An Adaptive Landable Area Determination Method and System Based on Flight Status
By identifying the flight status of the UAV and switching the judgment parameters, the problem of judgment result drift during autonomous landing of UAVs was solved, and accurate landing determination was achieved under different states, thereby improving the safety and reliability of autonomous landing of UAVs.
Patent Information
- Application Number
- CN202611604001.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-16
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies fail to effectively distinguish different flight states during the autonomous landing process of drones, resulting in fixed judgment parameters. This leads to serious misjudgments and omissions in dynamic states, and the judgment results drift with flight state and operating conditions, making it impossible to accurately determine the landable area.
The adaptive landable area determination method based on flight status collects UAV attitude data, identifies flight status, and switches determination parameters, including variance threshold, thickness hard cap and scoring weight, according to different flight statuses to determine the landability of grid cells.
This improves the safety and reliability of autonomous drone landing, ensuring consistent and accurate landing determination under different flight states and operating conditions, and reducing misjudgments and missed judgments.
Smart Images

Figure CN123113992A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of autonomous landing technology for unmanned aerial vehicles (UAVs), and more specifically to an adaptive method and system for determining landable areas based on flight status. Background Technology
[0002] Before an unmanned aerial vehicle (UAV) can land autonomously, it needs to determine a safe landing area from data collected by its onboard sensors. The mainstream approach in current technology is to use camera images and laser point clouds for obstacle perception and terrain statistics. Within a map building cycle, the point cloud is classified into dynamic and static points. The accumulated static point cloud is statistically analyzed by grid to obtain a grid map layer with parameters such as elevation, elevation variance, elevation thickness span, elevation thickness standard deviation, and semantic category. Then, a fixed set of judgment thresholds and scoring weights is used to determine whether each grid cell is landable, thereby obtaining the landable area.
[0003] However, the raster statistics formed by point cloud accumulation exhibit systematic differences under different flight states of the UAV, such as hovering, slow level flight, cruise, climb, descent, and turning. During cruise or turning, motion artifacts and pose calculation errors increase the variance and thickness span of the raster cells. During climb, as the flight altitude increases, the elevation variance, thickness span, and thickness standard deviation of the same terrain on the raster cells also increase approximately linearly. In other words, the raster statistics not only differ in magnitude between different flight states, but also drift with the conditions of that state, such as flight altitude, even within the same state. The same terrain collected under different states and conditions should yield consistent landing determination results.
[0004] Existing technologies have the following problems: fixed judgment parameters, using the same set of thresholds and weights in hovering, cruising, turning, climbing, and descending states, without distinguishing between flight states; serious misjudgments and omissions in dynamic states—fixed thresholds will misclassify a large number of landable areas as non-landable, and mismatched scoring weights will also cause landable areas to be fragmented or missed; existing technologies do not utilize the UAV's own attitude data (speed, altitude) to perform periodic quantification and identification of motion states, neither identifying flight states nor modeling and normalizing the statistics of each state separately, causing the judgment results to drift with flight states and operating conditions. The same terrain may be judged as landable at low altitudes and non-landable at high altitudes, thus failing to achieve adaptive determination of landable area judgment parameters.
[0005] Therefore, it is necessary to design a new method to solve the problems of fixed judgment parameters, serious misjudgments and omissions in dynamic states, and judgment results drifting with operating conditions in the existing technology, so as to obtain consistent, accurate and robust landing determination for the same terrain under different flight states and operating conditions, thereby improving the safety and reliability of UAV autonomous landing. Summary of the Invention
[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide an adaptive landing zone determination method and system based on flight status.
[0007] To achieve the above objectives, the present invention adopts the following technical solution: an adaptive landable area determination method based on flight state, comprising: During the map building cycle, attitude data of the UAV is collected, and a grid map corresponding to the map building cycle is obtained. The grid map includes multiple grid cells, and each grid cell carries terrain statistics. The flight features of the UAV during the map building cycle are extracted based on the attitude data, and the flight state of the UAV during the map building cycle is identified based on the flight features, wherein the flight state is one of a variety of preset flight states; Based on the flight state, select a corresponding set of judgment parameters from multiple sets of preset judgment parameters that correspond one-to-one with the various preset flight states; Based on the selected set of determination parameters and the terrain statistics of each grid cell, a landing capability determination is made for each grid cell to obtain a landing area. The preset flight states include hovering, slow level flight, cruise, climb, descent, and turning.
[0008] The further technical solution is as follows: the attitude data includes the three-axis velocity, yaw angle and altitude of the UAV recorded at sampling time during the map building cycle; The flight characteristics include horizontal speed characteristics, vertical speed characteristics, yaw rate characteristics, and altitude change characteristics, wherein the vertical speed characteristics include the absolute mean vertical speed characteristics and the mean vertical speed band sign characteristics.
[0009] The further technical solution is as follows: the step of extracting the flight features of the UAV during the map building period based on the attitude data includes: Based on the triaxial velocities at each sampling time, calculate the horizontal velocity characteristic, the absolute mean vertical velocity characteristic, and the sign mean vertical velocity characteristic. The yaw angles at adjacent sampling times are unwound to obtain the yaw angle difference, and the yaw rate characteristic is calculated based on the yaw angle difference. The height change characteristics are calculated based on the height at the beginning and end of the map construction cycle sampling time.
[0010] The further technical solution is as follows: the multiple preset flight states include hovering, slow level flight, cruise, climb, descent and turn; The step of identifying the flight status of the UAV within the map building cycle based on the flight characteristics includes: If the absolute average value of the vertical velocity exceeds the vertical velocity threshold, or the absolute value of the altitude change characteristic exceeds the altitude change threshold, then it is determined to be vertical motion, and the flight state is determined to be climbing or descending based on the direction of vertical motion. If the absolute average value of the vertical speed does not exceed the vertical speed threshold, and the absolute value of the altitude change feature does not exceed the altitude change threshold, then when the horizontal speed feature exceeds the cruise speed threshold, the flight state is determined to be cruise. If the absolute mean value of the vertical speed does not exceed the vertical speed threshold, and the absolute value of the altitude change characteristic does not exceed the altitude change threshold, then when the yaw rate characteristic exceeds the yaw rate threshold, the flight state is determined to be a turn. If the absolute average value of the vertical speed does not exceed the vertical speed threshold, and the absolute value of the altitude change feature does not exceed the altitude change threshold, and the horizontal speed feature exceeds the hovering speed threshold, then the flight state is determined to be slow level flight. If the absolute average value of the vertical velocity does not exceed the vertical velocity threshold, and the absolute value of the altitude change feature does not exceed the altitude change threshold, the flight state is determined to be hovering. The step of determining the flight state as either ascent or descent based on the vertical direction of motion includes: If the absolute value of the height change feature exceeds the height change threshold, then the ascent or descent is determined according to the sign of the height change feature; If the absolute value of the height change feature does not exceed the height change threshold, the ascent or descent is determined based on the sign of the mean value of the vertical speed band sign.
[0011] The further technical solution is as follows: each set of judgment parameters includes a variance threshold, a thickness hard cap, a score boundary, and multiple scoring weights; In each set of judgment parameters corresponding to different flight states, the variance threshold, the thickness hard cap, the score boundary, and the scoring weight are independently valued, and each set of judgment parameters is obtained by fractal statistical calibration of actual flight data under the corresponding flight state.
[0012] The further technical solution is as follows: the topographic statistics include static elevation variance, static elevation thickness span and static elevation thickness standard deviation, and the grid cell also carries a semantic category; Determining whether each of the grid cells is landable includes: A grid cell that meets any of the following conditions is considered a hard hole: The semantic category is non-landable; The square root of the static elevation variance exceeds the variance threshold; The static elevation thickness span exceeds the thickness hard cap.
[0013] The further technical solution is as follows: Landability determination is performed on each of the grid cells, and the method further includes: For grid cells not identified as hard holes, according to Calculate the landing score of the grid cell. Where r, p, and q are the square root of the static elevation variance, the static elevation thickness span, and the static elevation thickness standard deviation, respectively. , , For the multiple scoring weights, , , The lower the landability score, the more suitable the grid cell is for landing, and the lower the landing score is for all flight states, the more suitable the grid cell is for landing. Grid cells whose landability score is less than the score threshold are determined to pass the landability determination.
[0014] A further technical solution is as follows: before calculating the landing score of the grid cell, the method further includes: According to the fractional model corresponding to the identified flight state, the terrain statistics of the grid cell are converted to the baseline working condition of the flight state, and the landing score is calculated. The fractional models corresponding to different flight states are independent of each other, and the conversion is applied to the calculation of the landing score. Among them, the split-state model corresponding to the climb in various preset flight states is a linear model with the average altitude recorded at each sampling time within the map construction period as the independent variable. The conversion process includes normalizing the terrain statistics according to a preset reference altitude.
[0015] The further technical solution is as follows: the landing-ready determination of each of the grid cells to obtain the landing-ready area includes: A grid cell whose Chebyshev distance from the hard hole does not exceed a set number is marked as a near-hard hole cell, wherein the Chebyshev distance is the maximum value of the absolute values of the differences between the horizontal and vertical indices of the two grid cells; The grid cells that simultaneously meet the following conditions are summarized and output as the landable area: It was not identified as the hard hole mentioned above; Not marked as the near-hard hole unit; The landability score is less than the score threshold.
[0016] The present invention also provides an adaptive landable area determination system based on flight status, comprising: The data acquisition unit is used to acquire the attitude data of the UAV during the map building cycle and obtain the grid map corresponding to the map building cycle. The grid map includes multiple grid cells, and each grid cell carries terrain statistics. A state recognition unit is used to extract the flight features of the UAV during the map building cycle based on the attitude data, and to identify the flight state of the UAV during the map building cycle based on the flight features, wherein the flight state is one of a variety of preset flight states; The parameter selection unit is used to select a set of judgment parameters from multiple sets of preset judgment parameters that correspond one-to-one with the multiple preset flight states according to the flight state. The area determination unit is used to determine whether each of the grid cells is landable based on the selected set of determination parameters and the terrain statistics of each grid cell, so as to obtain the landable area. The preset flight states include hovering, slow level flight, cruise, climb, descent, and turning.
[0017] The advantages of this invention compared to existing technologies are as follows: During the autonomous landing process of a UAV, this invention collects the UAV's attitude data in units of one map building cycle and obtains a grid map containing multiple grid cells carrying terrain statistics corresponding to that cycle. Flight features are extracted based on the attitude data, and the UAV is identified as belonging to one of several preset flight states within that cycle. Accordingly, a set of preset judgment parameters corresponding to each flight state is selected for determining the landability of each grid cell. Before calculating the landability score, the terrain statistics are further converted to the baseline conditions of that state according to the fractional model corresponding to the identified flight state, eliminating... The systemic influence of changes in operating conditions within the same state is analyzed, and the landable area is output through a dual condition combining the exclusion of near-hard hole units and the landing score being lower than the score boundary. As a result, the flight state recognition drives the adaptive switching of judgment parameters, matching the variance threshold, thickness hard cap, scoring weight, and score boundary with the systematic differences in grid statistics under various flight states. This overcomes the shortcomings of existing technologies, such as fixed judgment parameters, serious misjudgments and omissions under dynamic states such as cruise and turning, and the drift of judgment results with operating conditions. It enables consistent, accurate, and robust landing determination of the same terrain under different flight states and operating conditions, thereby significantly improving the safety and reliability of UAV autonomous landing.
[0018] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. Attached Figure Description
[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0020] Figure 1 A flowchart illustrating the adaptive landable area determination method based on flight status provided in an embodiment of the present invention; Figure 2 A schematic block diagram of an adaptive landable area determination system based on flight status provided in an embodiment of the present invention; Figure 3 A schematic block diagram of a computer device provided for an embodiment of the present invention. Detailed Implementation
[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0022] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.
[0023] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.
[0024] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.
[0025] Please see Figure 1 , Figure 1 This is a flowchart illustrating the adaptive landable area determination method based on flight state provided in an embodiment of the present invention. This adaptive landable area determination method based on flight state is applied in a server. In this embodiment, when the UAV performs an autonomous landing mission, the UAV's onboard platform is equipped with a camera, a lidar, and an attitude sensor. The camera is used to acquire ground images, providing input for detection and semantic segmentation; the lidar is used to acquire lidar point clouds, providing geometric data for grid map construction; and the attitude sensor is used to acquire the UAV's three-axis velocity, yaw angle, and altitude, providing input for flight status recognition.
[0026] Camera images undergo target detection and semantic segmentation to obtain detection boxes for dynamic obstacles and semantic segmentation masks indicating whether the obstacle is landable or not. LiDAR collects point clouds, which are projected onto the camera images for dynamic / static classification and semantic annotation. Points projected into the detection boxes of dynamic obstacles are classified as dynamic points and do not participate in the accumulation of the static raster map, thus creating holes in that area of the static raster map. Points with valid projections but not covered by the detection boxes are assigned a semantic category (landable, non-landable, or unknown) based on the semantic segmentation mask and are accumulated into a static point cloud within a map building cycle. The accumulated static point cloud is statistically analyzed by raster to form a raster map containing static elevation, static elevation variance, static elevation thickness span, static elevation thickness standard deviation, and semantic category.
[0027] In this embodiment, a hole refers to an area in the corresponding region of a static raster map that has no effective statistics because dynamic points do not participate in static accumulation.
[0028] In this embodiment, a static point cloud refers to a set of laser points whose projection is valid and not covered by the detection box after removing dynamic points, and which are accumulated within a map building cycle; correspondingly, a dynamic point refers to a laser point whose projection falls into the dynamic obstacle detection box.
[0029] Specifically, the projection of laser point cloud onto camera image is achieved based on the pre-calibrated extrinsic matrix of camera and lidar and the camera intrinsic parameters, and the time synchronization mechanism ensures the time alignment of image frame and point cloud frame; target detection and semantic segmentation can be implemented using deep learning networks known in the art, such as YOLO series networks for target detection and DeepLab series networks for semantic segmentation, or other detection and segmentation models with equivalent functions can be used, which are not limited in this embodiment.
[0030] During the autonomous landing process of the UAV, an adaptive decision-making link based on flight status is added on top of the grid map: at the end of a map building cycle, attitude features are extracted and flight status is identified, and a landing area is determined accordingly. One set of decision parameters is selected from multiple sets to determine the landability of each grid cell in the grid map and output the landable area. The above data link includes the following processing stages: detection and semantic segmentation stage: target detection and semantic segmentation are performed on camera images to obtain dynamic obstacles and semantic results of landability / non-landability; point cloud classification and accumulation stage: the laser point cloud is dynamically / statically classified and assigned semantic categories based on the semantic results, accumulating static point cloud within the cycle; grid map building stage: the accumulated static point cloud is statistically analyzed by grid to obtain static elevation, static elevation variance, static elevation thickness span, static elevation thickness standard deviation, and semantic category layers; flight status recognition stage: the data within the cycle is read... Attitude data is used to extract horizontal velocity, vertical velocity, yaw rate, and altitude change features, and one of six flight states is identified according to preset rules. In the parameter switching stage, based on the identified flight state, the corresponding hard threshold, scoring boundary, and scoring weight are selected from multiple sets of parameters. In the single-cell landing determination stage, the selected parameters are used to determine the landing area cell by cell on the raster map (hard hole removal, and statistical modeling and scoring based on state before scoring). The landing area is the final result of the determination and is provided to the downstream landing planning module for planning the landing position and trajectory. It should be noted that the above detection and semantic segmentation, point cloud classification and accumulation, and raster map construction stages are the data sources and preprocessing stages of this method.
[0031] In other embodiments, the aforementioned camera can be replaced with other imaging sensors such as depth cameras or infrared cameras; the lidar can be replaced with other sensors capable of acquiring three-dimensional geometric data, such as millimeter-wave radar, structured light sensors, or binocular stereo vision sensors; the attitude sensor can be a combination of an inertial measurement unit, a satellite positioning receiver, and a barometric altimeter, or it can directly read the navigation data output by the UAV flight control system. The present invention does not limit the specific type of sensor, as long as it can obtain images, three-dimensional point clouds, and UAV attitude data.
[0032] It should be noted that the adaptive decision link described in this embodiment is deployed during the autonomous landing phase of the UAV. The flight status recognition and decision parameter switching are both used for the landing area determination during the autonomous landing phase, rather than for flight monitoring, fault diagnosis or perception resource scheduling. This allows the limited onboard computing resources to be concentrated on the core objective of landing safety.
[0033] For ease of understanding, the main symbols involved in this embodiment and their meanings are shown in Tables 1 to 3.
[0034] Table 1. Attitude-side symbols Table 2. Characteristic Symbols Table 3. Grid and decision side symbols Figure 1 This is a flowchart illustrating the adaptive landing zone determination method based on flight state provided in an embodiment of the present invention. This method is applied to the autonomous landing process of a UAV and can be deployed on the UAV's onboard computing unit or on ground station computing equipment communicating with the UAV, such as... Figure 1 As shown, the method includes the following steps S110 to S140.
[0035] S110. Collect the attitude data of the UAV during the map building cycle and obtain the grid map corresponding to the map building cycle, wherein the grid map includes multiple grid units, and each grid unit carries terrain statistics.
[0036] In this embodiment, the map building cycle refers to the time window during which accumulated laser point clouds over a period of time are aggregated into a raster map, for example, several seconds.
[0037] In this embodiment, attitude data refers to the motion data of the UAV recorded by the attitude sensor at the sampling time during the map building cycle. The attitude data includes the three-axis velocity, yaw angle and altitude of the UAV recorded at the sampling time during the map building cycle.
[0038] In this embodiment, the raster map refers to the gridded map obtained by statistically analyzing the static point cloud accumulated during the map construction cycle, including map layers such as static elevation, static elevation variance, static elevation thickness span, static elevation thickness standard deviation, and semantic category.
[0039] In this embodiment, a grid cell refers to a single grid in a raster map, carrying statistical quantities such as elevation, variance, thickness span, thickness standard deviation, and semantic category.
[0040] In this embodiment, topographic statistics refer to statistical indicators that reflect the topographic undulation and roughness obtained from static point cloud statistics on the raster cell, including static elevation variance, static elevation thickness span, and static elevation thickness standard deviation.
[0041] In this embodiment, semantic category refers to the category of landing, non-landing, or unknown that a point cloud is assigned based on a semantic segmentation mask.
[0042] Specifically, the duration of the map building cycle can be configured according to the UAV's flight speed, sensor frame rate, and onboard computing resources, for example, any value between 1 second and 10 seconds; the resolution of the grid unit is, for example, 0.5 meters, or any value between 0.1 meters and 1 meter depending on the landing accuracy requirements, and this invention does not limit it.
[0043] Therefore, by synchronously acquiring attitude data and grid maps on a map building cycle basis, flight status recognition and landing area determination are aligned on the same time reference, providing a data basis for subsequent periodic switching of determination parameters.
[0044] S120. Extract the flight features of the UAV during the map building cycle based on the attitude data, and identify the flight state of the UAV during the map building cycle based on the flight features, wherein the flight state is one of a variety of preset flight states.
[0045] In this embodiment, flight status refers to the movement mode of the UAV within the current map building cycle, which is divided into six categories: hovering, slow level flight, cruise, climb, descent, and turning, and is distinguished based on horizontal speed, vertical speed / altitude change, and yaw rate characteristics.
[0046] The preset flight states include hovering, slow level flight, cruise, climb, descent, and turning.
[0047] Specifically, the flight characteristics include horizontal speed characteristics, vertical speed characteristics, yaw rate characteristics, and altitude change characteristics, wherein the vertical speed characteristics include the absolute mean vertical speed characteristic and the mean vertical speed band sign characteristic.
[0048] The horizontal velocity characteristic refers to the average horizontal velocity magnitude at each sampling time within the period, denoted as .
[0049] In this embodiment, the vertical velocity feature refers to the feature that characterizes the strength of vertical motion within a period, including the absolute mean vertical velocity feature and the vertical velocity band sign mean feature; the absolute mean vertical velocity feature is the average value of the absolute vertical velocity at each sampling time within the period, denoted as... The mean characteristic of the vertical velocity band symbol refers to the average value of the vertical velocity band symbol at each sampling time within the period, denoted as . The symbol used to determine ascent / descent. For climbing, It is decreasing.
[0050] In this embodiment, the yaw rate characteristic refers to the measure of the rate of change of the yaw angle within a period, denoted as ω.
[0051] In this embodiment, the height change feature refers to the height difference between the sampling time at the end of the period and the sampling time at the beginning of the period, denoted as .
[0052] In one embodiment, extracting the flight features of the UAV during the map building cycle based on the attitude data includes: Based on the triaxial velocities at each sampling time, calculate the horizontal velocity characteristic, the absolute mean vertical velocity characteristic, and the sign mean vertical velocity characteristic. The yaw angles at adjacent sampling times are unwound to obtain the yaw angle difference, and the yaw rate characteristic is calculated based on the yaw angle difference. The height change characteristics are calculated based on the height at the beginning and end of the map construction cycle sampling time.
[0053] Specifically, suppose that N attitude samples are recorded during the map building cycle, and the three-axis velocity components of the t-th sample are: , , (Unit: m / s), yaw angle: (Unit: rad), altitude: (Unit: m), t=1,…,N, then each flight characteristic is calculated according to the following formula: ; ; ; ; ; ; The periodic sampling duration (in seconds). The yaw angle at the t-th and t-1-th sampling times The difference after unwinding (in rad); , These represent the heights at the end of the period and the beginning of the period, respectively.
[0054] In this embodiment, the yaw angle difference refers to the difference in yaw angles between adjacent sampling times after ±π unwinding; unwinding refers to converting the yaw angle difference to (-π / 2π). , Processing of the interval. By adjusting the yaw angle... Unwinding can prevent the yaw angle from being too high. When a change occurs at the boundary, the yaw rate feature is incorrectly amplified, thereby improving the accuracy of turning state recognition.
[0055] In other embodiments, the above-mentioned mean features can also be replaced by robust statistics such as median, quantile or weighted mean to reduce the impact of individual outlier sampling on feature extraction; the sampling frequency can be fixed within the period or dynamically adjusted according to computing resources, and this embodiment does not limit this.
[0056] The various preset flight states include hovering, slow level flight, cruise, climb, descent, and turning; in this embodiment, vertical motion refers to the combined motion of climb and descent, serving as an intermediate result in priority determination.
[0057] In one embodiment, the above-mentioned identification of the UAV's flight state within the map building cycle based on the flight characteristics includes determining the priority order of vertical movement, cruise, turning, slow level flight, and hovering in sequence. Each determination adopts a strict threshold rule (if the threshold is equal, no trigger is made), and one determination settles one state. If the absolute average value of the vertical velocity exceeds the vertical velocity threshold, or the absolute value of the altitude change characteristic exceeds the altitude change threshold, then it is determined to be vertical motion, and the flight state is determined to be climbing or descending based on the direction of vertical motion. If the absolute average value of the vertical speed does not exceed the vertical speed threshold, and the absolute value of the altitude change feature does not exceed the altitude change threshold, then when the horizontal speed feature exceeds the cruise speed threshold, the flight state is determined to be cruise. If the absolute mean value of the vertical speed does not exceed the vertical speed threshold, and the absolute value of the altitude change characteristic does not exceed the altitude change threshold, then when the yaw rate characteristic exceeds the yaw rate threshold, the flight state is determined to be a turn. If the absolute average value of the vertical speed does not exceed the vertical speed threshold, and the absolute value of the altitude change feature does not exceed the altitude change threshold, and the horizontal speed feature exceeds the hovering speed threshold, then the flight state is determined to be slow level flight. If the absolute average value of the vertical velocity does not exceed the vertical velocity threshold, and the absolute value of the altitude change feature does not exceed the altitude change threshold, the flight state is determined to be hovering. The step of determining the flight state as either ascent or descent based on the vertical direction of motion includes: If the absolute value of the height change feature exceeds the height change threshold, then the ascent or descent is determined according to the sign of the height change feature; For climbing, To decrease; If the absolute value of the height change feature does not exceed the height change threshold, the ascent or descent is determined based on the sign of the mean value of the vertical speed band sign. For climbing, It is decreasing.
[0058] The aforementioned priority order reflects a clear determination of overlapping states: for example, when a drone cruises while turning, the cruise determination takes precedence over the turning determination, so the state is determined as cruise; when a drone climbs / descends while yawing, the vertical motion determination takes precedence, so the state is determined as climb / descend. Thus, each map building cycle can uniquely determine a flight state, avoiding ambiguity in state identification and ensuring the determinism and repeatability of subsequent parameter switching.
[0059] In a set of examples, the vertical velocity threshold m / s, height change threshold Cruise speed threshold m / s, yaw rate threshold rad / s, hovering speed threshold m / s, the above values are for illustrative purposes only and are not intended to limit this embodiment.
[0060] In other embodiments, the aforementioned judgment thresholds can be readjusted based on the UAV model, power system response characteristics, and sensor accuracy; the aforementioned priority order can also be adjusted according to the application scenario. For example, in scenarios that focus more on turning conditions, turning judgment can be placed before cruise judgment; in addition, a hysteresis interval can be set for each threshold outside of the strict greater than rule. The hysteresis interval refers to the hysteresis margin interval set for each judgment threshold, that is, the flight feature needs to exceed the threshold by a certain margin before triggering the state switch, and needs to fall below the threshold by a certain margin before switching back to the original state, so as to avoid the flight state switching back and forth when the feature fluctuates near the threshold, and further improve the stability of state recognition.
[0061] S130. Select a set of judgment parameters from multiple preset judgment parameters that correspond one-to-one with the multiple preset flight states according to the flight state.
[0062] In this embodiment, the determination parameters refer to a set of parameters used to determine the descent capability of a grid cell.
[0063] Specifically, each set of the judgment parameters includes a variance threshold, a thickness hard cap, a score boundary, and multiple scoring weights; The variance threshold is the threshold at which the square root of the static elevation variance of a raster cell is deemed unlandable. Thickness hard cap refers to the threshold at which the static elevation thickness span of a grid cell is deemed undescending. The score threshold refers to the upper limit of the score required for a landing to pass the judgment; it is denoted as . Multiple scoring weights refer to the weights of each statistical item in the scoring formula, denoted as... , , The normalized denominator is the divisor used to convert each statistical quantity into a dimensionless term, denoted as . , , .
[0064] In each set of judgment parameters corresponding to different flight states, the variance threshold, the thickness hard cap, the score boundary, and the scoring weight are independently valued, and each set of judgment parameters is obtained by fractal statistical calibration of actual flight data under the corresponding flight state.
[0065] Among them, the three normalized denominators , , For sharing across all flight states, the example value is... =0.1m、 =0.4m =0.1m.
[0066] In one embodiment, the fractional statistical calibration process for each set of judgment parameters includes: collecting multiple real-flight map construction cycle data of the UAV under six flight states; constructing a raster map for each cycle and labeling the landability true value of each raster cell; for each flight state, statistically analyzing the terrain statistics distribution of known landable raster cells, determining the variance threshold and thickness hard cap for that state according to preset quantiles, and determining the scoring boundary and scoring weight for that state with the landability true value as supervision and the goal of minimizing the weighted sum of false positive and false negative rates; and solidifying the six sets of judgment parameters obtained from the calibration into an airborne parameter table for lookup and retrieval during runtime using flight state as the index. This ensures that the statistical characteristics of each set of judgment parameters match the corresponding flight state, and the calibration process can be periodically updated as real-flight data accumulates.
[0067] Taking the hovering state as the baseline state, its value is: variance threshold. =0.1m, thickness hard cap =0.5m, weight of flat term =1.0, span term weight =0.5, standard deviation term weight =0.3, score cutoff =1.0. A set of example values for the six flight states are shown in Table 4. These values are only examples and are not intended to limit this embodiment.
[0068] Table 4. Example Values for Various Flight Status Determination Parameters The variation patterns of each state parameter are as follows: hovering and slow level flight use the tightest hard threshold and baseline weighting ratio, with a scoring cutoff of 1.0; cruising uses the widest hard threshold, and only the flatness item is retained in the scoring. The scoring boundary is slightly tighter (0.9); climb and descent use a hard threshold between hovering and cruising, with the same weight as hovering. Climb has the tightest scoring boundary (0.6) and incorporates the pre-scoring statistical fractional model (see step S140 for details), while descent has a wider scoring boundary (1.21). The hard threshold for turning is the same as for climb and descent, with the widest scoring boundary (1.31). Note that the scoring boundary is an upper bound: the score s must be less than... Since it has only just passed, the larger the boundary, the more lenient it is.
[0069] The switching mechanism is as follows: after the flight status is identified, the corresponding status is retrieved by looking up a table using that status as an index. , , and , , This serves as the single-cell judgment parameter for the current cycle. As a result, the hard threshold, scoring boundary, and scoring weight used for landing area judgment adaptively switch with the flight state, so that the statistics that are raised due to motion artifacts and attitude calculation errors in dynamic states such as cruise and turning are no longer subject to fixed tight thresholds for misjudgment, thereby reducing misjudgment and missed judgment of landing areas. At the same time, the strictest judgment is maintained in static states such as hovering to ensure landing safety.
[0070] In other embodiments, multiple sets of decision parameters may not be switched by looking up tables according to discrete states, but may be continuously generated by a parameter function with flight characteristics as independent variables, such as interpolating the variance threshold with horizontal velocity characteristics as independent variables; multiple sets of decision parameters may also be calibrated according to UAV model, sensor configuration or environmental conditions such as season and lighting, and selected according to environmental conditions during operation. This invention does not limit this.
[0071] S140. Based on the selected set of determination parameters and the terrain statistics of each grid cell, landability determination is performed on each grid cell to obtain a landable area.
[0072] In this embodiment, the landable area refers to the set of grid cells that are identified as suitable for safe landing of the UAV after being determined as landable, and is provided to the downstream landing planning module for planning the landing position and landing trajectory.
[0073] In one embodiment, determining the landability of each of the grid cells includes: A grid cell that meets any of the following conditions is considered a hard hole: The semantic category is non-landable; The square root of the static elevation variance exceeds the variance threshold; that is, r > τ_flat. The static elevation thickness span exceeds the thickness hard cap. That is, p > τ_span; In this embodiment, a hard hole refers to a grid cell that is determined to be non-landable by a hard threshold. Grid cells identified as hard holes are directly discarded and do not participate in subsequent scoring.
[0074] In one embodiment, determining the landability of each of the grid cells further includes: For grid cells not identified as hard holes, according to Calculate the landing score of the grid cell. Where r, p, and q are the square root of the static elevation variance, the static elevation thickness span, and the static elevation thickness standard deviation, respectively. , , For the multiple scoring weights, , , The lower the landability score, the more suitable the grid cell is for landing, and the lower the landing score is for all flight states, the more suitable the grid cell is for landing. Grid cells whose landability score is less than the score threshold are determined to pass the landability determination.
[0075] Essentially, it is roughness cost. The lower the value, the flatter the terrain and the more suitable it is for landing. The judgment is based on the principle that the lower the value, the better, rather than the higher the value, the better. In this embodiment, roughness cost refers to the cost index that characterizes the roughness of the terrain by converting the terrain statistics of the grid cells into a normalized denominator and then weighting and summing them according to the scoring weights. The lower the value, the flatter the terrain and the more suitable it is for landing.
[0076] In one embodiment, before calculating the landability score of the grid cell, the method further includes: According to the fractional model corresponding to the identified flight state, the terrain statistics of the grid cell are converted to the baseline working condition of the flight state, and the landing score is calculated. The fractional models corresponding to different flight states are independent of each other. The conversion is applied to the calculation of the landing score without changing the hard hole determination.
[0077] Among them, the split-state model corresponding to the climb in various preset flight states is a linear model with the average altitude recorded at each sampling time within the map construction period as the independent variable. The conversion process includes normalizing the terrain statistics according to a preset reference altitude.
[0078] In this embodiment, the fractional model refers to a model established for a certain type of flight state, describing the variation of the terrain statistics of the grid cells with the operating conditions under that type of flight state; the operating conditions refer to physical quantities characterizing the operating conditions of that type of flight state, such as the flight altitude during climb; the reference operating condition refers to the operating condition used as a conversion reference under that type of flight state. The modeling methods for each state are independent of each other and can be different. Among them, the fractional model corresponding to climb in various preset flight states is a linear model with the average altitude recorded at each sampling time within the map construction period as the independent variable. The conversion process includes normalizing the terrain statistics according to the preset reference altitude. In this embodiment, the linear model of the independent variable refers to a model with the periodic average altitude Δ as the independent variable, and the terrain statistics change approximately linearly with Δ. Its slope is obtained by statistical fitting of the climb flight cycle. Other states can adopt linear or nonlinear models depending on their statistical laws. For example, the statistics in hovering and slow level flight states do not change with significant operating conditions, and their fractional model can adopt a constant model, that is, the terrain statistics are directly substituted without conversion. Scoring; the fractional model corresponding to the cruise state can be a linear model with horizontal speed characteristics as independent variables, and the terrain statistics are normalized according to a preset reference speed; the fractional model corresponding to the turning state can be a linear model with yaw rate characteristics as independent variables, and the terrain statistics are normalized according to a preset reference yaw rate; the slope of each of the above models is also obtained by statistical fitting of the actual flight cycle of the corresponding state.
[0079] Specifically, during ascent, as flight altitude increases, the elevation variance, thickness span, and thickness standard deviation of the same terrain on the grid cell increase approximately linearly; taking a reference altitude a0=20m as the baseline, the example value for the three slopes is k. r =0.0261m -1 k p =0.0401m -1 k q =0.0765m -1 In a specific example, the conversion is done by r′=r / (1+k) r ·(ā-a0)), p′=p / (1+k p ·(ā-a0)), q′=q / (1+k q ·(ā-a0)) normalizes the current statistic to the reference height, then substitutes r′, p′, and q′ into the equation. Rate it; In other embodiments, the above conversion can also be achieved using a piecewise linear model, a quadratic curve model, or a lookup table interpolation model that is segmented by height. The present invention does not limit the specific form of the fractional model.
[0080] Therefore, by performing statistical modeling and normalization before scoring, the systematic influence of operating condition differences (such as flight altitude during climb) on the statistics is eliminated, so that the same terrain can be judged consistently at low and high altitudes and under different flight conditions, avoiding the judgment results from drifting with operating conditions; at the same time, since the statistical modeling only applies to the calculation of the landability score and does not change the hard hole judgment, the safety bottom line is not relaxed.
[0081] In one embodiment, determining whether each of the grid cells is landable to obtain a landable area includes: Grid cells whose Chebyshev distance from the hard hole does not exceed a predetermined number are marked as near-hard hole cells, where the Chebyshev distance is the maximum of the absolute values of the differences between the horizontal and vertical indices of two grid cells; that is... ; The grid cells that simultaneously meet the following conditions are summarized and output as the landable area: It was not identified as the hard hole mentioned above; Not marked as the near-hard hole unit; The landability score is less than the score threshold.
[0082] In this embodiment, a near-hard hole cell refers to a grid cell whose Chebyshev distance from a hard hole does not exceed a predetermined number; the predetermined number refers to the Chebyshev distance threshold that defines the range of a near-hard hole cell, and is measured in grid cells, for example, 2 grid cells. At a grid resolution of 0.5 meters, this is approximately 1 meter. Units near hard holes are not included in the scoring and are excluded from the output of the landable area; even if a unit near a hard hole scores low, it is directly excluded from the output due to the second condition mentioned above. Thus, a safety buffer zone is formed around the hard hole, preventing the drone from landing on the edge of the hard hole and touching the dangerous area due to positioning errors, thereby improving the safety of the output landable area.
[0083] In other embodiments, the above-mentioned set quantity can be 1, 2, 3 or other values depending on the positioning accuracy and body size of the UAV; Chebyshev distance can also be replaced by other grid distance measurements such as Euclidean distance or Manhattan distance; the range of the safety buffer zone can also be adaptively adjusted according to the flight state, for example, a larger set quantity can be used in the cruise state, and the present invention does not limit this.
[0084] In summary, steps S110 to S140 constitute a complete closed loop for single-cell landing determination: first, hard hole elimination ensures a safety baseline; then, split-state modeling eliminates operational drift; next, scoring is performed based on the weight of state switching and the normalized denominator; finally, landing areas are output under two conditions: near-hard hole cell elimination and score boundary. This allows single-cell determination to form a complete and implementable closed loop under dynamic state parameters, and the output landing areas are complete, contiguous, and can be directly used by downstream landing planning modules.
[0085] Therefore, the above method overcomes the shortcomings of existing technologies, such as fixed landing area determination parameters, lack of differentiation between flight states, serious misjudgments and omissions in dynamic states, and drift of determination results with operating conditions. It provides an adaptive landing area determination method and system based on flight state, which identifies the flight state of the UAV in units of map building cycle and adaptively switches the landing area determination parameters accordingly, thereby improving the accuracy and robustness of landing area determination in each flight state.
[0086] Specifically, the following technical problems need to be addressed: How to quantify and identify the flight state (hovering, slow level flight, cruise, climb, descent, and turn) of the UAV during autonomous landing by utilizing its attitude trajectory within a single map building cycle, and use this as the basis for switching landing area determination parameters; How to adaptively switch the hard threshold, scoring weight, and scoring boundary for single-cell landing determination based on the identified flight state, so as to accurately and robustly determine the landing area even in dynamic states such as cruise and turning, reducing misjudgments and omissions; How to model the grid statistics for each flight state separately and normalize them according to their working conditions, eliminating the systematic influence of working condition differences (such as flight altitude during climb) on the statistics, so that the same terrain can obtain consistent landing determination results under different flight states and working conditions.
[0087] Specifically, using map building cycles as units, the drone's own attitude data is used to quantify and identify its flight state. Based on this, the hard threshold, scoring boundary, and scoring weight used for landing area determination are switched to match the determination parameters with the current flight state. This ensures accurate and robust determination of landing areas even in dynamic states such as cruising and turning, reducing misjudgments and omissions. Furthermore, by performing fractional modeling of grid statistics under each flight state before scoring and converting them to the baseline condition, the systematic influence of condition differences (such as flight altitude during climb) on the statistics is eliminated. This ensures that the same terrain yields consistent landing determination results under different flight states and conditions. At the same time, the landing area is output under two conditions: near-hard hole cell exclusion and landing score below the scoring boundary. This ensures the safety and integrity of the output results and provides reliable input for downstream landing location and trajectory planning.
[0088] The aforementioned method, during the autonomous landing of the UAV, collects the UAV's attitude data in a map building cycle and obtains a grid map containing multiple grid cells carrying terrain statistics corresponding to that cycle. Flight features are extracted based on the attitude data, and the UAV is identified as belonging to one of several preset flight states within that cycle. Accordingly, a set of preset judgment parameters corresponding to each flight state is selected to determine the landability of each grid cell. Before calculating the landability score, the terrain statistics are further converted to the baseline conditions of that state according to the fractional model corresponding to the identified flight state, eliminating variations in conditions within the same state. The systemic impact of quantitative changes is assessed, and the landable area is output under the dual conditions of excluding near-hard hole units and the landable score being below the scoring threshold. As a result, the flight state recognition drives the adaptive switching of judgment parameters, matching the variance threshold, thickness hard cap, scoring weight, and scoring threshold with the systematic differences in grid statistics under various flight states. This overcomes the shortcomings of existing technologies, such as fixed judgment parameters, serious misjudgments and omissions under dynamic states like cruise and turning, and the drift of judgment results with different operating conditions. This ensures that the same terrain can obtain consistent, accurate, and robust landable judgments under different flight states and operating conditions, thereby significantly improving the safety and reliability of UAV autonomous landing.
[0089] Figure 2 This is a schematic block diagram of an adaptive landable area determination system 300 based on flight state provided in an embodiment of the present invention. Figure 2 As shown, corresponding to the above-described adaptive landable area determination method based on flight state, the present invention also provides an adaptive landable area determination system 300 based on flight state. This adaptive landable area determination system 300 includes a unit for executing the above-described adaptive landable area determination method based on flight state, and the system can be configured in a server. Specifically, please refer to... Figure 2 The adaptive landable area determination system 300 based on flight status includes a data acquisition unit 301, a status recognition unit 302, a parameter selection unit 303, and an area determination unit 304.
[0090] The data acquisition unit 301 is used to acquire the attitude data of the UAV during the map building cycle and obtain the grid map corresponding to the map building cycle, wherein the grid map includes multiple grid cells, and each grid cell carries terrain statistics; the state recognition unit 302 is used to extract the flight characteristics of the UAV during the map building cycle based on the attitude data, and to identify the flight state of the UAV during the map building cycle based on the flight characteristics, wherein the flight state is one of a variety of preset flight states; the parameter selection unit 303 is used to select a set of judgment parameters corresponding to a set of preset judgment parameters based on the flight state; the area determination unit 304 is used to determine the landability of each grid cell based on the selected set of judgment parameters and the terrain statistics of each grid cell to obtain a landable area.
[0091] In one embodiment, the state recognition unit 302 is used to calculate the horizontal velocity feature, the absolute mean vertical velocity feature, and the mean vertical velocity band sign feature based on the three-axis velocities at each sampling time; to untangle the yaw angles of adjacent sampling times to obtain the yaw angle difference, and to calculate the yaw rate feature based on the yaw angle difference; and to calculate the altitude change feature based on the altitudes at the beginning and end of the map construction cycle sampling times.
[0092] In one embodiment, the state recognition unit 302 is configured to determine vertical motion if the absolute average value of the vertical speed feature exceeds a vertical speed threshold, or the absolute value of the altitude change feature exceeds an altitude change threshold, and determine the flight state as climb or descent based on the direction of vertical motion; if the absolute average value of the vertical speed feature does not exceed a vertical speed threshold, and the absolute value of the altitude change feature does not exceed an altitude change threshold, and the horizontal speed feature exceeds a cruise speed threshold, then the flight state is determined as cruise; if the absolute average value of the vertical speed feature does not exceed a vertical speed threshold, and the absolute value of the altitude change feature does not exceed an altitude change threshold, and the yaw rate feature exceeds a yaw rate threshold, then the flight state is determined as turning; if the absolute average value of the vertical speed feature does not exceed a vertical speed threshold, and the absolute value of the altitude change feature does not exceed an altitude change threshold, and the yaw rate feature exceeds a yaw rate threshold, then the flight state is determined as turning; if the absolute average value of the vertical speed feature does not exceed a vertical speed threshold, and the absolute value of the altitude change feature does not exceed an altitude change threshold, and the absolute value of the horizontal speed feature exceeds a yaw rate threshold, then the flight state is determined as turning; if the absolute average value of the vertical speed feature does not exceed a vertical speed threshold, and the absolute value of the altitude change feature exceeds an altitude change ... then the flight state is determined as turning. If the mean value of the horizontal speed characteristic does not exceed the vertical speed threshold, and the absolute value of the altitude change characteristic does not exceed the altitude change threshold, then the flight state is determined to be slow level flight if the horizontal speed characteristic exceeds the hovering speed threshold; if the absolute mean value of the vertical speed characteristic does not exceed the vertical speed threshold, and the absolute value of the altitude change characteristic does not exceed the altitude change threshold, then the flight state is determined to be hovering; wherein, determining the flight state as climb or descent based on the vertical direction of motion includes: if the absolute value of the altitude change characteristic exceeds the altitude change threshold, then determining climb or descent based on the sign of the altitude change characteristic; if the absolute value of the altitude change characteristic does not exceed the altitude change threshold, then determining climb or descent based on the sign of the mean value of the vertical speed band sign.
[0093] In one embodiment, the region determination unit 304 is configured to determine a grid cell that meets any of the following conditions as a hard hole: the semantic category is non-landable; the square root of the static elevation variance exceeds the variance threshold; the static elevation thickness span exceeds the thickness hard cap.
[0094] In one embodiment, the region determination unit 304 is used to determine the grid cells that are not determined to be hard holes according to... Calculate the landing score of the grid cell. Where r, p, and q are the square root of the static elevation variance, the static elevation thickness span, and the static elevation thickness standard deviation, respectively. , , For the multiple scoring weights, , , The lower the landability score, the more suitable the grid cell is for landing, and the lower the landability score is for landing. Grid cells with landability scores less than the score threshold are determined to pass the landability determination.
[0095] In one embodiment, the system further includes a conversion unit, used to convert the terrain statistics of the grid cell to the baseline condition of the flight state according to the fractional model corresponding to the identified flight state, and calculate the landing score. The fractional models corresponding to different flight states are independent of each other, and the conversion is applied to the calculation of the landing score. The fractional model corresponding to the climb in various preset flight states is a linear model with the average altitude recorded at each sampling time within the map construction period as the independent variable. The conversion process includes normalizing the terrain statistics according to a preset reference altitude.
[0096] In one embodiment, the region determination unit 304 is used to mark grid cells whose Chebyshev distance from the hard hole does not exceed a set number as near-hard hole cells, wherein the Chebyshev distance is the maximum value of the absolute values of the differences between the horizontal and vertical indices of two grid cells; and to summarize and output the grid cells that simultaneously meet the following conditions as the landable region: not judged as the hard hole; not marked as near-hard hole cells; and the landable score is less than the score boundary.
[0097] It should be noted that those skilled in the art can clearly understand that the specific implementation process of the above-mentioned adaptive landable area determination system 300 based on flight status and each unit can be referred to the corresponding description in the foregoing method embodiments. For the sake of convenience and brevity, it will not be repeated here.
[0098] The aforementioned adaptive landable area determination system 300 based on flight status can be implemented as a computer program, which can be used in, for example... Figure 3 It runs on the computer device shown.
[0099] Please see Figure 3 , Figure 3 This is a schematic block diagram of a computer device provided in an embodiment of this application. The computer device 500 can be a server, wherein the server can be a standalone server or a server cluster composed of multiple servers.
[0100] See Figure 3 The computer device 500 includes a processor 502, a memory, and a network interface 505 connected via a system bus 501. The memory may include a non-volatile storage medium 503 and internal memory 504.
[0101] The non-volatile storage medium 503 may store an operating system 5031 and a computer program 5032. The computer program 5032 includes program instructions that, when executed, cause the processor 502 to perform an adaptive landable area determination method based on flight status.
[0102] The processor 502 provides computing and control capabilities to support the operation of the entire computer device 500.
[0103] The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute an adaptive landing area determination method based on flight status.
[0104] This network interface 505 is used for network communication with other devices. Those skilled in the art will understand that... Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device 500 to which the present application is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0105] The processor 502 is used to run a computer program 5032 stored in a memory to implement all the steps of the adaptive landable area determination method based on flight status.
[0106] It should be understood that in the embodiments of this application, the processor 502 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.
[0107] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program includes program instructions and can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the embodiments of the above methods.
[0108] Therefore, the present invention also provides a storage medium. This storage medium may be a computer-readable storage medium. The storage medium stores a computer program, wherein when executed by a processor, the computer program causes the processor to perform all the steps of the adaptive landable area determination method based on flight state.
[0109] The storage medium can be any computer-readable storage medium capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), magnetic disk, or optical disk.
[0110] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.
[0111] In the embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of each unit is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.
[0112] The steps in the method of this invention can be adjusted, merged, or reduced in order according to actual needs. The units in the system of this invention can be merged, divided, or reduced according to actual needs. Furthermore, the functional units in the various embodiments of this invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.
[0113] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.
[0114] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
Claims
1. An adaptive landable area determination method based on flight state, characterized in that, include: During the map building cycle, attitude data of the UAV is collected, and a grid map corresponding to the map building cycle is obtained. The grid map includes multiple grid cells, and each grid cell carries terrain statistics. The flight features of the UAV during the map building cycle are extracted based on the attitude data, and the flight state of the UAV during the map building cycle is identified based on the flight features, wherein the flight state is one of a variety of preset flight states; Based on the flight state, select a corresponding set of judgment parameters from multiple sets of preset judgment parameters that correspond one-to-one with the various preset flight states; Based on the selected set of determination parameters and the terrain statistics of each grid cell, a landing capability determination is made for each grid cell to obtain a landing area. The preset flight states include hovering, slow level flight, cruise, climb, descent, and turning.
2. The adaptive landable area determination method based on flight state according to claim 1, characterized in that, The attitude data includes the three-axis velocity, yaw angle, and altitude of the UAV recorded at sampling times during the map building cycle; The flight characteristics include horizontal speed characteristics, vertical speed characteristics, yaw rate characteristics, and altitude change characteristics, wherein the vertical speed characteristics include the absolute mean vertical speed characteristics and the mean vertical speed band sign characteristics.
3. The adaptive landable area determination method based on flight state according to claim 2, characterized in that, The step of extracting the flight features of the UAV during the map building period based on the attitude data includes: Based on the triaxial velocities at each sampling time, calculate the horizontal velocity characteristic, the absolute mean vertical velocity characteristic, and the sign mean vertical velocity characteristic. The yaw angles at adjacent sampling times are unwound to obtain the yaw angle difference, and the yaw rate characteristic is calculated based on the yaw angle difference. The height change characteristics are calculated based on the height at the beginning and end of the map construction cycle sampling time.
4. The adaptive landable area determination method based on flight state according to claim 2 or 3, characterized in that, The preset flight states include hovering, slow level flight, cruise, climb, descent, and turning; The step of identifying the flight status of the UAV within the map building cycle based on the flight characteristics includes: If the absolute average value of the vertical velocity exceeds the vertical velocity threshold, or the absolute value of the altitude change characteristic exceeds the altitude change threshold, then it is determined to be vertical motion, and the flight state is determined to be climbing or descending based on the direction of vertical motion. If the absolute average value of the vertical speed does not exceed the vertical speed threshold, and the absolute value of the altitude change feature does not exceed the altitude change threshold, then when the horizontal speed feature exceeds the cruise speed threshold, the flight state is determined to be cruise. If the absolute mean value of the vertical speed does not exceed the vertical speed threshold, and the absolute value of the altitude change characteristic does not exceed the altitude change threshold, then when the yaw rate characteristic exceeds the yaw rate threshold, the flight state is determined to be a turn. If the absolute average value of the vertical speed does not exceed the vertical speed threshold, and the absolute value of the altitude change feature does not exceed the altitude change threshold, and the horizontal speed feature exceeds the hovering speed threshold, then the flight state is determined to be slow level flight. If the absolute average value of the vertical velocity does not exceed the vertical velocity threshold, and the absolute value of the altitude change feature does not exceed the altitude change threshold, the flight state is determined to be hovering. The step of determining the flight state as either ascent or descent based on the vertical direction of motion includes: If the absolute value of the height change feature exceeds the height change threshold, then the ascent or descent is determined according to the sign of the height change feature; If the absolute value of the height change feature does not exceed the height change threshold, the ascent or descent is determined based on the sign of the mean value of the vertical speed band sign.
5. The adaptive landable area determination method based on flight state according to claim 2 or 3, characterized in that, Each set of judgment parameters includes a variance threshold, a thickness hard cap, a score boundary, and multiple scoring weights; In each set of judgment parameters corresponding to different flight states, the variance threshold, the thickness hard cap, the score boundary, and the scoring weight are independently valued, and each set of judgment parameters is obtained by fractal statistical calibration of actual flight data under the corresponding flight state.
6. The adaptive landable area determination method based on flight state according to claim 5, characterized in that, The topographic statistics include static elevation variance, static elevation thickness span, and static elevation thickness standard deviation; the grid cell also carries a semantic category. Determining whether each of the grid cells is landable includes: A grid cell that meets any of the following conditions is considered a hard hole: The semantic category is non-landable; The square root of the static elevation variance exceeds the variance threshold; The static elevation thickness span exceeds the thickness hard cap.
7. The adaptive landable area determination method based on flight state according to claim 6, characterized in that, The method for determining the landability of each of the grid cells also includes: For grid cells not identified as hard holes, according to Calculate the landing score of the grid cell. Where r, p, and q are the square root of the static elevation variance, the static elevation thickness span, and the static elevation thickness standard deviation, respectively. , , For the multiple scoring weights, , , The lower the landability score, the more suitable the grid cell is for landing, and the lower the landing score is for all flight states, the more suitable the grid cell is for landing. Grid cells whose landability score is less than the score threshold are determined to pass the landability determination.
8. The adaptive landable area determination method based on flight state according to claim 7, characterized in that, Before calculating the landability score of the grid cell, the method further includes: According to the fractional model corresponding to the identified flight state, the terrain statistics of the grid cell are converted to the baseline working condition of the flight state, and the landing score is calculated. The fractional models corresponding to different flight states are independent of each other, and the conversion is applied to the calculation of the landing score. Among them, the split-state model corresponding to the climb in various preset flight states is a linear model with the average altitude recorded at each sampling time within the map construction period as the independent variable. The conversion process includes normalizing the terrain statistics according to a preset reference altitude.
9. The adaptive landable area determination method based on flight state according to claim 7 or 8, characterized in that, The step of determining whether each of the grid cells is landable to obtain a landable area includes: A grid cell whose Chebyshev distance from the hard hole does not exceed a set number is marked as a near-hard hole cell, wherein the Chebyshev distance is the maximum value of the absolute values of the differences between the horizontal and vertical indices of the two grid cells; The grid cells that simultaneously meet the following conditions are summarized and output as the landable area: It was not identified as the hard hole mentioned above; Not marked as the near-hard hole unit; The landability score is less than the score threshold.
10. An adaptive landable area determination system based on flight state, characterized in that, include: The data acquisition unit is used to acquire the attitude data of the UAV during the map building cycle and obtain the grid map corresponding to the map building cycle. The grid map includes multiple grid cells, and each grid cell carries terrain statistics. A state recognition unit is used to extract the flight features of the UAV during the map building cycle based on the attitude data, and to identify the flight state of the UAV during the map building cycle based on the flight features, wherein the flight state is one of a variety of preset flight states; The parameter selection unit is used to select a set of judgment parameters from multiple sets of preset judgment parameters that correspond one-to-one with the multiple preset flight states according to the flight state. The area determination unit is used to determine whether each of the grid cells is landable based on the selected set of determination parameters and the terrain statistics of each grid cell, so as to obtain the landable area. The preset flight states include hovering, slow level flight, cruise, climb, descent, and turning.
