Unmanned aerial vehicle following landing method based on UWB and visual dynamic fusion positioning

Abstract

The invention relates to the technical field of unmanned aerial vehicles, and discloses an unmanned aerial vehicle following landing method based on UWB and visual dynamic fusion, and the method comprises the steps: obtaining a first relative position deviation and a second relative position deviation; obtaining a first positioning weight according to the first relative position deviation, and obtaining a second positioning weight according to the first positioning weight; performing weighted fusion on the first positioning weight, the first relative position deviation, the second positioning weight and the second relative position deviation to serve as fusion deviation; according to the fusion deviation, a target position point where the unmanned aerial vehicle needs to go in the next step is generated; and circularly executing the steps. According to the technical scheme, the problems of sudden change of control instructions, flight jitter and tracking lag caused by simple switching of UWB and visual positioning or fixed weight fusion in the prior art are solved.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) technology, and in particular relates to a UAV follow-landing method based on UWB and visual dynamic fusion. Background Technology

[0002] Visual positioning technology relies on cameras to acquire external information and uses image processing to estimate the pose error of the UAV relative to the target platform in real time. Its advantages lie in its high-precision and high-stability positioning capabilities at close range, and its ability to perceive the relative position and attitude changes of the target platform in real time, which is particularly crucial for the accurate landing of UAVs. However, visual positioning also has significant drawbacks. First, the camera's field of view is limited, and visual tracking will be interrupted when the target is out of the field of view or obstructed. Second, visual algorithms are greatly affected by changes in lighting, flight attitude, and the camera's own angle, and accuracy will decrease in low-light or high-dynamic scenarios, especially for airborne cameras fixed to the UAV, whose field of view is particularly sensitive to flight attitude and angle during rapid movement. In addition, the number of feature points decreases at longer distances, reducing positioning accuracy.

[0003] In contrast, UWB positioning technology complements it. UWB positioning provides location information using Time-of-Flight (TOF) ranging between base stations and tags. Its advantages include providing global location information with global coverage, unaffected by ambient light and flight attitude. Furthermore, UWB technology boasts strong penetration and high anti-interference capabilities, low power consumption, and ease of deployment, making it suitable for guiding and locating distant targets using UAVs. However, UWB positioning suffers from poor dynamic stability, especially when UAVs or unmanned vehicles are moving at high speeds, leading to positioning drift and dynamic errors that fail to meet the requirements for precise landing. Moreover, its update frequency is often lower than that of vision systems.

[0004] The fusion of UWB and visual positioning technologies can achieve complementary advantages. Existing UWB and visual fusion positioning methods often employ simple static switching or fixed weight strategies, i.e., hard switching between different positioning methods based on confidence thresholds, or simply assigning fixed weight coefficients to different positioning methods. This leads to abrupt changes in UAV control commands during sensor dominance switching, causing flight jitter. Simultaneously, traditional methods fail to dynamically adjust fusion weights based on key factors such as distance, failing to fully utilize UWB data for smooth complementarity within the effective visual range, and lacking effective redundancy when vision fails. Ultimately, this results in insufficient accuracy, smoothness, and environmental adaptability for the UAV during following and landing. Based on this, this patent proposes a UAV following and landing method based on dynamic fusion of UWB and vision. Summary of the Invention

[0005] The purpose of this invention is to provide a method and device for unmanned aerial vehicle (UAV) landing based on the fusion of UWB positioning and visual positioning, thereby solving the problems of sudden changes in control commands, flight jitter, and tracking lag caused by simple switching or fixed weight fusion of UWB and visual positioning in the prior art.

[0006] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0007] This application discloses a method for drone follow-landing based on UWB and visual dynamics fusion, including the following steps: The first relative position deviation between the UAV and the ground platform is obtained using the visual positioning method, and the second relative position deviation between the UAV and the ground platform is obtained using the UWB positioning method. The first positioning weight is obtained based on the first relative position deviation, and the second positioning weight is obtained based on the first positioning weight. The first positioning weight, the first relative position deviation, the second positioning weight, and the second relative position deviation are weighted and fused together to form the fusion deviation. Based on the fusion deviation, the target location point to be visited by the drone is generated next; Repeat the above steps.

[0008] Some solutions also include the following steps: Determine whether the distance between the drone's onboard camera and the ground platform exceeds the field of view threshold, and then proceed with the following judgment: If not, the drone is controlled to follow the movement of the ground platform and land on the ground platform based on the fused deviation obtained by weighting and fusing the first positioning weight, the first relative position deviation, the second positioning weight, and the second relative position deviation. If so, the second relative position deviation is directly used as the fusion deviation, and the UAV is controlled to follow the movement of the ground platform and land on the ground platform based on the fusion deviation.

[0009] In some schemes, the specific methods for obtaining the first relative positional deviation between the UAV and the ground platform based on visual positioning include: An identification code is set on a ground platform, and the drone's onboard camera detects the position and attitude information of the identification code in real time. The first relative position deviation is obtained by using the position and orientation information of the identification code.

[0010] In some schemes, the specific methods for obtaining the second relative position deviation between the UAV and the ground platform based on the UWB positioning method include: The global positions of the UAV and the ground platform are obtained using the base station, and the absolute coordinates of the UAV and the ground platform are calculated separately. The second relative position deviation is obtained based on the absolute coordinates of the UAV and the absolute coordinates of the ground platform.

[0011] In some schemes, the specific formula for obtaining the first positioning weight based on the first relative position deviation is as follows: W V = M K F × distance In the formula, M is the upper limit adjustment parameter for visual positioning weight, and W... V K is the first positioning weight. F is the rate of change of weight with distance, where distance is the distance between the UAV and the ground platform in a two-dimensional plane.

[0012] In some schemes, the specific formula for obtaining the second positioning weight based on the first positioning weight is as follows: W U = 1 W V In the formula, W V W is the first positioning weight. U This is the second positioning weight.

[0013] In some schemes, the specific formula for obtaining the fusion bias is: e F (t)= W V· e V (t)+ W U· e U (t) if diastance ≤ diastance v e F (t)= e U (t) if diastance>diastance v In the formula, e F (t) represents the fusion bias, e V (t) represents the first relative position deviation, e U (t) represents the second relative position deviation, distance V This is the field of view threshold.

[0014] In some schemes, the specific methods for generating the next target location point for the UAV based on the fusion bias include: The controller converts the fusion deviation into a control variable. Specific target points are generated through incremental position control.

[0015] In some schemes, the steps involve incremental position control, and specific methods for generating specific target points include: The current actual position of the UAV in the global coordinate system is obtained in real time, and the fused deviation is input to the controller to obtain the speed control value; Multiplying the speed control quantity by the control period yields a small position displacement vector within the current period. The calculated position displacement vector is added to the current position vector of the UAV to generate the specific target point for the next control cycle.

[0016] In some schemes, the identification code includes a first identification code and a second identification code, with the second identification code nested in the center of the first identification code.

[0017] The technical solution adopted in this invention can achieve the following beneficial effects: This application presents a UAV landing method based on the fusion of UWB and visual positioning. By dynamically weighting and fusing the high precision of visual positioning with the global advantages of UWB positioning, it effectively overcomes the inherent limitations of single sensors in terms of field of view, dynamic stability, and environmental interference resistance. This enables UAVs to achieve high-precision, low-hysteresis, and stable and reliable automatic following and precise landing beyond visual range in complex environments. Furthermore, through dynamic weighted fusion and incremental position control, this application effectively solves the problems of abrupt control command changes, flight jitter, and tracking lag caused by simple switching or fixed-weight fusion of UWB and visual positioning in existing technologies. It achieves a smooth transition and adaptive complementarity of positioning advantages throughout the entire following and landing process, significantly improving the stability, continuity, and robustness of precise landing and mobile following in complex environments. Attached Figure Description

[0018] Figure 1 This is an overall flowchart of the UAV following landing method based on UWB and visual dynamic fusion of the present invention; Figure 2 This is a dynamic weighted control framework diagram of the UAV following landing method based on UWB and visual dynamic fusion of the present invention; Figure 3 This is a schematic diagram of the identification code of this invention. Detailed Implementation

[0019] The following detailed description of embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0020] This application discloses several embodiments of a UAV following landing method based on the fusion of UWB and visual dynamics, such as... Figures 1-2 As shown, it includes the following steps: S100. Obtain the first relative position deviation between the UAV and the ground platform using the visual positioning method, and obtain the second relative position deviation between the UAV and the ground platform using the UWB positioning method. S200. Obtain the first positioning weight based on the first relative position deviation, and then obtain the second positioning weight based on the first positioning weight. S300: The first positioning weight, the first relative position deviation, the second positioning weight, and the second relative position deviation are weighted and fused together to form the fusion deviation; S400: Based on the fusion deviation, generate the target location point that the UAV should go to next; S500, repeat the above steps.

[0021] In step S100, as is preferred in this embodiment, the ground platform is an unmanned vehicle that can be used for drone landing, and the same applies below.

[0022] In step S100, the specific method for obtaining the first relative position deviation between the UAV and the ground platform based on the visual positioning method includes: S110. Set up an identification code on the ground platform, and the UAV's onboard camera will detect the position and attitude information of the identification code in real time. S120. The first relative position deviation is obtained using the position and attitude information of the identification code.

[0023] In this embodiment, the identification code is preferably an ArUco QR code. An ArUco QR code is a square two-dimensional identification mark based on Hamming code design. It has a one-square-width black border and contains a binary matrix arranged according to specific encoding rules, where black represents 0 and white represents 1. This border design not only improves detection efficiency but also helps resist occlusion and background interference. By performing parity checks and redundant encoding on the encoding bits of different columns, highly reliable ID recognition and error correction can be achieved.

[0024] The drone's onboard camera acquires image data in real time, and the images are processed using the ArUco detection module in the OpenCV library. In the preprocessing stage, image grayscale conversion, edge detection, and subpixel fitting are performed to improve the ArUco recognition rate. In the pose estimation stage, the `cv::aruco::detectMarkers` function is used to detect the ArUco QR code ID in the image. Then, the PNP (Perspective-n-Point) algorithm provided by the OpenCV library is used to convert the two-dimensional image coordinates and camera intrinsic parameters (obtained through calibration) into a rotation vector (R) in three-dimensional space. x R y R zTranslation vector (T) x T y T z This attitude information is used to calculate the relative position of the drone and the center of the QR code in the drone's coordinate system, which is then used for subsequent follow-up control.

[0025] The identification code includes a first identification code and a second identification code, with the second identification code nested in the center of the first identification code. Both the first and second identification codes are ArUco QR codes. If a single ArUco QR code were used, as the drone's altitude decreases, its field of view gradually shrinks. When the QR code cannot be fully displayed within the field of view, the visual tracking algorithm loses its control over the drone. Therefore, by nesting the first and second identification codes, the algorithm prioritizes recognizing the smaller ArUco QR code. When the smaller ArUco QR code is invisible or lost, it automatically switches to recognizing the larger ArUco QR code at a greater distance. When the drone is at high altitude, the smaller ArUco QR code cannot be recognized, so the larger ArUco QR code is used for positioning. As the drone descends, the smaller ArUco QR code is gradually recognized, thus avoiding the positioning loss problem caused by the incomplete display of the larger ArUco QR code within the field of view. This algorithm can maintain stable visual tracking capabilities at different distances and altitudes.

[0026] As a preferred embodiment, the nested ArUco QR code is as follows: Figure 3 As shown, the outer QR code has a side length of 15cm, and the inner QR code has a side length of 2.143cm. Both nested QR codes use 7 × 7 ArUco QR codes. Excluding the black border, the internal binary matrix size is 5 × 5. The large QR code has an ID of 57, and the small QR code has an ID of 7. Since the large QR code's binary matrix has an odd number of squares, the center pixel box can be replaced with the small QR code. As long as the black pixels of the small QR code account for more than 50%, it can be considered a black pixel, thus not affecting the ID recognition of the large QR code. Figure 3 The marker shown is affixed to the top of the ground platform to provide the drone with the relative pose information needed for following and landing. The drone is controlled to fly to the vicinity of the ground platform, and its onboard monocular camera is used to detect the position and attitude information of the ArUco QR code in real time, laying the foundation for following and landing on the mobile platform.

[0027] After completing the detection of the position and orientation information of the identification code, since this application focuses on relative motion in the horizontal plane (x and y directions), the translation vector T extracted from the orientation estimation is selected. x and T y The x and y direction deviations e used to reflect the center of the field of view of the airborne camera and the center of the QR code fixed to the unmanned vehicle. x and ey The first relative position deviation is fed as input into a decoupled two-dimensional PID controller to enable the UAV to horizontally follow the mobile platform. Simultaneously, the UAV's altitude z remains constant during the following process and decreases linearly during landing.

[0028] In step S100, the specific method for obtaining the second relative position deviation between the UAV and the ground platform according to the UWB positioning method includes: S130. Use the base station to obtain the global position of the UAV and the ground platform, and calculate the absolute coordinates of the UAV and the absolute coordinates of the ground platform respectively. S140. Based on the absolute coordinates of the UAV and the absolute coordinates of the ground platform, the second relative position deviation is obtained.

[0029] This embodiment preferably uses the Time-of-Flight (TOF) algorithm for UWB positioning of the drone. The TOF algorithm directly calculates the distance between the tag and the base station by measuring the round-trip time of the signal, and uses a circular or spherical intersection position analysis algorithm to solve for the target coordinates. First, four base stations are used as the centers of a sphere, and the distance between the center of the sphere and the target tag is used as the radius. The distance is calculated by multiplying the signal arrival time from the tag to the base station by the speed of electromagnetic wave propagation, thus obtaining the global position of the tag-equipped drone and unmanned vehicle.

[0030] After acquiring the global position information of the drone and the unmanned vehicle, the global position information of the drone and the unmanned vehicle is obtained through the pose topic callback. Then, the relative position of the drone and the unmanned vehicle in different motion scenarios is obtained, namely the absolute coordinates of the drone and the absolute coordinates of the ground platform. Based on the absolute coordinates of the drone and the absolute coordinates of the ground platform, the second relative position deviation is obtained.

[0031] In step S200, the specific formula for obtaining the first positioning weight based on the first relative position deviation is as follows: W V = M K F × distance In the formula, M is the upper limit adjustment parameter for visual positioning weight, and W... V K is the first positioning weight. F is the rate of change of weight with distance, where distance is the distance between the UAV and the ground platform in a two-dimensional plane.

[0032] In step S200, the specific formula for obtaining the second positioning weight based on the first positioning weight is as follows: W U = 1 W V In the formula, WU This is the second positioning weight.

[0033] In step S200, by dynamically adjusting the weights of the two positional deviations during the following process using the above formula, the continuity and smoothness of the dynamic following process after fusion can be effectively guaranteed, and the instability of the flight state caused by the transitional discontinuity of the control method can be avoided.

[0034] In step S300, the first positioning weight, the first relative position deviation, the second positioning weight, and the second relative position deviation are weighted and fused together, and the specific formula for the fusion deviation is as follows: e F (t)= W V· e V (t)+ W U· e U (t) In the formula, e F (t) represents the fusion bias, e V (t) represents the first relative position deviation, e U (t) represents the second relative position deviation.

[0035] Because drone-borne cameras have a field-of-view threshold distance V When the drone is too far from the ground platform, visual positioning cannot function properly. Therefore, in order to ensure that the drone can better follow the ground platform and land on it, when the distance between the drone and the ground platform in the two-dimensional plane is greater than the field of view threshold, the following formula applies: e F (t)= e U (t) Therefore, step S300 also includes the step of determining whether the distance between the UAV's onboard camera and the ground platform exceeds the field of view threshold, and then proceeding to the following determination: If not, the fusion deviation obtained by weighted fusion of the first positioning weight and the second positioning weight is used to control the UAV to follow the movement of the ground platform and land on the ground platform. If so, the second relative position deviation is directly used as the fusion deviation, and the UAV is controlled to follow the movement of the ground platform and land on the ground platform based on the fusion deviation.

[0036] In summary, the formula is: e F (t)= W V· e V (t)+ W U· e U (t) if diastance ≤ diastance v e F(t)= e U (t) if diastance>diastance v The first relative position deviation e obtained by the visual positioning method V The weights of (t) are linearly negatively correlated with the distance, and the second relative position deviation e obtained by the UWB global positioning method is... U The weights of (t) are linearly positively correlated with distance. When the distance between the drone and the unmanned vehicle exceeds the maximum field of view of the onboard camera... V At that time, e U (t) represents the fusion bias, at which point UWB positioning is relied upon entirely for long-distance guidance and coarse tracking; when the distance between the UAV and the unmanned vehicle is less than the maximum field of view of the onboard camera... V In this process, the weight of visual positioning gradually increases as the distance decreases. Through this dynamic weight allocation mechanism, the system can smoothly transition the dominance of each sensor while maintaining system stability. At the same time, redundancy ensures that if one sensor fails, the UAV can still rely on another effective sensor to complete the following task.

[0037] In step S400, the specific method for generating the target location point that the UAV will go to next based on the fusion deviation includes: S410. The fusion deviation is converted into a control quantity through the controller; S420 generates specific target points through incremental position control.

[0038] In step S410, a controller intelligently converts the current error into a speed control command to guide the UAV's flight. The controller consists of two parts working in tandem: a PID feedback control and a feedforward control. The PID controller handles current and past errors: its proportional term responds instantly to the current fused deviation; the larger the deviation, the stronger the output control force. The integral term accumulates historical deviations to completely eliminate minor steady-state errors, ensuring the UAV accurately aligns with the target center. The derivative term monitors the trend of deviation changes, acting as a damper to suppress potential overshoot and oscillations, thus ensuring smooth and stable flight. Simultaneously, the feedforward controller, independent of the current deviation, directly utilizes real-time speed information from the UAV system to pre-calculate a compensation control quantity. Finally, this controller adds the PID output (error-based feedback force) to the feedforward output (target motion trend-based predictive force), synthesizing a comprehensive speed vector control command that balances accuracy and speed.

[0039] In step S420, the speed control command generated in S410 is converted into a specific target position point in the global coordinate system that the UAV flight controller can directly execute. This process is called incremental position control, and specifically includes the following steps: S421. Real-time acquisition of the UAV's current actual position in the global coordinate system, and inputting the fusion deviation into the controller to obtain the speed control quantity; S422. Multiply the speed control quantity by the control cycle to obtain the small position displacement vector within the current cycle; S423. Add the calculated position displacement vector to the current position vector of the UAV to generate the specific target point for the next control cycle.

[0040] The core principle of step S420 is a tiny forward movement based on the current state: the system first acquires the UAV's real-time position in the global coordinate system, then multiplies the velocity command by the control cycle to obtain a tiny position increment that should occur within the current cycle. This increment indicates how far the UAV should move from its current position in which direction. Finally, the system adds this position increment to the current position to calculate the absolute coordinate target point that the UAV needs to reach in the next control cycle. This method cleverly decomposes the complex continuous velocity tracking problem into a series of discrete, tiny position setpoint tracking problems, thus enabling perfect integration with the powerful inner-loop position controller of the flight controller. The flight controller is responsible for the lowest-level stable flight, achieving decoupling of the control tasks and ultimately ensuring that the UAV can smoothly and accurately track dynamic targets.

[0041] Specifically, step S420 has the following formula: u x ( t )= K p e x ( t )+ K i ∫ e x ( t ) dt + K d de x ( t ) / dt —Formula 1 u y ( t )= K p e y ( t )+ K i ∫ e y ( t ) dt + K d de y ( t ) / dt —Formula 2 Target x ( t )= P x ( t 1)+ u x ( t ) dt —Formula 3 Target y ( t )= P y ( t 1)+ u y ( t ) dt —Formula 4 After completing target attitude recognition, the next key task is to design a real-time controller based on positional deviation to keep the drone directly above the center of the QR code and ultimately achieve a smooth landing. Since this embodiment focuses on relative motion in the horizontal plane (x and y directions), the translation vector T extracted from attitude estimation is selected. x and T y The x and y direction deviations e used to reflect the center of the field of view of the airborne camera and the center of the QR code fixed to the unmanned vehicle. x and e y The input is fed into a decoupled two-dimensional PID controller to enable the UAV to horizontally follow the mobile platform.

[0042] The design formulas for the PID controller are shown in Formula 1 and Formula 2. The control signal u is calculated based on the deviation between the center of the QR code and the center of the image. x and u y Then, an incremental position control algorithm is used to dynamically update the target point. Target x ， Target y As shown in Formulas 3 and 4, where P x and P y u represents the actual coordinates of the drone. x and u y The product of the control period dt and the target position of the UAV in the x and y directions is adjusted as the position increment, so that the UAV always follows the target directly above the QR code.

[0043] It should be noted that K i K p K d These are the three gain parameters of the PID controller. K p It is the proportional term, directly responding to the current error, K. i It is an integral term that accumulates historical errors and addresses persistent small deviations. K d It is a differential term that predicts future trends and prevents oscillations.

[0044] Some embodiments of this application also disclose a drone control device that uses a drone following landing method based on UWB and visual dynamic fusion, including the drone body, a chassis platform and at least four UWB base stations.

[0045] The drone is equipped with an onboard camera; the ground platform is preferably an unmanned vehicle that can move on the ground; the ground platform is equipped with an identification code for the onboard camera to identify, preferably a nested ArUco QR code, for visual positioning of the drone.

[0046] At least four UWB base stations are required for UWB positioning of the drone itself.

[0047] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A UAV following landing method based on UWB and visual dynamic fusion, characterized in that, Includes the following steps: The first relative position deviation between the UAV and the ground platform is obtained using the visual positioning method, and the second relative position deviation between the UAV and the ground platform is obtained using the UWB positioning method. The first positioning weight is obtained based on the first relative position deviation, and the second positioning weight is obtained based on the first positioning weight. The first positioning weight, the first relative position deviation, the second positioning weight, and the second relative position deviation are weighted and fused together to form the fusion deviation. Based on the fusion deviation, the target location point to be visited by the drone is generated next; Repeat the above steps.

2. The UAV following landing method based on UWB and visual dynamic fusion according to claim 1, characterized in that, It also includes the following steps: Determine whether the distance between the drone's onboard camera and the ground platform exceeds the field of view threshold, and then proceed with the following judgment: If not, the drone is controlled to follow the movement of the ground platform and land on the ground platform based on the fused deviation obtained by weighting and fusing the first positioning weight, the first relative position deviation, the second positioning weight, and the second relative position deviation. If so, the second relative position deviation is directly used as the fusion deviation, and the UAV is controlled to follow the movement of the ground platform and land on the ground platform based on the fusion deviation.

3. The UAV following landing method based on UWB and visual dynamic fusion according to claim 1, characterized in that, The specific methods for obtaining the first relative position deviation between the UAV and the ground platform based on the visual positioning method include: An identification code is set on a ground platform, and the drone's onboard camera detects the position and attitude information of the identification code in real time. The first relative position deviation is obtained by using the position and orientation information of the identification code.

4. The UAV following landing method based on UWB and visual dynamic fusion according to claim 1, characterized in that, The specific method for obtaining the second relative position deviation between the UAV and the ground platform based on the UWB positioning method includes: The global positions of the UAV and the ground platform are obtained using the base station, and the absolute coordinates of the UAV and the ground platform are calculated separately. The second relative position deviation is obtained based on the absolute coordinates of the UAV and the absolute coordinates of the ground platform.

5. The UAV following landing method based on UWB and visual dynamic fusion according to claim 2, characterized in that, The specific formula for obtaining the first positioning weight based on the first relative position deviation is as follows: W V = M K F × distance In the formula, M is the upper limit adjustment parameter for visual positioning weight, and W V K is the first positioning weight. F is the rate of change of weight with distance, where distance is the distance between the UAV and the ground platform in a two-dimensional plane.

6. The UAV following landing method based on UWB and visual dynamic fusion according to claim 5, characterized in that, The specific formula for obtaining the second positioning weight based on the first positioning weight is as follows: IN U = 1 IN V In the formula, W V W is the first positioning weight. U This is the second positioning weight.

7. The UAV following landing method based on UWB and visual dynamic fusion according to claim 6, characterized in that, The specific formula for obtaining the fusion bias is: e F (t)= W V· e V (t)+ W U· e U (t) if diastance≤diastance v and F ( t ) = e U (t) if distance=distance v In the formula, e F (t) represents the fusion bias, e V (t) represents the first relative position deviation, e U (t) represents the second relative position deviation, distance V This is the field of view threshold.

8. The UAV following landing method based on UWB and visual dynamic fusion according to claim 7, characterized in that, The specific method for generating the next target location point for the UAV based on the fusion deviation includes: The controller converts the fusion deviation into a control variable. Specific target points are generated through incremental position control.

9. The UAV following landing method based on UWB and visual dynamic fusion according to claim 8, characterized in that, The specific methods for generating specific target points through incremental position control include: The current actual position of the UAV in the global coordinate system is obtained in real time, and the fused deviation is input to the controller to obtain the speed control value; Multiplying the speed control quantity by the control period yields a small position displacement vector within the current period. The calculated position displacement vector is added to the current position vector of the UAV to generate the specific target point for the next control cycle.

10. The UAV following landing method based on UWB and visual dynamic fusion according to claim 3, characterized in that, The identification code includes a first identification code and a second identification code, with the second identification code nested in the center of the first identification code.

Images