Landing control method based on four-rotor unmanned aerial vehicle under mobile platform

Abstract

The present application relates to the technical field of unmanned aerial vehicle, specifically relates to four rotor unmanned aerial vehicle landing control method based on mobile platform, including: S1, before unmanned aerial vehicle landing, carries out environment safety evaluation, and utilizes multimode perception algorithm detection and identification mobile landing platform;S2, based on the identification result, through visual solution and state estimation real-time evaluation platform relative pose, and adopts feedforward-feedback compound control to realize dynamic tracking and heading pre-alignment;S3, according to real-time relative state planning dynamic descent trajectory, and through hierarchical closed-loop controller control unmanned aerial vehicle along the trajectory accurate descent;S4, in the landing phase, multi-source ground detection fusion decision is used, confirms after ground execution system reset, this method realizes from target identification, real-time tracking, trajectory tracking to the full-automatic closed-loop control of safe ground contact, has the beneficial effect of landing process self-adaptation, strong anti-disturbance ability, high landing precision and good reliability.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) technology, and more specifically to a landing control method for a quadcopter UAV based on a mobile platform. Background Technology

[0002] With the expanding applications of drones in mobile scenarios such as logistics, inspection, and ship take-off and landing, achieving autonomous and precise landing of drones on mobile platforms has become a key challenge. Traditional landing methods have shortcomings in dynamic target tracking, anti-interference, and accurate ground contact determination, making it difficult to meet the high requirements of real-time performance, robustness, and safety for mobile landing. Therefore, there is an urgent need for a closed-loop landing solution that integrates perception, tracking, and control throughout the entire process.

[0003] Traditional UAV landing methods are primarily designed for static ground targets, relying on absolute positioning from global navigation satellite systems or pre-defined static visual markers. However, in mobile platform scenarios, these methods face significant limitations: First, the relative motion between the platform and the UAV leads to error accumulation and insufficient update frequency when relying solely on GPS positioning, making it difficult to meet the requirements of high-precision real-time tracking. Second, the platform's own motion (such as acceleration, deceleration, and turning) and potential environmental disturbances (such as wind disturbances and ocean wave fluctuations) introduce complex dynamic uncertainties, making traditional static closed-loop control strategies prone to lag and oscillations, resulting in unstable tracking or even landing failure. Furthermore, existing methods for ground contact detection often rely on a single sensor (such as an altimeter or contact switch), which is susceptible to impact and vibration interference during dynamic landing, posing a risk of misjudgment or missed detection, affecting landing safety and reliability.

[0004] Currently, although research has introduced visual tracking or optical flow technology to assist in moving target tracking, problems such as easy loss of markers, large computational latency, and insufficient fusion of multi-source information still exist in complex real-world environments. Furthermore, most solutions treat trajectory planning and control tracking as relatively independent processes, lacking tight coupling based on real-time state prediction in dynamic environments, resulting in insufficient adaptability and robustness of trajectory tracking. Therefore, the industry urgently needs an integrated solution for autonomous landing of mobile platform UAVs that can span the entire "perception-decision-control-confirmation" process, possessing strong environmental adaptability, high real-time performance, and high reliability. Summary of the Invention

[0005] In view of the above-mentioned shortcomings of the existing technology, the present invention provides a landing control method for quadcopter UAVs based on a mobile platform, which can effectively solve the problems of accurate tracking and reliable contact in dynamic environments in the background technology.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The present invention provides a landing control method for a quadcopter unmanned aerial vehicle based on a mobile platform, comprising: S1. When the drone is about to land, conduct a safety assessment of the landing environment and detect and identify the mobile landing platform.

[0007] S2. Real-time assessment of the relative attitude of the mobile landing platform, and dynamic tracking and heading pre-alignment.

[0008] S3. Generate the dynamic descent trajectory of the UAV to be landed, use this trajectory as an instruction, and use the actual relative state feedback in real time as the control error to form a closed loop to control the UAV to be landed to descend along the dynamic descent trajectory.

[0009] S4. Perform landing and touchdown detection on the drone to be landed, and reset the system after landing.

[0010] Preferably, the safety assessment of the landing environment is carried out as follows: Upon receiving the landing command, the drone to be landed first enters the landing preparation phase, where it is awakened and initializes the corresponding key subsystems.

[0011] Once the hardware is ready, the drone performs a rapid scan of the landing area below based on multi-sensor information to assess whether there are any unplanned static obstacles or dynamic threats, and to check whether the data communication link with the mobile landing platform meets the stability requirements. If there are no unplanned static obstacles or dynamic threats, and the data communication link with the mobile landing platform meets the stability requirements, the drone awaiting landing is allowed to attempt landing.

[0012] Preferably, the detection and identification of the mobile landing platform involves the following process: Acquire a real-time video stream below the drone to be landed and make an initial area proposal.

[0013] The system performs efficient feature point extraction and optical flow calculation in parallel to segment the outline of the mobile landing platform as an independent motion unit from the moving background.

[0014] Once the unmanned aerial vehicle (UAV) system is about to land, it enters a multi-level confirmation and precise positioning phase. The bounding box detected by vision and the motion area segmented by optical flow are spatiotemporally fused and verified. After confirming the target, the unique platform ID and physical size parameters encoded within the cooperative identifier are parsed, and the precise three-dimensional position and yaw angle of the mobile landing platform relative to the UAV are calculated in real time.

[0015] Preferably, the real-time evaluation of the relative pose of the mobile landing platform is carried out as follows: The onboard camera of the drone waiting to land continuously detects the cooperative markers on the mobile landing platform in each frame of the image, taking the center of the marker as the origin, based on the physical size of the marker and the three-dimensional coordinates of its four corner points in the world coordinate system.

[0016] By establishing the correspondence between the corner pixel coordinates detected in the image and the three-dimensional coordinate points, the pose of the camera relative to the labeled coordinate system can be solved.

[0017] Preferably, the dynamic tracking and heading pre-alignment process is as follows: Based on the real-time assessment of relative pose and velocity, feedback adjustments are made to the errors in relative position and yaw angle.

[0018] The motion state of the mobile landing platform, shared through the communication link, is directly injected into the control command as a feedforward quantity.

[0019] After the injection is controlled, the heading pre-alignment of the UAV to be landed is designed as a path tangent tracking process, which calculates the tangent angle of its motion direction in real time based on the motion trajectory of the mobile landing platform.

[0020] Control the yaw angle of the drone to be landed to track this dynamically changing tangent angle.

[0021] Preferably, the specific process of generating the dynamic descent trajectory of the drone to be landed is as follows: Based on the real-time relative state between the UAV to be landed and the mobile landing platform, and combined with the prediction of the future movement of the mobile landing platform, a time-parameterized trajectory from the current relative position to the target landing point is planned, wherein the target landing point is associated with the mobile landing platform.

[0022] The initial conditions of the trajectory are determined based on the real-time relative state, and the boundary conditions of the trajectory are determined at least based on the motion prediction of the mobile landing platform.

[0023] Based on the initial conditions and the boundary conditions, the desired motion trajectory of the UAV to be landed in the horizontal and vertical directions is generated, so that the desired motion trajectory can guide the UAV to be landed toward the expected future position of the mobile landing platform.

[0024] The desired vertical trajectory is planned to satisfy the boundary conditions for a soft landing, which include zero relative altitude and zero vertical velocity at the expected landing time. The dynamic descent trajectory is periodically or triggered during the descent based on the updated real-time relative state and motion prediction.

[0025] Preferably, the process of controlling the drone to descend along a dynamic descent trajectory is as follows: The dynamic descent trajectory is used as a reference input for the control system.

[0026] The real-time relative status between the drone to be landed and the mobile landing platform is obtained as feedback.

[0027] Based on the reference input and the feedback, control commands are generated to drive the movement of the drone to be landed.

[0028] The control command is executed to make the actual movement trajectory of the drone to be landed follow the dynamic descent trajectory.

[0029] By calculating the error between the reference input and the feedback, and combining it with the feedforward control quantity, low-level commands are generated to control the attitude and thrust of the UAV to be landed.

[0030] Preferably, the landing and touchdown detection of the drone to be landed is performed as follows: Based on information from multiple sensors on the drone to be landed, it is determined whether preset conditions associated with a ground contact event are met.

[0031] In response to the fulfillment of the preset conditions, it is determined that the ground contact event has occurred.

[0032] Judgments based on information from multiple sources include: It processes sensor data from the inertial measurement unit, dynamic system, and relative state estimator in parallel or collaboratively.

[0033] Based on the sensor data, one or more independent detection flags related to ground contact are triggered.

[0034] The ground contact event is confirmed based on the fusion decision of the one or more independent detection markers.

[0035] Preferably, the system reset of the drone to be landed is performed as follows: After confirming that the grounding event has occurred, a system reset operation is performed, the operation including: The active tracking control of the UAV to be landed is terminated, its power output is adjusted to a safe idling state, its flight control mode is switched to safe hold mode, and its internal control logic and status flags related to this landing mission are reset.

[0036] The technical solution provided by this invention has the following advantages compared with the known prior art: 1. In the landing target detection and identification process, the embodiments of the present invention utilize the synergistic application of multi-modal perception algorithms and multi-level confirmation mechanisms to significantly improve the robustness and initialization accuracy of target acquisition in complex dynamic environments. This enables the detection and segmentation of the mobile platform from two independent dimensions: "appearance features" and "motion characteristics," effectively overcoming the problem of identifier loss caused by changes in illumination, background interference, or partial occlusion. The subsequent spatiotemporal fusion verification process further eliminates interference from similar moving objects such as ground moving shadows, ensuring that the associated target ID and initial pose are highly reliable when the subsequent tracking algorithm is activated, thus laying an accurate data foundation for the entire landing process.

[0037] 2. In the relative state assessment and dynamic tracking process, the embodiments of the present invention utilize sensor fusion based on extended Kalman filtering and introduce a feedforward-feedback composite control structure, which facilitates smooth and low-latency tracking of the high-speed, high-maneuverability motion of the mobile platform. It not only integrates visual pose calculation and the UAV's own IMU data, but also integrates the motion state shared by the mobile platform's communication, constructing a more comprehensive and accurate relative motion model. This design significantly reduces the phase delay and steady-state error of horizontal tracking, ensuring that the UAV can still hover stably directly above the platform in dynamic scenarios such as platform acceleration and turning, creating near-static favorable conditions for subsequent vertical landing.

[0038] 3. In the dynamic descent trajectory planning and control process of this invention, the trajectory generation of the platform motion prediction is coupled with the hierarchical closed-loop tracking control architecture, which is conducive to achieving accurate, smooth and adaptive trajectory tracking. The trajectory planner not only takes the current relative state as the starting point, but also integrates the prediction of the platform's future speed or position, so that the planned trajectory is forward-looking and guides the UAV to move to the platform's expected landing point in advance. Essentially, it combines open-loop prediction with closed-loop feedback. The outer loop position controller uses the combined effect of the feedforward acceleration of the trajectory derivative and the state estimation feedback, while the inner loop attitude controller efficiently solves the acceleration command into the underlying execution command, ensuring that the UAV can closely follow the time-varying dynamic trajectory. Even if the platform motion changes during the descent, the system can respond quickly and finally achieve a "soft landing" with less impact, effectively protecting the UAV and platform equipment.

[0039] 4. In the final landing and touchdown confirmation process, the embodiments of the present invention utilize multi-source detectors based on quantized thresholds and fusion decision logic to achieve highly reliable and safe determination of the touchdown moment. This method eliminates the vulnerability of a single sensor, simultaneously monitors inertial impact characteristics, motor load changes, and altitude / speed stagnation, and sets clear physical thresholds, greatly improving the confidence of the touchdown judgment. Landing success is only confirmed when multiple physical evidences support it simultaneously, thereby avoiding the risk of "crash" or "go-around" that may be caused by misjudgment and ensuring the final safe closed loop of the landing process. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0041] Figure 1 This is a schematic diagram of the system structure connection of the present invention. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0043] The present invention will be further described below with reference to embodiments.

[0044] Please see Figure 1 As shown, the landing control method for a quadcopter UAV based on a mobile platform includes at least the following: S1. When the drone is about to land, conduct a safety assessment of the landing environment and detect and identify the mobile landing platform.

[0045] In one specific embodiment, the safety assessment of the landing environment is carried out as follows: Upon receiving the landing command, the UAV to be landed first enters the landing preparation phase. The UAV to be landed is awakened and initializes the corresponding key subsystems, which include an onboard computing unit, sensors, and data streams. The landing control program is loaded onto the onboard computing unit, the sensors are powered on for self-test, and calibration and alignment with the data stream are performed.

[0046] Once the hardware is ready, the drone performs a rapid scan of the landing area below based on multi-sensor information to assess whether there are any unplanned static obstacles or dynamic threats, and to check whether the data communication link with the mobile landing platform meets the stability requirements. If there are no unplanned static obstacles or dynamic threats, and the data communication link with the mobile landing platform meets the stability requirements, the drone awaiting landing is allowed to attempt landing.

[0047] It should be noted that airborne computing units include flight controllers and onboard computers; sensors include visual cameras, inertial measurement units (IMUs), ultrasonic / laser ranging modules, and GPS / RTK receivers; multi-sensor information includes wide-angle cameras or downward-looking depth sensors; static obstacles include ground debris and temporary facilities; and dynamic threats include personnel and vehicles.

[0048] In one specific embodiment, the detection and identification of the mobile landing platform is carried out as follows: The drone to be landed activates its multi-modal perception algorithm optimized for landing, captures a real-time video stream below the drone using an onboard RGB camera, performs preliminary region proposal using a lightweight deep convolutional neural network (such as an improved YOLO or SSD architecture), and simultaneously, a downward-looking grayscale camera performs efficient feature point extraction and optical flow calculation in parallel. By analyzing the motion consistency of feature points in the scene, the outline of the mobile landing platform as an independent motion unit is segmented from the moving background.

[0049] Subsequently, the unmanned aerial vehicle (UAV) system enters a multi-level confirmation and precise positioning phase. The algorithm performs spatiotemporal fusion verification of the bounding box detected by vision and the motion area segmented by optical flow. After confirming the target, the UAV parses the platform's unique ID and physical size parameters encoded within the cooperative identifier, and uses known dimensions, camera intrinsic parameters, and the PnP algorithm to calculate the precise three-dimensional position and yaw angle of the mobile landing platform relative to the UAV in real time.

[0050] It should be noted that lightweight deep convolutional neural networks (such as improved YOLO or SSD architectures) are specifically optimized for training on high-contrast cooperative identifiers (such as nested concentric circles or specific QR code variants) on mobile platforms, enabling them to quickly locate suspected target regions in complex backgrounds.

[0051] It should be noted that the process of "real-time calculation of the precise three-dimensional position and yaw angle of the mobile landing platform relative to the UAV" runs at a high frequency (e.g., 30Hz), and is associated with the platform ID upon the first successful identification, and the subsequent tracking algorithm is then activated.

[0052] In the landing target detection and identification process, the embodiments of this invention utilize a multi-modal perception algorithm and a multi-level confirmation mechanism to significantly improve the robustness and initialization accuracy of target acquisition in complex dynamic environments. This enables the detection and segmentation of the mobile platform from two independent dimensions: "appearance features" and "motion characteristics," effectively overcoming the problem of identifier loss caused by changes in illumination, background interference, or partial occlusion. The subsequent spatiotemporal fusion verification process further eliminates interference from similar moving objects such as ground shadows, ensuring that the associated target ID and initial pose are highly reliable when the subsequent tracking algorithm is activated, thus laying an accurate data foundation for the entire landing process.

[0053] S2. Based on the mobile landing platform identified by the UAV to be landed, the relative pose of the mobile landing platform is evaluated in real time, and dynamic tracking and heading pre-alignment are performed.

[0054] In one specific embodiment, the real-time evaluation of the relative pose of the mobile landing platform is carried out as follows: The onboard camera of the UAV to be landed continuously detects the cooperative marker on the mobile landing platform in each frame of the image. Taking the center of the marker as the origin, based on the physical size of the marker and the three-dimensional coordinates of its four corner points in the world coordinate system, the correspondence between the corner pixel coordinates detected in the image and the three-dimensional coordinate points is established, and the perspective n-point algorithm is called to solve the pose of the camera relative to the marker coordinate system. It should be noted that this solution process is expressed as solving for a rotation matrix R and a translation vector T, where T directly represents the three-dimensional position (X, Y, Z) of the marker center in the camera coordinate system, while R includes the pose angle of the marker relative to the camera.

[0055] By transforming the coordinates, the relative position and yaw angle between the center of mass of the UAV to be landed and the center of the mobile landing platform can be obtained.

[0056] It should be noted that if the rotation matrix R_cm and translation vector T_cm (whose components represent the coordinates of the marker center in the camera coordinate system) relative to the marker coordinate system (denoted as {M}) have been calculated using the PnP algorithm, and the camera's position on the UAV is known to be fixed, the transformation from the camera coordinate system to the UAV's body coordinate system (denoted as {B}, with the origin at the UAV's centroid) is obtained through calibration as R_bc and T_bc. Simultaneously, the marker is precisely installed at the center of the mobile landing platform, and its coordinate system {M} completely coincides with the platform coordinate system {P}. Therefore, the transformation of the UAV's body coordinate system {B} relative to the platform coordinate system {P} can be obtained through the following chained coordinate transformation: T_bp = R_cm^T (T_bc-T_cm) (This is a simplified representation of the vector transformation; the complete expression is a homogeneous matrix multiplication). The calculated translation vector T_bp represents the three-dimensional coordinates of the UAV's center of mass in the coordinate system of the moving landing platform, directly providing the relative position. The relative yaw angle can be obtained from the final composite rotation matrix R_bp (derived from R_bc). The value of R_cm is extracted and obtained by calculating its rotation angle around the vertical axis (Z-axis) (e.g., using the formula: atan2(R_bp[1,0],R_bp[0,0])).

[0057] In one specific embodiment, the dynamic tracking and heading pre-alignment process is as follows: Based on the real-time evaluated relative attitude and velocity, the UAV to be landed activates a dynamic tracking controller with a feedforward-feedback composite structure. The controller uses the error of relative position and yaw angle for feedback adjustment (such as PID control), and also directly injects the motion state (such as speed and angular velocity) of the mobile landing platform shared through the communication link as a feedforward quantity into the control command.

[0058] After the injection is controlled, the heading pre-alignment of the UAV to be landed is designed as a path tangent tracking process. The tangent angle of the direction of motion of the mobile landing platform is calculated in real time based on its motion trajectory (calculated from its velocity vector or shared path points). The flight control system will control the UAV's yaw angle to track this dynamically changing tangent angle. Before landing, the longitudinal axis (nose direction) of the UAV to be landed is always precisely aligned with the direction of motion of the platform.

[0059] It should be noted that the specific process for calculating the heading angle (i.e., the tangent angle in the platform's direction of motion) that the UAV needs to track is as follows: Assume that through the communication link, the UAV to be landed obtains the instantaneous velocity vector of the mobile landing platform in the horizontal plane (based on the world coordinate system or the local coordinate system shared by the UAV), denoted as: , and These are the platform's velocity components in the east and north directions, and the tangent angle in that direction of motion. The target yaw angle (i.e., the azimuth angle) can be obtained by calculating the azimuth angle of the velocity vector. The core formula is: , It is a four-quadrant arctangent function, which can determine the correct angle (range) based on the signs of the two velocity components. (Radians), this angle represents the orientation of the platform's motion relative to a reference east axis. For example, if the platform is moving north in a straight line at a constant speed, then... Calculated The drone uses radians (i.e., 90 degrees) as its target yaw angle for tracking. In actual control, to ensure a smooth response, it typically adjusts... Perform low-pass filtering, and with Current yaw angle of the drone The difference is used as input, and a proportional or proportional-integral controller is used to generate a yaw rate command, which drives the longitudinal axis of the UAV fuselage (nose direction) to align with the platform's movement direction in real time.

[0060] In the relative state assessment and dynamic tracking process, this invention utilizes sensor fusion based on extended Kalman filtering and introduces a feedforward-feedback composite control structure. This facilitates smooth, low-latency tracking of the mobile platform's high-speed, highly maneuverable motion. It not only integrates visual pose calculation and the UAV's own IMU data but also incorporates its own motion state shared from the mobile platform's communication, constructing a more comprehensive and accurate relative motion model. This design significantly reduces phase delay and steady-state error in horizontal tracking, ensuring that the UAV can still hover stably directly above the platform during dynamic scenarios such as platform acceleration and turning, creating near-static favorable conditions for subsequent vertical landing.

[0061] S3. Based on the real-time relative state of the mobile landing platform, generate a dynamic descent trajectory for the UAV to be landed. Use this trajectory as a command and the actual relative state fed back in real time as the control error to form a closed loop, and control the UAV to be landed to descend along the dynamic descent trajectory.

[0062] In one specific embodiment, the process of generating the dynamic descent trajectory of the drone to be landed is as follows: the trajectory planner outputs the current time as S2. Given the initial conditions and considering the motion prediction of the mobile platform, a path is planned from the current relative position to the target landing point (with the origin of the mobile landing platform coordinate system). The time-parameterized trajectory: ,in , For the estimated landing time, Represented as relative position, Expressed as relative velocity, This represents the speed of the mobile landing platform itself.

[0063] Using a polynomial trajectory (such as cubic or quintic), plan a cubic trajectory that satisfies the boundary conditions in the vertical direction (z-axis, downward is positive): The boundary conditions include initial conditions and termination conditions. The initial conditions are: The termination condition is: .

[0064] The coefficients can be obtained by solving the above system of equations. to Trajectory planning in the horizontal direction (x, y axes) requires coupling with the platform's motion prediction. The planner sends the platform's future velocity prediction via a communication link. Or integrate location prediction to make the planning The drone is guided to move ahead of the platform to its expected future position, rather than just its current position.

[0065] In one specific embodiment, controlling the drone to descend along a dynamic descent trajectory involves the following process: planning the dynamic descent trajectory... and its first derivative (velocity reference) The input to the controller is used as the outer loop (position / velocity control) and the inner loop (attitude / throttle control). The outer loop position controller is used to calculate the acceleration command required to track the trajectory. It compares the reference state with the real-time estimated state fed back by S2 to generate a three-dimensional acceleration vector. , ,in , representing feedforward acceleration, derived from the second derivative of trajectory planning, enabling the system to better track time-varying commands. and The real-time relative position and velocity output by the state estimator (such as EKF) in step S2 are used as feedback quantities. and It is a diagonal positive definite matrix, with proportional and differential gains respectively, which are determined by adjusting the controller parameters. The minus sign in the formula represents the calculation error: the reference value minus the estimated value.

[0066] The acceleration command is converted into the UAV's low-level execution command, based on the current UAV attitude (roll angle measured by the onboard IMU). Pitch angle ) and expected acceleration By combining gravity compensation, the desired attitude angle (roll angle) required to achieve this acceleration is calculated. , looking up Total tension .

[0067] The simplified solution relationship is as follows (under the small angle approximation): ,in: It is gravitational acceleration. It refers to the mass of the drone (a known system parameter). This is the current yaw angle of the UAV, obtained by combining the relative yaw angle estimation from step S2 with the platform's heading information. It is used for coordinate transformation. The high-bandwidth attitude controller in the inner loop (usually a PID or more advanced algorithm) will operate at an extremely high frequency to control the motor differential speed, thus controlling the actual attitude of the UAV. Rapid tracking while throttle command controls total thrust tracking .

[0068] It should be noted that the state estimator in step S2 continues to operate throughout the descent process, providing the outer loop controller with the latest state information. and The outer loop controller calculates new values ​​based on this feedback and trajectory reference values. The inner loop adjusts the attitude and throttle accordingly. This closed loop of "trajectory planning → outer loop control → inner loop control → motor drive → body motion → state estimation (S2) → feedback" operates at a high speed of hundreds of hertz, ensuring that the UAV can closely track the dynamically changing desired trajectory and ultimately achieve a precise and smooth landing.

[0069] In the dynamic descent trajectory planning and control process, this invention, through the coupling of platform motion prediction trajectory generation and hierarchical closed-loop tracking control architecture, facilitates accurate, smooth, and adaptive trajectory tracking. The trajectory planner not only starts from the current relative state but also integrates predictions of the platform's future speed or position, making the planned trajectory forward-looking and guiding the UAV to move to the platform's expected landing point in advance. Essentially, it combines open-loop prediction with closed-loop feedback. The outer-loop position controller utilizes the combined effect of feedforward acceleration from the trajectory derivative and state estimation feedback, while the inner-loop attitude controller efficiently solves the acceleration command into the underlying execution command, ensuring that the UAV can closely follow the time-varying dynamic trajectory. Even if the platform's motion changes during descent, the system can respond quickly, ultimately achieving a "soft landing" with minimal impact, effectively protecting the UAV and platform equipment.

[0070] S4. Based on closed-loop control execution, the system performs landing and touchdown detection on the drone to be landed, and resets the system after landing.

[0071] In one specific embodiment, the landing and touchdown detection of the drone to be landed is performed as follows: when the downlink trajectory altitude component planned in step S3... Below the preset proximity threshold At this point, the drone enters the final landing phase. During this phase, the horizontal position controller remains active to track the platform, but the vertical controller switches to a more conservative, constant low-speed descent mode, setting a very small vertical velocity reference value. (For example (0.1m / s) to achieve a gentle ground contact.

[0072] Inertial detector: Monitors the specific force (specific force) vector measured by the UAV IMU. During flight, vector It mainly includes the gravitational component. At the moment of impact, due to the sudden change in supporting force, its modulus or component will undergo characteristic changes, defining the change in acceleration. ,when In continuous The threshold was continuously exceeded within each control cycle. (For example, 2 m / s) will trigger the inertial ground contact indicator. .

[0073] Power system detector: Monitors the average command current of the motor based on the ESC feedback. When hovering or descending, Maintaining a relatively stable value Upon grounding, a sudden increase in load causes a momentary rise in current. The rate of change of current is defined as follows: ,when Exceeding the threshold If it remains in place for a short period of time, the power contact signal will be triggered. .

[0074] Altitude Change Detector: Monitors the relative altitude provided by the state estimator in step two. and its rate of change When satisfied (e.g., 0.05m) and If the speed (e.g., 0.02 m / s) exceeds a certain duration, it indicates that the motion has stopped near the ground, triggering the height ground contact marker. .

[0075] Final touchdown confirmation Follow majority voting or key combination logic, such as: That is, inertial detection must be confirmed simultaneously with at least one other detector to improve reliability and avoid misjudgment due to noise from a single sensor (such as body shaking or airflow disturbance).

[0076] when If step 1 is confirmed, the flight controller will immediately execute the command to reduce the power of each motor to a safe idle speed, exit the active position control mode of step S3, switch to pure attitude self-stabilization mode (if the platform is stationary) or lock the current throttle, send the "landing successful" status word to the host computer through the communication link, reset the internal landing state machine, and prepare for the next mission.

[0077] in and Direct measurement from IMU sensor and ESC, and The real-time state estimator output from step S2. and These are pre-calibrated constants (gravitational acceleration, average hovering current). , , , , and The system thresholds and parameters were determined through experiments and set based on factors such as the drone's weight, landing gear's vibration damping characteristics, and sensor noise levels.

[0078] In one specific embodiment, the system reset of the UAV to be landed is performed as follows: The flight control system first immediately stops the trajectory tracking control loop in step three and sends a command to all motors to smoothly descend to a predetermined safe idle speed to eliminate residual thrust and allow the weight of the UAV to be fully supported by the landing gear; then, the flight mode is automatically switched to manual attitude hold or fully locked state, and the position control loop is disconnected; at the same time, a confirmation signal containing the code "landing completed" is sent to the host computer and the mobile landing platform through the communication link; finally, all temporary status flags and queues of this landing mission are cleared internally, the relevant accumulated errors in the sensor fusion algorithm are reset, and the entire landing control logic is returned to the initial standby state that can accept new commands at any time, completing the landing closed loop.

[0079] It should be noted that the flight control system refers to the core hardware computing unit (such as flight control boards like Pixhawk and DJIA3) installed on the UAV to be landed and the embedded software running on it. This system directly connects to and manages all airborne sensors (IMU, barometer, GPS, etc.) and actuators (ESC and motors), and runs the state estimation algorithm in step S2, the hierarchical control law in step S3, and the ground contact detection logic in step four in real time.

[0080] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement any module of an information management system for a flatbed printing machine.

[0081] In the final landing and touchdown confirmation process, this invention utilizes a multi-source detector based on quantized thresholds and fusion decision logic to achieve highly reliable and safe determination of the touchdown moment. This method eliminates the vulnerability of a single sensor, simultaneously monitors inertial impact characteristics, motor load changes, and altitude / speed stagnation, and sets clear physical thresholds, greatly improving the confidence of the touchdown determination. Landing success is only confirmed when multiple physical evidences support it simultaneously, thereby avoiding the risk of "crash" or "go-around" that may result from misjudgment and ensuring the final safe closed loop of the landing process.

[0082] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the protection scope of the technical solutions of the embodiments of the present invention.

Claims

1. A landing control method for a quadcopter UAV based on a mobile platform, characterized in that, include: S1. When the drone is about to land, conduct a safety assessment of the landing environment and detect and identify the mobile landing platform. S2. Real-time assessment of the relative attitude of the mobile landing platform, and dynamic tracking and heading pre-alignment; S3. Generate the dynamic descent trajectory of the drone to be landed, use this trajectory as an instruction, and use the actual relative state fed back in real time as the control error to form a closed loop, and control the drone to be landed to descend along the dynamic descent trajectory. S4. Perform landing and touchdown detection on the drone to be landed, and reset the system after landing.

2. The landing control method for a quadcopter UAV based on a mobile platform according to claim 1, characterized in that, The specific process for conducting a safety assessment of the landing environment is as follows: Upon receiving the landing command, the drone awaiting landing first enters the landing preparation phase, where it wakes up and initializes the corresponding key subsystems. Once the hardware is ready, the drone performs a rapid scan of the landing area below based on multi-sensor information to assess whether there are any unplanned static obstacles or dynamic threats, and to check whether the data communication link with the mobile landing platform meets the stability requirements. If there are no unplanned static obstacles or dynamic threats, and the data communication link with the mobile landing platform meets the stability requirements, the drone awaiting landing is allowed to attempt landing.

3. The landing control method for a quadcopter UAV based on a mobile platform according to claim 2, characterized in that, The specific process for detecting and identifying the mobile landing platform is as follows: Acquire a real-time video stream below the drone to be landed and make an initial area proposal; The parallel operation of efficient feature point extraction and optical flow calculation is used to segment the outline of the mobile landing platform as an independent motion unit from the moving background. Once the unmanned aerial vehicle (UAV) system is about to land, it enters a multi-level confirmation and precise positioning phase. The bounding box detected by vision and the motion area segmented by optical flow are spatiotemporally fused and verified. After confirming the target, the unique platform ID and physical size parameters encoded within the cooperative identifier are parsed, and the precise three-dimensional position and yaw angle of the mobile landing platform relative to the UAV are calculated in real time.

4. The landing control method for a quadcopter UAV based on a mobile platform according to claim 3, characterized in that, The real-time evaluation of the relative pose of the mobile landing platform is carried out as follows: The onboard camera of the drone waiting to land continuously detects the cooperative markers on the mobile landing platform in each frame of the image, with the center of the marker as the origin, based on the physical size of the marker and the three-dimensional coordinates of its four corner points in the world coordinate system. By establishing the correspondence between the corner pixel coordinates detected in the image and the three-dimensional coordinate points, the pose of the camera relative to the labeled coordinate system can be solved.

5. The landing control method for a quadcopter UAV based on a mobile platform according to claim 4, characterized in that, The specific process for dynamic tracking and heading pre-alignment is as follows: Based on the real-time assessment of relative pose and velocity, feedback adjustment is made to the errors of relative position and yaw angle; The mobile landing platform's own motion state, shared through the communication link, is directly injected into the control command as a feedforward quantity. After the control injection is completed, the heading pre-alignment of the UAV to be landed is designed as a path tangent tracking process, which calculates the tangent angle of its motion direction in real time based on the motion trajectory of the mobile landing platform. Control the yaw angle of the drone to be landed to track this dynamically changing tangent angle.

6. The landing control method for a quadcopter UAV based on a mobile platform according to claim 5, characterized in that, The specific process for generating the dynamic descent trajectory of the drone to be landed is as follows: Based on the real-time relative state between the UAV to be landed and the mobile landing platform, and combined with the prediction of the future movement of the mobile landing platform, a time-parameterized trajectory from the current relative position to the target landing point is planned, wherein the target landing point is associated with the mobile landing platform. The initial conditions of the trajectory are determined based on the real-time relative state, and the boundary conditions of the trajectory are determined at least based on the motion prediction of the mobile landing platform. Based on the initial conditions and the boundary conditions, the desired motion trajectory of the UAV to be landed in the horizontal and vertical directions is generated, so that the desired motion trajectory can guide the UAV to be landed toward the expected future position of the mobile landing platform. The desired vertical trajectory is planned to satisfy the boundary conditions for a soft landing, which include zero relative altitude and zero vertical velocity at the expected landing time. The dynamic descent trajectory is periodically or triggered during the descent based on the updated real-time relative state and motion prediction.

7. The landing control method for a quadcopter UAV based on a mobile platform according to claim 6, characterized in that, The specific process of controlling the drone to descend along a dynamic descent trajectory is as follows: The dynamic descent trajectory is used as a reference input for the control system; The real-time relative status between the drone to be landed and the mobile landing platform is obtained as feedback. Based on the reference input and the feedback, control commands are generated to drive the movement of the UAV to be landed. The control command is executed to make the actual motion trajectory of the drone to be landed track the dynamic descent trajectory; By calculating the error between the reference input and the feedback, and combining it with the feedforward control quantity, low-level commands are generated to control the attitude and thrust of the UAV to be landed.

8. The landing control method for a quadcopter UAV based on a mobile platform according to claim 7, characterized in that, The specific process for performing landing and touchdown detection on the drone to be landed is as follows: Based on information from multiple sensors on the drone to be landed, it is determined whether preset conditions associated with a ground contact event are met. In response to the fulfillment of the preset conditions, it is determined that the ground contact event has occurred; Judgments based on information from multiple sources include: It processes sensor data from the inertial measurement unit, dynamic system, and relative state estimator in parallel or in a coordinated manner. Based on the sensor data, trigger one or more independent ground-contact-related detection flags; The ground contact event is confirmed based on the fusion decision of the one or more independent detection markers.

9. The landing control method for a quadcopter UAV based on a mobile platform according to claim 8, characterized in that, The specific process for resetting the system of the drone to be landed is as follows: After confirming that the grounding event has occurred, a system reset operation is performed, the operation including: The active tracking control of the UAV to be landed is terminated, its power output is adjusted to a safe idling state, its flight control mode is switched to safe hold mode, and its internal control logic and status flags related to this landing mission are reset.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the system according to any one of claims 1 to 9.

Images