Vehicle-mounted unmanned aerial vehicle slope take-off and landing control method and control system

Abstract

This invention discloses a method and control system for controlling the take-off and landing of a vehicle-mounted unmanned aerial vehicle (UAV) on a slope, belonging to the field of vehicle-mounted UAV control technology. The invention acquires point cloud data through a terrain perception module, fits a ground plane, and establishes a dynamic take-off and landing coordinate system based on this plane. Based on this coordinate system, the component of gravity on the slope is calculated, generating a feedforward thrust compensation amount and superimposing it onto the motor commands, enabling the UAV to achieve force balance before unlocking. During take-off, the UAV climbs vertically in the dynamic take-off and landing coordinate system and smoothly transitions the control target to the world coordinate system through linear interpolation to achieve stable hovering. During landing, it smoothly switches back to the dynamic take-off and landing coordinate system bound to the vehicle's posture, fusing relative navigation information to achieve precise landing. This invention fundamentally eliminates the risk of overturning during slope take-off and landing, improves control stability and environmental adaptability, and significantly enhances the operational capabilities of vehicle-mounted UAVs in complex terrain.

Description

Technical Field

[0001] This invention relates to the field of vehicle-mounted unmanned aerial vehicle (UAV) control technology, specifically to a method and control system for controlling the take-off and landing of a vehicle-mounted UAV on a slope. Background Technology

[0002] As drones and vehicles increasingly collaborate in various scenarios (such as field exploration, emergency command, and military reconnaissance), vehicle-mounted drones need to be able to be rapidly deployed and retrieved in any terrain where the vehicle is parked. Among these challenges, ramp takeoff and landing is one of the core technological hurdles currently facing.

[0003] Currently, methods for ramp takeoff and landing of drones include, for example, the patent with patent number "CN112987780A" entitled "Multirotor Aircraft Ramp Takeoff Method and Multirotor Aircraft Ramp Landing Method". This method involves the aircraft using its own sensors to obtain the tilt angle of the fuselage before takeoff, i.e., the tilt angle 'a' of the ramp where the aircraft is positioned. If 'a' is greater than a first angle, the aircraft remains stationary and does not take off. If 'a' is less than or equal to the first angle, the aircraft gradually adjusts the power of all rotors to the first power. Based on the images captured by the FPV camera, the aircraft is analyzed to determine if sideslip has occurred. If sideslip is detected, the aircraft shuts off the power of all rotors and enters an alarm state. If sideslip is not detected and the aircraft remains stationary, the aircraft immediately adjusts the power of all rotors to the second power. Based on the images captured by the FPV camera, the aircraft is analyzed to determine if takeoff has occurred. If not, the aircraft shuts off the power of all rotors and enters an alarm state. If so, the aircraft adjusts the tilt angle of the fuselage after takeoff based on sensor data and enters normal flight mode.

[0004] This method can control the takeoff of a drone when it is on a slope, but this control method has a major problem. The method is based on a safety judgment, that is, it will not take off when the slope is steep. This judgment method obviously does not meet the usage requirements of vehicle-mounted drones.

[0005] Moreover, current traditional drone takeoff and landing are based on the horizontal plane or the "world horizontal plane" defined by the inertial navigation system (INS). When a vehicle is parked on a slope, if the drone still takes off based on the horizontal plane, its initial attitude will have a huge angle with the real ground (slope), which can easily lead to lateral slippage or even overturning and collision at the moment of takeoff.

[0006] Furthermore, on a slope, gravitational acceleration produces a continuous component in the direction parallel to the slope. Traditional PID position or speed control models do not dynamically compensate for this, causing the UAV to consume a large amount of extra energy to resist the descent during takeoff / landing, resulting in severe control loop coupling, violent attitude oscillations, and poor hovering stability.

[0007] During the landing phase, the drone needs to land precisely on the roof platform. On a slope, the roof platform's plane is not level with the horizontal plane. Traditional methods require the vehicle to level itself or the drone to perform complex visual servo alignment. The former places high demands on the vehicle, while the latter is unreliable under complex lighting and texture conditions, easily leading to landing impacts, landing gear damage, or docking failures. Summary of the Invention

[0008] The purpose of this application is to address the shortcomings of the aforementioned background technology and provide a method and control system for controlling the take-off and landing of vehicle-mounted unmanned aerial vehicles on slopes.

[0009] The technical solution of this application is: a method for controlling the take-off and landing of a vehicle-mounted unmanned aerial vehicle on a slope, comprising, Based on the point cloud data of the terrain around the vehicle, a local ground plane is fitted in the vehicle coordinate system, and a dynamic take-off and landing coordinate system based on this plane is established. Based on the dynamic take-off and landing coordinate system and vehicle attitude information, the ground normal vector in the world coordinate system is calculated, and a feedforward thrust compensation amount is generated to offset the gravity component along the slope. The feedforward thrust compensation amount is then superimposed on the UAV motor commands, and the UAV is unlocked. Vertical takeoff is performed in the dynamic takeoff and landing coordinate system, and after reaching the preset altitude, the position control target coordinate system is smoothly transitioned from the dynamic takeoff and landing coordinate system to the world coordinate system to achieve hovering; Upon receiving the landing command, the control coordinate system is smoothly switched from the world coordinate system back to the dynamic take-off and landing coordinate system bound to the current vehicle position and orientation, and high-precision landing is performed by integrating relative navigation information.

[0010] According to the method for controlling the take-off and landing of a vehicle-mounted unmanned aerial vehicle on a ramp provided in this application, the method for establishing a dynamic take-off and landing coordinate system based on the plane includes: acquiring point cloud data by collecting data of the area surrounding the vehicle using a binocular camera or lidar; performing plane fitting on the point cloud data using a random sampling consensus algorithm to obtain the ground plane equation and its unit normal vector; defining the Z-axis direction of the dynamic take-off and landing coordinate system based on the unit normal vector, the X-axis being the projection of the vertical axis of the vehicle coordinate system onto the fitted ground plane, and the Y-axis being determined according to the right-hand rule.

[0011] According to the vehicle-mounted unmanned aerial vehicle (UAV) ramp takeoff and landing control method provided in this application, the calculation method for the feedforward thrust compensation includes: calculating the feedforward thrust compensation according to the following formula. in: —Feedforward thrust compensation; —The rotation matrix from the world coordinate system to the UAV body coordinate system; —Ground unit normal vector in the world coordinate system; —Drone quality; —The vector of gravitational acceleration, directed vertically downwards.

[0012] According to the vehicle-mounted unmanned aerial vehicle (UAV) ramp takeoff and landing control method provided in this application, the method for smoothly transitioning the position control target coordinate system from the dynamic takeoff and landing coordinate system to the world coordinate system includes: mixing the target position in the dynamic takeoff and landing coordinate system and the world coordinate system through linear interpolation. in: —The target position is a combination of the dynamic take-off and landing coordinate system and the world coordinate system; —Smooth transition from 1 to 0; —Target location in the world coordinate system; —Target position in dynamic take-off and landing coordinate system; —Transformation matrix from dynamic takeoff and landing coordinate system to world coordinate system.

[0013] According to the vehicle-mounted UAV ramp take-off and landing control method provided in this application, before the UAV lands, a real-time safety assessment is performed on the terrain within a predetermined radius around the vehicle. If there is an area that meets the safety threshold, a landing command is issued; otherwise, an unsuitable landing alarm is triggered.

[0014] According to the vehicle-mounted UAV ramp take-off and landing control method provided in this application, the safety assessment method includes: dividing the area around the vehicle into multiple fan-shaped sub-regions, calculating the slope index, flatness index, obstacle density index and material index for each sub-region, and weighting them to synthesize a comprehensive safety score; when the comprehensive safety score of any sub-region is not lower than a preset threshold and the slope and obstacle density both meet the constraints, it is determined that it can land directly.

[0015] According to the vehicle-mounted UAV ramp take-off and landing control method provided in this application, when it is determined that the landing is not suitable, the UAV takes off to expand the reconnaissance range, constructs a terrain safety heat map, identifies candidate areas that meet the conditions of size, slope, flatness and obstacle-free access, and calculates a comprehensive recommendation index based on safety score, distance from vehicle, area area and passage cost to recommend the best or second best landing point to the user.

[0016] According to the vehicle-mounted UAV ramp take-off and landing control method provided in this application, the UAV continuously tracks the vehicle and the position of the recommended point while the vehicle is driving towards the recommended landing point, and monitors the changes in the surrounding environment of the recommended point in real time. If new dangerous factors are detected, the recommendation results are dynamically updated.

[0017] According to the vehicle-mounted UAV ramp take-off and landing control method provided in this application, after the vehicle arrives at the recommended landing point, the terrain of the new location is rescanned, the dynamic take-off and landing coordinate system is reconstructed, the gravity feedforward compensation parameters are updated, and after verifying the terrain safety, the standard landing procedure is executed to complete the precise docking.

[0018] This application also relates to a vehicle-mounted unmanned aerial vehicle (UAV) ramp takeoff and landing control system, including: The terrain perception module is used to acquire point cloud data around the vehicle; The dynamic take-off and landing coordinate system construction module is used to fit the ground plane and establish a dynamic take-off and landing coordinate system; The gravity compensation calculation module is used to generate the feedforward thrust compensation amount; The coordinate system transition control module is used to manage the smooth switching between the dynamic take-off and landing coordinate system and the world frame; The terrain safety assessment module is used to calculate multi-dimensional safety indicators and determine landing suitability; The collaborative optimization and recommendation module is used to search, rate, and recommend alternative landing points; The precision landing control module is used to integrate relative navigation information to perform a high-precision landing; The vehicle-to-vehicle communication interface is used to synchronize vehicle position and orientation information and dynamic take-off and landing coordinate system parameters.

[0019] The advantages of this application are: 1. This application creatively introduces a dynamic take-off and landing coordinate system. By sensing the terrain in real time and establishing a coordinate system based on the real slope, it fundamentally solves the problem of mismatch between the attitude of the UAV and the ground when taking off and landing on the slope, ensuring the physical alignment at the moment of take-off and landing, and eliminating sideslip and overturning caused by the error of the reference. 2. This application does not rely on traditional post-correction PID control, but calculates the feedforward thrust compensation amount through a precise physical model, and actively counteracts the adverse effects of gravity before the control command is issued. This pre-control strategy enables the UAV to achieve a force balance state on the slope, significantly reduces the coupling and oscillation of the control loop, improves hovering stability, and reduces energy consumption. 3. This application perfectly connects the three different stages of takeoff (dynamic system), hovering (world system), and landing (dynamic system) through a unique coordinate system smooth transition mechanism; through algorithms such as linear interpolation, it ensures the continuity of control commands and avoids flight attitude jumps and oscillations caused by sudden changes in coordinate system, providing smooth and reliable control quality for the UAV's full-process autonomous flight on slopes. 4. This application not only solves the control problem of "how to land", but also intelligently solves the decision-making problems of "whether to land" and "where to land". Through multi-dimensional terrain safety assessment and UAV-vehicle collaborative optimization mechanism, when the original location is not suitable for landing, the system can autonomously expand the search range, comprehensively evaluate multiple factors such as safety, distance, and area, recommend the best alternative landing point for the user, and can dynamically update the plan according to environmental changes. This makes the whole system highly adaptable and capable of completing tasks in unstructured environments such as the wild and mountains. Attached Figure Description

[0020] Figure 1 The method for controlling the takeoff of a vehicle-mounted unmanned aerial vehicle (UAV) as described in this application; Figure 2 The present application discloses a method for controlling the landing of a vehicle-mounted unmanned aerial vehicle. Detailed Implementation

[0021] The embodiments of this application are described in detail below, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0022] In the description of this application, it should be understood that the terms "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0023] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0024] The present application will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0025] This application relates to a method for controlling the takeoff and landing of a vehicle-mounted unmanned aerial vehicle (UAV) on a slope, used to control the UAV's takeoff and landing when the vehicle is on a slope. The control method of this application abandons the traditional horizontal reference and uses the RANSAC algorithm to fit the terrain surrounding the vehicle in real time, establishing a dynamic coordinate system based on the actual slope, thus eliminating the risk of takeoff rollover at the source. Based on a physical model, it accurately calculates and superimposes thrust compensation to actively offset the component of gravity along the slope, achieving a force balance state on the slope and significantly improving hovering stability. During takeoff and landing, linear interpolation is used to achieve seamless switching between the dynamic frame and the world frame, ensuring the continuity and smoothness of flight attitude. A multi-dimensional safety score is performed on the terrain; if landing in place is not possible, the UAV takes off to expand the search, recommending the best alternative landing point based on multi-objective optimization, and possessing dynamic tracking and update capabilities.

[0026] Specifically, when the vehicle is parked on the slope and the drone is ready to take off, such as Figure 1 and 2 As shown, the following steps can be followed: S1. Establish a dynamic takeoff and landing coordinate system The vehicle's surrounding terrain is scanned by onboard sensors (such as lidar or cameras) to obtain point cloud data. In the vehicle coordinate system, these point cloud data are fitted to a plane to obtain the ground plane equation of the area where the vehicle is located. Based on this fitted plane, a dynamic take-off and landing coordinate system is established, which is updated in real time according to the vehicle's attitude and terrain. This dynamic take-off and landing coordinate system takes the plane normal vector as the Z-axis, which represents the ideal take-off / landing direction of the UAV. S2, Feedforward Gravity Compensation and Unlocking Based on the dynamic take-off and landing coordinate system established in step S1 and the current attitude information of the vehicle, the ground normal vector in the world coordinate system is calculated; based on this normal vector, the mass of the UAV and the gravitational acceleration, a feedforward thrust compensation amount is calculated to offset the gravity along the slope component; this compensation amount is superimposed on the motor speed command of the UAV, and the UAV is unlocked after the compensation amount is stably loaded. S3, Smooth Coordinate System Transition Takeoff The UAV executes vertical take-off commands in the dynamic take-off and landing coordinate system, meaning that the UAV's motion control is entirely based on the slope plane. After the UAV climbs to the preset safe altitude, it initiates a coordinate system transition program to smoothly transition the target coordinate system of position control from the dynamic take-off and landing coordinate system to the world coordinate system (i.e., the North-East-Earth coordinate system), and finally achieves stable hovering in the world coordinate system. S4, Reverse Transition Landing Once the drone receives the landing command, it initiates a reverse transition procedure, smoothly switching the control coordinate system from the world coordinate system back to the dynamic take-off and landing coordinate system that is currently bound to the vehicle's pose in real time. In this coordinate system, relative navigation information such as vision, RTK-GPS, or UWB is integrated to guide the drone to land precisely and gently on the vehicle's take-off and landing platform.

[0027] The core principle of this application lies in reference reconstruction and physical feedforward. The fundamental reason for the failure of traditional methods is the mismatch between their control reference (horizontal plane) and the physical environment (slope). This application dynamically adjusts the control reference to be consistent with the physical environment by sensing the environment in real time, thus fundamentally eliminating reference deviation. At the same time, the component of gravity along the slope is a deterministic disturbance that can be accurately calculated. This disturbance is directly canceled by feedforward compensation, so that the feedback controller only needs to deal with uncertain disturbances, which greatly reduces the control difficulty.

[0028] This application completely eliminates the problem of the angle between the attitude and the ground at the moment of takeoff by using a dynamic coordinate system, thus eliminating the risk of sideslip and overturning; feedforward compensation enables the UAV to achieve a force balance state even on slopes, making takeoff and hovering more stable and energy-efficient; the control method of this application covers all key steps from takeoff preparation to final landing, with complete logic, providing a solid foundation for subsequent refined improvements.

[0029] In some embodiments of this application, the method for constructing the dynamic take-off and landing coordinate system in step S1 above has been optimized. Specifically, it can be performed according to the following steps: S11. Using a binocular camera or lidar installed on the vehicle, scan the area around the vehicle (e.g., within a radius of 5-10 meters) to obtain high-density three-dimensional point cloud data. S12. The Random Sample Consensus (RANSAC) algorithm is used to perform plane fitting on the acquired point cloud data. The RANSAC algorithm, through an iterative approach, robustly estimates the optimal ground plane equation, Ax + By + Cz + D = 0, from the point cloud containing noise and outliers (such as weeds and small stones), and calculates the unit normal vector n of the plane. g ; S13. Construct a dynamic takeoff and landing coordinate system (DL system) based on this normal vector, with the Z-axis (Z... DL ): Defined as the unit normal vector n of the fitted plane. g The direction is perpendicular to the ground and upwards; X-axis (X) DL ): Take the projection vector of the vertical axis of the vehicle coordinate system (usually the vehicle's forward direction) onto the fitting plane; this projection ensures that the forward direction of the UAV is aligned with the forward direction of the vehicle on the slope plane; Y-axis ((Y DLAccording to the right-hand rule, it is determined by multiplying the Z-axis by the X-axis, i.e., Y = Z × X.

[0030] Thus, a dynamic take-off and landing coordinate system centered on the vehicle and perfectly adapted to the sloping terrain was constructed.

[0031] The introduction of the RANSAC algorithm is key to ensuring the robustness of the coordinate system. In complex field environments, sensor data inevitably contains noise and invalid points. RANSAC can effectively identify and remove these outliers, ensuring that the fitted plane truly represents the macroscopic ground available for UAV take-off and landing, rather than a local plane disturbed by tiny objects. Projecting the vehicle's longitudinal axis as the X-axis is to achieve natural alignment between the UAV and the vehicle in the direction of motion, laying the foundation for subsequent accurate tracking and landing.

[0032] The RANSAC algorithm in this embodiment ensures that even in real environments with disturbances such as weeds and gravel, it can fit a stable and reliable ground plane, providing an accurate environmental model for the entire control method. The definition of the dynamic take-off and landing coordinate system, especially the determination of the X-axis, establishes a geometric link between the vehicle and the UAV, so that the orientation of the UAV during take-off and landing is naturally consistent with that of the vehicle, simplifying the complexity of relative pose calculation.

[0033] In other embodiments of this application, the method for calculating the feedforward thrust compensation in step S2 above has been optimized. Specifically, after obtaining the ground unit normal vector in the world coordinate system... Then, the airborne flight controller calculates in real time the thrust compensation amount that needs to be superimposed on the motor commands according to the following formula: in: —The feedforward thrust compensation is a thrust vector expressed in the body coordinate system to counteract the gravitational component; —The rotation matrix from the world coordinate system to the UAV body coordinate system, and the rotation matrix from the world coordinate system {W} to the UAV body coordinate system {b}, are determined in real time by the current attitude of the UAV (IMU data); —The ground unit normal vector in the world coordinate system {W}; —Drone quality; —The gravitational acceleration vector, vertically downwards in the world coordinate system, has a magnitude of approximately 9.8m. 2 / s.

[0034] In this embodiment, the feedforward thrust compensation is calculated as the projection length of the gravitational acceleration vector onto the ground normal vector. Multiplying this by the normal vector yields the component vector of gravitational acceleration in the direction perpendicular to the ground. Multiplying this vector by the mass gives the force of gravity in the direction perpendicular to the ground. This force is the root cause of the UAV's downward gliding tendency on the slope. The formula calculates the expression of the force to be offset in the world coordinate system, then transforms it to the current dynamic take-off and landing coordinate system using a rotation matrix. Finally, taking the negative sign means that an equal and opposite force needs to be applied to offset it.

[0035] The formula in this embodiment is a precise mathematical expression of Newtonian mechanics in this application scenario, transforming complex control problems into clear physical quantity calculations, ensuring the accuracy and effectiveness of compensation. By accurately offsetting known gravitational disturbances, feedback controllers such as PID controllers only need to focus on handling unknown disturbances such as wind resistance and model uncertainty, greatly simplifying control parameter tuning and improving the dynamic response performance and robustness of the system. The variables in the formula are all quantities that can be measured in real time or are known, with low computational load, making it easy to implement real-time calculations on embedded platforms.

[0036] In a further embodiment of this application, the coordinate system smoothing transition method in step S3 above has been optimized. Specifically, the target position in the dynamic take-off and landing coordinate system and the world coordinate system is mixed by linear interpolation. in: —The target position, which is a combination of the dynamic take-off and landing coordinate system and the world coordinate system, is input to the position controller at time t; —A time-varying mixing factor that smoothly decreases from 1 (at the start of takeoff) to 0 during the takeoff transition phase; for example, it could be designed as follows: =1-t / T Where T is the preset total transition time (e.g., 2 seconds), or is designed as a function related to the current height; —The target position in the world coordinate system, and the expected final hovering target position in the world coordinate system; —Target position in dynamic takeoff and landing coordinate system, the desired takeoff target position in dynamic takeoff and landing coordinate system (usually a certain height ascended along the Z-axis). —The transformation matrix from the dynamic takeoff and landing coordinate system to the world coordinate system is used to transform the target point in the dynamic takeoff and landing system to the world system.

[0037] In the initial stage of the transition, λ(t)≈1, the mixed position is mainly determined by the target in the dynamic take-off and landing system, and the UAV climbs vertically; as λ(t) gradually decreases, the target weight in the world coordinate system gradually increases, and the UAV begins to smoothly correct its horizontal position until λ(t)=0, the mixed position is completely equal to the hovering target point in the world coordinate system, and the transition is completed.

[0038] This embodiment uses a time-varying factor λ(t) to linearly superimpose control targets in two different coordinate systems. This is equivalent to allowing the control reference point to drift from one coordinate system to another, rather than jumping. This continuous and gradual process ensures the continuity of control commands and avoids control quantity jumps caused by sudden changes in the target.

[0039] The linear interpolation method in this embodiment is simple and reliable, effectively eliminating the impact of coordinate system switching, ensuring the continuity of UAV attitude and position, and greatly improving flight quality and passenger experience. The λ(t) function can be flexibly designed according to the specific UAV dynamic characteristics and environmental requirements (such as linear, S-curve, etc.), providing engineers with the freedom to adjust the softness or hardness of the transition process. This hybrid method is not only applicable to takeoff transition, but also to the landing reverse transition process in claim 1, simply by reversing the transition direction of λ(t).

[0040] In a preferred embodiment of this application, step S4 described above has been optimized, specifically as follows: Figure 2 As shown, this embodiment constructs a complete intelligent safe landing process, especially suitable for situations where a vehicle cannot land safely in place. Specifically, it includes the following steps: S41. Preliminary Security Assessment When the drone is preparing to land, it first scans the area around the vehicle (e.g., a radius of 20-50 meters); the area is then divided into multiple fan-shaped sectors (e.g., one sector every 30°, for a total of 12); for each sector, the slope, flatness (root mean square of residual), obstacle density, and material properties based on visual / laser reflectivity (e.g., grass, hard road surface, water surface, etc.) are calculated. a. Slope index S i in: S i —Slope index; —The maximum slope angle of the i-th region; —Preset safe slope threshold (e.g., 15°); b. Flatness index F i in: F i —Smoothness index; —Standard deviation of elevation for region i; —Maximum permissible standard deviation of elevation; c. Obstacle density index O i in: O i —Obstacle density index; —The number of obstacles above the safe height; —The area of ​​the i-th region; —Maximum permissible obstacle density; d. Material specifications M i M i ={1.0 (hard, flat surface), 0.6 (short grass), 0.3 (soft sand), 0.1 (water / muddy)}; Overall safety score calculation: SafetyScore i =w 1 ⋅(2-S i )+w 2 ⋅(1-F i )+w 3 ⋅(1-O i )+w 4 ⋅M i in w 1 +w 2 +w 3 +w 4 =1 Weighting coefficients (generally: w 1 =0.4, w 2 =0.3,w 3 =0.2, w 4 =0.1).

[0041] Set a safety threshold t safe (e.g., 1.2); S42, Direct Landing Decision If at least one sector's overall score exceeds the set safety threshold, and both the slope and obstacle density of that sector meet the safety constraints, then... SafetyScore i ≥ t safe and S i <1 and O i If the value is less than 0.5, the system determines that the sector is suitable for direct landing and sends a landing command to the drone. S43, Collaborative Optimization Recommendation If no sector meets the conditions (all areas) SafetyScore i < t safe If the location is deemed unsuitable for landing, the drone increases its altitude to 30 meters and uses its onboard visual or lidar system to scan a larger surrounding area (e.g., a radius of 100-200 meters) to construct a 3D terrain map and generate a terrain safety heat map. From the heat map, all candidate areas meeting the following criteria are identified: size (e.g., diameter > 2 meters), slope, flatness, and accessibility. For each candidate area, a landable area identification calculation is performed. Connectivity analysis can be used to identify continuous regions that meet the following conditions: { i Region ≤ i safe , s z ≤σ max No obstacles (or small obstacles that can be avoided), area diameter ≥D min ( D min =3× (drone wingspan) Optimal recommendation rating model: For each candidate region j, calculate the comprehensive recommendation index. R j : R j =α⋅SafetyScore j / SafetyScoremax +β⋅(1-d j / d max )+γ⋅A j / A max +δ⋅C j in: d j —The distance from the area center to the vehicle's current location; A j —Effective area of ​​the region; C j —Traffic cost coefficient (considering the terrain complexity of the route the vehicle takes); Weight a+b+c+d=1 (General value:) α=0.4,β=0.3,γ=0.2,δ=0.1 ); The system will select the 1-2 areas with the highest overall recommendation index as the optimal / second-best landing points, with the highest-scoring areas being the most suitable. R j For regions with a value >0.8, the second-best recommendation (suboptimal) is: 0.6 < R j For areas with a value ≤0.8, their location information and reasons for recommendation will be sent to the ground station or driver terminal; S44, Dynamic Tracking and Updates While the driver is driving the vehicle to the recommended point, the drone continuously monitors from the air; it uses onboard vision algorithms to monitor the recommended point and its surrounding environment in real time; if a new obstacle is detected or the terrain changes (such as other vehicles stopping), the system immediately re-evaluates the point; if the safety score drops below the threshold, the update mechanism is triggered, and step S43 is re-executed to dynamically push new recommendation results to the user. S45, New Site Verification and Landing Once the vehicle reaches the new recommended landing point, the UAV re-scans the current location at low altitude; then it executes the method described in the previous embodiment again: reconstructing the dynamic take-off and landing coordinate system, updating the feedforward gravity compensation parameters according to the new terrain, and re-verifying the terrain safety; after everything is verified to be correct, it executes the standard high-precision landing procedure to complete the landing.

[0042] This embodiment constructs a closed-loop feedback loop of perception, evaluation, decision-making, action, and re-perception. When the initial action (landing in place) is obstructed, the system automatically expands the perception range, introduces multi-objective optimization decision-making, and continuously perceives environmental changes during the execution of new actions to ensure the success of the final action. This reflects the system's high degree of autonomy and adaptability.

[0043] The system in this embodiment possesses the ability to make autonomous decisions and collaboratively plan in complex environments, greatly expanding the application scenarios of vehicle-mounted drones, especially in unknown or dynamic environments. A comprehensive safety scoring and recommendation index model ensures that the recommended landing point is the optimal solution after weighing multiple dimensions (safety, efficiency, and cost). A continuous dynamic monitoring and update mechanism enables the system to adapt to environmental changes and avoid decision-making errors caused by information lag. By giving the final choice to the user while providing strong data support, the system achieves an organic combination of human and machine intelligence.

[0044] The vehicle-mounted unmanned aerial vehicle (UAV) ramp takeoff and landing control method of this application can be implemented in practice as follows: Phase 1: Takeoff Preparation and Autonomous Takeoff Terrain perception and modeling: After the vehicle stops, the terrain perception module is activated to scan the surrounding environment using the vehicle-mounted LiDAR and obtain a 3D point cloud; the dynamic take-off and landing coordinate system construction module uses the RANSAC algorithm to process the point cloud, fits the ground plane, and establishes a dynamic take-off and landing coordinate system (DL system) based on this plane; this coordinate system is sent to the UAV through the vehicle-to-machine communication interface. Gravity feedforward compensation: After the UAV receives the dynamic take-off and landing coordinate system, the gravity compensation calculation module calculates the feedforward thrust compensation amount based on the coordinate system normal vector and its own attitude, and loads it into the motor command to make the UAV in a static equilibrium state on the slope. Smooth takeoff transition in coordinate system: The UAV is unlocked; the coordinate system transition control module is activated, and the mixing factor λ starts from 1; the UAV takes off vertically in the dynamic takeoff and landing coordinate system; as the altitude increases, λ gradually decreases, and the UAV smoothly transitions to the hovering point in the world coordinate system, completing the takeoff phase; Phase Two: Mission Execution and Landing Decisions Mission execution: The UAV performs reconnaissance, tracking, and other flight missions in a world coordinate system; In-situ landing assessment: Mission accomplished, ready for recovery; the drone flies back to the vicinity of the vehicle; the terrain safety assessment module quickly scans and assesses the area around the vehicle, calculating the safety score for each sector; Decision branches: Scenario A (Safe): If there are sectors that meet the safety score, the system determines that the landing can proceed directly to stage four; Situation B (Unsafe): If all sectors fail to meet the standards, the system determines that it is not suitable to land in place and enters Phase 3; Phase 3: Collaborative Optimization and Dynamic Programming Expanding reconnaissance and mapping: The collaborative optimization recommendation module takes over control; the UAV ascends to a safe altitude to scan a larger area around it, builds a terrain safety heat map, and identifies all potential candidate landing points; Intelligent recommendation: The module performs multi-objective optimization (safety, distance, area, etc.) on candidate points, calculates a comprehensive recommendation index, and recommends the best 1-2 points to the ground driver terminal, while planning a navigation path from the current location to the recommended points; Dynamic tracking: The drone continuously monitors the vehicle from the air as it travels to the recommended point; if the environment of the recommended point changes, the system immediately re-evaluates and dynamically updates the recommendation results to ensure that the information is accurate in real time. Arrival at the new site and re-verification: After the vehicle arrives at the recommended point, the drone performs another low-altitude scan; the dynamic take-off and landing coordinate system construction module and the gravity compensation calculation module update all parameters based on the location, and the terrain safety assessment module re-verifies the safety of the new location; Phase Four: Precise and Safe Landing Reverse smooth transition: After successful verification, the coordinate system transition control module initiates the reverse transition; the control target smoothly switches from the world coordinate system back to a new dynamic take-off and landing coordinate system that is bound in real time to the current vehicle position and posture; Integrated relative navigation landing: The precision landing control module is activated, integrating visual / RTK / UWB and other relative navigation information to guide the UAV to land precisely and smoothly on the vehicle landing platform in a direction perpendicular to the new ground plane under the new dynamic take-off and landing coordinate system; Process complete: The motor stops, and the task is finished.

[0045] In addition, this application also relates to a vehicle-mounted unmanned aerial vehicle (UAV) ramp take-off and landing control system, which mainly includes the following eight modules: Terrain perception module: Deployed on vehicles and / or drones, it includes LiDAR, binocular cameras and corresponding data processing units, and is used to acquire high-precision point cloud and image data around the vehicle in real time. Dynamic take-off and landing coordinate system construction module: Receives data from the terrain perception module and vehicle pose information, uses algorithms such as RANSAC to fit the ground plane, and establishes and publishes a dynamic take-off and landing coordinate system in real time; Gravity compensation calculation module: Deployed in the UAV flight control computer, it receives dynamic coordinate system information and UAV attitude information, calculates the feedforward thrust compensation amount in real time according to the above formula, and superimposes it into the motor control command; Coordinate system transition control module: manages the coordinate system switching logic of the UAV during takeoff and landing, and realizes smooth hybrid interpolation of the target position; Terrain safety assessment module: Analyzes the data from the terrain perception module, calculates indicators such as slope, flatness, and obstacle density of each sub-region, and generates safety scores and terrain heat maps; Collaborative optimization and recommendation module: When landing on the spot is not feasible, it plans the reconnaissance path of the UAV, identifies candidate areas, calculates the recommendation index based on the multi-objective optimization algorithm, generates and recommends the best / second best landing point to the user, and has a dynamic update function; Precision landing control module: During the landing phase, it integrates relative navigation information from sensors such as vision, RTK, or UWB to perform high-precision trajectory tracking and landing control in a dynamic takeoff and landing coordinate system; Vehicle-to-everything (V2X) communication interface: responsible for the high-speed, low-latency data link between the vehicle and the drone, synchronizing key information such as vehicle attitude, dynamic take-off and landing coordinate system parameters, safety assessment results, and landing commands.

[0046] The system in this application modularizes and integrates complex control methods; each module is responsible for an independent and cohesive function, and the modules communicate and exchange data through standardized interfaces; this architecture makes the system have good maintainability, scalability and portability; the vehicle-to-machine communication interface is the nervous system of the entire system, ensuring the synchronization and consistency of air-ground coordination.

[0047] The foregoing has shown and described the basic principles, main features, and advantages of this application. Those skilled in the art should understand that this application is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of this application. Various changes and modifications can be made to this application without departing from the spirit and scope thereof, and all such changes and modifications fall within the scope of this application as claimed. The scope of protection of this application is defined by the appended claims and their equivalents.

Claims

1. A method for controlling the take-off and landing of a vehicle-mounted unmanned aerial vehicle on a slope, characterized in that: include, Based on the point cloud data of the terrain around the vehicle, a local ground plane is fitted in the vehicle coordinate system, and a dynamic take-off and landing coordinate system based on this plane is established. Based on the dynamic take-off and landing coordinate system and vehicle attitude information, the ground normal vector in the world coordinate system is calculated, and a feedforward thrust compensation amount is generated to offset the gravity component along the slope. The feedforward thrust compensation amount is then superimposed on the UAV motor commands, and the UAV is unlocked. Vertical takeoff is performed in the dynamic takeoff and landing coordinate system, and after reaching the preset altitude, the position control target coordinate system is smoothly transitioned from the dynamic takeoff and landing coordinate system to the world coordinate system to achieve hovering; Upon receiving the landing command, the control coordinate system is smoothly switched from the world coordinate system back to the dynamic take-off and landing coordinate system bound to the current vehicle position and orientation, and high-precision landing is performed by integrating relative navigation information.

2. The method for controlling the take-off and landing of a vehicle-mounted unmanned aerial vehicle on a slope according to claim 1, characterized in that: The method for establishing a dynamic take-off and landing coordinate system based on this plane includes: acquiring point cloud data by collecting data about the area surrounding the vehicle using a binocular camera or lidar; performing plane fitting on the point cloud data using a random sampling consensus algorithm to obtain the ground plane equation and its unit normal vector; defining the Z-axis direction of the dynamic take-off and landing coordinate system based on the unit normal vector, the X-axis being the projection of the vertical axis of the vehicle coordinate system onto the fitted ground plane, and the Y-axis being determined according to the right-hand rule.

3. The method for controlling the take-off and landing of a vehicle-mounted unmanned aerial vehicle on a slope according to claim 1, characterized in that: The calculation method for the feedforward thrust compensation includes: calculating the feedforward thrust compensation according to the following formula. in: —Feedforward thrust compensation; —The rotation matrix from the world coordinate system to the UAV body coordinate system; —Ground unit normal vector in the world coordinate system; —Drone quality; —The vector of gravitational acceleration, directed vertically downwards.

4. The method for controlling the take-off and landing of a vehicle-mounted unmanned aerial vehicle on a slope according to claim 1, characterized in that: The method for smoothly transitioning the target coordinate system from the dynamic take-off and landing coordinate system to the world coordinate system includes: mixing the target positions in the dynamic take-off and landing coordinate system and the world coordinate system using linear interpolation. in: —The target position is a combination of the dynamic take-off and landing coordinate system and the world coordinate system; —Smooth transition from 1 to 0; —Target location in the world coordinate system; —Target position in dynamic take-off and landing coordinate system; —Transformation matrix from dynamic takeoff and landing coordinate system to world coordinate system.

5. The method for controlling the take-off and landing of a vehicle-mounted unmanned aerial vehicle on a slope according to claim 1, characterized in that: Before the drone lands, a real-time safety assessment is conducted on the terrain within a predetermined radius around the vehicle. If an area that meets the safety threshold is found, a landing command is issued; otherwise, an unsuitable landing alarm is triggered.

6. The method for controlling the take-off and landing of a vehicle-mounted unmanned aerial vehicle on a slope according to claim 5, characterized in that: The safety assessment method includes: dividing the area around the vehicle into multiple fan-shaped sub-regions, calculating the slope index, flatness index, obstacle density index, and material index for each sub-region, and weighting them to synthesize a comprehensive safety score; when the comprehensive safety score of any sub-region is not lower than a preset threshold, and the slope and obstacle density both meet the constraints, it is determined that the vehicle can land directly.

7. The method for controlling the take-off and landing of a vehicle-mounted unmanned aerial vehicle on a slope according to claim 5, characterized in that: When a landing is deemed unsuitable, the drone takes off to expand its reconnaissance range, constructs a terrain safety heat map, identifies candidate areas that meet the requirements of size, slope, flatness, and accessibility, and calculates a comprehensive recommendation index based on safety score, distance to the vehicle, area, and travel cost to recommend the best or second-best landing point to the user.

8. The method for controlling the take-off and landing of a vehicle-mounted unmanned aerial vehicle on a slope according to claim 7, characterized in that: As the vehicle travels to the recommended landing point, the drone continuously tracks the vehicle's position relative to the recommended point and monitors changes in the surrounding environment in real time. If any new hazards are detected, the drone dynamically updates the recommendations.

9. A method for controlling the take-off and landing of a vehicle-mounted unmanned aerial vehicle on a slope according to claim 7 or 8, characterized in that: After the vehicle arrives at the recommended landing point, the terrain of the new location is rescanned, the dynamic take-off and landing coordinate system is reconstructed, the gravity feedforward compensation parameters are updated, and after verifying the safety of the terrain, the standard landing procedure is executed to complete the precise docking.

10. A vehicle-mounted unmanned aerial vehicle (UAV) ramp takeoff and landing control system implementing the method as described in any one of claims 1 to 9, characterized in that, include: The terrain perception module is used to acquire point cloud data around the vehicle; The dynamic take-off and landing coordinate system construction module is used to fit the ground plane and establish a dynamic take-off and landing coordinate system; The gravity compensation calculation module is used to generate the feedforward thrust compensation amount; The coordinate system transition control module is used to manage the smooth switching between the dynamic take-off and landing coordinate system and the world frame; The terrain safety assessment module is used to calculate multi-dimensional safety indicators and determine landing suitability; The collaborative optimization and recommendation module is used to search, rate, and recommend alternative landing points; The precision landing control module is used to integrate relative navigation information to perform a high-precision landing; The vehicle-to-vehicle communication interface is used to synchronize vehicle position and orientation information and dynamic take-off and landing coordinate system parameters.

Images