] Current open status; current occupied status; estimated release time when a candidate landing site is occupied;

[0058] Whether the candidate landing site is an emergency landing site, the source of the candidate landing site data, and the update time of the candidate landing site data.

[0059] Candidate landing point identifiers are used to distinguish and track candidate landing points across different decision-making cycles. Candidate landing point type is used as the base qualification score for subsequent candidate landing point determination; the longitude, latitude, and altitude of the candidate landing point are used to calculate the three-dimensional spatial distance from the UAV's current position to the candidate landing point; runway length is used to determine the runway length factor and whether the runway meets the UAV's minimum takeoff and landing requirements; runway orientation is used to determine the wind direction matching factor in conjunction with the current wind direction.

[0060] The current open status indicates whether the candidate landing site is currently permitted for drone landing. The current occupancy status and estimated release time are used to determine the occupancy penalty factor for the candidate landing site. Whether the candidate landing site is an emergency landing site is used to determine its emergency type attribute. The source and update time of the candidate landing site data are used to determine the data confidence factor.

[0061] In this embodiment, candidate landing point data can come from official ground stations, airborne visual recognition modules, or data shared by other UAVs through inter-UAV communication.

[0062] The official ground station can send basic attributes and real-time status information of official airports, standard take-off and landing sites, and registered temporary take-off and landing sites to drones. The onboard visual recognition module can identify temporary take-off and landing sites, emergency landing sites, or other areas suitable for landing based on ground images acquired by the drone. Other drones can share their acquired landing site locations and status information via V2V communication.

[0063] When receiving candidate landing site data, the landing site selection system simultaneously records the corresponding data source and update time. This information is then combined with the reliability of the data source and the time difference between the data and the current time to determine the confidence level of the candidate landing site data.

[0064] Real-time drone status data describes the drone's flight status, power status, and equipment status at the start of the current decision-making cycle. Real-time drone status data includes at least:

[0065] The drone's current location, current altitude, current ground speed, and current heading;

[0066] Remaining energy data that is compatible with the energy form of drones;

[0067] Current power unit operating status, including the number of power units currently in normal operation and the operating or fault status of each power unit; flight time; current airborne alarm status.

[0068] The current position of the UAV includes its current longitude and latitude, which, together with its current altitude, are used to determine the spatial relationship between the UAV and each candidate landing point. The current ground speed is used to calculate the flight time required for the UAV to reach the candidate landing point and is also included in the energy consumption calculation for the cruise segment. The current heading is used to calculate the target heading required for the UAV to reach the candidate landing point, as well as the turning angle between the current heading and the target heading.

[0069] The remaining energy data is used to determine the remaining available energy and remaining flight time of the UAV. For fuel-powered UAVs, the remaining energy data includes at least the remaining fuel quantity, and may also include the remaining electrical power or remaining support time to power the flight control module, navigation module, communication module, and other key airborne equipment; for electric-powered UAVs, the remaining energy data includes the remaining electrical power or remaining available electrical energy. During subsequent reachability verification, the total energy consumption required for the UAV to reach each candidate landing point is calculated using energy dimensions matching the UAV's energy form. S310 and subsequent reachability calculations in this main implementation method take a fuel-powered multi-engine UAV with at least one engine capable of operation as an example; for electric-powered or single-engine UAVs, implementation can proceed after configuring the corresponding power status indicators, emergency classification rules, and staged energy models according to the corresponding aircraft type. In the case of complete power unit failure, this main implementation method does not generate a valid powered endurance calculation result or a valid reachability result, and terminates the subsequent processing within this main implementation method in accordance with the dedicated state of complete power failure in S430.

[0070] The operating status of the power unit reflects the current operational capability of the UAV's power system and includes at least the number of operational power units required for subsequent emergency level determination and phased energy consumption calculation, as well as the normal, abnormal, faulty, or degraded status of each power unit. Flight time is used to construct the real-time status information employed by the dynamic weighting model. The current airborne alarm status reflects at least one or more of the following: power unit malfunction, insufficient fuel or battery power, flight control malfunction, navigation or communication malfunction, forced landing warning, and serious flight malfunction.

[0071] In one specific implementation, the UAV's current position, current altitude, current ground speed, and current heading are provided by the UAV's positioning interface and flight control interface; remaining energy data, power unit operating status, and current onboard alarm status, adapted to the UAV's energy form, are provided by the UAV's health monitoring interface. For fuel-powered UAVs, the power unit operating status may include the number of currently operating engines and the operating or fault status of each engine.

[0072] Environmental data is used to describe the external environmental conditions that affect landing site selection and drone flight safety. Environmental data includes at least current wind speed, current wind direction, and current visibility, and may include weather warnings and terrain warnings.

[0073] Current wind speed and direction are used to determine the wind direction matching factor for candidate landing sites and are subsequently used to calculate the impact of wind on flight energy consumption. Current visibility is used to determine the emergency response level of the UAV.

[0074] When the UAV system provides weather or terrain warnings related to flight safety, the landing point selection system simultaneously acquires the corresponding warning data and uses it as environmental status data for subsequent emergency level determination.

[0075] For cases where at least one decision cycle has been executed, the landing point selection system reads the identifier of the current target landing point from the navigation module and matches it with the latest status data of the current target landing point in the current decision cycle. The current target landing point is the landing point that was finally determined in the previous decision cycle and is currently being used by the navigation module for trajectory planning and flight guidance.

[0076] In subsequent steps, the landing point selection system re-determines the proposed landing point based on the data of the current decision cycle, and compares the proposed landing point with the current target landing point to determine whether it is necessary to change the target landing point currently being used by the navigation module.

[0077] If this method is executed for the first time and the navigation module does not yet have a current target landing point, the current target landing point will be marked as empty. In this case, the proposed landing point determined in subsequent steps can be identified as the initial target landing point after undergoing appropriate reachability, maneuverability, and range margin checks.

[0078] After completing the above data acquisition, S100 outputs the candidate landing point set for the current decision cycle, the candidate landing point data corresponding to each candidate landing point, the UAV real-time status data, environmental data, the current target landing point identifier, and the latest status data of the current target landing point obtained by matching the identifier; when there is no current target landing point during the first execution, the corresponding identifier and status data are empty.

[0079] in:

[0080] Candidate landing site data, the current location of the UAV and environmental data are sent to S200 to determine seven numerical basic characteristics of each candidate landing site, emergency landing site attributes and basic risk values;

[0081] The drone's real-time status data and environmental data are sent to the S300 to determine the current emergency level; when the current emergency level is normal, early warning, or emergency, it is used to generate dynamic weights for the current decision-making cycle; when the current emergency level is extremely urgent, it is used to form an extremely urgent rapid path identifier.

[0082] The location and altitude data of candidate landing sites, as well as the location, flight status and remaining energy data of the UAV, are sent to the S400 to calculate the total energy consumption and total flight time required for the UAV to reach each candidate landing site.

[0083] The current target landing point identifier and the latest status data obtained by matching the identifier are retained in S700 to determine whether the current target landing point is still valid, and to make a hysteresis change determination when the automatic decision path is applied.

[0084] Thus, S100 forms the unified input data required for subsequent candidate landing point evaluation, dynamic weight generation, reachability verification, and target landing point switching judgment.

[0085] S200 determines the basic characteristic data, emergency landing point attributes, and basic risk values ​​of each candidate landing point based on candidate landing point data, real-time UAV status data, and environmental data.

[0086] The basic feature data includes feature data representing the basic qualifications, distance, runway length, wind direction matching, availability, occupancy status, and data confidence of candidate landing sites. The basic risk value is determined based on the availability, occupancy status, wind direction matching, runway length, and data confidence of the candidate landing sites. Specifically, the feature data representing the occupancy status is determined based on the current occupancy status of the candidate landing sites and the validity of the corresponding data, and serves as the basic information in the candidate landing site evaluation process.

[0087] For the candidate landing point set obtained by S100, the landing point selection system performs feature calculations on each candidate landing point in the set. For the i-th candidate landing point, seven numerical basic features are determined according to basic attributes, geometric attributes, and state attributes, and emergency landing point attribute labels are determined based on the candidate landing point data.

[0088] After determining the current emergency level and corresponding emergency type bonus in S300, the complete eight-dimensional feature vector formed by the i-th candidate landing point before entering S400 is denoted as:

[0089]

[0090] in, Indicates basic qualification score. Represents the distance factor. This represents the runway length factor. Indicates the wind direction matching factor. Represents availability factor, Indicates the penalty factor for occupancy. This represents the confidence factor of the data. This indicates an emergency type bonus.

[0091] S210 determines the basic qualification score based on the candidate landing site type and support capabilities.

[0092] The basic qualification score characterizes the official qualifications of candidate landing sites, as well as their communication, navigation, firefighting, and on-site support capabilities. The landing site selection system reads candidate landing sites. The landing site type is determined, and the basic qualification score is determined based on the pre-established correspondence between the landing site type and the basic qualification score. .

[0093] In one specific implementation:

[0094] Candidate landing sites When a airport is designated as a regular airport or a standard take-off and landing point, basic qualifications will be divided. It is set at 100 points;

[0095] Candidate landing sites When used as a temporary take-off and landing point or an emergency landing site in the field, basic qualifications will be divided. The score is set at 60.

[0096] Candidate landing sites When designating an emergency landing site, basic qualifications will be categorized based on the site's navigation, communication, firefighting, rescue, or ground support capabilities. The score is set at 40 to 80 points.

[0097] The above scores reflect the routine support level of candidate landing sites. Candidate landing site type and emergency landing site attribute are evaluated separately: emergency landing sites are one type of candidate landing site and are used to determine the basic qualification score; whether a candidate landing site participates in emergency type bonuses and top-priority screening as an emergency landing site with emergency support capabilities is determined by the emergency landing site attribute marker.

[0098] After completing S210, the basic qualifications will be divided into P. base,i This serves as the first numerical basic feature of candidate landing point i.

[0099] S220 determines the distance factor based on the UAV's current location and candidate landing point locations.

[0100] The distance factor characterizes the impact of the spatial distance between the UAV's current location and candidate landing points on landing point priority. The closer a candidate landing point is to the UAV, the less flight time and energy the UAV typically needs to reach that landing point, and therefore the higher the distance factor.

[0101] The landing site selection system first considers the UAV's current location and candidate landing sites. The horizontal spherical distance is calculated from the location, and then the three-dimensional spatial distance is determined by combining the height difference between the UAV and the candidate landing point.

[0102] Let the current latitude and longitude of the drone be respectively... and Candidate landing sites The latitude and longitude are respectively and Before performing calculations, latitude and longitude are converted from degrees to radians.

[0103] The Haversine formula is used to calculate the horizontal spherical distance between the UAV's current position and the candidate landing point i. A commonly used expression for this is:

[0104]

[0105]

[0106] in, This represents an intermediate quantity in the Haversine calculation process; Indicates the current location of the drone and the candidate landing sites. The horizontal spherical distance between them, in meters; The radius represents the average radius of the Earth, which is taken as 6,371,000 meters in this embodiment.

[0107] Let the current altitude of the drone be... Candidate landing sites The height is The height difference between the two is:

[0108]

[0109] Further, based on the horizontal spherical distance and altitude difference, the current position of the UAV and the candidate landing point are determined. The three-dimensional spatial distance between them:

[0110]

[0111] in, Indicates the distance from the drone's current location to the candidate landing point. The three-dimensional spatial distance is expressed in meters.

[0112] Obtain three-dimensional spatial distance Then, it is converted into a distance factor using an exponential decay method:

[0113]

[0114] in, Indicates candidate landing sites Distance factor; This represents the distance attenuation reference constant.

[0115] Through the exponential decay mechanism described above, the distance factor does not drop directly to zero when the candidate landing site is relatively far away, but rather decreases continuously with increasing spatial distance. This result can reflect the distance advantage of nearby candidate landing sites, while also preserving the possibility of participating in subsequent evaluations of candidate landing sites that are farther away but may have better runway, support, or environmental conditions.

[0116] After completing S220, output the horizontal spherical distance, three-dimensional spatial distance di, and distance factor P of the candidate landing point i. dist,i Among these, the distance factor serves as the second numerical fundamental feature; the three-dimensional spatial distance will also be used by the S400 to calculate the flight time and energy consumption required to reach the candidate landing point, and by the S500 to perform distance range filtering under different emergency levels. The three-dimensional spatial distance is a geometric distance determined based on the horizontal spherical distance and altitude difference, and is not equivalent to the actual flight path length after considering detours.

[0117] S230, determine the runway length factor based on the runway length of the candidate landing sites.

[0118] The runway length factor characterizes the impact of runway length at candidate landing sites on the landing safety margin and taxiing tolerance of UAVs. The longer the runway, the greater the deceleration, correction, and taxiing space the UAV typically has during landing; however, once the runway length exceeds the reference length required for UAV landing, further increasing the runway length no longer improves the score for this factor.

[0119] Candidate landing sites The length of the runway is The minimum safe runway length for the drone adapted to this embodiment is The full score reference value for runway length is Then the runway length factor Determined according to the following segmentation relationship:

[0120]

[0121] in, , and The unit for all values ​​is meters.

[0122] In one specific implementation:

[0123]

[0124]

[0125] When candidate landing sites When the runway length is less than 500 meters, the runway length factor is set to 0, and the candidate landing point is marked as not meeting the minimum runway length requirement of the UAV in this embodiment.

[0126] When the track length exceeds 2000 meters, the track length factor is set to 100 points, and no additional points will be added for this item as the track continues to lengthen.

[0127] For candidate landing sites with runway lengths less than the minimum safe runway length, in addition to setting the runway length factor to 0, a runway length non-compliance flag is generated. This flag is used in subsequent candidate set processing to prevent candidate landing sites that do not meet the minimum takeoff and landing conditions from continuing to participate in the final selection based solely on higher scores of other features.

[0128] After completing S230, the runway length factor P will be adjusted. len,i As the third numerical basic feature of candidate landing point i, the runway length is not satisfied and the marker is retained for subsequent screening.

[0129] S240 determines the wind direction matching factor based on the runway landing direction of the candidate landing point and the current wind vector direction.

[0130] The wind direction matching factor is used to characterize the degree of matching between the runway landing direction and the current wind vector direction when the UAV lands along the runway landing direction of the candidate landing point.

[0131] In this implementation, a higher wind direction matching factor corresponds to a headwind landing; the wind direction matching factor decreases when the crosswind increases; and the wind direction matching factor can be negative when landing with the wind, thus incurring a penalty in the overall priority calculation.

[0132] The wind direction obtained by S100 is the meteorological wind direction, indicating the direction from which the wind is blowing. The meteorological wind direction, wind vector direction, and runway landing direction used in this calculation are all pre-converted to the same angular reference and normalized to a range of 0° to 360°, with true north as 0° and increasing clockwise. The meteorological wind direction is denoted as Wmet, and the actual wind vector direction is denoted as Wvec, both satisfying the following conditions:

[0133]

[0134] Subsequent calculations involving runway wind direction matching and airway wind speed projection all use the wind vector direction Wvec. When the data source provides meteorological wind direction, only one 180° transformation is performed according to the aforementioned relationship. When the data source directly provides the wind vector direction, it is used directly after completing the same angle reference transformation and angle normalization, without performing a 180° transformation.

[0135] When the ground station, airport operations system, or airspace control system has already designated the currently permitted runway landing direction, the designated directional landing direction is adopted. When the input only contains runway axis data, and the input data clearly indicates that both opposite directions of the same runway are permitted as candidate landing directions in this decision cycle, the headwind angles corresponding to the two directions are calculated separately, and the direction with the smaller headwind angle is determined as the runway landing direction currently used for evaluation. When neither a designated directional landing direction is obtained nor it can be confirmed that both opposite directions are permitted as candidate landing directions in this decision cycle, the reason for insufficient runway direction data is recorded, no valid wind direction matching factor or headwind angle is formed, and no valid results are output to subsequent calculations that depend on the wind direction matching factor or headwind angle.

[0136] Let H be the current directional runway landing heading used by candidate landing point i. runway,i The current wind vector direction is W. vec Both use the same angular reference, with true north as 0°, and azimuth angles ranging from 0° to 360° in a clockwise direction. runway,i This indicates the directional heading of the UAV when it lands along the corresponding runway. It is not the same as the current heading of the UAV, nor is it the same as the expected flight heading used by S440 for wind field projection along the route.

[0137] Considering that the azimuth angle has a periodicity of 360°, first determine the minimum azimuth angle between the runway landing direction and the wind vector direction:

[0138]

[0139] Where, Δ i This represents the minimum azimuth angle between the runway landing direction and the wind vector direction, and its value ranges from 0° to 180°.

[0140] Further determine the degree of deviation of this angle relative to the headwind and tailwind conditions:

[0141]

[0142] in, Indicates the angle of deflection against the wind.

[0143] When θ i When θ = 0°, it indicates that the runway landing direction is opposite to the wind vector direction, forming a headwind condition; when θ i When θ = 90°, it indicates the formation of a crosswind condition; when θ i When the angle is 180°, it means that the runway landing direction is the same as the wind vector direction, forming a tailwind condition.

[0144] Calculate the wind direction matching factor based on the headwind yaw angle:

[0145]

[0146] in, Indicates candidate landing sites Wind direction matching factor.

[0147] Therefore, the wind direction matching factor in this embodiment ranges from -100 to 100. Negative values ​​are used to ensure that downwind conditions have a negative impact in subsequent comprehensive priority calculations, rather than truncating all unfavorable wind directions to zero.

[0148] After completing S240, if the runway direction data is valid, the wind direction matching factor is used as the fourth numerical basic feature of the candidate landing point. The runway landing direction, wind vector direction, and the same headwind angle θi are transmitted to S520 along with the candidate landing point data for wind direction deviation threshold screening in the early warning level emergency state. S520 directly reuses the θi without recalculating the other headwind angle.

[0149] S250 determines the availability factor based on the current openness of candidate landing sites.

[0150] The availability factor is used to characterize whether a candidate landing site is in a state that allows drones to land in the current decision cycle.

[0151] For candidate landing sites The landing site selection system reads the current availability status obtained by S100 and determines the availability factor based on the current availability status. :

[0152]

[0153] in, This indicates that the candidate landing sites are currently available. 0 indicates that the candidate landing site is currently unavailable.

[0154] Although the availability factor is included as one of the eight features in the subsequent comprehensive priority calculation, it is also a veto-qualifier. When candidate landing sites... When the availability factor is 0, the landing site selection system marks the candidate landing site as an unavailable candidate landing site and excludes it from the candidate sites for subsequent decision-making.

[0155] In other words, the handling of availability factors involves two levels:

[0156] First, record the current open state of the candidate landing site in the feature data so that the subsequent calculation process can retain and track the state;

[0157] Second, before entering the comprehensive priority comparison, reachability verification and multi-target selection, currently unavailable candidate landing points are directly excluded to avoid them being re-selected due to their proximity, long runway, or other high feature scores.

[0158] For candidate landing sites whose current openness cannot be confirmed, they are not directly presumed to be open. The landing site selection system may continue to retain the data record of the candidate landing site, but will not treat it as a currently available candidate landing site until it obtains status information that can prove that it is permitted to land.

[0159] After completing S250, output the availability factor P of candidate landing site i. avail,i And the corresponding availability verification results. The availability factor is the fifth numerical basic feature; the availability verification results are used to determine whether the candidate landing site can proceed to subsequent processing.

[0160] S260, determine the occupancy penalty factor based on the current occupancy status and expected release time of the candidate landing site.

[0161] The occupancy penalty factor is used to characterize the adverse effects on the current UAV landing decision when a candidate landing site is occupied by another aircraft.

[0162] The occupancy penalty factor is a negative evaluation metric. In subsequent comprehensive priority calculations, the weighted result corresponding to the occupancy penalty factor is deducted from the weighted results of other positive features.

[0163] For candidate landing sites The landing site selection system first reads the current occupancy status of the candidate landing site.

[0164] When candidate landing sites When the space is not occupied by other aircraft, the occupancy penalty factor will be determined as follows:

[0165]

[0166] At this time, the candidate landing site will not incur a penalty for being occupied.

[0167] When candidate landing point i is already occupied by another aircraft, the remaining time from the current decision time until the candidate landing point is expected to be reused is further obtained, and this remaining time is recorded as the expected release remaining time. If the input data uses an absolute estimated release time, then the absolute estimated release time and the current decision time are first converted to the same time base, and then the remaining estimated release time is determined.

[0168] Let the maximum acceptable waiting time be The maximum penalty for occupying space is To ensure that the penalty for longer expected waiting times is greater, the occupancy penalty factor is determined according to the following segmented relationship:

[0169]

[0170] in:

[0171] Indicates candidate landing sites The penalty factor for occupancy is expressed in points.

[0172] This represents the estimated remaining release time for candidate landing point i, calculated from the current decision time, in seconds.

[0173] This indicates the maximum acceptable waiting time, in seconds.

[0174] This indicates the maximum penalty limit, expressed in points.

[0175] In one specific implementation:

[0176]

[0177]

[0178] In other words, the maximum penalty for occupying space is 50 points, and the maximum acceptable waiting time is 180 seconds.

[0179] When a candidate landing site is occupied, and the estimated remaining release time is missing, expired, non-numerical, negative, or cannot use the same time base as the current decision time, the corresponding reason is recorded, the estimated remaining release time is marked as invalid, and the occupancy penalty factor is set to the upper limit of the occupancy penalty. The invalid state does not equate to the candidate landing site being correctly calculated as unreachable; after obtaining valid updated data in subsequent decision cycles, the estimated remaining release time is re-determined and the occupancy penalty factor is recalculated.

[0180] After completing S260, the penalty factor P will be used. occ,i As the sixth numerical basic feature of candidate landing point i, it is used as a deduction item in the subsequent comprehensive priority calculation; at the same time, the occupancy status of the candidate landing point, the expected remaining release time trelease,i and its validity flag are retained for S450 to determine the occupancy waiting time and the corresponding equivalent fuel consumption.

[0181] S270, determine the data confidence factor based on the source and update time of the candidate landing site data.

[0182] The data confidence factor is used to characterize the reliability of candidate landing site data. Data confidence considers both the reliability of the data source and the freshness of the data.

[0183] For candidate landing sites The landing site selection system reads the data source corresponding to the candidate landing site data and determines the source reliability coefficient based on the data source. .

[0184] In one specific implementation:

[0185]

[0186] The data provided by the official ground station has high authority and reliable status updates, so the source reliability coefficient is set to 1.0; the airborne visual recognition results may be affected by the shooting angle, image clarity and recognition error, so the source reliability coefficient is set to 0.9; the data shared by other UAVs through V2V may have communication transmission and status update delays, so the source reliability coefficient is set to 0.8.

[0187] Further, the update time of the candidate landing site data is read, and the data time difference is determined based on the difference between the current time in the current decision cycle and the update time:

[0188]

[0189] in:

[0190] Indicates candidate landing sites The data is the time difference from the current time, in seconds;

[0191] This indicates the current moment in the current decision-making cycle;

[0192] This indicates the data update time for candidate landing point i.

[0193] When the data has just been updated Smaller; the longer the data remains unchanged, The larger.

[0194] Based on the source reliability coefficient and data time difference, the data confidence factor is determined using an exponential decay method:

[0195]

[0196] in:

[0197] Indicates candidate landing sites The data confidence factor is expressed in points.

[0198] This represents the reliability coefficient of the candidate landing site data source;

[0199] This indicates the time difference between the data and the current moment.

[0200] This represents the data confidence decay time constant, in seconds.

[0201] In one specific implementation:

[0202]

[0203] When candidate landing site data comes from official ground stations and has just been updated:

[0204]

[0205]

[0206] The corresponding data confidence factor is:

[0207]

[0208] When the official ground station data has not been updated for 30 seconds:

[0209]

[0210] That is, the data confidence factor is approximately 37 points.

[0211] When a newly updated piece of data originates from airborne visual recognition, its data confidence factor is:

[0212]

[0213] When newly updated data originates from V2V sharing between other drones, its data confidence factor is:

[0214]

[0215] As the data update time gradually increases from the current moment, the confidence level of data from different sources decreases exponentially. Therefore, even if a candidate landing site has good basic attributes, its confidence contribution in subsequent comprehensive priority calculations will decrease if its openness, occupancy, or environmental status data has not been updated for a long time.

[0216] In this embodiment, the data confidence factor is used to correct the impact of candidate landing point data quality on the landing point selection result. However, a low data confidence level does not necessarily mean that the candidate landing point is unavailable. Whether a candidate landing point is open is still determined by the availability check in S250.

[0217] When multiple sources provide data for the same candidate landing site, data from sources with higher reliability coefficients are prioritized. If the reliability coefficients are the same, data with a more recent update time is used. If the time difference between data from a higher reliability source and the current time exceeds a preset validity period, data from other sources that are still valid and have a more recent update time are used. The data confidence factor is determined based on the finally selected data source and update time.

[0218] After completing S270, the data confidence factor P conf,i This serves as the seventh numerical fundamental feature for candidate landing point i.

[0219] S280, determine the emergency landing point attributes of the candidate landing points, and determine the emergency type bonus after determining the current emergency level in S300.

[0220] The emergency type bonus is used to improve the evaluation results of emergency landing sites with emergency support capabilities when the drone is in an emergency state.

[0221] It should be noted that the emergency landing point attribute of the candidate landing point and the current emergency level of the drone are two different data objects:

[0222] The emergency landing point attribute is used to indicate whether a candidate landing point itself is an emergency landing point with emergency support capabilities;

[0223] The current emergency level indicates whether the drone is in a normal, warning, emergency, or critical state during the current decision-making cycle.

[0224] The emergency landing point attributes are determined by the candidate landing point data obtained by S100, while the current emergency level is determined by S300 based on the real-time status data and environmental data of the UAV.

[0225] For candidate landing sites Define emergency landing point attribute markers :

[0226]

[0227] When candidate landing sites It has the emergency support capability required for the forced landing of a malfunctioning drone, and when its landing point data is marked as an emergency landing point, the emergency landing point attribute is marked as 1; otherwise, the emergency landing point attribute is marked as 0.

[0228] Since emergency type bonuses only take effect in emergency scenarios, the S200 phase first determines the emergency attributes of candidate landing sites and does not directly apply unconditional bonuses to them before the current emergency level is determined.

[0229] After S300 determines the current emergency level, the emergency type bonus for candidate landing point i is determined based on the current emergency level and the emergency landing point attribute marker.

[0230] Its basic relationship can be expressed as:

[0231]

[0232] in, This indicates the preset emergency bonus value corresponding to the current emergency level.

[0233] To avoid the same emergency landing point attribute being awarded repeatedly in S200 and S500, these 30 points are used as an emergency type bonus under the emergency level. The specific values ​​are determined, rather than being independently added together after the subsequent comprehensive priority calculation.

[0234] Therefore, in one specific implementation of the emergency level:

[0235]

[0236] Under normal and warning levels, the emergency type bonus is 0; under emergency level, the emergency type bonus for emergency landing points is 30 points, and for non-emergency landing points it is 0 points; under extremely urgent level, the distance-based fast path is used, and the emergency type bonus is not used for sorting.

[0237] After completing S280, output candidate landing sites. Emergency landing point attribute marker After S300 determines the current emergency level, an emergency type bonus is further obtained based on this attribute marker. .

[0238] S285, determine the basic risk value based on the safety-related characteristics of the candidate landing sites.

[0239] The baseline risk value is used to characterize the comprehensive landing risk of candidate landing sites in terms of current openness, occupancy, runway wind conditions, runway length, and data reliability, and is used by the subsequent dynamic weighting model and multi-objective optimization process.

[0240] The landing site selection system converts availability factors, occupancy penalty factors, wind direction matching factors, runway length factors, and data confidence factors into directionally consistent safety components. The safer the candidate landing site, the higher its corresponding safety component.

[0241] In one specific implementation, the availability factor is mapped from 0 or 1 to 0-100 points; the occupancy penalty factor is converted inversely to an occupancy safety component according to the upper limit of the occupancy penalty; the wind direction matching factor of -100 to 100 points is linearly converted to 0-100 points; and the runway length factor and data confidence factor are directly used as the corresponding safety components.

[0242] Let the five directional safety components of candidate landing point i be S in sequence. 1,i To S 5,i The corresponding risk assessment weights are α1 to α5, then the overall safety score S safe,i =α1S 1,i +α2S 2,i +α3S 3,i +α4S 4,i +α5S 5,i , where each α m ≥0 and α1+α2+α3+α4+α5=1. The risk assessment weights can be preset according to the UAV model, mission scenario and historical operation data; in one specific implementation, all five weights can be set to 0.2, or a non-equal weight setting obtained by calibration based on historical operation data can be used. The above settings are not limited to unique values.

[0243] The difference between 100 and the aforementioned comprehensive safety score is determined as the baseline risk value for the candidate landing site, i.e., R. base,i =100-S safe,i A higher overall safety score corresponds to a lower baseline risk value; conversely, a lower overall safety score corresponds to a higher baseline risk value.

[0244] The basic risk value is not added as a new dimension in the eight-dimensional feature vector. S300 calculates the average risk of candidate landing points based on the basic risk value of each candidate landing point, and S600 directly uses the basic risk value of a single candidate landing point as the landing risk target, thereby avoiding the repeated establishment of inconsistent risk assessment relationships in different steps.

[0245] After completing S285, output the overall safety score and basic risk value of each candidate landing site.

[0246] S290 summarizes the basic characteristic data of candidate landing sites, emergency landing site attributes, and basic risk values.

[0247] After processing by S210 to S285, the landing point selection system obtains the following seven numerical basic characteristics of candidate landing point i and one emergency landing point attribute marker:

[0248] Basic qualification points ;

[0249] Distance factor ;

[0250] Runway length factor ;

[0251] Wind direction matching factor ;

[0252] Availability factor ;

[0253] Occupy penalty factor ;

[0254] Data confidence factor ;

[0255] Emergency landing point attribute marker .

[0256] In addition to the seven numerical basic features and one emergency landing point attribute marker mentioned above, S285 also forms the basic risk value of candidate landing points.

[0257] The first seven items are numerical features already determined in this step; the eighth item is first determined in the form of the emergency attribute of the candidate landing site itself, and after S300 obtains the current emergency level, it is converted into an emergency type bonus that is actually effective in the current decision-making cycle. The basic risk value is an auxiliary evaluation result used by S300 and S600 and does not constitute a new dimension in the eight-dimensional feature vector.

[0258] Therefore, before proceeding to the S400 integrated priority calculation, the complete eight-dimensional numerical eigenvector of candidate landing point i is determined as follows:

[0259]

[0260] The S200 outputs seven numerical basic characteristics, basic risk values, availability verification results, runway length non-compliance markers, three-dimensional spatial distances, and emergency landing point attribute markers for each candidate landing point.

[0261] in:

[0262] The seven numerical basic features, together with the emergency type determined by S300 based on the current emergency level, form a complete eight-dimensional feature vector, which is then combined with the dynamic weights generated by S300 in S400 to calculate the comprehensive priority of candidate landing sites.

[0263] The baseline risk value is used by S300 to calculate the average risk of candidate landing sites, and by S600 directly as the landing risk target for a single candidate landing site;

[0264] Availability verification results and runway length not meeting the requirements are used to exclude candidate landing sites that do not meet the basic usage conditions;

[0265] Three-dimensional spatial distance is used for calculating the flight time and energy consumption of the S400, and for screening the emergency range of the S500;

[0266] Emergency landing point attribute markers are used to determine emergency type bonuses and for filtering S500 emergency and critical levels.

[0267] The S300 determines the current emergency level based on real-time status data and environmental data of the UAV, and generates eight-dimensional dynamic weights under the normal, early warning and emergency levels.

[0268] The S300 is used to determine the emergency level for landing site selection in the current decision-making cycle based on the UAV's flight status, power status, equipment status, and external environment status. For normal, early warning, and emergency levels, an eight-dimensional dynamic weight is generated based on the corresponding emergency level; for extremely urgent level, a fast path identifier is output and the DDPG online fine-tuning and eight-dimensional comprehensive priority calculation are skipped.

[0269] In this embodiment, the candidate landing site processing strategy represents the processing method for candidate landing sites determined according to the current emergency level. Different emergency levels correspond to different candidate landing site processing procedures. Specifically, when the current emergency level is non-urgent, the candidate landing site processing strategy includes generating dynamic evaluation weights for the current decision-making cycle and determining the processing path of the comprehensive priority of each candidate landing site based on the basic feature data and the dynamic evaluation weights. When the current emergency level is urgent, the candidate landing site processing strategy includes forming an urgent-level fast path identifier and skipping the processing path determined by the comprehensive priority.

[0270] This implementation method employs a dynamic weight generation approach that combines "tiered baseline weights with adaptive fine-tuning through reinforcement learning." The baseline weights corresponding to the emergency level are used to determine the overall decision-making tendency for landing point selection in the current scenario. Subsequent reinforcement learning models continuously fine-tune these baseline weights based on the current specific state, rather than completely replacing them.

[0271] S310 determines the current emergency level based on real-time drone status data and environmental data.

[0272] The landing point selection system reads the current airborne alarm status from the real-time status data of the UAV acquired by S100, as well as the current wind speed, current visibility, weather alarm, and terrain alarm from the environmental data, and determines the emergency level of the current decision-making cycle according to the preset emergency level judgment rules.

[0273] Record the current emergency level as Its value is:

[0274]

[0275] in:

[0276] This indicates the normal level, i.e., Level 0;

[0277] This indicates the warning level, i.e., Level 1;

[0278] This indicates the emergency level, i.e., Level 2;

[0279] This indicates the express level, or Level 3.

[0280] 1. Normal Level 0

[0281] When the drone system has no alarms, all engines are in normal working condition, weather conditions meet the requirements for normal flight, terrain alarms are not triggered, and relevant equipment is operating normally, the current emergency level is set to Level 0 (normal).

[0282] 2. Warning Level 1

[0283] When a warning-level alarm is detected, the current emergency level will be set to Level 1 (early warning level).

[0284] Warning-level alarms include one or more of the following situations:

[0285] The current wind speed exceeds 15 m / s;

[0286] Current visibility is less than 3000m;

[0287] Battery charge is below 20%;

[0288] A single engine is showing abnormal signs;

[0289] The system generates a warning-level alarm equivalent to the severity of the aforementioned state.

[0290] The abnormal symptoms of the engine indicate that the engine has not yet been determined to be completely faulty, but the health monitoring interface has detected that its operating parameters deviate from the normal range and generated corresponding warning information.

[0291] 3. Emergency Level 2

[0292] When an emergency alert is detected, the current emergency level is determined to be Level 2.

[0293] Emergency alerts include one or more of the following scenarios:

[0294] A single engine completely failed;

[0295] The current wind speed exceeds 25 m / s;

[0296] Current visibility is less than 1000m;

[0297] Trigger terrain alarm;

[0298] The remaining fuel level was determined to be low;

[0299] The system generates an emergency alert of equal severity to the aforementioned conditions.

[0300] Whether the remaining fuel level is low can be determined based on the fuel alarm status output by the drone's health monitoring module. This implementation does not limit the uniform threshold value corresponding to low fuel level, nor does it set a fixed percentage; instead, it is based on the fuel alarm rules preset for the current drone model.

[0301] 4. Limited Express Level 3

[0302] When a fatal alarm is detected, the current emergency level is set to Level 3.

[0303] Fatal warnings include one or more of the following situations:

[0304] Two or more engines malfunction;

[0305] The remaining fuel is about to run out;

[0306] Triggering an emergency landing warning;

[0307] A serious flight malfunction occurred;

[0308] The system generates a fatal alarm of similar severity to the above-mentioned state.

[0309] 5. Determining the level when multiple triggering conditions exist simultaneously.

[0310] In one specific implementation, when multiple alarm states correspond to different emergency levels, the highest emergency level is determined as the current emergency level. Its expression is:

[0311]

[0312] in:

[0313] This indicates the number of alarm states detected in the current decision-making cycle;

[0314] Indicates the first The emergency level corresponding to each alarm status;

[0315] This indicates the final determined current emergency level.

[0316] After completing S310, output the current emergency level. This emergency level has at least the following subsequent uses:

[0317] When the current emergency level is normal, warning, or emergency, it is used to determine the eight-dimensional baseline weights in S320.

[0318] When the current emergency level is normal, warning, or emergency, it is used as a state variable in the subsequent DDPG state vector;

[0319] Used to determine whether the emergency landing point attribute generates an emergency type bonus in the current cycle;

[0320] Used for S500 to select candidate point filtering rules and decision paths corresponding to the current level.

[0321] S320: When the current emergency level is normal, warning, or emergency, the eight-dimensional baseline weights are determined based on the current emergency level.

[0322] The landing point selection system reads the eight baseline weights corresponding to the current emergency level output by S310 from a preset four-level baseline weight table. The table retains the baseline weights for the "Urgent Level" level to accommodate implementations that require recording the "Urgent Level" scoring status. In this main implementation, the "Urgent Level" level does not invoke subsequent DDPG online fine-tuning and comprehensive priority calculation.

[0323] The eight-dimensional baseline weight vector corresponding to the current level is denoted as:

[0324] in:

[0325] This represents the baseline weight of the basic qualification score;

[0326] Represents the baseline weight of the distance factor;

[0327] This represents the baseline weight of the runway length factor;

[0328] This represents the baseline weight of the wind direction matching factor;

[0329] The baseline weighting of the availability factor;

[0330] This represents the baseline weight of the penalty factor.

[0331] The baseline weights of the data confidence factors;

[0332] This represents the baseline weight for the emergency type bonus.

[0333] The baseline weights for each emergency level are shown in the table below.

[0334] Weight Dimension Normal Level 0 Warning Level 1 Emergency Level 2 Limited Express Level 3 Basic qualification weight 0.15 0.10 0.05 0.05 Distance weight 0.20 0.25 0.40 0.50 runway length weight 0.15 0.15 0.10 0.05 Wind direction matching weight 0.15 0.20 0.10 0.05 Availability weight 0.10 0.10 0.15 0.15 Occupy penalty weight 0.10 0.10 0.10 0.10 Data confidence weight 0.10 0.08 0.05 0.05 Emergency bonus weighting 0.05 0.02 0.05 0.05

[0335] For each emergency level, the sum of the weights of the eight benchmarks is 1, that is:

[0336]

[0337] Specifically, the weights for the normal level are:

[0338] 0.15 + 0.20 + 0.15 + 0.15 + 0.10 + 0.10 + 0.10 + 0.05 = 1

[0339] The weighted sum of the warning levels is: 0.10 + 0.25 + 0.15 + 0.20 + 0.10 + 0.10 + 0.08 + 0.02 = 1;

[0340] The weighted sum of the emergency level is: 0.05 + 0.40 + 0.10 + 0.10 + 0.15 + 0.10 + 0.05 + 0.05 = 1;

[0341] The weighted sum of the express and special express grades is: 0.05 + 0.50 + 0.05 + 0.05 + 0.15 + 0.10 + 0.05 + 0.05 = 1.

[0342] As the emergency level is raised from normal to extremely urgent, the weights of various benchmarks show a changing relationship that matches the emergency landing requirements.

[0343] The baseline weight of the distance factor is gradually increased from 0.20 in the normal level to 0.50 in the urgent level. This is because the importance of reaching the landing point as quickly as possible continues to increase as the drone's condition deteriorates. In the urgent level, the distance weight accounts for half of the total baseline weight.

[0344] The baseline weight for the basic qualification score has been reduced from 0.15 for the normal level to 0.05 for the emergency and express levels; the baseline weight for the runway length factor has also been reduced from 0.15 for the normal level to 0.05 for the express level. This indicates that under severe emergency conditions, the requirements for airport qualifications and conventional support conditions such as longer runways are relatively reduced.

[0345] The baseline weight of the wind direction matching factor is increased to 0.20 in the warning level to reflect the importance of runway orientation matching with wind direction when environmental anomalies such as wind speed or visibility occur; in the emergency and emergency levels, as the demand for rapid landing further increases, this weight is reduced to 0.10 and 0.05, respectively.

[0346] The baseline weight of the availability factor has been increased from 0.10 for the normal and warning levels to 0.15 for the emergency and critical levels. Although currently unavailable candidate landing sites have been vetoed by S250, the availability weight remains in the baseline weight system to maintain the complete eight-dimensional evaluation structure.

[0347] The baseline weight of the occupancy penalty factor is 0.10 across all four levels. The baseline weight of the data confidence factor gradually decreases from 0.10 for the normal level to 0.05 for the emergency and critical levels, indicating that as the urgency level increases, the system pays more attention to distance and current availability under time constraints.

[0348] For the Normal, Warning, and Emergency levels, the aforementioned baseline weights define the basic direction of dynamic weight adjustments. Subsequent weight adjustments output by the reinforcement learning model are fine-tuned based on the baseline weights for the corresponding level, rather than generating a new set of weights unrelated to the current emergency level. The Extremely Urgent level uses the fast path specified in S540 and does not invoke the aforementioned dynamic weight fine-tuning.

[0349] S200 has already determined emergency landing point attribute markers for each candidate landing site. However, whether the emergency type bonus actually takes effect depends on the current emergency level determined by S310.

[0350] In this embodiment:

[0351] When the current emergency level is Level 0 (normal level), the emergency landing bonus is not activated.

[0352] When the current emergency level is Level 1 (early warning level), the emergency type bonus is 0, and the candidate points are processed according to the early warning level selection rules.

[0353] When the current emergency level is Level 2 and the candidate landing sites Emergency landing point attribute marker At that time, the emergency type bonus for the candidate landing site was set at 30 points;

[0354] When the current emergency level is Level 3, the S500 will only retain emergency landing points within a preset distance range and skip the regular weighted scoring. Therefore, it will not rely on the emergency type bonus to re-sort emergency landing points.

[0355] Therefore, in one specific implementation of Level 2 emergency response:

[0356]

[0357] The aforementioned 30 points are specific values ​​for the emergency type bonus in the eight-dimensional feature system. They are multiplied with the corresponding emergency type dynamic weight in S410 and then used in the comprehensive priority calculation. No additional 30 points are added to the same candidate landing point outside of the comprehensive priority calculation result.

[0358] After completing S320, when the current emergency level is normal, warning, or emergency, the landing point selection system outputs the current emergency level, the corresponding eight-dimensional baseline weight vector, the actual emergency type bonus for each candidate landing point, and the current emergency level identifier for use by S500; when the current emergency level is extremely urgent, it outputs the current emergency level and the extremely urgent fast path identifier, but does not output the baseline weight and emergency type bonus used for comprehensive scoring in this cycle.

[0359] S330 constructs a twelve-dimensional state vector based on the current flight status, environmental status, and candidate landing point set when the current emergency level is normal, early warning, or emergency.

[0360] After obtaining the current emergency level and its corresponding eight-dimensional baseline weights, the landing point selection system further constructs a state vector for adaptive fine-tuning in reinforcement learning.

[0361] The state vector is used to describe the UAV's flight state, power state, external environment state, and overall state of candidate landing points related to the landing point weight adjustment during the current decision cycle.

[0362] Let the current decision cycle be denoted as the th cycle. For each decision cycle, its twelve-dimensional state vector is denoted as:

[0363] Among them, the one above the symbol This indicates that the corresponding state variable has been normalized to the range of 0 to 1.

[0364] The twelve state variables are as follows:

[0365] The data includes the drone's current ground speed, current altitude, remaining energy status, power unit operating status, current wind speed, current wind direction, current visibility, current emergency level, number of candidate landing sites, average distance to candidate landing sites, average risk value of candidate landing sites, and drone's flight time.

[0366] 1. Current ground speed

[0367] The current ground speed is obtained from S100 and denoted as .

[0368] This state variable reflects the actual speed at which the UAV is currently moving along the ground. The current ground speed affects both the time required to reach the candidate landing point and the energy consumption estimate during the cruise phase.

[0369] 2. Current altitude

[0370] The current altitude is obtained from S100 and denoted as .

[0371] This state variable reflects the current vertical position of the drone. The relationship between the current altitude and the altitude of the candidate landing point will affect whether a climb or descent is needed and the corresponding energy consumption.

[0372] 3. Remaining Energy Status

[0373] In this main implementation, taking a fuel-powered UAV as an example, the remaining energy state before normalization is represented as the remaining fuel quantity, denoted as . .

[0374] This state variable is used to reflect the current remaining propulsion energy state of the UAV and influences the dynamic weight model's tendency to adjust between safety and efficiency.

[0375] For drones powered by electricity, the state dimension can be replaced with the remaining power or remaining available power before normalization without increasing the state dimension, and then normalized according to the corresponding effective range before being input into the dynamic weight model.

[0376] 4. Power unit operating status

[0377] The operating state of the power unit is denoted as N. power,t .

[0378] This state variable reflects the current state of the power units that the UAV can provide power normally. In the fuel-powered multi-engine UAV of this main embodiment, this state variable is represented by the number of engines that are currently working normally. For other power types, equivalent state indices that can characterize the number of working power units, the degree of failure, or degradation can be used and normalized before being input into the dynamic weight model. When a power unit experiences an anomaly, failure, or degradation, the dynamic weight model can correspondingly increase the importance of factors such as distance, availability, or emergency landing point.

[0379] 5. Current wind speed

[0380] The current wind speed is obtained from S100 and is denoted as .

[0381] This state variable is used to reflect the impact of external airflow on the UAV's flight energy consumption, track maintenance, and landing safety.

[0382] 6. Current wind direction

[0383] The current wind direction is obtained from S100 and is denoted as .

[0384] This state variable, along with the runway orientation of the candidate landing site, jointly influences the wind direction matching evaluation. During formal normalization, the current wind direction is mapped to the range of 0 to 1 according to the system's preset angle normalization rules.

[0385] 7. Current visibility

[0386] Current visibility is recorded as .

[0387] This state variable reflects the visual conditions and weather safety level of the UAV's current flight environment. When visibility decreases, the dynamic weight model can adjust the weights of factors such as distance, runway conditions, and data confidence accordingly.

[0388] 8. Current Emergency Level

[0389] The current emergency response level is determined using S310. .

[0390] Its preset coding range is 0, 1, 2 or 3, which correspond to normal level, warning level, emergency level and urgent level respectively; in this main implementation, S330 is only executed in normal level, warning level or emergency level, and urgent level directly enters the fast path.

[0391] The current emergency level, as an input dimension in the state vector, is used to enable the reinforcement learning model to understand the overall security status in this round of decision-making; simultaneously, the current emergency level is also used by S320 to determine the baseline weights. Their functions differ:

[0392] The baseline weights are used to predetermine the overall decision-making direction at the current level;

[0393] The emergency level in the state vector is used by the reinforcement learning model to continuously fine-tune the overall decision direction in conjunction with other real-time states.

[0394] 9. Number of candidate landing sites

[0395] The number of candidate landing sites retained during the current decision-making cycle is denoted as .

[0396] In one implementation, this number is the number of candidate landing sites that are still retained after S100 is obtained and after S250 availability processing.

[0397] The number of candidate landing sites reflects the choice space for the current landing decision. When the number of candidate landing sites is small, the model can reduce the requirements for comfort or general support conditions; when the number of candidate landing sites is large, the model can conduct a more detailed evaluation among more candidate options.

[0398] When the current emergency level is normal, warning, or emergency, and the number of candidate landing sites is currently zero, the average distance, average risk value, and eight-dimensional dynamic weight of the candidate landing sites are not calculated. The reason for the empty candidate set is recorded, and a "no unique proposed landing site has been formed" status is generated. In this decision-making cycle, the processing steps S340 to S695, which rely on the candidate landing site set, are not executed further. The status, the reason for its formation, and the current target landing site identifier are sent to S700 for current target technical status determination.

[0399] 10. Average distance to candidate landing sites

[0400] Based on the three-dimensional spatial distances of each candidate landing point already calculated by S220, calculate the average distance of the current candidate landing point:

[0401]

[0402] in:

[0403] This represents the average three-dimensional spatial distance between candidate landing sites within the current decision-making cycle;

[0404] This represents the three-dimensional spatial distance from the current position of the UAV to the i-th candidate landing point during the t-th decision cycle;

[0405] This indicates the number of candidate landing sites currently retained.

[0406] The average distance to candidate landing sites is used to reflect the overall distance between the current set of candidate landing sites and the UAV.

[0407] 11. Average risk value of candidate landing sites

[0408] The average risk value of the current candidate landing sites is denoted as... .

[0409] In this embodiment, the basic risk values ​​of each candidate landing site already determined in S285 are directly read and calculated:

[0410] in:

[0411] This represents the risk value of the i-th candidate landing site within the current decision-making cycle;

[0412] This represents the average risk value of the current set of candidate landing sites.

[0413] The baseline risk values ​​for each candidate landing site have already been uniformly determined in S285 based on availability, occupancy status, wind direction matching, runway length, and data confidence. This step only averages the baseline risk values ​​and does not reset another set of risk weights or risk mapping relationships.

[0414] The average risk value is only used to describe the overall risk level of the current set of candidate landing sites and does not directly replace the comprehensive priority or landing risk target calculated for individual candidate landing sites.

[0415] 12. Flight duration

[0416] The cumulative flight time of the UAV from the start of this mission to the current decision-making cycle is recorded as follows: .

[0417] This state variable is used to reflect the time progress of the missions performed by the drone, and together with the remaining energy state and equipment operating status, it describes the current endurance and equipment load.

[0418] Since the physical meanings, value ranges, and units of the twelve state variables are different, each state variable is normalized to the range of 0 to 1 before being input into the DDPG model.

[0419] In one implementation, the landing point selection system pre-sets the corresponding effective range for each continuous state variable based on the deployed UAV model, operating environment, and mission scope, and normalizes it according to the corresponding effective range.

[0420] For the Term continuous state quantity Its normalization result can be expressed as:

[0421]

[0422] in:

[0423] Represents the normalized state quantity;

[0424] This represents the actual value of the current state variable;

[0425] This indicates the preset lower limit of the state variable;

[0426] This indicates the preset upper limit of the state variable.

[0427] When the actual value is lower than the preset lower limit, the normalization result can be set to 0; when the actual value is higher than the preset upper limit, the normalization result can be set to 1. When the preset upper limit of a certain continuous state quantity is equal to the preset lower limit, the state quantity is set to the preset normalization value to avoid the normalization denominator being 0.

[0428] The normalization upper and lower limits of the above state variables are preset by the performance range and mission environment of the adapted UAV, and the specific values ​​are not fixed in this implementation.

[0429] For the current emergency level, it can be normalized according to its level range of Level 0 to Level 3; for the current wind direction, it can be normalized according to the preset angle mapping rules; for the number of candidate landing points, average distance, and average risk value, it can be normalized according to the candidate point capacity, mission distance range, and risk value range configured by the corresponding system.

[0430] After normalization, a twelve-dimensional state vector is obtained. And use it as the input for the DDPG model's current online inference.

[0431] S340: When the current emergency level is normal, warning, or emergency, the twelve-dimensional state vector is input into the DDPG model to generate an eight-dimensional weight adjustment.

[0432] This implementation uses the Deep Deterministic Strategy Gradient Algorithm (DDPG) to establish a dynamic weight model.

[0433] The DDPG model is suitable for continuous action spaces. In this implementation, all required weight adjustments are continuous values; therefore, the DDPG model is used to output eight-dimensional continuous weight adjustments based on the twelve-dimensional real-time state.

[0434] The DDPG model includes an Actor network and a Critic network.

[0435] The Actor network is used to generate eight weight adjustments based on the current twelve-dimensional state vector.

[0436] The input to the Actor network is:

[0437]

[0438] The output of the Actor network is an eight-dimensional action vector:

[0439]

[0440] in:

[0441] Adjustment amount for basic qualification weighting;

[0442] This is the distance weight adjustment amount;

[0443] This is the runway length weight adjustment amount;

[0444] Adjustment amount for wind direction matching weight;

[0445] This is the adjustment amount for availability weight;

[0446] Adjustment amount for penalty weighting;

[0447] Adjustment amount for data confidence weights;

[0448] Adjust the weighting of emergency response types.

[0449] The above eight weight adjustments correspond one-to-one with the eight benchmark weights determined in S320.

[0450] During offline training, the reward function is used to evaluate whether the weight adjustment of the Actor network output is conducive to forming a safe, efficient and stable landing point decision.

[0451] The total reward of the current training process is denoted as... ,but:

[0452]

[0453] in:

[0454] Indicates a safety reward;

[0455] Indicates a reward for efficiency;

[0456] This indicates a stable reward.

[0457] 1. Safety Rewards

[0458] When the landing point selected based on the current weight meets the accessibility requirements, a positive safety reward can be given based on the degree to which the basic risk value of the selected candidate landing point is reduced relative to the preset reference risk value. The specific mapping relationship is pre-configured by the training model.

[0459] When a landing point that is unreachable due to insufficient range is selected based on the current weight, a negative reward of -10 is given:

[0460]

[0461] The efficiency reward is negatively correlated with the time required to reach the selected landing point. The shorter the arrival time, the higher the efficiency reward.

[0462] Stable rewards are used to limit drastic changes in weights between two adjacent decision cycles.

[0463] The landing point selection system calculates the change in weight between two consecutive decisions in the dynamic weight vector. The smaller the change in weight, the higher the stability reward; when the change in weight exceeds 0.1, a negative stability reward is given.

[0464] After offline training is completed, the trained dynamic weight model is deployed on the drone's onboard computing equipment.

[0465] During the actual flight of the UAV, the landing point selection system reacquires the latest twelve-dimensional state variables according to the preset inference cycle, completes the normalization to form the current state vector, and inputs the current state vector into the trained Actor network.

[0466] The Actor network outputs the eight-dimensional weight adjustment for the current decision period based on the latest state:

[0467]

[0468] The online inference phase primarily utilizes the trained Actor network to generate weight adjustments. The Critic network is mainly used during the offline training phase to evaluate the combination of states and actions and guide the optimization of the Actor network.

[0469] In one specific implementation, the dynamic weight model performs online inference every 10 seconds. Each inference regenerates an eight-dimensional weight adjustment based on the latest acquired flight status, environmental status, and candidate landing site set.

[0470] The 10-second period is the dynamic weight update cycle, and it is not required that the overall decision cycle of S100 to S800 be exactly the same as it. When the overall decision cycle is shorter than the dynamic weight update cycle, the most recently generated eight-dimensional dynamic weight is used between two adjacent dynamic weight updates.

[0471] When the current emergency level changes between two adjacent decision cycles, the baseline weight of the corresponding level is immediately reread and dynamic weight update is triggered; when the emergency level changes to Level 3, the system directly enters the Level 3 fast path and no longer calls DDPG online inference.

[0472] S350: When the current emergency level is normal, early warning or emergency, the eight-dimensional weight adjustment amount is combined with the eight-dimensional baseline weight corresponding to the current emergency level to obtain the eight-dimensional dynamic weight of the current decision cycle.

[0473] S320 has determined an eight-dimensional baseline weight vector based on the current emergency level, and S340 has generated an eight-dimensional weight adjustment amount through the trained Actor network based on the current twelve-dimensional state vector.

[0474] The first The eight-dimensional baseline weight vector corresponding to each decision cycle is denoted as:

[0475]

[0476] The Actor network in the first The eight-dimensional weight adjustment amount output in each decision cycle is denoted as:

[0477]

[0478] Each of the eight weight adjustment parameters corresponds one-to-one with one of the eight baseline weights. The landing point selection system applies each weight adjustment parameter to its corresponding baseline weight, forming an eight-dimensional weight vector to be normalized:

[0479]

[0480] Right now:

[0481]

[0482] in:

[0483] This indicates the first of the eight evaluation dimensions. item;

[0484] Indicates the first In the decision-making cycle, the first The weights to be normalized corresponding to each evaluation dimension;

[0485] Indicates the current emergency level. The baseline weight of each evaluation dimension;

[0486] This represents the weight adjustment amount output by the Actor network for this evaluation dimension.

[0487] The weight adjustment can be positive or negative. When the weight adjustment is positive, the importance of the corresponding evaluation dimension relative to the baseline weight is increased under the current specific flight and environmental conditions; when the weight adjustment is negative, the importance of the corresponding evaluation dimension relative to the baseline weight is decreased.

[0488] However, the increase or decrease mentioned is based on the baseline weight corresponding to the current emergency level. Therefore, the current emergency level is still used to define the overall decision-making tendency, and the DDPG model only continuously adjusts this tendency according to the current specific situation.

[0489] The above explanation is only used to clarify the relationship between the weight adjustment amount and the baseline weight, and does not imply that the specific value that the Actor network must output in a certain state is predetermined in this embodiment. The specific weight adjustment amount is generated by the trained Actor network based on the twelve-dimensional state vector.

[0490] Since the values ​​of the eight weights to be normalized may be arbitrary real numbers, the landing point selection system performs Softmax normalization on the eight weights to be normalized.

[0491] The initial dynamic weights obtained after Softmax normalization are denoted as ω̃. k,t ,but:

[0492]

[0493] in:

[0494] ω̃ k,t This represents the initial dynamic weight corresponding to the k-th evaluation dimension in the t-th decision-making cycle;

[0495] This represents the corresponding weight to be normalized;

[0496] Index representing the eight evaluation dimensions;

[0497] Represented by natural constant An exponential function with base 0.

[0498] After Softmax normalization, all initial dynamic weights are positive and satisfy the following conditions:

[0499]

[0500] Therefore, Softmax normalization converts the eight weights to be normalized into the initial weight allocation result.

[0501] This implementation also requires that the final dynamic weight of each item be no less than 0.02 and no more than 0.6, in order to avoid the weight of a certain evaluation dimension being too low and thus completely losing its function, or the weight of a certain evaluation dimension being too high and causing the landing point decision to rely too much on a single indicator.

[0502] The Softmax function can make the initial dynamic weights of each item positive and make their sum 1, but it cannot naturally guarantee that each weight is in the range of 0.02 to 0.6 under all input conditions.

[0503] Therefore, based on Softmax normalization, the landing point selection system also performs preset range constraint processing on the initial dynamic weights to ensure that the final dynamic weights satisfy:

[0504]

[0505] And simultaneously satisfy:

[0506]

[0507] in, This indicates that after normalization and preset range constraints are completed, the first... In the decision-making cycle, the first The final dynamic weights corresponding to each evaluation dimension.

[0508] The final dynamic weights are determined using boundary constraint processing.

[0509] The landing point selection system performs boundary constraint processing on the eight initial dynamic weights obtained from Softmax, so that the processing result simultaneously satisfies the lower limit of a single weight, the upper limit of a single weight, and the weight and constraint.

[0510] The result after boundary constraint processing satisfies 0.02≤ω k,t ≤0.6, and Σk=1..8ω k,t =1.

[0511] The boundary constraint processing can employ iterative boundary adjustment and residual weight redistribution, or other equivalent methods that enable the range of individual weights to be satisfied simultaneously with the weights and constraints. The specific processing method is determined based on the system configuration.

[0512] The boundary constraint processing method adopted is not limited to one, as long as it maintains a one-to-one correspondence between the eight weights and the eight candidate landing point evaluation features, and ensures that the final dynamic weight of each item is within the range of 0.02 to 0.6 and the sum of the eight final dynamic weights is 1.

[0513] In actual implementation, the weighting method used only needs to ensure that:

[0514] The number of weight dimensions remains at eight.

[0515] The eight weights correspond one-to-one with the eight candidate landing site evaluation features;

[0516] Each dynamic weight shall be no less than 0.02 and no more than 0.6;

[0517] The sum of the eight dynamic weights is 1; and

[0518] The processed weights should maintain the relative distribution relationship formed by the baseline weights and the Actor network adjustment as much as possible.

[0519] The above processing, while satisfying the upper and lower limits of weights and the weight sum constraints, tries to maintain the relative distribution relationship formed by the baseline weight corresponding to the emergency level and the adjustment amount of the Actor network, so that the reinforcement learning model can make continuous adjustments according to the current real-time state, while avoiding the concentration of dynamic weights on a single evaluation dimension.

[0520] After completing the above processing, the first The eight-dimensional dynamic weight vector for each decision cycle is:

[0521]

[0522] in:

[0523] This represents the dynamic weight of the basic qualification score;

[0524] Represents the dynamic weight of the distance factor;

[0525] Represents the dynamic weight of the runway length factor;

[0526] This represents the dynamic weight of the wind direction matching factor;

[0527] Represents the dynamic weights of the availability factor;

[0528] This represents the dynamic weight of the penalty factor.

[0529] Represents the dynamic weights of the data confidence factors;

[0530] This represents the dynamic weighting of the emergency type bonus.

[0531] The order of the eight dynamic weights is consistent with the order of the complete eight-dimensional feature vector composed of the seven numerical basic features formed by S200 and the emergency type bonus formed by S300 based on the current emergency level. It serves as the comprehensive scoring coefficient determined by combining the evaluation emphasis of each item and the predetermined original score range.

[0532] Within the current decision-making cycle, the landing site selection system generates a set of eight-dimensional dynamic weights. This set of dynamic weights is used collectively to evaluate all candidate landing sites in this round, without generating separate weights for each candidate landing site.

[0533] In other words, for the first Any candidate landing point within a decision cycle All use the same group The eight-dimensional feature vector of the candidate landing site itself Perform the corresponding calculations.

[0534] The differences in evaluation results among different candidate landing sites stem from differences in their basic qualifications, distance, runway length, wind direction matching, openness, occupancy, data confidence, and emergency landing site attributes, rather than from the system generating different weighted evaluation criteria for different candidate sites.

[0535] Using the same set of dynamic weights to evaluate the current set of candidate landing sites allows for horizontal comparison of all candidate landing sites in this round under the same emergency state and the same decision-making tendency.

[0536] When entering the next decision cycle, the landing point selection system reacquires the latest UAV status, environmental status, and candidate landing point set data, and redetermines the emergency level, baseline weight, and weight adjustment amount, thereby obtaining the new dynamic weights corresponding to the next decision cycle.

[0537] After completing S300, output the current emergency level; for non-urgent levels, output eight-dimensional dynamic weights; for urgent levels, output fast path identifiers; when the candidate set is empty, output the empty set status and send it directly to S700.

[0538] In S400, under normal, warning, and emergency levels, the comprehensive priority is calculated based on the complete eight-dimensional feature vector of each candidate landing point and the eight-dimensional dynamic weight of the current decision cycle, and accessibility verification is performed on the candidate landing points that meet the basic usage conditions. Under the urgent level, the comprehensive priority calculation is skipped, and a set of candidate landing points to be verified is formed based on the availability verification results formed by S200 and the minimum runway length condition, and accessibility verification is performed directly.

[0539] S400 includes comprehensive priority calculation and reachability verification. Comprehensive priority is used to characterize the relative priority of candidate landing points under the current emergency level, the real-time status of the UAV, and the environmental conditions; reachability verification is used to determine whether the UAV can reach the corresponding candidate landing point with a reserved safety margin.

[0540] S410: When the current emergency level is normal, warning, or emergency, calculate the overall priority of each candidate landing site.

[0541] The landing site selection system reads the seven numerical basic features generated by S200, and reads the emergency type bonus and the eight-dimensional dynamic weight of each candidate landing site generated by S300, in order to form a complete eight-dimensional feature vector of each candidate landing site and calculate the comprehensive priority.

[0542] For a candidate landing point i in the t-th decision period, its complete eight-dimensional feature vector is denoted as:

[0543]

[0544] in:

[0545] Indicates basic qualification score;

[0546] Represents the distance factor;

[0547] Indicates the runway length factor;

[0548] Indicates the wind direction matching factor;

[0549] Indicates availability factor;

[0550] Indicates the penalty factor for occupancy;

[0551] Indicates the confidence factor of the data;

[0552] This indicates an emergency type bonus.

[0553] The eight-dimensional dynamic weight vector for the current decision-making cycle is:

[0554]

[0555] The eight dynamic weights maintain a one-to-one correspondence with the eight candidate landing site features, and satisfy the following:

[0556]

[0557] All candidate landing sites within the current decision-making cycle are evaluated using the same set of eight-dimensional dynamic weights.

[0558] Candidate landing sites The overall priority is denoted as It is determined according to the following formula:

[0559]

[0560]

[0561] in:

[0562] This represents the overall priority of candidate landing point i in the t-th decision cycle;

[0563] This represents the weighted result corresponding to the basic qualifications;

[0564] This represents the weighted result corresponding to the distance;

[0565] This represents the weighted result corresponding to the runway length;

[0566] This indicates the weighted result corresponding to the wind direction matching;

[0567] This represents the weighted result corresponding to the current available state;

[0568] This indicates the weighted penalty corresponding to the occupancy status;

[0569] This represents the weighted result corresponding to the data confidence level;

[0570] This represents the weighted result corresponding to the emergency type attribute.

[0571] The basic qualification items reflect the official qualifications of the candidate landing site, as well as its navigation, communication, fire protection, and on-site support capabilities. The higher the emergency level, the lower the overall weight of the basic qualifications, reducing the system's reliance on standard qualifications and comfort conditions during emergencies.

[0572] The distance term is used for relative optimization among candidate landing points that meet the basic conditions of distance range or accessibility. The three-dimensional spatial distance in S200 is used for distance range determination and accessibility calculation; the distance factor obtained by exponential decay is used for relative ranking in S410. The closer the distance, the higher the distance factor, and the greater the positive contribution to the overall priority.

[0573] The runway length item is used to reflect the landing safety margin and taxiing space corresponding to the runway length. When the runway length is less than the minimum safe length, S230 sets the runway length factor to 0 and generates a runway length non-compliance mark, and the corresponding candidate landing point is directly excluded according to the basic usage conditions; among the candidate landing points that meet the minimum safe length, the runway length factor is used to reflect the runway tolerance difference.

[0574] The wind direction matching term reflects the degree of matching between the runway landing direction of the candidate landing point and the current wind direction. Its value ranges from -100 to 100: close to 100 under headwind conditions, close to 0 under crosswind conditions, and can be negative under tailwind conditions. Therefore, although the wind direction matching term is connected by a plus sign in the comprehensive priority formula, when the wind direction matching factor is negative, the corresponding weighted result naturally has a negative contribution, and there is no need to set a separate tailwind penalty term.

[0575] The availability factor is either 0 or 1. Candidate landing sites with an availability factor of 0 are directly excluded according to the S250 veto rule and do not enter the subsequent comprehensive comparison and reachability verification. Qualified candidate landing sites that enter the comprehensive priority comparison are usually all set to 1, and their weighted contributions are the same. They do not undertake the function of continuous ranking and are only retained to maintain the eight-dimensional evaluation interface and status expression.

[0576] The occupancy penalty factor reflects the adverse impact of candidate landing sites being occupied by other aircraft and waiting for their release on the current landing decision. The occupancy penalty item is a deduction item in the overall priority; no points are deducted for occupancy status when a candidate landing site is not occupied, and the longer the expected remaining time for release when occupied, the greater the occupancy penalty.

[0577] In one specific implementation, when a candidate landing point is occupied, the occupancy penalty is calculated based on the estimated remaining release time determined in S260, and the overall priority is reduced. When the estimated remaining release time reaches or exceeds the maximum acceptable waiting time, the occupancy penalty is set to the upper limit of the penalty. S410 does not calculate the estimated flight time corresponding to the candidate landing point, nor does it determine the actual occupancy waiting time. The occupancy status, estimated remaining release time, and validity of the candidate landing point are determined separately by S450 in conjunction with the estimated flight time excluding the occupancy waiting item to determine the waiting time and equivalent fuel consumption. The occupancy penalty is only used for ranking and does not change the open status of the candidate landing point, nor does it constitute a hard constraint on accessibility on its own.

[0578] The data confidence factor reflects the reliability and freshness of the candidate landing site data. The more reliable the data source and the closer the data update time is to the current moment, the higher the data confidence factor; when the data has not been updated for a long time, the data confidence factor decreases exponentially. A lower data confidence factor reduces the evaluation advantage of the corresponding candidate landing site, but does not automatically render it unusable.

[0579] The emergency type bonus is used to increase the ranking priority of emergency landing sites with emergency support capabilities in the corresponding emergency scenario. Whether a candidate landing site is an emergency landing site is determined by the emergency landing site attribute marker formed in S200, and whether the emergency type bonus is effective and its value is determined by the current emergency level determined in S300.

[0580] In one specific implementation of Level 2 emergency response, when the candidate landing site is an emergency landing site:

[0581]

[0582] When the candidate landing site is not an emergency landing site:

[0583]

[0584] In the current specific implementation, the 30 points are used as an emergency type bonus in the eight-dimensional feature system, and multiplied by the corresponding emergency type dynamic weight to participate in the comprehensive priority calculation. No additional 30 points are added after the comprehensive priority calculation is completed.

[0585] In this implementation, the basic qualification score, distance factor, runway length factor, and data confidence factor are mainly set to 0-100 points, the wind direction matching factor is set to -100-100 points, the availability factor is set to 0 or 1, the occupancy penalty factor is set to 0-50 points and is used as a deduction item, and the emergency type bonus is set to 0 or 30 points in a specific implementation of the emergency level.

[0586] In the comprehensive priority calculation, the eight features directly participate in the comprehensive score according to their respective original value ranges. Before and after the comprehensive priority calculation, the eight candidate landing point scoring features are no longer uniformly normalized.

[0587] Since the scoring scales for each item are not entirely the same, the eight dynamic weights are comprehensive scoring coefficients determined by combining the evaluation emphasis of each item and the original score range. They do not represent the dimensionless pure importance ratio that deviates from the original score range. The comprehensive priority is not limited to the range of 0 to 100. Within the same decision-making cycle, each candidate landing point is relatively ranked using the same feature definition, original scoring caliber, dynamic weights, and comprehensive priority formula, and the S700 is used to compare the priority difference between the proposed landing point and the current target landing point.

[0588] In order to make a hysteresis change determination between the proposed landing point and the current target landing point in S700, when the current target landing point is still included in the current candidate landing point set and the current feature data that meets the preset validity requirements can be obtained, the comprehensive priority of the current target landing point in this decision cycle is recalculated according to the same feature calculation method and the same set of dynamic weights as other candidate landing points.

[0589] If current status data that meets the preset validity requirements cannot be obtained for the current target landing point, the validity of each data item actually used in this cycle, including open status, minimum usage conditions, reachability recalculation, and comprehensive priority calculation, will be judged according to the existing or preset timeliness requirements of the corresponding data. When the data has not been updated, only the most recent data whose timestamp is still within the validity period of the data item, still corresponds to the current target landing point, and whose source, semantics, and dimensions are still applicable and have not been negated by updated data will be used; no uniform retention time limit will be set for different data items, and the comprehensive priority, comprehensive priority difference, or dynamic weight of the previous cycle will not be directly used. When all necessary scoring data are valid, the calculation will be recalculated according to the feature definition, dynamic weight, and S410 comprehensive priority formula of this decision cycle. If there is a clear result indicating that the target landing point is unavailable due to an open state, does not meet the minimum usage conditions, or is unreachable after recalculation based on valid current data, the current target landing point is determined to be technically failed. If all the hard conditions can be confirmed to be met but the comprehensive priority cannot be effectively calculated, the output state is that the current target comparison data is invalid and the technical conditions can be confirmed to be met, and no valid current target comprehensive priority is output. If the hard conditions themselves cannot be confirmed and there is no clear evidence of failure, the output state is that the current target comparison data is invalid and the technical validity cannot be confirmed, and no valid current target comprehensive priority is output. When the current target landing point is technically failed, it will no longer be used as the hysteresis comparison benchmark, and the proposed landing point formed in this decision cycle will enter the subsequent maneuver feasibility and range margin verification according to the existing S700 rules.

[0590] A higher overall priority indicates a higher relative priority of the candidate landing point within the current eight-dimensional dynamic evaluation system. However, the overall priority does not directly constitute a target switching instruction. Candidate landing points must undergo subsequent reachability verification, hierarchical screening, multi-target optimization, and target switching safety verification before they can be determined as the final target landing point.

[0591] After completing S410, the landing site selection system outputs the following for the current decision cycle: Normal, Warning, and Emergency levels.

[0592] Overall priority of each candidate landing site;

[0593] The overall priority of the current target landing point when current feature data can be obtained;

[0594] Availability verification results for each candidate landing site;

[0595] The runway lengths of the candidate landing sites do not meet the requirements of the markings;

[0596] The set of valid candidate landing sites to participate in subsequent reachability calculations.

[0597] The overall priority of each candidate landing site continues to be used in S500 and S700. The availability verification result and the runway length not meeting the requirements are used to exclude candidate landing sites that do not meet the basic conditions. The set of valid candidate landing sites enters S420 and its subsequent processing.

[0598] When the current emergency level is Level 3, the landing site selection system skips S410 and does not generate a comprehensive priority for site selection or hysteresis judgment; it forms a set of candidate landing sites for accessibility verification based on the availability verification results formed in S200 and the minimum runway length condition, and directly enters the phased energy consumption, flight time and accessibility verification in S420 to S460.

[0599] S420 determines the equivalent phased flight parameters for reaching each candidate landing point based on the spatial relationship between the UAV's current position and the candidate landing points, as well as the corresponding flight speed parameters.

[0600] For normal, warning, and emergency landing levels, S420 receives the set of valid candidate landing sites output by S410; for extremely urgent levels, S420 receives the set of candidate landing sites to be verified, determined based on the availability verification result generated by S200 and the minimum runway length condition. The candidate landing sites entering S420 will be collectively referred to as candidate landing sites to be verified.

[0601] For candidate landing sites to be verified The landing site selection system reads the following data:

[0602] Current altitude of the drone ;

[0603] Candidate landing site altitude ;

[0604] Horizontal spherical distance from the current position of the drone to the candidate landing site ;

[0605] Three-dimensional spatial distance from the current position of the drone to the candidate landing site ;

[0606] Current ground speed of the drone ;

[0607] Preset climb speed ;

[0608] Preset descent speed ;

[0609] When the flight control module is able to provide the actual level flight segment length, it also reads the actual level flight segment length.

[0610] The current altitude and ground speed of the UAV are provided by the flight control interface; the landing surface altitude of the candidate landing point is provided by the candidate landing point data; and the horizontal spherical distance and three-dimensional spatial distance are obtained by S220.

[0611] Before calculating the net altitude difference, the current altitude of the UAV and the altitude of the candidate landing surface are converted to the same altitude reference so that the two can be directly subtracted.

[0612] The preset climb rate and preset descent rate are stage speed parameters adapted to the current drone model. They are used to calculate the duration of the corresponding stage and are provided by the flight control module according to the current flight mode and drone performance parameters, or preset according to the flight performance parameters of the deployed drone. In the duration calculation, both the preset climb rate and preset descent rate use positive speed values.

[0613] Candidate landing sites The difference between the altitude and the current altitude of the drone is recorded as:

[0614]

[0615] in:

[0616] Indicates the drone's current altitude and candidate landing sites The net height difference between landing surface heights, in meters;

[0617] Indicates candidate landing sites The landing surface height, in meters;

[0618] This indicates the drone's current altitude, in meters.

[0619] Determine the climbing height difference based on the net height difference:

[0620]

[0621] in, This indicates that the drone has reached the candidate landing site. The non-negative elevation gain that needs to be overcome.

[0622] The difference in descent altitude is denoted as:

[0623]

[0624] when hour, This indicates the existence of a descent phase, where the amount of descent is the absolute value of the difference in descent height.

[0625]

[0626] In this embodiment, the horizontal spherical distance from the current position of the UAV to the candidate landing point calculated in S220 is used as the estimated value of the level flight distance during the cruise segment, that is:

[0627]

[0628] in:

[0629] Indicates candidate landing sites The corresponding cruising distance, in meters;

[0630] This represents the horizontal spherical distance calculated by S220 based on the latitude and longitude of the UAV and the candidate landing points.

[0631] If the flight control system has already generated the actual flight path from the current position to the candidate landing point, and can obtain the length of the level flight segment within that path, then the actual level flight segment length is used as... When the actual level flight segment length is not obtained, the horizontal spherical distance is used as an estimate.

[0632] when:

[0633]

[0634] At that time, the duration of the ascent segment is determined based on the difference in elevation and the preset ascent speed.

[0635]

[0636] in:

[0637] Indicates the duration of the ascent phase, in seconds;

[0638] This indicates the climb rate used in the drone's endurance estimation, and its value is greater than 0, expressed in meters per second.

[0639] The duration of the cruise segment is determined based on the level flight distance of the cruise segment and the effective cruise ground speed used for range estimation in this decision cycle:

[0640]

[0641] in:

[0642] This indicates that the drone has reached the candidate landing site. The corresponding cruise segment duration, in seconds;

[0643] This indicates the cruising distance in meters.

[0644] This represents the effective cruising ground speed used for range estimation in this decision-making cycle, expressed in meters per second.

[0645] The predicted cruise ground speed provided by the flight control module based on the current flight status is used first; if no predicted cruise ground speed is provided, the current ground speed is used. When the used ground speed is not greater than the preset minimum effective speed, the corresponding reason is recorded, the validity of the phase parameter of the corresponding candidate landing point is marked as invalid, no effective cruise segment duration is formed, and the invalidation mark of the phase parameter and the reason are sent to S430.

[0646] when At that time, the duration of the descent segment is determined based on the absolute value of the descent altitude and the preset descent speed.

[0647]

[0648] in:

[0649] Indicates the duration of the descent segment, in seconds;

[0650] This indicates the descent speed used in the drone's endurance estimation, and its value is greater than 0, expressed in meters per second.

[0651] After completing S420, for each candidate landing site to be verified Landing point selection system output:

[0652] Net height difference ;

[0653] Climbing height difference ;

[0654] Descent height difference ;

[0655] cruise phase level flight distance ;

[0656] Climb duration ;

[0657] cruise duration ;

[0658] Duration of descent ;

[0659] Climb rate, effective cruising ground speed, and descent rate are used for range estimation;

[0660] Validity marker for stage parameters.

[0661] The aforementioned stage geometric parameters, stage durations, and stage parameter validity flags are sent to S430. Only when a stage parameter validity flag is valid is the stage duration available for subsequent aggregation by S450; if a stage parameter validity flag is invalid, S430 generates an energy consumption calculation invalid flag based on the flag and transmits the energy consumption calculation invalid flag to S450.

[0662] The S430 calculates energy consumption during the climb, cruise, and descent phases based on parameters for each flight phase.

[0663] S420 has determined the climb difference, cruise distance, descent difference, and duration of each flight phase for each candidate landing site to be verified.

[0664] S430 further reads energy consumption calculation parameters such as the number of currently operating engines, the current mass of the UAV, climb efficiency, fuel energy density, cruise fuel consumption rate, and descent fuel consumption rate, and calculates the energy consumption required for the UAV to reach the candidate landing point during the climb, cruise, and descent phases based on the stage parameters and energy consumption calculation parameters.

[0665] S430 first reads the stage parameter validity flag output by S420. When the stage parameter validity flag is invalid, it records the specific reason transmitted by S420, generates an invalid energy consumption calculation flag for the candidate landing point, does not perform subsequent stage energy consumption calculations, does not output the valid cruise segment basic energy consumption to S440, and sends the invalid energy consumption calculation flag and reason to S450; only when the stage parameter validity flag is valid, does it continue to perform the following verification of the number of currently operating engines and energy consumption calculation parameters.

[0666] Before performing energy consumption calculations during the execution phase, the number of currently operating engines is verified. When the number of currently operating engines is 0, the powered endurance calculation of this main implementation method is terminated according to the "all power failures" branch described later, and the cruise fuel consumption rate and descent fuel consumption rate are not required. When the number of currently operating engines is greater than 0, the validity of the UAV's current mass, climb efficiency, fuel energy density, cruise fuel consumption rate, and descent fuel consumption rate are verified. If any parameter to be verified is missing, exceeds the preset valid range, or its dimensions cannot be unified, the corresponding reason is recorded, an invalid energy consumption calculation flag is generated, and the output of the valid basic energy consumption results of the candidate landing point to the S440 wind field correction node and the S450 summary node is stopped.

[0667] In this main implementation, a fuel-powered UAV is used as an example, and the energy consumption of each flight phase is uniformly converted into fuel consumption mass. The term "phase energy consumption" as used below refers to the fuel consumption mass of the corresponding flight phase, so as to be directly compared with the UAV's current remaining fuel. For electrically powered UAVs, a phase electrical energy model matching its power system can be used to determine the corresponding energy consumption. The parameters and calculation relationships used are pre-configured according to the specific power system.

[0668] For candidate landing point i, the basic energy consumption for the three flight phases is denoted as follows:

[0669] Energy consumption during the ascent phase ;

[0670] Cruise segment energy consumption ;

[0671] Energy consumption during descent .

[0672] The above three energy consumption items use the same fuel mass unit.

[0673] When candidate landing sites When the landing surface altitude is higher than the current altitude of the drone, the landing point selection system estimates the corresponding fuel consumption during the climb based on the increased gravitational potential energy during the climb.

[0674] when:

[0675]

[0676] At that time, the energy consumption during the climb phase is:

[0677]

[0678] in:

[0679] This indicates that the drone has reached the candidate landing site. Required fuel consumption for the climb phase;

[0680] This indicates the difference in elevation, expressed in meters.

[0681] This indicates the current mass of the drone, in kilograms.

[0682] This represents the acceleration due to gravity, measured in meters per second squared.

[0683] This represents the climb efficiency, which is a dimensionless parameter.

[0684] This indicates the fuel energy density, expressed in joules per kilogram.

[0685] The product of the climb altitude difference, the current mass of the UAV, and the gravitational acceleration represents the increase in gravitational potential energy required to overcome the climb altitude difference; after dividing by the climb efficiency and fuel energy density, the corresponding fuel consumption mass is obtained, with the unit being kilograms.

[0686] In one specific implementation, the climbing efficiency is taken as:

[0687]

[0688] This parameter indicates that 70% of the energy released by the fuel is used to form the corresponding gravitational potential energy, while the remainder is consumed by engine efficiency losses, transmission losses, aerodynamic drag, and other flight losses.

[0689] when:

[0690]

[0691] At this time, the drone does not need to perform a climb based on the net altitude difference between the starting point and the candidate landing point, therefore:

[0692]

[0693] drone quality The current mass is calculated using the current decision cycle and determined based on the UAV model parameters, current payload status, and remaining fuel status. This mass is used as a fixed approximation within the current decision cycle, and is re-determined based on the updated status in the next decision cycle.

[0694] The cruise segment energy consumption is used to characterize the amount of fuel consumed by the UAV as it approaches the candidate landing point from its current position along the level flight direction.

[0695] S420 has identified candidate landing sites. Corresponding cruising distance:

[0696]

[0697] And the duration of the cruise segment:

[0698]

[0699] Assume the fuel consumption rate per unit time of a single engine during cruise is The number of engines currently operating normally is: Then the candidate landing sites The corresponding cruise segment energy consumption is:

[0700]

[0701] Substituting the cruising time into the equation, we get:

[0702]

[0703] in:

[0704] Indicates fuel consumption during the cruising segment;

[0705] This indicates the cruising distance in meters.

[0706] This indicates the effective cruising ground speed determined by S420, in meters per second.

[0707] This indicates the cruise fuel consumption rate per unit time for a single working engine under the current number of engines, flight load, speed, and fault degradation state. This fuel consumption rate does not represent the total cruise fuel consumption rate of the entire aircraft.

[0708] This indicates the number of currently operating engines.

[0709] Before calculation, convert the cruise segment duration and cruise fuel consumption rate to matching time units; when the cruise segment duration is in seconds and the cruise fuel consumption rate is in kilograms per second, the unit of the obtained cruise segment fuel consumption is kilograms.

[0710] The single working engine's cruising fuel consumption rate The fuel consumption rate can be directly provided by the engine monitoring interface. When provided directly, the parameter semantics should be clearly defined as the current operating condition cruise fuel consumption rate of a single working engine, and it should meet the requirements of numerical validity and dimensional uniformity. If the engine monitoring interface does not directly provide the fuel consumption rate of a single engine, it can be determined based on the engine performance parameters and the current number of engines, load, speed, and fault degradation status. When an engine fault causes changes in the current number of engines, load, speed, or fault degradation status, the corresponding current operating condition cruise fuel consumption rate of a single engine should be re-obtained or determined, and not replaced by a fixed rated cruise fuel consumption rate. If only the total cruise fuel consumption rate of the entire engine is provided, or if the fuel consumption rate of a single engine or the entire engine cannot be confirmed, or if a valid current operating condition cruise fuel consumption rate of a single engine cannot be formed from either source, an invalid energy consumption calculation flag should be generated, and this input should not be multiplied by the current number of working engines.

[0711] The base cruise fuel consumption rate for S440 wind field correction corresponds to the same base cruise condition as the reference cruise airspeed. This base cruise fuel consumption rate already reflects the effects of the current number of engines, load, speed, and fault degradation status, but does not include the airspeed deviation caused by the current wind field. The resulting base fuel consumption for the cruise segment already includes the aforementioned power state effects. S440 only performs airspeed correction for the current wind field based on this base value, without readjusting the number of currently operating engines or repeatedly adding the aforementioned power state correction.

[0712] Number of currently operating engines The real-time status data acquired using S100 only includes engines that are operating normally and providing flight power.

[0713] When the number of currently operating engines is greater than 0, calculations are performed according to the powered cruise and reduced fuel consumption formulas of this step. When the number of currently operating engines is 0, it is identified as a failure of all power units, the cause of all power unit failures is recorded, and a state of inapplicability of this main implementation method and termination of powered range calculation is generated; cruise fuel consumption rate and reduced fuel consumption rate are not required, the aforementioned powered cruise and reduced fuel consumption formulas are not used, effective basic energy consumption, effective total energy consumption, or effective reachable candidate set are not formed, and the effective results under this main implementation method are not passed to S440, S450, S460, and subsequent ordinary S500 to S800 selection chains, and the subsequent processing of this decision cycle within this main implementation method is terminated. The failure of all power units and the emergency level 3 are independent judgment dimensions; when the emergency level is 3 and at least one engine is still operating, the existing emergency path and powered range verification continue to be executed.

[0714] When the landing surface altitude of candidate landing point i is lower than the current altitude of the UAV, the UAV needs to perform a descent flight.

[0715] The S420 has determined the duration of the descent segment based on the absolute value of the descent altitude and the preset descent speed:

[0716]

[0717] Let the fuel consumption rate per unit time of a single working engine in the descent state be qdescent.

[0718] Then when At that time, the energy consumption during the descent phase is:

[0719]

[0720] Substituting the duration of the descent segment, we get:

[0721]

[0722] in:

[0723] This indicates the fuel consumption during the descent phase;

[0724] Represents a non-positive signed difference in descent, the absolute value of which represents the actual amount of descent, in meters;

[0725] This indicates the magnitude of the descent speed determined by S420, and its value is greater than 0, with the unit being meters per second;

[0726] This indicates the rate of fuel consumption reduction per unit time for a single working engine in the current descent state, as used in the formula of this article. This rate of fuel consumption does not represent the total rate of fuel consumption reduction for the entire engine.

[0727] This indicates the number of currently operating engines.

[0728] when:

[0729] At that time, there is no descent phase caused by the net height difference, therefore During the descent phase, the drone can reduce engine thrust, so the fuel consumption rate of a single engine during descent is lower than the cruise fuel consumption rate.

[0730] In one specific implementation, the reduced fuel consumption rate is set to 60% of the cruising fuel consumption rate, that is:

[0731]

[0732] Substituting this into the energy consumption formula for the descent phase, we get:

[0733]

[0734] This relationship is used to reflect that fuel consumption per unit time is lower than during cruising after reducing throttle during the descent phase, and is only one specific implementation method, not limited to a unique fixed relationship. The reduced fuel consumption rate adopts the same single-engine caliber and validity requirements as the cruising fuel consumption rate: the engine monitoring interface can directly provide a semantically clear, numerically valid, and dimensionlessly consistent single-engine current descent fuel consumption rate; if not directly provided, it can be determined based on engine performance parameters and current power status, or in the above specific implementation method, based on 60% of the current cruising fuel consumption rate. If only the total reduced fuel consumption rate of the entire engine is provided, the single-engine or whole-engine caliber cannot be confirmed, or a valid single-engine reduced fuel consumption rate cannot be formed, an invalid energy consumption calculation mark is generated, and this input is not multiplied by the number of currently operating engines.

[0735] After calculating the energy consumption for the climb, cruise, and descent phases, the landing point selection system adds up the fuel consumption for the three phases to obtain the baseline energy consumption without considering the influence of wind fields:

[0736]

[0737] in:

[0738] This represents the base fuel consumption for the drone to reach candidate landing point i.

[0739] This indicates the fuel consumption during the climb phase;

[0740] Indicates fuel consumption during the cruising segment;

[0741] This indicates the fuel consumption during the descent phase.

[0742] Based on the net altitude difference determined by S420, at most one of the fuel consumption during the climb and descent phases will be positive; the fuel consumption during the cruise phase will be determined based on the cruising distance.

[0743] Basic energy consumption The impact of wind field on airspeed, air resistance, and fuel consumption has not yet been included. The next step will be to calculate drag-corrected energy consumption based on the current wind speed, wind direction, and flight heading corresponding to the candidate landing point.

[0744] After completing S430, output the energy consumption of each stage, the validity flag of the energy consumption calculation and the corresponding invalidation reasons, for use by S440 and S450.

[0745] S440 determines the energy consumption correction amount corresponding to the wind field impact based on the current wind speed, wind direction, and the expected flight course corresponding to the candidate landing point.

[0746] S430 has determined the baseline fuel consumption for the cruise segment without considering the wind field effect for candidate landing point i and generated an energy consumption calculation validity marker; only when the marker is valid, S440 continues to determine the energy consumption correction amount corresponding to the wind field effect.

[0747] When any of the parameters is missing, exceeds the preset effective range, or the dimensions cannot be unified, the corresponding reason for the missing parameter, exceeding the effective range, or the inconsistency of dimensions is recorded, an invalid wind field correction calculation mark is generated, and the output of the valid wind field correction result of the candidate landing point to the S450 aggregation node is stopped.

[0748] Determine the expected flight heading corresponding to the candidate landing sites.

[0749] For candidate landing point i, before the actual flight path is obtained, the landing point selection system will determine the expected flight heading by the azimuth angle directly pointing from the current position of the UAV to the candidate landing point, and record it as:

[0750]

[0751] in, It is expressed in azimuth, with true north as 0°, and values ​​ranging from 0° to 360° in a clockwise direction.

[0752] Once the actual flight path has been obtained, the wind field projection parameters are determined according to the heading information corresponding to that path.

[0753] S440 uses the meteorological wind direction and wind vector direction determined by S240. When velocity vector projection is involved, the wind vector direction representing the actual direction of air blowing is used; when the data source directly provides the wind vector direction, the 180° conversion is not performed repeatedly.

[0754] Determine the angle between the wind direction and the expected flight course.

[0755] Based on the expected flight heading Hflight,i corresponding to the candidate landing point i and the wind vector direction Wvec, determine the minimum azimuth angle between them:

[0756]

[0757] in:

[0758] ψ i This indicates the angle between the wind vector direction and the expected flight path of the drone;

[0759] The value range is 0° to 180°.

[0760] Calculate the projected component of wind speed along the expected flight path.

[0761] Let the current wind speed be:

[0762]

[0763] Wind speed at candidate landing sites The corresponding projected component of the expected flight heading is denoted as:

[0764]

[0765] but:

[0766]

[0767] in:

[0768] : Represents the projected component of wind speed along the expected flight path, in meters per second;

[0769] : Indicates the current wind speed, in meters per second;

[0770] : Indicates the angle between the wind vector direction and the expected flight heading.

[0771] Airspeed is determined by the projection of ground speed and wind speed.

[0772] Under the one-dimensional projection relationship of the expected flight heading, the projections of ground speed, airspeed, and wind speed satisfy the following:

[0773]

[0774] Therefore, candidate landing sites The corresponding airspeed is:

[0775]

[0776] in:

[0777] : Indicates that the drone is flying toward the candidate landing site The corresponding airspeed at that time;

[0778] Vgnd: Represents the effective cruising ground speed determined by S420 for this decision cycle;

[0779] : Represents the longitudinal projection component of wind speed, with positive for downwind and negative for headwind.

[0780] Airspeed validity check

[0781] The landing point selection system reads at least the minimum allowable airspeed corresponding to the current UAV model; when the model configuration provides the maximum allowable airspeed, it also reads the corresponding maximum allowable airspeed and compares the calculated airspeed with the allowable airspeed range.

[0782] When the calculated airspeed meets the allowable airspeed conditions corresponding to the current aircraft type and fault state, the airspeed is used for subsequent wind farm energy consumption correction; if the conditions are not met, the current expected cruise state is marked as invalid.

[0783] When the flight control module or navigation module is able to provide a valid predicted cruise ground speed again, the cruise segment duration in S420 is updated, and the cruise segment base fuel consumption in S430, the airspeed and wind field energy consumption correction in S440, and the total energy consumption and total flight time in S450 are recalculated accordingly; if a valid predicted cruise state cannot be obtained, the flight conditions for the corresponding candidate landing point are marked as infeasible.

[0784] Determine the energy consumption correction amount corresponding to the wind field influence based on the air speed ratio.

[0785] In the range estimation model of this embodiment, the estimation relationship of cruise fuel consumption as a function of the cube of airspeed ratio is used to correct the base fuel consumption during the cruise segment for wind field.

[0786] Let the reference cruising airspeed of the UAV be:

[0787]

[0788] The reference cruise airspeed is provided by the flight performance parameters of the current UAV model and corresponds to the same basic cruise operating condition as the single-unit basic cruise fuel consumption rate used by the S430; the applicable airspeed range of the estimation relationship is pre-configured according to the model performance parameters.

[0789] Candidate landing sites The corresponding wind farm energy consumption correction is:

[0790]

[0791] in:

[0792] : Indicates the fuel consumption correction amount corresponding to the wind field impact;

[0793] This indicates the fuel consumption during the cruise phase without considering the impact of wind fields.

[0794] : Indicates flying towards the candidate landing site The corresponding airspeed at that time;

[0795] : Indicates the reference cruising airspeed.

[0796] Determine the cruise segment energy consumption after considering the influence of wind field.

[0797] Adding the base cruise energy consumption to the wind field energy consumption correction, we obtain the cruise energy consumption considering the wind field's influence:

[0798]

[0799] in, This indicates the fuel consumption during the cruise segment after taking into account the influence of wind fields.

[0800] When the fuel consumption of the basic cruise segment is non-negative, and both the airspeed used and the reference cruise airspeed are within the preset effective range, the fuel consumption of the cruise segment after considering the influence of the wind field remains non-negative.

[0801] Scope of wind field correction

[0802] The cubic wind field correction is calculated based on the baseline fuel consumption during the cruise phase and is not applied repeatedly to the climb and descent phases.

[0803] This implementation uses the wind speed and wind vector direction obtained in the current decision cycle as the equivalent constant wind field parameters corresponding to the candidate flight path.

[0804] After completing S440, output the expected flight heading, cruise segment energy consumption after wind field correction, and calculation validity markers for S450 to summarize.

[0805] S450 summarizes the total fuel consumption and estimated total flight time required for the UAV to reach each candidate landing site to be verified.

[0806] For candidate landing sites that are valid for verification, both the S430 energy consumption calculation validity marker and the S440 wind field correction calculation validity marker are valid. S450 reads the fuel consumption during the climb phase, the fuel consumption during the cruise phase after considering the wind field effect, the fuel consumption during the descent phase, and the duration of each phase determined by S420. It also reads the candidate landing site occupancy status, the estimated remaining release time trelease,i, and its validity marker, which are retained by S260.

[0807] The total fuel consumption and estimated total flight time are used by S460 to implement hard constraints on rangeability, and do not in themselves indicate that the candidate landing sites are already reachable.

[0808] When the S430 energy consumption calculation validity flag or the S440 wind field correction calculation validity flag is invalid, the effective total fuel consumption of the candidate landing point is not calculated, the summary calculation invalid flag is output, and the output of the effective summary result of the candidate landing point to the S460 reachability verification node is stopped; the summary calculation invalid state is distinguished from the unreachable state that does not meet the reachability constraints after normal calculation.

[0809] Summary of flight fuel consumption excluding occupancy and waiting items

[0810] In this main implementation, the energy consumption at each stage has been converted into fuel consumption mass. The cruise fuel consumption output by S440, taking into account the wind field effect, is:

[0811]

[0812] Therefore, the fuel consumption of the UAV when it reaches candidate landing point i, excluding waiting periods, is calculated as follows:

[0813]

[0814] Where Eflight,i represents the flight fuel consumption of the UAV when it reaches candidate landing point i, excluding occupancy and waiting items; Eclimb,i Indicates fuel consumption during the climb phase; E cruise_wind,i This indicates the fuel consumption during the cruise phase after taking into account the influence of wind fields; E descent,i This indicates the fuel consumption during the descent phase.

[0815] The cruise fuel consumption figure after considering the wind field effect already includes the base cruise fuel consumption and the wind field energy consumption correction, so these two items will not be added again. The wind field energy consumption correction can be positive, zero, or negative, but the cruise fuel consumption figure after considering the wind field effect and the flight fuel consumption excluding occupancy and holding items remain non-negative.

[0816] The flight fuel consumption, excluding occupancy and waiting items, is expressed in units of mass consistent with the current remaining available fuel of the UAV read by S460. For electrically powered UAVs, a stage-specific energy calculation model adapted to their power system can be used, and the energy consumption for each stage can be summarized according to this step; the specific stage-specific energy calculation parameters and model are determined based on the electric power system used. This main implementation method will continue to explain the summarization relationship of occupancy and waiting items using fuel consumption mass as an example.

[0817] Summary of estimated flight times excluding waiting periods

[0818] S420 has determined the climb, cruise, and descent durations corresponding to the candidate landing sites. The landing site selection system adds up the actual equivalent phase durations to obtain the estimated flight time excluding occupancy and waiting periods.

[0819]

[0820] Among them, T flight,i This represents the estimated flight time for the drone to reach candidate landing point i, excluding occupancy and waiting periods; t climb,i t cruise,i and t descent,i These represent the durations of the climb, cruise, and descent phases, respectively.

[0821] Based on the equivalent phased relationship of the net altitude difference determined by S420, at most one of the climb phase duration and the descent phase duration is greater than zero; the actual phase duration is added to the cruise phase duration to obtain the estimated flight time excluding occupancy and waiting items.

[0822] In this embodiment, the cruise duration is calculated based on the cruise level flight distance and the effective cruise ground speed determined in S420; the influence of the wind field on the power consumption required to maintain the ground speed is applied to the cruise fuel consumption by the airspeed change determined in S440.

[0823] When S440 triggers the predicted cruise ground speed update, S450 uses the recalculated phase duration and phase fuel consumption to form T. flight,i and E flight,i .

[0824] Determine the occupancy waiting time and its equivalent fuel consumption.

[0825] S450, based on the occupancy status retained by S260 and the estimated remaining release time t, release,i The validity flag of the candidate landing site determines the occupied waiting item. If a candidate landing site is not occupied, a waiting time t will be allocated. wait,i And occupancy waiting equivalent fuel consumption E wait,i All values ​​are set to 0; at this point, there is no need to obtain the fuel consumption rate per unit time corresponding to the occupancy waiting period.

[0826] If a candidate landing site is already occupied, and the estimated remaining release time is missing, expired, non-numerical, or negative, or cannot use the same time base as the current decision time, S450 records the corresponding reason passed by S260, marks the summary calculation of this candidate landing site as invalid, and does not form a valid E. total,i and T total,i Nor will it output a valid summary result to S460; the state is not considered an unattainable result after normal calculation.

[0827] If a candidate landing site is already occupied and the remaining estimated release time is valid, S450 will add the remaining estimated release time to the estimated flight time T excluding the occupancy waiting period. flight,i Compare the results and determine the waiting time using the following formula:

[0828]

[0829] Among them, t wait,i This indicates the equivalent additional waiting time caused by the occupancy. The estimated remaining release time is no greater than T. flight,i At that time, t wait,i =0; Expected remaining release duration is greater than T flight,i At that time, t wait,i The difference between the two is given. This relationship is used for reachability estimation and does not necessarily require the UAV to adopt a specific circling path, waiting airspace, or specific control maneuver.

[0830] When t wait,i When E is 0, wait,i The value is set to 0, and the summary calculation of this candidate landing site is not invalidated due to the missing or invalid fuel consumption rate per unit time corresponding to the waiting period.

[0831] When t wait,iWhen the value is greater than 0, the equivalent unit time fuel consumption rate q of the entire aircraft corresponding to the occupancy waiting time can be obtained from a valid data interface or a performance configuration compatible with the current drone model. wait,i In this case, the equivalent fuel consumption for occupancy and waiting is determined according to the following formula:

[0832]

[0833] Where, q wait,i The equivalent unit-time fuel consumption rate for the entire machine is adapted to the occupancy and waiting equivalent operating conditions and the current power status. The dimension is fuel mass / time, and it is no longer multiplied repeatedly according to the number of engines. The above source is a conditional, non-unique value method and does not imply that a specific module in the existing system will necessarily output this parameter.

[0834] When t wait,i Greater than 0, while q wait,i If the value is missing, expired, non-numerical, non-finite, or non-positive, or if its dimensions or corresponding equivalent operating conditions do not match, record the corresponding reason, mark the occupied waiting fuel consumption and the summary calculation of this candidate landing point as invalid, and do not form a valid E. total,i and T total,i Nor will it output a valid summary result to S460; the state is not considered an unattainable result after normal calculation.

[0835] When the occupancy waiting item calculation is valid, the final total fuel consumption and estimated total flight time are generated for S460 to read:

[0836]

[0837]

[0838] E total,i and T total,i These represent the final total fuel consumption and the estimated total flight time, respectively, which have been included in the effective occupancy waiting items. The occupancy waiting time and equivalent fuel consumption are only aggregated once in this step. S460 directly reads the final total and does not repeatedly add up the occupancy waiting items.

[0839] Boundary of final total calculation

[0840] This step calculates the estimated total flight time based on the equivalent phase parameters determined by S420. For non-occupancy waiting items such as approaches and detours, those already reflected in calculations based on actual segment lengths, phase durations, or existing routes are not included again. Items confirmed as essential in the current candidate landing site assessment, whose energy or time requirements can be determined through existing segments, phase parameters, or effective data interfaces, are included in the corresponding segment or phase requirements. Items that are only potential, not yet actually triggered, or not yet confirmed as essential are not fictitious additional segments, maneuvers, or consumption amounts.

[0841] If a candidate landing site is still occupied and its estimated arrival time has not yet been released based on the estimated flight time, the occupancy waiting time and its effectively quantifiable equivalent fuel consumption are explicitly included in the final estimated total flight time and total fuel consumption, according to the above relationship. The occupancy waiting item is only included once in this step. For additional items that are confirmed to be necessary for the current route but are neither included in existing segment or phase requirements nor can effective energy or time requirements be obtained, the corresponding reason for data loss is recorded, and the summary calculation of this candidate landing site is marked as invalid, thus not forming a valid E. total,i and T total,i Furthermore, it does not output valid summary results to S460; the 20% safety reserve of S460 does not replace the necessary requirement for the missing data. This step does not add specific maneuver models such as approach or detour.

[0842] The estimated total flight time determined by S450 is used for the initial assessment of the reachability of candidate landing sites and does not replace the feasibility verification of the target switching maneuver performed by S700 for the proposed landing site.

[0843] Validation of summary results

[0844] Before outputting the summary results, the landing point selection system verification E is performed. flight,i T flight,i t wait,i E wait,i And finally E total,i and T total,i The system checks whether the applicable validity conditions are met and verifies whether the total fuel consumption and remaining available fuel, as well as the estimated total flight time and remaining flight time, use consistent units of measurement. If negative values, non-numerical values, non-finite values, inconsistent units of measurement, or invalid pending inputs exist, the corresponding reasons are recorded, an invalid summary calculation flag is output, and the output of valid summary results for the candidate landing point to the S460 reachability verification node is stopped.

[0845] After completing S450, output the total fuel consumption, estimated total flight time, occupancy waiting items, and summary validity flags for S460 to perform dual reachability verification.

[0846] S460 performs dual hard reachability checks on each valid candidate landing point based on the UAV's current remaining energy and remaining flight time, using both energy consumption and time constraints, to form a reachability candidate set.

[0847] S450 has determined the total fuel consumption, including valid occupancy wait items without repeatedly adding the wait items, for each valid candidate landing site to be verified based on the aggregated calculation:

[0848]

[0849] And the estimated total flight time:

[0850]

[0851] The S460 reads the latest remaining energy and remaining flight time corresponding to the current decision cycle and verifies the update time, numerical validity, and dimensional consistency of the corresponding data. Only candidate landing sites that simultaneously meet the total fuel consumption constraint and the estimated total flight time constraint are determined as reachable candidate landing sites.

[0852] The reachability check is a hard constraint that can be vetoed; other evaluation results for candidate landing sites cannot compensate for the inability to reach them due to range limitations.

[0853] Obtain the current remaining energy of the drone

[0854] The current remaining energy of the drone is denoted as:

[0855]

[0856] The remaining energy amount is determined by the UAV health monitoring module based on the remaining energy data obtained by S100 that is compatible with the UAV's energy form.

[0857] In this main embodiment, taking a fuel-powered drone as an example, the remaining energy is represented by the current remaining fuel mass, and uses the same mass unit as the total fuel consumption obtained in S450.

[0858] For electrically powered drones or drones using other energy forms, a stage energy consumption model adapted to the corresponding power system is adopted, and the total energy consumption corresponding to the candidate landing point and the current remaining energy are given the same dimensions before constraint judgment is implemented.

[0859] Obtain the current remaining flight time of the drone

[0860] Record the remaining flight time of the drone as:

[0861]

[0862] In this main implementation, the health monitoring module outputs the remaining flight time that the current remaining fuel can support:

[0863]

[0864] And the remaining time that the batteries of critical onboard equipment can support:

[0865]

[0866] S460 takes the smaller of the two values ​​as the current remaining flight time:

[0867]

[0868] If either the propulsion fuel or the battery of the critical airborne equipment reaches a state where it can no longer support the flight mission, the effective remaining flight time of the drone will be limited; the specific prediction model used inside the health monitoring module to determine the remaining support time of the two items is not included in the calculation of this step.

[0869] Check remaining battery life parameters

[0870] The S460 checks whether the remaining energy and remaining flight time are the latest valid values ​​corresponding to the current decision cycle, and satisfies the following conditions: remaining energy is greater than 0, remaining flight time is greater than 0, total fuel consumption has the same dimension as remaining energy, and the expected total flight time and remaining flight time use the same time unit.

[0871] In cases where data is missing, non-numerical, exceeds the preset valid range, update time does not meet the requirements, or units cannot be unified, an invalid flag for the remaining range parameters and an invalid flag for the reachability calculation will be output. This situation will not be directly equated to unreachable after normal calculation.

[0872] Before applying safety factor constraints, verify the safety reserve criteria used in the health monitoring module's output data. The current remaining energy quantity E in this step... remaining and current remaining flight time T remaining All values ​​represent the complete current availability at the time of S460 execution, before the application of the 20% safety reserve in this step. Only when the semantics of the data source parameters clearly conform to the stated criteria are they substituted into the 0.8 double hard constraint described later. If the interface only provides a net value that has already deducted the same 20% safety reserve, even if it can be confirmed that the reserve has been deducted, the net value will not be substituted into the 0.8 formula again. Without adding a back-calculation or conversion model, the reason why it does not meet the interface requirements of this implementation method is recorded, and an invalid flag for the remaining range parameters and an invalid flag for reachability calculation are output, without forming a valid reachability candidate point result.

[0873] Set the safety factor for reachability verification.

[0874] Set the safety level:

[0875]

[0876] In one specific implementation:

[0877]

[0878] The safety factor is used to limit the planned fuel consumption and flight time for flying to candidate landing sites, and must not consume all of the UAV's current remaining endurance. After all currently necessary and effectively determinable energy and time requirements have been taken into account, the 20% is used to form an initial screening margin for candidate landing sites according to a predetermined double hard constraint of 0.8; this margin is used to address general uncertainties in the current estimate and does not replace necessary consumption that has been omitted or lacks data. 0.8 is applied only once in the initial screening of candidate landing sites in S460 and does not form the 80% net margin passed to S740; S740 performs a 10% endurance margin check based on the latest complete current available amount at its execution time, without multiplying by 0.8 again, and the 20% and 10% are not added or continuously subtracted.

[0879] Implement hard constraints on total fuel consumption

[0880] The fuel consumption constraint for candidate landing point i is defined as follows:

[0881]

[0882] The fuel consumption constraint adopts a strict less than relationship; when the total fuel consumption reaches exactly 80% of the current remaining fuel, the fuel consumption constraint result is 0.

[0883] Implement hard constraints on total estimated flight time

[0884] The flight time constraint result for candidate landing point i is defined as follows:

[0885]

[0886] The flight time constraint also adopts a strict less than relationship; when the expected total flight time reaches exactly 80% of the current remaining flight time, the flight time constraint result is 0.

[0887] Determining the reachability results of candidate landing sites

[0888] Define the reachability tagged candidate landing point i as:

[0889]

[0890] When both the fuel consumption constraint and the flight time constraint are 1, the reachability flag is 1, and the candidate landing point passes the reachability check; when either constraint is 0, the reachability flag is 0, and the candidate landing point fails the reachability check.

[0891] Candidate landing sites that fail the reachability check will be excluded.

[0892] For candidate landing sites that fail the reachability check at the normal, warning, and emergency levels, the original integrated priority generated by S410 is retained in the decision log. The effective integrated priority for subsequent screening is set to 0, and the candidate landing site is excluded from subsequent emergency level screening, multi-target selection, and effective ranking.

[0893] For urgent landings, candidate landing sites that fail the reachability check are directly marked as unreachable and excluded, and the overall priority is not set to 0.

[0894] Forming a reachable candidate set

[0895] The set of candidate landing sites to be verified for entering S460 is denoted as:

[0896]

[0897] Among them, the set of candidate landing sites to be verified under normal, warning and emergency levels comes from S410; the set of candidate landing sites to be verified under the emergency level comes from the availability verification results and runway length minimum condition verification results formed by S200, and the range estimation has been completed through S420 to S450.

[0898] Candidate landing points that are valid for reachability calculations and simultaneously pass the dual hard constraints are grouped into a reachability candidate set:

[0899] In this embodiment, the reachability verification includes performing hard energy consumption and hard time constraints on the total energy consumption and estimated total flight time required for the UAV to reach the candidate landing point, respectively. Only candidate landing points whose reachability calculations are valid and which pass both the hard energy consumption and hard time constraints are determined as reachable candidate results. The reachable candidate results correspond to the reachable candidate set and the candidate landing point information stored therein.

[0900]

[0901] All candidate landing sites in the reachable candidate set are retained together:

[0902] Candidate landing site identifiers;

[0903] Reachability markers;

[0904] Total fuel consumption;

[0905] Estimated total flight time;

[0906] Three-dimensional spatial distance;

[0907] Runway length;

[0908] Currently open;

[0909] Emergency landing point attributes.

[0910] For normal, warning, and emergency levels, the eight-dimensional feature vector, original comprehensive priority, effective comprehensive priority, and headwind deflection angle are further retained for subsequent screening data. For the extremely urgent level, the comprehensive priority for sorting is not required, but the three-dimensional spatial distance and emergency landing point attributes are retained for the S500 to perform 5km range screening and distance sorting.

[0911] The difference between candidate point initial screening redundancy and target switching margin

[0912] The S460 uses a 20% overall range redundancy to perform initial reachability screening of candidate landing sites; the S740 performs a secondary range safety confirmation based on the latest status and the additional maneuver costs of target switching when the proposed landing site is about to replace the current target landing site. The two processes operate at different stages and calculation points, and are not simply added together.

[0913] The case where the reachable candidate set is empty.

[0914] When there are no candidate landing sites that simultaneously satisfy both hard constraints:

[0915]

[0916] When there are no candidate landing points that are valid in the aggregate calculation and simultaneously pass the dual hard constraints of energy consumption and time, the reachable candidate set is marked as empty. The corresponding reasons for invalid upstream aggregation, invalid remaining endurance parameters, or unreachable status after normal verification are recorded, and a "no unique proposed landing point has been formed" status is generated. The safety factor of 0.8, availability conditions, minimum runway length, fuel consumption constraints, or estimated total flight time constraints are not automatically relaxed. The status, reasons, and current target landing point identifier are sent to S700.

[0917] After completing S460, output the reachable candidate set and the energy consumption and time verification results of each candidate landing point; the empty set and Neng=0 state are processed according to the aforementioned dedicated route.

[0918] S500 performs tiered screening of the available candidate set based on the current emergency level and determines the subsequent decision-making path.

[0919] S460 has formed the reachable candidate set for the current decision-making cycle:

[0920]

[0921] For normal, warning, and emergency levels, reachable candidate points carry data such as an eight-dimensional feature vector, original comprehensive priority, effective comprehensive priority, estimated total flight time, total fuel consumption, and basic risk value. For extremely urgent levels, reachable candidate points are not required to have a comprehensive priority and mainly carry three-dimensional spatial distance, emergency landing point attributes, and previous reachability and endurance verification results.

[0922] S310 has determined the current emergency level:

[0923]

[0924] The landing point selection system applies corresponding additional screening conditions to the reachable candidate set based on the current emergency level, and determines whether to continue into the S600 multi-target preferred path, or skip the S600 and adopt the emergency-level distance rapid sorting path.

[0925] S510: When the current emergency level is Level 0 (normal level), no additional filtering conditions are added to the reachable candidate set.

[0926] when:

[0927]

[0928] In this case, the reachable candidate set formed by S460 is directly used as the normal-level candidate set for further optimization:

[0929]

[0930] The term "normal level" means that no additional screening conditions are added, which means that no additional thresholds related to the emergency level are added on top of the availability, minimum runway conditions, and range reachability verifications already completed for S200 and S460.

[0931] When the candidate set for optimization at the normal level is not empty, it is sent to S600 to continue multi-objective optimization.

[0932] S520, when the current emergency level is warning level 1, is selected according to runway length and headwind angle.

[0933] when:

[0934]

[0935] At that time, based on the reachable candidate set, only candidate landing points with a runway length of not less than 800m and a headwind angle of not more than 60° are retained. For candidate landing point i, its runway length and headwind angle are denoted as: and .

[0936] The candidate set for early warning level is as follows:

[0937]

[0938] Among them, the headwind deflection angle θ i The same result determined by S240 for the corresponding runway and its directional landing direction, and transmitted along with the candidate landing point data, is directly read and reused. No further angle is recalculated in S520 based on meteorological wind direction, wind vector direction, or runway direction. Candidate landing points meet the screening criteria when the runway length is exactly 800m or the headwind angle is exactly 60°. The 500m in S230 is used for minimum takeoff and landing condition verification, and the 800m in S520 is used to further retain candidate landing points with greater runway margin under warning levels.

[0939] When the candidate set for early warning is not empty, it is sent to S600 to continue multi-objective optimization.

[0940] S530: When the current emergency level is Level 2, filtering is performed based on distance range and current availability.

[0941] when: At that time, candidate landing points whose three-dimensional spatial distance from the UAV's current position to the candidate landing point is no greater than 20km and are currently available are retained from the reachable candidate set. The corresponding conditions are:

[0942]

[0943]

[0944] The candidate set for emergency priority is as follows:

[0945]

[0946] The three-dimensional spatial distance uses the calculation results from S220. Since the candidate landing points in the reachable candidate set have already passed the availability verification, the availability status condition is used to clarify the status of candidate points that the emergency level screening depends on, without adding new availability scores.

[0947] The 30-point emergency type bonus for emergency landing sites has been determined in S300 based on the current emergency level and emergency landing site attributes as follows:

[0948]

[0949] And passed in S410:

[0950]

[0951] As the eighth feature participating in the comprehensive priority calculation, S530 will no longer add the 30 points to the comprehensive priority result.

[0952] When the emergency-level candidate set is not empty, it is sent to S600 to continue multi-objective optimization.

[0953] S540: When the current emergency level is Level 3 (Urgent), only emergency landing points within a preset distance range are retained and sorted directly by distance.

[0954] when: At that time, only candidate landing points with a three-dimensional spatial distance of no more than 5km and possessing the emergency landing point attribute are retained from the reachable candidate set. The corresponding conditions are:

[0955]

[0956]

[0957] The emergency candidate set is as follows:

[0958]

[0959] Ordinary airports, standard take-off and landing points, or temporary take-off and landing points, regardless of other evaluation characteristics, will not be included in the emergency candidate set if they are not marked as emergency landing points; emergency landing points with a three-dimensional spatial distance exceeding 5km will also not be included in the emergency candidate set.

[0960] Limited Express Class, sorted by distance from nearest to farthest.

[0961] When the emergency candidate set is not empty, the S600 multi-target Pareto optimization process is not invoked; instead, the candidates are directly sorted in ascending order of three-dimensional spatial distance. Let M be the number of candidate landing sites in the emergency candidate set, and the sorting result be denoted as: And satisfy: .

[0962] Among them, C (1) d represents the nearest emergency landing point to the drone's current location. (1) This indicates the three-dimensional spatial distance. When only one candidate landing point in the sorting results achieves the minimum three-dimensional spatial distance, that candidate landing point is determined as the only candidate result under the distance fast path and sent to S695 for aggregation to form the proposed landing point: .

[0963] After the unique candidate results are summarized in S695 to form the proposed landing point, they enter S700 to perform turning angle, altitude change rate and switching range margin verification. The non-dominated sorting, Pareto front extraction, entropy weight method, comprehensive proximity or manual Pareto selection in S600 are not performed, nor is the 10-point or 5-point hysteresis judgment based on the comprehensive priority difference performed.

[0964] When the current emergency level is extremely urgent, the candidate landing points that have been screened by level are used to form extremely urgent distance path results according to their distance; the extremely urgent distance path results are used to represent the unique candidate result with the smallest distance or the result with the same distance formed under the extremely urgent distance fast sorting path.

[0965] When the three-dimensional spatial distance between two or more emergency landing points is the smallest, the system outputs a candidate set of emergency-level distances and distance-level markers. It does not automatically add total fuel consumption, base risk value, data confidence, candidate landing point identifiers, or other secondary sorting conditions. It does not form a unique proposed landing point and forms a "no unique proposed landing point" status.

[0966] S550 determines the subsequent decision-making path based on the current emergency level and the results of the tiered screening.

[0967] when: And the corresponding candidate set for optimization: When not in flight, the candidate set to be optimized is sent to S600, and a Pareto candidate solution set is formed for the three minimization objectives of total expected flight time, total fuel consumption and basic risk value, and corresponding candidate results are formed according to automatic mode or manual mode.

[0968] When the current emergency level is extremely urgent and the emergency candidate set is not empty, the process does not proceed to S600; when a unique minimum distance result is formed, the unique minimum distance candidate result and its associated data are sent to S695 for aggregation, and S695 forms the proposed landing point; when there are results with the same distance, the status of "no unique proposed landing point formed", the set of the same candidate and the reason are sent to S695 for aggregation, and S695 outputs the result to S700 according to the corresponding result type.

[0969] Therefore, as the emergency level increases, the decision-making path changes from multi-objective optimization involving all reachable points to additional safety condition screening, nearby candidate point screening, and rapid sorting of emergency landing point distances for the most urgent level.

[0970] Case where the tiered filtering results are empty.

[0971] When the candidate set for the normal, warning, or emergency level is empty: Or, if the emergency candidate set is empty: .

[0972] Output the status that no candidate landing point meets the conditions under the current level and the classification screening result is marked as empty. Record the reason for not meeting the corresponding classification conditions. Do not generate a proposed landing point based on this empty set and form the status of "no unique proposed landing point formed". Do not automatically relax the runway length, headwind angle, distance range, availability, total fuel consumption, estimated total flight time or other safety constraints. Send the status, reason and current target landing point identifier to S700.

[0973] After completing S500, non-urgent processes output the candidate set to be optimized and proceed to S600; urgent processes output the distance path results and proceed to S695; empty set status is directly processed and proceeds directly to S700.

[0974] S600 performs multi-objective non-dominated sorting on the candidate set to be optimized, forms a Pareto candidate solution set, and generates automatic decision candidate results or manual selection results according to the current decision-making mode.

[0975] When the current emergency level is Level 0 (normal), Level 1 (early warning), or Level 2 (urgent), and the candidate set for optimization output by S500 is not empty, the landing point selection system receives the candidate set for optimization and performs multi-target optimization for each candidate landing point.

[0976] Let the set of candidates to be optimized in the current decision-making cycle be denoted as:

[0977]

[0978] in:

[0979] This represents the set of candidates to be optimized corresponding to the current emergency level;

[0980] Indicates the first One candidate landing site;

[0981] This indicates the number of candidate landing sites in the pool of candidates to be selected.

[0982] The candidate landing points in the candidate set to be optimized have been:

[0983] Current open status verification;

[0984] Minimum runway length verification;

[0985] Dual hard constraints: total fuel consumption and estimated total flight time;

[0986] Additional screening options corresponding to the current emergency level.

[0987] Therefore, S600 no longer determines whether candidate landing sites are basically reachable. Instead, it further balances the conflicting relationships between the expected total flight time, total fuel consumption, and basic risk value among candidate landing sites that have already met the aforementioned hard constraints.

[0988] S600 is not applicable to Level 3 express trains. For express trains, the distance is already sorted in S540 according to the three-dimensional spatial distance from near to far, and the corresponding results are sent to S695 for aggregation before entering S700.

[0989] S600 reads the current decision mode identifier, which indicates whether the current decision cycle uses an automatic decision mode or a manual decision mode. Urgent cases are not controlled by the decision mode identifier and are processed according to the fast path determined in S500.

[0990] The specific sources, switching conditions, and switching permissions for automatic and manual decision-making modes are determined based on the actual flight control and operation terminal configuration.

[0991] S610 sets the estimated total flight time required to reach the candidate landing site as the arrival time target.

[0992] For candidate landing point C in the candidate set to be optimized i The estimated total flight time determined by S450 is taken as the target arrival time for the candidate landing site, and denoted as:

[0993]

[0994] in:

[0995] f1(i) represents the candidate landing point C i The corresponding target arrival time;

[0996] T total,i This indicates that the drone will fly from its current location to candidate landing point C. i Total estimated flight time required.

[0997] The arrival time target directly adopts the estimated total flight time output by S450, and S600 does not add, delete or change its constituent items.

[0998] The arrival time objective is to minimize the time required for arrival, i.e.:

[0999]

[1000] All other things being equal, the shorter the estimated total flight time for a candidate landing site, the better its performance in terms of arrival time objective.

[1001] However, a shorter arrival time does not necessarily mean that the candidate landing site is superior to other candidate landing sites in terms of overall performance across the three objectives. For example, a candidate landing site that is closer to the target may have a higher landing risk; therefore, the proposed landing site cannot be determined directly based solely on the minimum value of f1(i).

[1002] S620, the total fuel consumption required to reach the candidate landing point is determined as the flight energy consumption target in this main embodiment.

[1003] For candidate landing site C i The total fuel consumption determined by S450 is taken as the flight energy consumption target for this candidate landing site, and denoted as:

[1004]

[1005] in:

[1006] f2(i) represents the candidate landing point C i The corresponding flight energy consumption target value;

[1007] E total,i This indicates that the drone will fly from its current location to candidate landing point C. i Total fuel consumption required.

[1008] In this main implementation, E total,i The fuel mass unit is used, consistent with the remaining fuel quantity; for UAVs using other forms of energy, the flight energy consumption target is the total energy consumption that matches its power system.

[1009] The flight energy consumption target is to minimize it, that is:

[1010]

[1011] All other things being equal, the lower the total fuel consumption required to reach a candidate landing point, the better the performance in terms of flight energy consumption.

[1012] All candidate landing sites in the candidate set have already met the S460's hard constraints on fuel consumption. Therefore, the role of the flight energy consumption target is not to re-evaluate the reachability of candidate landing sites, but rather to compare their energy consumption among multiple reachable candidate landing sites.

[1013] In other words:

[1014] Total fuel consumption in S460 is a veto-type accessibility constraint.

[1015] The total fuel consumption in the S620 falls under the flight energy consumption optimization objective in the Pareto multi-objective comparison.

[1016] The same total fuel consumption data has different technical applications in the two steps.

[1017] S630 reads the basic risk values ​​of candidate landing sites and determines the landing risk target.

[1018] S285 has already determined the basic risk value for each candidate landing site based on availability, occupancy status, wind direction matching, runway length, and data confidence. S600 does not repeat the risk weighting or re-execute the safety feature mapping.

[1019] For candidate landing point i in the candidate set to be optimized, its basic risk value R_i is directly determined as the landing risk target f_3(i).

[1020] f_3(i)=R_i

[1021] The smaller the baseline risk value, the lower the overall landing risk of the candidate landing site. Therefore, the landing risk objective is to minimize the risk.

[1022] The landing risk target and the integrated priority formed by S410 belong to different evaluation results. The integrated priority also considers basic qualifications, distance and emergency type bonuses, and is used for the hysteresis comparison of old and new targets in automatic mode; the landing risk target is only one of the three multi-objective optimization targets of S600.

[1023] After completing S610 to S630, a three-dimensional target vector F_i=[f_1(i),f_2(i),f_3(i)] is formed for the candidate landing point i, where all three targets are minimization targets.

[1024] S640, determine the Pareto dominance relationship between candidate landing sites.

[1025] For any two candidate landing sites in the candidate set to be optimized and The corresponding three-dimensional target vectors are as follows:

[1026]

[1027] and:

[1028]

[1029] Since all three objectives are better the smaller they are, when considering candidate landing points... Candidate landing sites are determined when the following two conditions are met. Pareto Dominance Candidate Landing Sites .

[1030] First, candidate landing sites for all three targets. All are no worse than candidate landing sites ,Right now:

[1031]

[1032] Second, on at least one target, candidate landing sites Strictly superior to candidate landing sites That is, there exists at least one target index. , so that:

[1033]

[1034] When both of the above conditions are met, it is denoted as:

[1035]

[1036] Among them, the symbol " "Indicates that the previous candidate landing site dominates the next candidate landing site."

[1037] The S650 uses the fast non-dominated sorting mechanism in NSGA-II to perform front-end stratification of candidate landing sites.

[1038] The landing site selection system adopts the fast non-dominated sorting mechanism in NSGA-II. Based on the aforementioned Pareto dominance relationship, the candidate landing sites in the candidate set to be optimized are divided into multiple non-dominated front layers.

[1039] This implementation uses the fast non-dominated sorting and crowding distance mechanism in NSGA-II to process existing candidate landing sites, without generating new landing site schemes through crossover, mutation, or multi-generation population iteration.

[1040] The candidate landing points that are not dominated by any other candidate landing points in the candidate pool to be optimized constitute the first front layer:

[1041]

[1042] in:

[1043] This represents the first frontier layer of the current decision-making cycle;

[1044] This indicates that there are no other candidate landing sites dominating the candidate landing site. .

[1045] Determine the Pareto candidate solution set

[1046] The candidate landing points in the first frontier layer are identified as the Pareto candidate solution set for the current decision cycle:

[1047]

[1048] in:

[1049]

[1050] This represents the Pareto candidate solution set for the current decision-making cycle.

[1051] S610 to S650 output the three-dimensional target vector of each candidate landing point, the non-dominated front layer, and the Pareto candidate solution set corresponding to the first front layer.

[1052] S660 calculates the congestion distance of each candidate landing site within each non-dominated front layer.

[1053] To characterize the distribution of candidate landing points in the three-dimensional target space, the landing point selection system calculates the congestion distance within each non-dominated front layer. The specific calculation process is illustrated below using the Pareto candidate solution set corresponding to the first front layer as an example; the second and subsequent front layers are calculated in the same manner.

[1054] The number of candidate landing sites in the Pareto candidate solution set is denoted as:

[1055]

[1056] For the candidate landing sites The total congestion distance is denoted as: ;

[1057] The Pareto candidate solutions are sorted according to the three objectives respectively;

[1058] For each candidate landing point in the Pareto candidate solution set, the corresponding three-dimensional target vector has been obtained:

[1059]

[1060] in:

[1061] f1(i) represents the arrival time target;

[1062] f2(i) represents the flight energy consumption target;

[1063] f3(i) represents the landing risk target.

[1064] All three objectives are minimization objectives.

[1065] The landing point selection system sorts the candidate landing points in the Pareto candidate solution set from smallest to largest according to the target value for each target dimension.

[1066] For the For each target, the sorted candidate landing points are denoted as follows:

[1067]

[1068] And satisfy:

[1069]

[1070] in:

[1071] When m=1, sort the flights by estimated total flight time from shortest to longest.

[1072] When m=2, sort by total fuel consumption from low to high;

[1073] When m=3, sort by basic risk value from low to high.

[1074] Determine the maximum and minimum values ​​for each target dimension.

[1075] For the Given a set of objectives, determine the minimum objective value within the current Pareto candidate solution set:

[1076]

[1077] And the maximum target value:

[1078]

[1079] in:

[1080] and These represent the shortest and longest estimated total flight times in the Pareto candidate solution set, respectively;

[1081] and These represent the minimum and maximum total fuel consumption, respectively.

[1082] and These represent the lowest and highest landing risks, respectively.

[1083] Because the three objectives have different dimensions:

[1084] The arrival time target is expressed in time units;

[1085] The flight energy consumption target in this main implementation method is expressed in fuel mass units;

[1086] Landing risk is assessed using a risk score.

[1087] Therefore, the original adjacent differences on the three targets cannot be directly added together.

[1088] This step uses the range of each objective:

[1089]

[1090] The adjacent differences in the target dimension are normalized so that the contributions of the three target dimensions to the congestion distance are all converted into dimensionless values.

[1091] Calculate the crowding distance contribution of non-boundary candidate landing points in a single target dimension.

[1092] For the One target, candidate landing sites. In the ascending order of this target, the non-boundary position is denoted as follows: The preceding and following adjacent candidate landing points are respectively:

[1093]

[1094] and:

[1095]

[1096] Candidate landing sites In the The contribution of crowding distance in each objective dimension is:

[1097]

[1098] in:

[1099] Indicates candidate landing sites In the Contribution of congestion distance across multiple target dimensions;

[1100] This represents the first adjacent candidate landing point in the sorting. One target value;

[1101] Indicates the th order of the next adjacent candidate landing sites in the sorting. One target value;

[1102] The denominator represents the range of target values ​​in the Pareto candidate solution set.

[1103] The total congestion distance of the candidate landing sites is obtained by summing them up.

[1104] Candidate landing sites The total crowding distance is obtained by summing the crowding distance contributions across the three target dimensions:

[1105]

[1106] Or it can be expressed as:

[1107]

[1108] in, It is a dimensionless numerical value.

[1109] For each objective dimension, the crowding distance of the boundary candidate solutions with the minimum and maximum objective values ​​is set to infinity or greater than the preset boundary value of the crowding distance of all non-boundary candidate solutions; when the maximum and minimum values ​​of a certain objective dimension are the same, the crowding distance contribution of that objective dimension is 0.

[1110] When the Pareto candidate solution set contains only one or two candidate landing points, the corresponding candidate landing point is treated as a boundary candidate solution; when the target value or congestion distance is equal, the equal relationship is preserved and no other evaluation quantity is introduced in S660 to break the equal relationship.

[1111] S670, candidate solutions are ranked based on the non-dominated frontier level and the crowding distance.

[1112] For all candidate landing sites after fast non-dominated sorting, the landing site selection system first sorts them according to the level of the non-dominated front.

[1113] Therefore, the sorting rules can be summarized as follows:

[1114] First, compare the non-dominated frontier levels, and give priority to the level with better level; if the levels are the same, give priority to the level with larger crowding distance.

[1115] In this embodiment, the subsequent automatic decision-making mode and manual decision-making mode are mainly used to process the first frontier layer, namely the Pareto candidate solution set.

[1116] Non-dominated frontier levels take precedence over congestion distance; within the same frontier level, levels are sorted from largest to smallest congestion distance. Congestion distance does not constitute a fourth optimization objective, nor does it directly determine the sole candidate result for the automatic mode. The automatic mode still determines the sole automatic candidate result based on the overall proximity of the S680.

[1117] In automatic decision-making mode, S680 determines the objective weights of the three objectives based on the target data of the Pareto candidate solution, and determines the unique automatic candidate result or outputs the parallel state based on the comprehensive closeness of the ideal solution.

[1118] S650 has determined the Pareto candidate solution set for the current decision-making cycle:

[1119]

[1120] in, This indicates the number of Pareto candidate solutions.

[1121] When the Pareto candidate solution set contains only one candidate landing point, that is:

[1122]

[1123] In the automatic decision-making mode, there is no need to calculate entropy weight, positive and negative ideal solutions, and comprehensive proximity. The unique candidate landing point is directly used as the only automatic candidate result in the automatic decision-making mode. In the manual decision-making mode, the candidate landing point is still displayed on the control terminal, and a valid manual selection result is formed after receiving a valid manual selection.

[1124] In the automatic decision-making mode, the landing point selection system does not use manually preset fixed weights for the three objectives. Instead, it uses the entropy weight method to determine the objective weights of the three objectives based on the actual distribution of the three objective values ​​in the current Pareto candidate solution set.

[1125] For the Given several objectives, determine the maximum objective value in the Pareto candidate solution set:

[1126]

[1127] And the minimum target value:

[1128]

[1129] in:

[1130]

[1131] Since all three objectives are minimization objectives, in one specific implementation, the objective values ​​are converted into positively standardized values ​​using the following relationship:

[1132]

[1133] in:

[1134] Indicates candidate landing sites In the Positive standardized values ​​on each target;

[1135] The value range is 0 to 1;

[1136] The smaller the original target value, the larger the standardized value.

[1137] When a certain objective has the same value in all Pareto candidate solutions, the standardized value corresponding to that objective is uniformly set to 1, so that the objective does not artificially introduce differences between candidate solutions.

[1138] For the For each target, candidate landing sites are calculated based on the standardized values ​​of each candidate landing site. The proportion under this objective:

[1139]

[1140] in:

[1141] Indicates candidate landing sites In the The proportion of standardized data for each target;

[1142] Indicates the Pareto candidate solution index;

[1143] For any target The sum of the weights of each candidate landing site is 1.

[1144] Right now:

[1145]

[1146] When a certain standardized value hour:

[1147]

[1148] In subsequent information entropy calculations, the following convention is adopted:

[1149]

[1150] This is to avoid performing logarithmic operations directly on 0.

[1151] When the number of Pareto candidate solutions satisfies:

[1152]

[1153] When, define the entropy normalization coefficient:

[1154]

[1155] For the The information entropy of the target is:

[1156]

[1157] in:

[1158] Indicates the first Information entropy of a target;

[1159] The value of is usually in the range of 0 to 1;

[1160] This indicates the proportion of candidate landing sites under the corresponding target.

[1161] According to the The information entropy of each target is used to determine its difference coefficient:

[1162]

[1163] in:

[1164] Indicates the first The difference coefficient of each objective;

[1165] The smaller the information entropy, the larger the difference coefficient;

[1166] The higher the information entropy, the smaller the difference coefficient.

[1167] The difference coefficient is used to represent the amount of effective discriminative information provided by the corresponding objective in the current Pareto candidate solution set.

[1168] The objective weights are determined based on the difference coefficients of the three objectives.

[1169] The first The objective weights of each objective are denoted as:

[1170]

[1171] but:

[1172]

[1173] in:

[1174] This indicates the objective weight of the landing time objective;

[1175] The objective weight representing the flight energy consumption target;

[1176] This indicates the objective weight of the landing risk target;

[1177] Indicates the index of the three targets.

[1178] Three objective weights are satisfied:

[1179]

[1180] and:

[1181]

[1182] The objective weights for the three objectives are determined by the data distribution of arrival time, total fuel consumption, and landing risk in the Pareto candidate solution set, and are independent of the eight-dimensional dynamic weights used by S300 to calculate the overall priority.

[1183] When the difference coefficients of the three objectives are all 0, the objective weights of the three objectives are set to equal values ​​to avoid the denominator of the weight calculation being 0.

[1184] Based on the objective weights of the three objectives and the positive standardized values ​​of each candidate landing site, the weighted standardized value is determined:

[1185]

[1186] in:

[1187] Indicates candidate landing sites In the Weighted standardized values ​​over each objective;

[1188] Indicates the positively standardized target value;

[1189] Indicates the first The objective weight of each objective.

[1190] For candidate landing sites This forms the corresponding weighted standardized vector:

[1191]

[1192] For the One objective is to determine the largest weighted standardized value among all Pareto candidate solutions:

[1193]

[1194] This leads to the ideal solution:

[1195]

[1196] For the The objective is to determine the minimum weighted standardized value among all Pareto candidate solutions:

[1197]

[1198] This leads to a negative ideal solution:

[1199]

[1200] For Pareto candidate solutions Calculate the Euclidean distance from its weighted normalized vector to the positive ideal solution:

[1201]

[1202] For a Pareto candidate solution Ci, calculate the Euclidean distance from its weighted normalized vector to the negative ideal solution:

[1203]

[1204] When the positive ideal solution and the negative ideal solution coincide, the distance from each candidate landing point to both the positive and negative ideal solutions is 0. The effective comprehensive proximity is not calculated, and instead, parallel candidate results that cannot be distinguished by automatic decision-making are output.

[1205] The overall proximity is calculated based on the distances of the candidate landing point to the positive and negative ideal solutions.

[1206] Candidate landing sites The overall similarity is recorded as:

[1207]

[1208] but:

[1209]

[1210] The comprehensive proximity score ranges from 0 to 1, with a larger value indicating that the candidate landing point is closer to the positive ideal solution and further away from the negative ideal solution. The comprehensive proximity score is only used in the automatic decision-making mode to determine the unique automatic candidate result from the Pareto candidate solution set; the comprehensive priority score is used for the hysteresis comparison of the new and old targets in S700; and the crowding distance is used to characterize the sparseness of the distribution of candidate solutions within the same frontier layer.

[1211] Identifying candidate outcomes in automated decision-making mode

[1212] The landing site selection system calculates the comprehensive proximity of each candidate landing site in the Pareto candidate solution set. When there is only one candidate landing site with the maximum comprehensive proximity, that candidate landing site is taken as the only automatic candidate result in the automatic decision-making mode.

[1213] Right now:

[1214]

[1215] When two or more candidate landing sites have the highest overall proximity score, the system outputs a candidate set with the same overall proximity score and an automatic decision-making marker for the candidate sites. In this implementation, the system does not add secondary sorting conditions such as overall priority, congestion distance, total fuel consumption, distance, or candidate landing site identifiers.

[1216] And it was determined that:

[1217]

[1218] in:

[1219] i∗ represents the index of the candidate landing point corresponding to the unique maximum comprehensive proximity score, C autocand,t This represents the unique automatic candidate result formed under the automatic decision-making mode in the current decision-making cycle.

[1220] This represents the unique automatic candidate result determined under the automatic decision-making mode in the current decision-making cycle.

[1221] The candidate landing point with the highest overall proximity is sent to S695 as the sole candidate result in the automatic decision-making mode for aggregation. After S695 forms the proposed landing point, it proceeds to S700 for execution.

[1222] A hysteresis comparison with the current target landing point;

[1223] Turning angle verification;

[1224] Verify the rate of increase or decrease.

[1225] Battery life margin verification after switching.

[1226] Therefore, this step is collectively referred to as "automatic decision candidate results"; only S695 forms a unique proposed landing point, and only after passing S700 can it become the final target landing point.

[1227] In this embodiment, the automatic decision-making candidate results formed based on the multi-objective evaluation results of the candidate solution set under the automatic decision-making mode are also referred to as automatic candidate results. The automatic candidate results are used to represent the candidate landing point results determined based on the candidate solution evaluation process.

[1228] S690, in the manual decision-making mode, forms a valid manual selection result based on the operator's valid selection.

[1229] The landing site selection system sends the Pareto candidate solution set and the identifier, type, estimated total flight time, total fuel consumption, basic risk value, comprehensive priority, congestion distance, runway conditions, wind direction matching, availability status, occupancy status and data confidence of each candidate landing site to the control terminal.

[1230] When an operator selects a candidate landing point from the Pareto candidate solution set, that candidate landing point is considered a valid human selection result under the manual decision-making mode. Manual selection is not subject to the mandatory constraint of the maximum comprehensive proximity value.

[1231] When the result of manual selection does not belong to the current Pareto candidate solution set, the selection is not considered a valid manual selection result, and the state of invalid manual selection is output. When no valid manual selection result is obtained, no new manual selection result is output, and the state of waiting for manual selection or missing manual selection result is output.

[1232] The valid selection results formed by the manual decision-making mode do not perform the 10-point positive switching threshold and 5-point back-switching threshold judgment based on the comprehensive priority difference, but should be sent to S695 for aggregation. After S695 forms the proposed landing point, it is used as the landing point to be switched into the S700's turning angle, vertical change rate and range margin safety verification.

[1233] In this embodiment, the effective human selection result formed based on the human selection received for the candidate solution set in the human decision-making mode is also referred to as the human selection result. The human selection result is used to represent the candidate landing point result formed based on the human selection of candidate landing points.

[1234] S695 summarizes the results of S500 express-level distance paths and S600 automatic and manual multi-target paths.

[1235] In the automatic decision-making modes of normal, early warning and emergency levels, when the Pareto candidate solution set contains only one candidate landing point, or there is only one candidate landing point corresponding to the maximum comprehensive proximity, the candidate landing point is output as the proposed landing point, and the automatic decision path identifier, comprehensive priority and comprehensive proximity data are output.

[1236] When the positive and negative ideal solutions coincide or the overall proximity of multiple candidate landing points is the highest, the automatic decision-making candidate set and corresponding parallel labels are output, and the unique proposed landing point that does not exist is not output.

[1237] In the manual decision-making mode, the corresponding candidate landing point is output as the proposed landing point only when a valid manual selection belonging to the current Pareto candidate solution set is received, and the manual selection path identifier is output; when the manual selection is invalid or no valid manual selection is obtained, the corresponding invalid or missing status is output, and no new manual proposed landing point is output.

[1238] In the express category, S695 receives the distance sorting results generated by S500. Only when there is a unique minimum distance result, the corresponding candidate landing point is output as the proposed landing point and the express category distance path identifier; when there are ties in the distance results, the set of ties in the distance candidate points and the status "no unique proposed landing point formed" are output.

[1239] When a unique proposed landing point is selected, the proposed landing point identifier, formation path identifier, original integrated priority and effective integrated priority, estimated total flight time, total fuel consumption, basic risk value, and current spatial and altitude relationships are sent to S700. When no unique proposed landing point is selected, the corresponding status, formation steps, parallel, empty set, reasons for invalidity or missing information, and the current target landing point identifier are sent to S700. Non-existent proposed landing point identifiers are not output. The proposed landing point does not directly overwrite the current target landing point in S600.

[1240] In this embodiment, unified aggregation includes uniformly receiving the urgent distance path results, the automatic candidate results, or the manual selection results. Upon receiving a unique valid result, the candidate landing point corresponding to the unique valid result is determined as the proposed landing point, while the formation path corresponding to the proposed landing point is retained. The unified aggregation result is used to represent the state of either a unique valid result formed after centralized processing or the state of not forming a unique proposed landing point; wherein, the formation path is used to represent the source path of the candidate result corresponding to the proposed landing point, including urgent distance paths, automatic decision paths, and manual decision paths.

[1241] The S700 performs hysteresis-based change judgment, maneuverability verification, and range margin verification based on the formation path of the proposed landing point and the current target landing point status to determine the final target landing point.

[1242] S700 receives two types of inputs: one is the unique proposed landing point path formed by S695 and its preceding calculation data; the other is the empty set state formed by S330, S460 or S500, and the "no unique proposed landing point formed" state output by S695 after summarizing the results of S540, S680 or S690, such as parallel, invalid or missing manual selections.

[1243] The landing point selection system reads the current target landing point identifier and its latest status stored in S100, and classifies the current target status into three categories: current target valid, current target technically unavailable, and current target comparison data invalid. First, it determines whether there are any explicit hard condition failures confirmed by valid data, such as unavailability of the open state, failure to meet minimum usage conditions, or unreachability after recalculation based on the current valid state. If any of these results exist, the current target is prioritized as technically unavailable. If no explicit hard condition failure results exist, it checks whether the open state, minimum usage conditions, and reachability can all be confirmed to be met based on valid data, and further determines whether the evaluation data for this decision cycle is complete and valid. If both the hard conditions and evaluation data are valid, and the comprehensive priority can be recalculated according to the feature definition, dynamic weights, and the S410 comprehensive priority formula for this decision cycle, the current target is determined to be valid. If the hard conditions can be confirmed to be met, but the status data is missing, the corresponding data has exceeded its existing or preset validity period, the scoring features are incomplete, or the comprehensive priority cannot be calculated temporarily, the current target comparison data is determined to be invalid and the technical conditions can be confirmed to be met. If the hard conditions themselves cannot be confirmed and there is no clear evidence of failure, the current target comparison data is determined to be invalid and the technical validity cannot be confirmed. Invalid comparison data is not directly equivalent to the current target being technically failed.

[1244] For automatic hysteresis paths that rely on the overall priority difference of the current target, if the current target comparison data is invalid and the technical conditions can be confirmed to be met, the automatic target change for this cycle is blocked, no landing point to be switched on the automatic path is formed, and the path does not enter S720, S730, and S740, but instead enters the target holding branch of S750. If the current target comparison data is invalid and the technical validity cannot be confirmed, hysteresis judgment is also not performed, no landing point to be switched on the automatic path is formed, and the path does not enter S720, S730, and S740, but instead enters the branch of S750 where no valid final target has been formed. Hysteresis bypasses that are technically invalid for the current target do not apply to either of the two invalid comparison data scenarios. The blocking rules only apply to automatic hysteresis paths that rely on the overall priority difference of the current target and do not change the existing bypass range for manual decision-making mode, express-level distance rapid path, initial target determination, and current target technical failure scenarios.

[1245] When S700 receives the "No unique proposed landing point has been formed" status, it first determines the technical status of the current target landing point according to the aforementioned rules; if the current target landing point identifier is empty, it directly forms the current target non-existent status. The "No unique proposed landing point has been formed" status does not enter S710, S720, S730, and S740, which depend on a unique proposed landing point. If the current target landing point is valid, or the current target comparison data is invalid but the technical conditions can be confirmed to be met, it proceeds to the target holding branch of S750; if the current target landing point is technically invalid, its technical validity cannot be confirmed, or the current target landing point does not exist, it proceeds to the "No valid final target has been formed" branch of S750.

[1246] S710 determines whether to perform a comprehensive priority hysteresis judgment and generates a hysteresis admission result.

[1247] When the proposed landing point is formed by the automatic decision-making mode under normal, early warning, or emergency levels, and the current target landing point is valid but the proposed landing point is different from the current target landing point, a comprehensive priority hysteresis judgment is performed according to the forward switching or back-switching scenarios, respectively. Both the proposed landing point and the current target landing point use the same feature definitions, original scoring caliber, the same set of dynamic weights, and the same S410 comprehensive priority formula for the current decision cycle to recalculate the original comprehensive priority; the 10-point forward switching threshold and the 5-point back-switching threshold apply to this predetermined comprehensive score caliber. The comprehensive priority difference judgment is only performed when both the current target technical conditions and comparison data are valid; if the current target comparison data is invalid but the technical conditions can be confirmed to be met, an automatic change blocking result is formed and the process is transferred to the target holding branch of S750; if the current target comparison data is invalid and the technical validity cannot be confirmed, an automatic change blocking result is formed and the process is transferred to the non-formed valid final target branch of S750.

[1248] The initial overall priority increment of the proposed landing site relative to the current target landing site is:

[1249]

[1250] For a forward handover where the proposed landing point is not the previous target landing point recorded before the most recent successful handover, the forward handover threshold is:

[1251]

[1252] The proposed landing site is qualified to proceed to subsequent safety checks when the original integrated priority increment satisfies the following formula:

[1253]

[1254] The previous target landing point is the current target landing point before the most recent successful target switch. The previous target record is only updated after the target switch is actually completed; the previous target is not updated when a proposed landing point, a landing point to be switched, or a switch fails in the current cycle is formed.

[1255] When the proposed landing point identifier is the same as the previous target landing point identifier recorded during the most recent successful target switch, and the previous target landing point is still valid in the current decision cycle, the current target change is identified as a rollback. The rollback threshold is:

[1256]

[1257] The previous target landing point is qualified to enter subsequent safety checks when the current period's original comprehensive priority increment relative to the current target landing point satisfies the following formula:

[1258]

[1259] When the corresponding original comprehensive priority increment does not reach the threshold of 10 or 5 points, a landing point to be switched is not generated, and the current valid target landing point is maintained. The threshold is only used to determine whether the proposed landing point is allowed to enter the subsequent safety verification, and does not directly trigger navigation target updates.

[1260] In scenarios involving manual decision-making, rapid path for emergency targets, initial target identification, and technical failure of the current target landing point, only the hysteresis judgment based on the comprehensive priority difference is bypassed; horizontal maneuvering, vertical maneuvering, and endurance margin checks are not bypassed. The corresponding proposed landing point is directly used as the landing point to be switched to in subsequent safety checks.

[1261] In this embodiment, a target hold or target update determination is performed based on the formation path of the proposed landing point and the status of the current target landing point. When the formation path of the proposed landing point is an automatically decided path and the current target landing point is valid, a hysteretic change determination is performed based on the combined priority of the proposed landing point and the current target landing point. When the formation path of the proposed landing point is an urgent distance path or a manually decided path, or when there is no valid current target landing point, a bypass access determination is performed. The bypass access determination indicates that for a proposed landing point with a corresponding formation path, if the corresponding access conditions are met, the subsequent judgment process of horizontal maneuver feasibility verification, vertical maneuver feasibility verification, and range margin verification will proceed; it does not indicate that the proposed landing point directly replaces the current target landing point.

[1262] When the current target landing point is valid and the identifier of the proposed landing point is the same as the identifier of the current target landing point, no target switching is required in this decision cycle, and the current target landing point is directly maintained; when the current target is technically invalid, the branch that maintains the same identifier is not applicable.

[1263] The hysteresis or bypass access result is recorded as:

[1264]

[1265] The S720 verifies the horizontal maneuvers required for target switching.

[1266] When the actual flight path to the landing point to be switched has not yet been obtained, the azimuth angle of the UAV's current position directly pointing to the landing point to be switched is used as the expected flight heading; when the actual flight path has been obtained, the heading information corresponding to that path at the beginning of the target switching phase is used. The expected flight heading is consistent with the heading caliber used for wind field correction in S440.

[1267] Normalize both the current heading of the UAV and the predicted flight heading to 0°–360°, and calculate the original heading difference:

[1268]

[1269] Determine the minimum azimuth difference between the two based on the original heading difference:

[1270]

[1271] Determine whether the minimum azimuth difference meets the maximum permissible turning angle of 30°, and generate a horizontal maneuver verification result:

[1272]

[1273] A horizontal maneuver pass is achieved when the minimum azimuth difference is no greater than 30°; a difference of exactly 30° is still considered a pass. The minimum azimuth difference refers to the minimum azimuth change within the heading plane, and is not equivalent to the aircraft roll angle, turning radius, cumulative turning angle of a complete curved track, or headwind deflection between the runway direction and the wind direction.

[1274] The S730 verifies the vertical maneuver required for target switching.

[1275] Prioritize reading the projected climb rate or projected descent rate generated by the flight control module or trajectory planning module for the landing point to be switched. If the module does not directly output the effective vertical change rate but provides an effective vertical adjustment time window, calculate it based on the current altitude, the target path altitude corresponding to the landing point to be switched, and the vertical adjustment time window.

[1276] The target path altitude corresponding to the landing point to be switched uses the same altitude reference as the current altitude. The vertical altitude difference is defined as:

[1277]

[1278] Specifically, the target path altitude is preferentially used based on the path altitude generated by the flight control module or trajectory planning module for target switching; in a specific implementation where path altitude data is unavailable, an equivalent estimation is performed using the landing surface altitude of the landing point to be switched. The vertical adjustment time window should meet the following requirements:

[1279]

[1280] When using the height difference and vertical adjustment time window for calculation, the required climb rate and required descent rate are as follows:

[1281]

[1282]

[1283] The maximum permissible rate of ascent and the maximum permissible rate of descent are respectively:

[1284]

[1285]

[1286] When climbing is required, the required descent rate is 0; when descending is required, the required climb rate is 0; when no vertical adjustment is required, both are 0. Vertical maneuvering passes the test when the required climb rate is no greater than 10 m / s and the required descent rate is no greater than 5 m / s; it is still considered passed if the corresponding thresholds are reached precisely.

[1287] Based on this, the vertical motion verification results are generated:

[1288]

[1289] The vertical maneuver verification results also record specific reasons such as the required climb rate exceeding the limit, the required descent rate exceeding the limit, invalid altitude parameters, invalid vertical adjustment time window, or failure to obtain a valid predicted vertical change rate.

[1290] When both horizontal and vertical maneuvers have been completed, an overall maneuver feasibility marker is formed:

[1291]

[1292] The S740 recalculates the range margin after target switching based on the latest real-time status.

[1293] Before actually updating the target, the landing point selection system recalculates the route flight requirements to the landing point to be switched, based on the phased endurance calculation method of S420 to S450, and on the latest position, altitude, ground speed, engine status, wind field, remaining energy and remaining flight time.

[1294] In the primary fuel-powered implementation, the recalculated total fuel consumption and estimated total flight time are as follows:

[1295]

[1296]

[1297] First, determine whether the route recalculation result already includes the additional steering and altitude adjustment costs incurred by target switching. If it does, the corresponding additional maneuver cost and additional maneuver time are set to 0. Setting it to 0 only means that the corresponding additional correction amount will not be added again, not that the additional maneuver itself has no cost. If the route recalculation result does not include, but the flight control module can output a valid additional steering or altitude adjustment cost separately, it is added as an additional correction amount once, and the basic climb, cruise, and descent costs already included in the route calculation must not be included again.

[1298] When it has been confirmed that the target switch requires additional steering, additional altitude adjustment or other additional maneuvers, but the route recalculation result does not include the corresponding additional maneuver consumption, and it is impossible to obtain effective additional fuel consumption or additional time, record the specific missing items and reasons, do not default the additional maneuver consumption and additional maneuver time to 0, do not use empirical values, fixed proportions or unsupported conservative estimates, and do not replace them with the 20% safety reserve of S460; mark this range margin calculation as invalid, determine the range margin verification result as failing, and set Bmargin,t to 0.

[1299] With the additional maneuvering parameters in effect, the total fuel consumption and estimated total flight time required to switch and reach the desired landing point are as follows:

[1300]

[1301]

[1302] In one specific implementation, when both the current remaining fuel and the current remaining flight time are greater than 0, the route recalculation result is valid, the necessary additional maneuvering consumption has been confirmed to be included in the route recalculation result or an effective additional correction can be obtained, and the fuel quantity and time quantity dimensions are consistent, the fuel margin ratio and time margin ratio are calculated separately:

[1303]

[1304]

[1305] The smaller of the two values ​​is determined as the overall range margin:

[1306]

[1307] In this specific implementation, the latest complete current available fuel quantity and the latest complete current remaining flight time before the S740 target switching verification time, the 10% margin judgment in this step, and the target switching verification time are used as the denominators for the corresponding margin ratios, and the smaller of the two margins is taken as the comprehensive range margin. The 0.8 times the amount in S460, the net value of the interface after deducting the same 20% safety reserve, or the value after further deduction from other unspecified reserves are not used as the denominator. The 0.8 constraint is not re-implemented in S740, and the 20% and 10% are not added together or continuously deducted.

[1308] The range margin check passes when the overall range margin is not less than 10% and all calculated parameters are valid; it fails when the overall range margin is less than 10% or any necessary calculated parameter is invalid. The corresponding results are:

[1309]

[1310] The range margin verification results record reasons such as a comprehensive range margin of less than 10%, invalid remaining energy parameters, invalid remaining flight time parameters, invalid route recalculation, or failure to determine additional maneuver consumption.

[1311] S750 determines the final target landing point and updates the target status.

[1312] For paths where landing points have already been identified for handover, the hysteresis or bypass access results, horizontal maneuver verification results, vertical maneuver verification results, and endurance margin verification results are used together to form a target handover permission marker:

[1313]

[1314] Specifically, when automatic decision-making is used and the current target is valid, the hysteresis admission result is determined by the judgment result of the 10-point positive switching threshold or the 5-point back-cutting threshold; when manual decision-making is used, for urgent-level distance fast paths, for the first time determining the target, or when the current target is technically invalid, the bypass admission result is 1. If the proposed landing point is the same as the valid current target, it belongs to an independent path that does not require switching and does not enter the product judgment of the target switching permission mark.

[1315] When the target switching permission flag is set to 1, the landing point to be switched is determined as the final target landing point for this decision cycle:

[1316]

[1317] When the current target is successfully switched to the landing point to be switched, the current target before the switch is recorded as the previous target landing point for the next cycle:

[1318]

[1319] Then update the landing point to be switched to the current target landing point for the next cycle:

[1320]

[1321] When a target is first identified, the previous target's landing point remains empty.

[1322] If the proposed landing point is the same as the valid current target, or if any target switching condition is not met but the current target landing point remains valid, the current target landing point will continue to be determined as the final target landing point for this decision cycle. Even if the current target does not meet the aforementioned complete definition of a valid current target due to invalid comparison data, but its openness, minimum usage conditions, and reachability can be confirmed to be met based on valid data, and no new target passing the applicable path and security verification is formed in this decision cycle, the current target landing point will also continue to be determined as the final target landing point for this decision cycle. When no unique proposed landing point is formed in this decision cycle, if the current target landing point is valid, or the current target comparison data is invalid but its openness, minimum usage conditions, and reachability can be confirmed to be met based on valid data, the current target landing point will also continue to be determined as the final target landing point for this decision cycle.

[1323]

[1324] If no successful target switch occurs, the current target landing point and the previous target landing point will remain unchanged in the next cycle:

[1325]

[1326]

[1327] If no sustainable current target landing point exists, and the proposed landing point fails the horizontal maneuver, vertical maneuver, or range margin checks, the proposed landing point will not be designated as the final target landing point. The status of no valid final target landing point being formed in the current decision cycle and the corresponding reason for the check failure will be output. If the current target comparison data is invalid and its technical validity cannot be confirmed, and no new target that has passed applicable safety checks is formed in this decision cycle, the original current target landing point or the proposed landing point will not be designated as the final target landing point. The status of no valid final target landing point being formed in the current decision cycle and the specific reason why its technical validity cannot be confirmed will be output. Turning angle, vertical change rate, and range margin thresholds will not be automatically relaxed.

[1328] If no unique proposed landing point is formed in this decision cycle, and there is no current target landing point that can be maintained, the original current target landing point or any candidate landing point that is not uniquely determined will not be determined as the final target landing point. The status of no valid final target landing point is output in the current decision cycle, and the corresponding reasons for not forming a unique proposed landing point and the inability to maintain the current target are recorded.

[1329] After completing S700, output the final target landing point or the status of the failure to form a valid final target landing point, and retain the proposed point formation path, current target status, applicable hysteresis and safety verification results, and target update reasons for S800 output and recording.

[1330] The S800 generates navigation guidance information based on the final target landing point, synchronizes the landing intention, records the decision log, and forms a periodic dynamic selection closed loop.

[1331] When the S700 forms a valid final target landing point, output the final target landing point for this decision cycle:

[1332]

[1333] When S700 outputs a status indicating that no valid final target landing point has been formed in the current decision cycle, S800 does not generate new target landing navigation guidance or target change broadcasts, but records the corresponding status, the formed path, and the reason for verification failure. The path still executes the log recording for this decision cycle and proceeds to the next decision cycle while the dynamic landing point selection function remains enabled.

[1334] S810 generates and outputs target landing navigation guidance information. The target position and distance use the coordinate and distance caliber of S220, the expected flight heading and heading adjustment amount use the caliber of S720, and the expected arrival information and range margin use the latest calculation results of S740; if any necessary parameter is invalid, an invalid calculation mark is output and no alternative value is generated.

[1335] The landing point selection system sends the target landing navigation guidance information Gt to the navigation module and the control terminal. When the final target changes, the navigation module updates the target landing point identifier and position, and executes existing track planning and flight guidance based on the new target landing point; when the final target does not change, the navigation module maintains the original target landing point and updates the heading, distance, estimated arrival information, and remaining flight information according to the latest status. The control terminal updates the corresponding displayed content in both cases.

[1336] S820, indicating the landing intent of a synchronized unmanned aerial vehicle.

[1337] When the final target landing point changes, a new flight plan summary containing the new final target landing point identifier, expected arrival information, and target landing status will be sent as a landing intention to the V2V communication module. The V2V communication module will then broadcast this information to surrounding UAVs through existing communication interfaces, enabling surrounding UAVs to consider the target landing arrangements of this UAV during their collaborative decision-making and conflict identification processes.

[1338] S830 records the decision log for the current decision-making cycle.

[1339] The decision log records the current decision cycle identifier or timestamp, current emergency level, summary of candidate landing point changes, current target before change, proposed landing point, final target or status of not forming a valid final target, proposed point formation path, current target validity status, original comprehensive priority difference when applicable, reachability verification results, hysteresis access results, horizontal and vertical maneuver verification results, endurance margin results, final switchover results and their reasons.

[1340] When the target remains unchanged, the log still records the current target landing point, the proposed landing point, the reason for not performing the switch, and the corresponding verification results; when no valid final target is formed, the log records the corresponding parallel, invalid, missing, or security verification failure status, thereby ensuring that the entire decision-making process is traceable.

[1341] Decision logs are used for post-mission analysis and can be combined with actual flight results as a data source for subsequent offline training or parameter optimization of the dynamic weight model. During online flight, the trained Actor network performs inference, and the deployed model parameters are not modified in real time based on the logs of individual decision cycles.

[1342] S840 forms a closed loop for periodic dynamic landing site selection.

[1343] After completing the navigation guidance output, landing intent synchronization, abnormal status output, and decision log recording applicable to this decision cycle, the decision cycle ends. In the next decision cycle, candidate landing point data, UAV real-time status data, and environmental data are reacquired, and steps S100 to S800 are executed repeatedly.

[1344] The decision cycle is repeated while the drone is in flight and the dynamic landing point selection function remains enabled; the cycle ends when the drone has completed landing, the flight mission has ended, or the dynamic landing point selection function has stopped.

[1345] After completing S800, the output includes the final target status (or the status of not forming a valid final target), navigation guidance information and validity markers, landing intent synchronization results, decision log, and the start or end status of the next decision cycle.

[1346] Second embodiment:

[1347] This embodiment provides a dynamic landing point selection system for executing the method described in the first embodiment. The system includes a data acquisition module, a feature determination module, an emergency classification and dynamic weight generation module, a comprehensive evaluation and reachability verification module, a classification screening module, a multi-target candidate result formation module, a proposed landing point aggregation module, a target switching determination module, and an output and closed-loop module. These modules are functional divisions and are not limited to independent hardware implementations.

[1348] The data acquisition module corresponds to S100 and is used to acquire candidate landing point data, UAV real-time status data and environmental data, and to acquire the identifier and latest status of the current target landing point when one exists. The corresponding data are provided to the subsequent feature determination, emergency classification, reachability verification and target switching determination modules.

[1349] The feature determination module corresponds to S200 and is used to determine seven numerical basic features, emergency landing point attributes, basic risk values, availability verification results, runway length non-compliance markers, distance, and headwind angle, and provide the corresponding results to the subsequent evaluation, screening, and multi-target candidate result formation modules; the specific determination method adopts the provisions of the first embodiment.

[1350] The emergency classification and dynamic weight generation module corresponds to S300 and is used to determine the current emergency level. It generates eight-dimensional dynamic weights for normal, early warning, or emergency levels, and a rapid path identifier for extremely urgent levels. When the current candidate set is empty, no effective dynamic weights dependent on that set are generated; instead, the empty set status, cause, and current target landing point identifier are directly provided to the target switching determination module.

[1351] The comprehensive evaluation and reachability verification module corresponds to S400. It is used to calculate the comprehensive priority in non-urgent situations and skip this calculation in urgent situations. It is also used to form the total energy consumption and estimated total flight time according to the phased parameters, energy consumption, wind field, and occupancy waiting criteria in the first embodiment, and to implement dual hard constraints of energy consumption and time. When the calculation is invalid or the reachable candidate set is empty, the corresponding status and reason are directly provided to the target switching determination module. When Neng=0, a powered endurance calculation termination state is formed, which is not merged into the S460 ordinary empty set and does not enter the subsequent ordinary selection chain.

[1352] The tiered screening module corresponds to S500 and is used to perform tiered screening according to the current emergency level: the non-urgent level non-empty waiting candidate set is provided to the multi-target candidate result formation module, and the distance and path results of the urgent level are provided to the proposed landing point summary module; when the screening result is empty, the conditions are not relaxed, but the empty set status and reason are directly provided to the target switching determination module.

[1353] The multi-objective candidate result generation module corresponds to S600. It is used to generate Pareto candidate solutions for non-urgent priority candidate sets with arrival time, flight energy consumption and landing risk as objectives. In automatic decision-making mode, it generates a unique automatic candidate result or a parallel state. In manual decision-making mode, it generates a valid manual selection result, an invalid state or a missing state. This module does not directly generate the proposed landing point.

[1354] The proposed landing point aggregation module corresponds to S695 and is used to uniformly receive the distance path results, automatic candidate results and manual selection results for urgent-level applications. It only forms a proposed landing point and retains its formation path when a unique valid result is obtained. When there are parallel, invalid or missing results, it only forms the status and reason for not forming a unique proposed landing point.

[1355] The target switching determination module corresponds to S700. It is used to receive the state of having a unique proposed landing point or not having formed a unique proposed landing point, and directly receives the empty set exception state from S330, S460, and S500. Based on the current target state and the formed path, it performs applicable hysteresis access, horizontal maneuver, vertical maneuver, and endurance margin checks to determine the result: target held, target updated, or no valid final target landing point formed. The relevant thresholds and safety reserve criteria adopt the provisions of the first embodiment.

[1356] The output and closed-loop modules correspond to the S800, used to generate navigation guidance information, synchronize landing intent, and record decision logs when a valid final target landing point is formed; when a valid final target is not formed, no new target navigation guidance or target change broadcast is generated, but the status, formation path, and failure reason are recorded. When the dynamic landing point selection function remains enabled, the next decision cycle is initiated.

[1357] The inputs, outputs, formulas, thresholds, and abnormal routing of the above modules are based on the first embodiment, without adding candidate sorting, parallel elimination, manual timeout, automatic substitution, flight control, trajectory planning, or external scheduling actions.

[1358] Third embodiment:

[1359] This embodiment provides an electronic device, including a processor and a memory, wherein the memory stores a computer program; when the processor executes the computer program, it implements the low-altitude unmanned aerial vehicle dynamic landing point selection method described in the first embodiment. The electronic device can be configured to implement the functions of each module of the dynamic landing point selection system described in the second embodiment.

[1360] Fourth embodiment:

[1361] This embodiment provides a computer-readable storage medium storing a computer program; when the computer program is executed by a processor, it implements the low-altitude unmanned aerial vehicle dynamic landing point selection method described in the first embodiment.

[1362] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of the invention (including the claims) is limited to these examples; under the concept of the invention, the technical features of the above embodiments or different embodiments can also be combined, and each step should be performed in accordance with the aforementioned data dependency relationship; for local processing that does not have a direct data dependency, it can be executed in parallel or the order can be adjusted without changing the input, output and decision relationship of each step, and there are many other variations of different aspects of the embodiments of the invention as described above, which are not provided in the details for the sake of brevity.

[1363] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for dynamic landing point selection of low-altitude unmanned aerial vehicles (UAVs), characterized in that, include: Acquire candidate landing point data, UAV real-time status data, and environmental data for the current decision-making cycle, and acquire the latest status data of the current target landing point when one exists; determine the basic characteristic data, emergency landing point attributes, and basic risk values ​​of each candidate landing point based on the candidate landing point data, the UAV real-time status data, and the environmental data. The current emergency level is determined based on the real-time status data of the UAV and the environmental data. A corresponding candidate landing point processing strategy is then determined based on the current emergency level. Specifically, when the current emergency level is non-urgent, dynamic evaluation weights for the current decision-making cycle are generated, and the comprehensive priority of each candidate landing point is determined based on the basic feature data and the dynamic evaluation weights. Accessibility verification is performed on each candidate landing point based on the energy consumption and flight time required for the UAV to reach each candidate landing point, as well as the UAV's remaining energy and flight time, resulting in reachable candidate results. These reachable candidate results are then graded and filtered according to the current emergency level, and an urgent-level distance path result is generated according to the candidate landing point processing strategy. Candidate results or manually selected results; the emergency distance path results, the automatic candidate results, or the manually selected results are uniformly summarized, and the proposed landing point and the formation path corresponding to the proposed landing point are determined based on the unified summary results; based on the formation path corresponding to the proposed landing point and the status of the current target landing point, target hold or target update determination is performed, and hysteresis change determination, maneuver feasibility verification, and endurance margin verification are performed when applicable to determine the final target landing point or the status of the current decision cycle in which no valid final target landing point has been formed; when the final target landing point is determined, target landing navigation guidance information is generated and output; when the dynamic landing point selection function is kept enabled, the next decision cycle is started.

2. The method for selecting dynamic landing points for low-altitude unmanned aerial vehicles according to claim 1, characterized in that, The basic feature data includes feature data that characterizes the basic qualifications, distance, runway length, wind direction matching, availability, occupancy status, and data confidence of the candidate landing sites; the basic risk value is determined based on the availability, occupancy status, wind direction matching, runway length, and data confidence of the candidate landing sites.

3. The method for selecting dynamic landing points for low-altitude unmanned aerial vehicles according to claim 1, characterized in that, The current emergency level includes normal level, early warning level, emergency level and extremely urgent level; when the current emergency level is normal level, early warning level or emergency level, the dynamic evaluation weight is generated according to the baseline weight of the corresponding emergency level and the drone status and environmental status of the current decision cycle. When the current emergency level is extremely urgent, an extremely urgent fast path identifier is generated and the determination of the overall priority is skipped.

4. The method for selecting dynamic landing points for low-altitude unmanned aerial vehicles according to claim 1, characterized in that, The reachability verification includes performing hard energy consumption and hard time constraints on the total energy consumption required for the UAV to reach the candidate landing point and the estimated total flight time, respectively. Only the candidate landing points whose reachability calculation is valid and which pass both the hard energy consumption and hard time constraints are determined as the reachable candidate results.

5. The method for selecting dynamic landing points for low-altitude unmanned aerial vehicles according to claim 1, characterized in that, When the current emergency level is extremely urgent, the candidate landing points selected through hierarchical screening are used to form the extremely urgent distance path result; when the current emergency level is normal, early warning, or emergency, a candidate solution set is formed with arrival time, flight energy consumption, and landing risk as multiple optimization objectives, and the automatic candidate result is formed based on the multi-objective evaluation results of the candidate solution set in automatic decision-making mode, or the manual selection result is formed based on the manual selection received for the candidate solution set in manual decision-making mode.

6. The method for selecting dynamic landing points for low-altitude unmanned aerial vehicles according to claim 1, characterized in that, The unified aggregation includes uniformly receiving the urgent distance path results, the automatic candidate results, or the manually selected results. When a unique valid result is received, the candidate landing point corresponding to the unique valid result is determined as the proposed landing point, while retaining the formation path of the proposed landing point. The proposed landing point does not directly replace the current target landing point, but is determined as the final target landing point based on the formation path and the status of the current target landing point, after target retention or target update determination, maneuver feasibility verification, and endurance margin verification.

7. The method for selecting dynamic landing points for low-altitude unmanned aerial vehicles according to claim 1, characterized in that, When the formation path of the proposed landing point is an automatically decided path and the current target landing point is valid, a hysteresis change decision is performed based on the combined priority of the proposed landing point and the current target landing point; when the formation path of the proposed landing point is an urgent distance path or a manually decided path, or when there is no valid current target landing point, a bypass access decision is performed; for proposed landing points that have been approved by the hysteresis change decision or the bypass access decision and are different from the current target landing point, horizontal maneuver feasibility verification, vertical maneuver feasibility verification, and endurance margin verification are performed, and when all the verifications are passed, the proposed landing point is determined as the final target landing point.

8. A dynamic landing point selection system, characterized in that, include: The data acquisition module is used to acquire candidate landing point data, UAV real-time status data and environmental data for the current decision-making cycle, and to acquire the latest status data of the current target landing point when a current target landing point exists; The feature determination module is used to determine the basic feature data, emergency landing point attributes, and basic risk values ​​of each candidate landing point based on the candidate landing point data, the UAV real-time status data, and the environmental data. The emergency classification and dynamic weight generation module is used to determine the current emergency level and the corresponding candidate landing point processing strategy based on the UAV real-time status data and the environmental data, and to generate dynamic evaluation weights for the current decision cycle when the current emergency level is non-urgent. The comprehensive evaluation and accessibility verification module is used to determine the comprehensive priority of each candidate landing point based on the basic feature data and the dynamic evaluation weights when the current emergency level is non-urgent, and to perform accessibility verification on each candidate landing point based on the energy consumption and flight time required for the UAV to reach each candidate landing point, as well as the UAV's remaining energy and remaining flight time, to form accessible candidate results. The hierarchical screening module is used to perform hierarchical screening of the reachable candidate results according to the current emergency level, and to form an emergency-level distance path result under the emergency-level distance path according to the candidate landing point processing strategy, or to provide the multi-objective candidate result forming module with the candidate results to be optimized under the multi-objective candidate preferred path; the multi-objective candidate result forming module is used to process the candidate results to be optimized according to multiple optimization objectives, and to form automatic candidate results or manual selection results according to the current decision-making mode; The proposed landing point aggregation module is used to uniformly receive the urgent distance path results generated by the hierarchical screening module and the automatic candidate results or manual selection results generated by the multi-target candidate result formation module, uniformly aggregate the received results, and determine the proposed landing points and the formation paths corresponding to the proposed landing points based on the uniform aggregation results. The target switching determination module is used to receive the proposed landing point and its formation path and the latest status data of the current target landing point obtained by the data acquisition module, and to perform target retention or target update determination based on the formation path and the status of the current target landing point, and to perform hysteresis change determination, maneuver feasibility verification and endurance margin verification when applicable, so as to determine the final target landing point or the status of the current decision cycle in which no valid final target landing point has been formed; The output and closed-loop module is used to generate and output target landing navigation guidance information when the final target landing point is determined, and to start the next decision cycle while the dynamic landing point selection function remains enabled.

9. An electronic device, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the low-altitude unmanned aerial vehicle dynamic landing point selection method as described in claim 1.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the low-altitude unmanned aerial vehicle dynamic landing point selection method as described in claim 1.

Images