Vehicle-mounted unmanned aerial vehicle landing control method and related equipment

Abstract

The invention discloses a vehicle-mounted unmanned aerial vehicle landing control method and related equipment, and relates to the technical field of vehicle-mounted unmanned aerial vehicles, and the method comprises the steps: obtaining first sensor data of a vehicle-mounted platform and second sensor data of a target unmanned aerial vehicle; determining a current fusion strategy from the multi-stage fusion strategies according to the relative distance between the target unmanned aerial vehicle and the vehicle-mounted platform; based on a current fusion strategy, performing fusion processing on the first sensor data and the second sensor data, and determining relative pose information of the target unmanned aerial vehicle relative to the vehicle-mounted platform; and based on the relative pose information, generating a landing control instruction for the target unmanned aerial vehicle to guide the target unmanned aerial vehicle to land on the vehicle-mounted platform in the vehicle driving process. The unmanned aerial vehicle can be guided to stably and accurately complete landing in the vehicle running process, and the stability and applicability of the landing task of the mobile platform are improved.

Description

Technical Field

[0001] This application relates to the field of vehicle-mounted unmanned aerial vehicle (UAV) technology, and more specifically, to a landing control method and related equipment for a vehicle-mounted UAV. Background Technology

[0002] With the continuous development of drone technology and intelligent transportation systems, drones are increasingly being used in logistics delivery, inspection and monitoring, and emergency rescue. To improve operational efficiency and expand application scenarios, utilizing vehicle-mounted platforms as mobile take-off and landing platforms for drones is gradually becoming an important development direction. Achieving autonomous landing of drones while the vehicle is in motion can effectively improve drone mobility and mission continuity, thereby enabling more flexible and efficient collaborative operations in complex environments. Therefore, how to achieve stable and accurate landing of drones while the vehicle is in motion has gradually become an important research topic in drone control technology.

[0003] In related technologies, the landing control of drones on mobile platforms typically relies on single sensor information or simple positioning methods to obtain the relative positional relationship between the drone and the vehicle-mounted platform, such as position estimation based on single visual or positioning information. However, during vehicle movement, due to factors such as continuous vehicle motion, environmental changes, and differences in sensor observation conditions, a single sensor often struggles to provide consistently accurate pose information, thus affecting the reliability of drone landing control. This results in the inability to flexibly adjust positioning and control methods according to environmental and state changes as the drone approaches the vehicle-mounted platform, thereby limiting the stability and accuracy of landing control. In other words, related technologies suffer from the problem of difficulty in achieving stable and accurate drone landing control during vehicle movement. Summary of the Invention

[0004] In the summary section of this application, the relevant technical solutions are described in general terms, and a series of simplified concepts are introduced. These concepts will be further elaborated in the detailed embodiments section. This summary section should not be construed as limiting the key or essential technical features of the claimed solutions, nor is it intended to limit the scope of protection of the claimed solutions.

[0005] The vehicle-mounted UAV landing control method and related equipment provided in this application can obtain highly reliable relative pose information by fusing multi-source sensor data from the vehicle platform and the UAV and dynamically selecting a multi-stage fusion strategy based on the relative distance. This generates landing control commands to guide the UAV to complete a stable and accurate landing while the vehicle is in motion, thereby improving the stability and applicability of mobile platform landing missions.

[0006] In a first aspect, this application provides a method for controlling the landing of a vehicle-mounted unmanned aerial vehicle (UAV), comprising: acquiring first sensor data from a vehicle-mounted platform and second sensor data from a target UAV; determining a current fusion strategy from a multi-stage fusion strategy based on the relative distance between the target UAV and the vehicle-mounted platform; performing fusion processing on the first sensor data and the second sensor data based on the current fusion strategy to determine the relative pose information of the target UAV relative to the vehicle-mounted platform; and generating a landing control command for the target UAV based on the relative pose information to guide the target UAV to land on the vehicle-mounted platform during vehicle operation.

[0007] In some implementations, the multi-stage fusion strategy includes a long-range return-to-home strategy, a mid-range approach strategy, a terminal alignment strategy, and an anti-disturbance touchdown strategy. Determining the current fusion strategy from the multi-stage fusion strategies based on the relative distance between the target UAV and the vehicle-mounted platform includes: determining the current fusion strategy as the long-range return-to-home strategy when the relative distance is greater than a first preset distance; determining the current fusion strategy as the mid-range approach strategy when the relative distance is less than or equal to the first preset distance and greater than a second preset distance; determining the current fusion strategy as the terminal alignment strategy when the relative distance is less than or equal to the third preset distance; and determining the current fusion strategy as the anti-disturbance touchdown strategy when the relative distance is less than or equal to the third preset distance.

[0008] In some implementations, the first sensor data includes ultra-wideband base station array data, visual marker data, vehicle motion state data, and first global navigation positioning data; the second sensor data includes second global navigation positioning data, ultra-wideband tag data, image acquisition data, and inertial measurement unit data. The step of fusing the first sensor data and the second sensor data based on the current fusion strategy to determine the relative pose information of the target UAV relative to the vehicle platform includes: when the current fusion strategy is the remote return-to-home strategy, performing data fusion based on the first global navigation positioning data, the inertial measurement unit data, and the second global navigation positioning data. The relative pose information is obtained through data fusion processing based on the ultra-wideband tag data, the ultra-wideband base station array data, the image acquisition data, and the inertial measurement unit data when the current fusion strategy is the mid-range approach strategy. Similarly, when the current fusion strategy is the end-effector alignment strategy, the relative pose information is obtained through data fusion processing based on the visual marker data, the image acquisition data, and the inertial measurement unit data. Finally, when the current fusion strategy is the disturbance-resistant landing strategy, the relative pose information is obtained through data fusion processing based on the inertial measurement unit data and the vehicle motion state data.

[0009] In some implementations, the step of performing data fusion processing based on the first global navigation positioning data, the inertial measurement unit data, and the second global navigation positioning data to obtain the relative pose information includes: determining the first relative position of the target UAV relative to the vehicle platform based on the first global navigation positioning data and the second global navigation positioning data; and fusing the first relative position and the inertial measurement unit data based on a preset filtering algorithm to obtain the relative pose information.

[0010] In some embodiments, the step of performing data fusion processing based on the UWTA tag data, the UWTA base station array data, the image acquisition data, and the inertial measurement unit data to obtain the relative pose information includes: determining a second relative position of the target UAV relative to the vehicle-mounted platform based on the UWTA tag data and the UWTA base station array data; performing visual inertial odometry assessment based on the image acquisition data and the inertial measurement unit data to obtain the visual inertial pose information of the target UAV; and fusing the second relative position and the visual inertial pose information based on a preset filtering algorithm to obtain the relative pose information.

[0011] In some implementations, the step of performing data fusion processing based on the visual marker data, the image acquisition data, and the inertial measurement unit data to obtain the relative pose information includes: identifying cooperative markers corresponding to the visual marker data based on the image acquisition data; calculating the cooperative markers using a perspective n-point algorithm to obtain the six-degree-of-freedom relative pose of the target UAV relative to the cooperative markers; determining the inertial pose information of the target UAV based on the inertial measurement unit data; and fusing the six-degree-of-freedom relative pose and the inertial pose information based on a preset filtering algorithm to obtain the relative pose information.

[0012] In some implementations, the step of performing data fusion processing based on the inertial measurement unit data and the vehicle motion state data to obtain the relative pose information includes: performing motion recursion based on the inertial measurement unit data to obtain the predicted motion trajectory of the target UAV; using the vehicle motion state data as a feedforward compensation amount to dynamically correct the predicted motion trajectory to obtain the compensated motion trajectory; and updating the state of the compensated motion trajectory based on a preset filtering algorithm to obtain the relative pose information.

[0013] Secondly, this application also provides a vehicle-mounted unmanned aerial vehicle (UAV) landing control device, comprising: a data acquisition unit for acquiring first sensor data of a vehicle platform and second sensor data of a target UAV; a strategy determination unit for determining a current fusion strategy from a multi-stage fusion strategy based on the relative distance between the target UAV and the vehicle platform; a pose determination unit for performing fusion processing on the first sensor data and the second sensor data based on the current fusion strategy to determine the relative pose information of the target UAV relative to the vehicle platform; and a landing control unit for generating landing control commands for the target UAV based on the relative pose information to guide the target UAV to land on the vehicle platform during vehicle operation.

[0014] Thirdly, this application also provides a vehicle-mounted drone, including: a memory and a processor, wherein the processor is used to execute a computer program stored in the memory to implement the steps of the vehicle-mounted drone landing control method described in the first aspect.

[0015] Fourthly, this application also provides a computer-readable storage medium storing computer-executable instructions or a computer program, which, when executed by a processor, implement the steps of the vehicle-mounted UAV landing control method described in the first aspect.

[0016] Fifthly, this application also provides a computer program product, including a computer program or computer-executable instructions, which, when executed by a processor, implement the steps of the vehicle-mounted unmanned aerial vehicle landing control method provided in the embodiments of this application.

[0017] In summary, this application simultaneously acquires sensor data from both sides of the vehicle platform and the UAV, and then fuses this data. This leverages the complementarity between different sensor information to improve the accuracy and reliability of relative pose information acquisition, enabling the UAV to more stably perceive its positional relationship with the vehicle platform. By determining the current fusion strategy from a multi-stage fusion strategy based on the relative distance between the UAV and the vehicle platform, the appropriate fusion method can be dynamically selected based on distance changes during landing, thereby improving the adaptability of positioning and control at different landing stages and ensuring that the UAV obtains effective pose information under various distance conditions. The fusion processing yields the relative pose information of the UAV relative to the vehicle platform, which is then used to generate landing control commands. This allows the UAV to perform motion control based on real-time pose relationships, guiding its landing trajectory and improving the stability of landing missions while the vehicle is in motion. By generating landing control commands for the UAV during vehicle movement, the UAV can complete landing operations on a mobile platform, thus expanding the application scenarios for UAV landing and improving the applicability of the method in mobile platform environments. In summary, this technical solution achieves landing guidance for UAVs during vehicle movement by acquiring multi-source sensor data, selecting multi-stage fusion strategies, and implementing landing control based on relative pose. This improves the information utilization efficiency and stability of the UAV landing control process. In conclusion, the vehicle-mounted UAV landing control method provided in this application fuses multi-source sensor data from the vehicle platform and the UAV, and dynamically selects a multi-stage fusion strategy based on relative distance to obtain highly reliable relative pose information. This generates landing control commands to guide the UAV to land stably and accurately during vehicle movement, improving the stability and applicability of mobile platform landing tasks. Attached Figure Description

[0018] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit this specification. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart illustrating a vehicle-mounted unmanned aerial vehicle (UAV) landing control method provided in an embodiment of this application; Figure 2 This is a schematic diagram of the composition structure of a vehicle-mounted unmanned aerial vehicle landing control device provided in an embodiment of this application; Figure 3 This is a schematic diagram of the composition structure of a vehicle-mounted drone provided in an embodiment of this application. Detailed Implementation

[0019] The terms used in the specification, claims, and drawings of this application, such as "first," "second," "third," "fourth," etc. (if any), are used to distinguish similar objects and not to describe a specific order or sequence. Therefore, it is to be understood that these terms can be used interchangeably where appropriate, allowing the described embodiments to be used in different orders, unless specifically required by the illustrations or description. Furthermore, the terms "is" and "has," and any variations thereof, are intended to cover, non-exclusively, all possible constituent elements. For example, a process, method, system, product, or apparatus comprising several steps or units is not necessarily limited to the steps or units explicitly listed, but may also include other steps or units not explicitly listed, or steps or units inherent to the process, method, product, or apparatus.

[0020] In this application, a "module" or "unit" refers to a computer program or part of a computer program that has a specific function and works in conjunction with other related parts to achieve a predetermined goal. These modules or units can be implemented by software, hardware (e.g., processing circuitry or memory), or a combination of both. One or more processors or memories can implement one or more modules or units. Furthermore, each module or unit can also be part of a larger module or unit.

[0021] The technical solutions of this application will be described in detail below with reference to the accompanying drawings of the embodiments. It should be noted that the described embodiments are only a part of this application, and not all embodiments. In the following description, the "some embodiments" mentioned are only a subset of all possible embodiments, which may be the same or different subsets, and different embodiments can be combined with each other without conflict.

[0022] Figure 1 This is a schematic flowchart illustrating a vehicle-mounted unmanned aerial vehicle (UAV) landing control method provided in an embodiment of this application. For example, see [link to example]. Figure 1 The vehicle-mounted drone landing control method provided in this application embodiment may include the following steps 101 to 104: Step 101: Acquire the first sensor data of the vehicle platform and the second sensor data of the target drone.

[0023] In some examples, the first sensor data is the sensing data collected by various sensing sensors on the vehicle platform side, used to assist the target UAV in completing landing positioning and control. It is the data source for the vehicle platform to perceive its own motion state and landing area characteristics. Various sensors on the vehicle platform side can collect their own sensing information in real time according to a preset collection frequency. The vehicle platform's onboard computing and communication unit performs formatting, data validity verification, and timestamp marking on the collected raw data to complete the acquisition and temporary storage of the first sensor data. For example, the first sensor data may include motion state data such as vehicle speed, angular velocity, and attitude angle, as well as sensing data related to vehicle positioning and landing area identification. The acquisition device can obtain vehicle motion data through the vehicle controller local area network bus and collect landing area-related data through dedicated onboard sensing sensors.

[0024] The target drone is an unmanned aerial vehicle that performs dynamic landing missions on vehicle-mounted platforms. Various models suitable for low-altitude, low-speed dynamic landing scenarios can be selected. For example, the target drone can be a quadcopter drone equipped with multi-source sensor modules, data fusion processing units, and flight control units. It has the ability to autonomously navigate, locate, and receive external control commands to complete landing maneuvers, and is the main body for performing vehicle-mounted dynamic landing missions. The second sensor data refers to the data collected by various onboard sensors of the target UAV for sensing its own position, attitude, and motion state. It is the basic data source for the target UAV to achieve autonomous positioning and navigation. The onboard multi-source sensor module of the target UAV can continuously collect various sensing information during flight at a preset acquisition frequency. The sensor fusion center module onboard the UAV performs time synchronization, data denoising, and spatial calibration verification on the collected raw data to complete the acquisition and caching of the second sensor data. For example, the second sensor data may include inertial motion data from the Global Navigation Satellite System (GNSS) and the Inertial Measurement Unit (IMU), as well as image data collected by the onboard visual sensors. The acquisition equipment consists of various navigation, inertial, and visual dedicated sensors onboard the UAV.

[0025] For example, after the vehicle-mounted UAV dynamic landing mission is initiated, various sensing sensors on the vehicle platform immediately and continuously collect data at a preset frequency. After preliminary processing by the vehicle-mounted computing and communication unit, the data is generated as the first sensor data, which is stored in real time and broadcast synchronously. After receiving the landing command from the ground or the vehicle platform, the target UAV immediately enters the full-condition acquisition mode, synchronously collects various flight sensing data and completes preliminary preprocessing to generate the second sensor data, which is then cached locally. The vehicle platform and the target UAV achieve bidirectional real-time interaction between the first and second sensor data through a low-latency communication link, ensuring the synchronization of the data in the time dimension and the integrity in the data dimension, providing accurate and comprehensive data source support for subsequent fusion processing.

[0026] By implementing step 101, the first sensor data of the vehicle platform and the second sensor data of the target UAV are obtained. Environmental perception information from both the vehicle platform and the UAV can be obtained simultaneously, thus providing a multi-source data foundation for subsequent position relationship calculation. By utilizing the complementarity between sensor information from different sources, the completeness and reliability of the method in obtaining the state information of the UAV and the vehicle platform can be improved, providing effective data support for subsequent fusion processing and landing control.

[0027] Step 102: Determine the current fusion strategy from the multi-stage fusion strategy based on the relative distance between the target UAV and the vehicle platform.

[0028] In some examples, the relative distance is the straight-line distance between the target UAV and the center of the preset landing area of ​​the vehicle platform in three-dimensional flight space. This distance serves as the basis for quantitatively determining the UAV's landing flight phase and matching the corresponding data fusion strategy. Its value changes dynamically in real time with the UAV's flight movement. It can be calculated in real time by the target UAV's sensor fusion center module based on the raw data collected by the positioning sensors at both ends of the vehicle platform and the target UAV, using a spatial distance calculation algorithm. The calculation process relies on the onboard computing unit to execute at high speed, and the calculation results are continuously updated according to a preset frequency. For example, the relative distance can be calculated using the two-end three-dimensional positioning coordinates collected by the Global Navigation Satellite System, or it can be calculated using Ultra Wide Band (UWB) ranging data. The calculation results are output in meters and can be presented as specific values ​​such as one hundred meters, fifty meters, and ten meters.

[0029] The multi-stage fusion strategy is a set of pre-defined sensor data fusion processing rules designed to adapt to the positioning needs of different flight stages during the dynamic landing of vehicle-mounted drones. Each strategy corresponds to different sensor data fusion logic, maximizing the technical advantages of different sensors at each landing stage. For example, corresponding fusion strategies are formulated for different landing stages, such as the drone flying from a long distance to the vehicle platform, gradually approaching the vehicle platform during the mid-range, and aligning with the landing area at the terminal stage. Each strategy clarifies the basic sensor data participation types, forming a complete multi-stage fusion strategy system. The current fusion strategy is a sensor data fusion strategy that is precisely matched from the pre-defined multi-stage fusion strategy set based on the target drone's real-time calculated relative distance to the vehicle platform, and is suitable for the drone's current flight state. For example, when the relative distance is in a relatively long range, the fusion strategy corresponding to the long-range flight stage is matched as the current fusion strategy; when the relative distance enters the mid-range range, it automatically switches to the fusion strategy corresponding to the mid-range flight stage. Each distance range uniquely matches a current fusion strategy.

[0030] For example, during the dynamic landing of a vehicle-mounted drone, the sensor fusion center module of the target drone continuously calculates the relative distance with the vehicle platform at a high frequency of hundreds of hertz, ensuring the real-time nature and accuracy of the distance data. The sensor fusion center module compares the real-time relative distance calculated each time with the distance judgment threshold preset in the system initialization phase to complete the accurate judgment of the landing flight phase. After the phase matching is completed through threshold comparison, the sensor fusion center module retrieves the corresponding strategy from the pre-stored multi-stage fusion strategy set as the current fusion strategy and loads the strategy into the data fusion processing execution unit in real time, preparing the rules for subsequent sensor data fusion processing steps.

[0031] By implementing step 102, the current fusion strategy is determined from the multi-stage fusion strategy based on the relative distance between the target UAV and the vehicle platform. As the UAV gradually approaches the vehicle platform, a suitable fusion method can be dynamically selected according to the distance change, so that the positioning and control strategy is more in line with the actual needs of the current landing phase and the adaptability of the method under different distance conditions is improved.

[0032] Step 103: Based on the current fusion strategy, perform fusion processing on the first sensor data and the second sensor data to determine the relative pose information of the target UAV relative to the vehicle platform.

[0033] In some examples, fusion processing is a comprehensive data processing process that integrates, calibrates, calculates, and optimizes multi-dimensional data from the vehicle platform's first sensor data and the target UAV's second sensor data, based on the current fusion strategy. The purpose of this process is to integrate the technical advantages of dual-end sensor data, eliminate the errors and limitations of single data, and provide an accurate and reliable data source for subsequent pose calculation. For example, fusion processing may include filtering and integrating global navigation satellite system positioning data and inertial measurement unit inertial data, collaborative calculation processing of ultra-wideband ranging data and visual image acquisition data, or comprehensive hierarchical integration processing of multiple types of positioning, inertial, and visual data.

[0034] Relative pose information is a set of spatial relative position information of the target UAV relative to the center of the predetermined landing area of ​​the vehicle platform in the vehicle coordinate system preset by the vehicle platform, and the UAV's own attitude information. The position information reflects the spatial relative orientation of the UAV and the landing area of ​​the vehicle platform, and the attitude information reflects the spatial angular state of the UAV itself. This information is the control data that guides the target UAV to complete the dynamic landing. The sensor fusion center module of the target UAV can fuse the data from the first sensor and the second sensor, and then call the spatial pose calculation model in the vehicle coordinate system to calculate the effective data after fusion processing to obtain the corresponding pose parameter set. The calculation results are updated in real time at a preset frequency and cached in the airborne data storage unit for flight control to call. For example, the relative position information may include the relative horizontal coordinate, relative vertical coordinate, and relative height in three-dimensional space, specifically presented as a relative horizontal coordinate of 0.1 meters, a relative height of 2 meters, etc. For example, the attitude information may include the roll angle, pitch angle, and yaw angle, specifically presented as a roll angle of 0°, a pitch angle of 0°, a yaw angle of 0°, etc. The combination of the above position and attitude values ​​is the complete relative pose information.

[0035] Based on the current fusion strategy, the process of fusing the first and second sensor data to determine the relative pose information of the target UAV relative to the vehicle platform can be achieved by the sensor fusion center module first parsing the rules of the current fusion strategy, extracting the sensor data types, data processing algorithms, and data weight allocation rules specified in the strategy, and then performing targeted filtering and dual-end data calibration on the first and second sensor data according to the rules. Subsequently, the matching algorithm is called to perform fusion processing on the calibrated valid data, and finally the relative pose information is calculated and output through the pose calculation model. For example, if the current fusion strategy is adapted to the UAV flying remotely to the vehicle platform, the fusion processing is performed by filtering the global navigation satellite system data and inertial measurement unit data from the dual-end data according to the rules of the strategy, and then the relative pose information in this flight state is calculated. If the current fusion strategy is adapted to the UAV approaching the vehicle platform, the fusion processing is performed by filtering the ultra-wideband data and visual acquisition data from both ends according to the corresponding rules, and then the corresponding relative pose information is calculated.

[0036] By implementing step 103, the first sensor data and the second sensor data are fused based on the current fusion strategy. This allows for the comprehensive utilization of multi-source sensor information, resulting in more accurate and stable pose estimation results. This determines the relative pose information of the target UAV relative to the vehicle platform, thereby improving the accuracy and stability of the spatial relationship calculation between the UAV and the vehicle platform.

[0037] Step 104: Based on the relative pose information, generate landing control commands for the target UAV to guide the target UAV to land on the vehicle platform while the vehicle is in motion.

[0038] In some examples, landing control commands are a standardized set of control commands generated by the target UAV's flight control unit based on real-time relative pose information to guide the UAV in adjusting its flight state and completing a dynamic landing task on a moving vehicle. These commands directly act on the UAV's power actuators, serving as the basis for adjusting the UAV's heading, speed, altitude, and attitude. Their generation and issuance change dynamically in real time as the relative pose information is updated. The target UAV's flight control unit can act as the executing entity for command generation, receiving relative pose information output from the sensor fusion center module in real time, and calling the preset flight control calculation algorithm within the flight control system to achieve the actual pose and ideal landing. The attitude deviation is calculated and the control quantity is solved. Then, the solved control quantity is converted into standardized instructions that can be recognized by the UAV's power actuator. The entire generation process is automatically completed by the airborne flight control system and continuously updated at a preset high frequency. For example, the landing control command can include heading adjustment command, altitude descent command, attitude maintenance command, and speed matching command. Specifically, it can be presented as adjusting the speed forward by 0.2 m / s towards the center of the landing area of ​​the vehicle platform, descending vertically at a rate of 0.1 m / s, maintaining the roll and pitch angles at 0°, and matching the vehicle platform's speed to maintain relative stillness. The command format is compatible with the receiving standard of the UAV's power actuator.

[0039] The real-time relative pose information of the target UAV relative to the vehicle platform, calculated and output by the sensor fusion center module, can be used as the sole core data basis. The deviation analysis between the actual pose and the preset ideal landing pose is completed by the dedicated algorithm of the target UAV flight control unit. The adjustment control quantities of each flight dimension of the UAV are calculated, and then the control quantities are converted into standardized landing control commands and sent to the power actuator. This drives the UAV to dynamically adjust its flight state in real time, gradually eliminate pose deviation, approach the preset landing area of ​​the vehicle platform, and finally achieve a precise and stable landing of the UAV on the moving vehicle. In the implementation process, the flight control unit can first load ideal attitude parameters and flight control calculation algorithms that match the dynamic landing scenario of the vehicle-mounted UAV during the system initialization phase. During the landing process, it receives relative attitude information in real time and establishes a deviation model between the actual attitude and the ideal attitude, calculating the deviation values ​​in dimensions such as heading, speed, altitude, and attitude. Then, the flight control calculation algorithm converts the deviation values ​​into executable control quantities for the power actuators. After format standardization and validity verification, landing control commands are generated and directly sent to the corresponding actuators. The entire process is completed by the airborne flight control system without human intervention. For example, if the relative attitude information shows that the target UAV is 1 meter to the right of the center of the landing area of ​​the vehicle platform and 5 meters above the ground, and there is a forward speed difference of 0.3 m / s between it and the moving vehicle platform, the flight control unit will generate a combination of landing control commands to adjust to the left by 0.1 m / s, descend vertically at a rate of 0.2 m / s, and increase the forward speed by 0.3 m / s. This command guides the UAV to gradually eliminate attitude and speed deviations and move closer to the landing area.

[0040] For example, the flight control unit of the target UAV receives relative pose information output by the sensor fusion center module at a high frequency in real time, and simultaneously retrieves the ideal pose parameters of the vehicle platform's preset landing area from the flight control system. The flight control unit uses a dedicated deviation calculation algorithm to accurately calculate the deviation values ​​between the actual relative pose and the ideal pose in various flight dimensions such as position, attitude, and velocity, and calculates the corresponding control quantities for each power actuator of the UAV based on the deviation values. After converting the calculated control quantities into standardized landing control commands that are compatible with the receiving standards of the power actuators, the flight control unit immediately sends the commands to the UAV's propellers, servos, and other power actuators. The UAV's power actuators adjust their working state in real time according to the received landing control commands, driving the UAV to dynamically correct the pose deviation from the vehicle platform, gradually approaching and finally landing accurately and smoothly in the preset landing area of ​​the vehicle platform during the journey.

[0041] By implementing step 104, a landing control command for the target UAV is generated based on the relative pose information, enabling the UAV to perform motion control according to the real-time acquired pose relationship, thereby guiding the UAV's landing trajectory and enabling the UAV to complete a stable landing on the vehicle platform while the vehicle is in motion, thus improving the stability and controllability of the landing process.

[0042] In summary, this application embodiment, by simultaneously acquiring sensor data from both sides of the vehicle platform and the drone and performing fusion processing, leverages the complementarity between different sensor information to improve the accuracy and reliability of relative pose information acquisition, enabling the drone to more stably perceive its positional relationship with the vehicle platform. By determining the current fusion strategy from a multi-stage fusion strategy based on the relative distance between the drone and the vehicle platform, a suitable fusion method can be dynamically selected based on distance changes during landing, thereby improving the adaptability of positioning and control at different landing stages and ensuring that the drone can obtain effective pose information under different distance conditions. The relative pose information of the drone relative to the vehicle platform is obtained through fusion processing, and landing control commands are generated based on this information, enabling the drone to perform motion control based on real-time pose relationships, guiding the drone's landing trajectory, and improving the stability of the drone performing landing tasks while the vehicle is in motion. By generating landing control commands for the drone during vehicle movement, the drone can complete the landing operation on the mobile platform, thereby expanding the application scenarios of drone landing and improving the applicability of the method in mobile platform environments. In summary, the vehicle-mounted UAV landing control method provided in this application integrates multi-source sensor data from the vehicle platform and the UAV, and dynamically selects a multi-stage fusion strategy based on the relative distance to obtain highly reliable relative pose information, thereby generating landing control commands to guide the UAV to complete a stable and accurate landing while the vehicle is in motion, which can improve the stability and applicability of mobile platform landing tasks.

[0043] In some embodiments, the aforementioned multi-stage fusion strategy may include a long-range return-to-home strategy, a mid-range approach strategy, a terminal alignment strategy, and an anti-disturbance landing strategy; the aforementioned step 102 may include: determining the current fusion strategy as a long-range return-to-home strategy when the relative distance is greater than a first preset distance; determining the current fusion strategy as a mid-range approach strategy when the relative distance is less than or equal to the first preset distance and greater than a second preset distance; determining the current fusion strategy as a terminal alignment strategy when the relative distance is less than or equal to the second preset distance and greater than a third preset distance; and determining the current fusion strategy as an anti-disturbance landing strategy when the relative distance is less than or equal to the third preset distance.

[0044] In some examples, the long-range return-to-home strategy, mid-range approach strategy, terminal alignment strategy, and anti-disturbance touchdown strategy are sensor data fusion strategies customized for the four progressive flight stages of the dynamic landing of vehicle-mounted drones. Together, they form a complete system of multi-stage fusion strategies. Each strategy formulates exclusive data fusion rules based on the positioning accuracy, distance characteristics, and environmental adaptation requirements of the corresponding flight stage, which can maximize the adaptation to the positioning requirements of the drone from its long-range flight to the vehicle platform to its final touchdown. For example, the long-range return-to-home strategy is adapted to the stage when the drone and the vehicle platform are flying at a long distance relative to each other, focusing on meeting the fusion requirements of long-range coarse positioning; the mid-range approach strategy is adapted to the stage when the drone gradually approaches the vehicle platform, focusing on meeting the fusion requirements of mid-range relative positioning; the terminal alignment strategy is adapted to the stage when the drone approaches the landing area of ​​the vehicle platform, focusing on meeting the fusion requirements of terminal fine positioning; and the anti-disturbance touchdown strategy is adapted to the stage when the drone finally lands on the vehicle platform, focusing on meeting the fusion requirements of anti-disturbance precise touchdown. The four strategies form a progressive fusion rule system according to the landing process.

[0045] The first, second, and third preset distances are three spatial distance thresholds pre-set to divide the four flight stages of a vehicle-mounted drone's dynamic landing. These three values ​​decrease sequentially and serve as the quantitative basis for the target drone's sensor fusion center module to determine the relative distance range and match the corresponding fusion strategy. The specific values ​​of these three distance thresholds can be determined through field calibration and simulation testing, taking into account the sensor sensing range of the vehicle platform, the flight performance of the target drone, and the actual dynamic landing scenario. These values ​​are then entered into the threshold configuration library of the target drone's sensor fusion center module. These values ​​can be flexibly adjusted according to different landing scenarios, drone models, and vehicle platform types. For example, after calibration, the first preset distance can be set to fifty meters, and the second preset distance... The first preset distance can be set to 10 meters, and the second preset distance can be set to 3 meters. These three distances form a hierarchical distance judgment standard of 50 meters, 10 meters, and 3 meters, enabling precise division of the landing phase. For example, if the first preset distance is 50 meters, the second preset distance is 10 meters, and the third preset distance is 3 meters, when the calculated relative distance is 60 meters, it is determined to be a long-range return-to-home strategy because it is greater than the first preset distance. When the relative distance is 30 meters, it is determined to be a mid-range approach strategy because it is less than or equal to the first preset distance and greater than the second preset distance. When the relative distance is 5 meters, it is determined to be a terminal alignment strategy because it is less than or equal to the second preset distance and greater than the third preset distance. When the relative distance is 2 meters, it is determined to be an anti-disturbance touchdown strategy because it is less than or equal to the third preset distance.

[0046] Through the implementation of the above embodiments, the multi-stage fusion strategy is divided into a long-range return-to-home strategy, a mid-range approach strategy, a terminal alignment strategy, and an anti-disturbance touchdown strategy. The corresponding strategy is selected based on the relative distance between the UAV and the vehicle platform, enabling the UAV to adopt different positioning and control methods according to different distance stages during the landing process. This allows the vehicle-mounted UAV to adopt a more suitable fusion scheme in the long-range, mid-range, and short-range landing stages, thereby improving the positioning stability and control adaptability of the UAV as it gradually approaches the vehicle platform, and helping to solve the problem of unstable control of UAV landing while the vehicle is in motion.

[0047] In some embodiments, the aforementioned first sensor data may include ultra-wideband base station array data, visual marker data, vehicle motion state data, and first global navigation positioning data; the aforementioned second sensor data may include second global navigation positioning data, ultra-wideband tag data, image acquisition data, and inertial measurement unit data; the aforementioned step 103 may include: when the current fusion strategy is a long-range return-to-home strategy, performing data fusion processing based on the first global navigation positioning data, inertial measurement unit data, and second global navigation positioning data to obtain relative pose information; when the current fusion strategy is a mid-range approach strategy, performing data fusion processing based on ultra-wideband tag data, ultra-wideband base station array data, image acquisition data, and inertial measurement unit data to obtain relative pose information; when the current fusion strategy is an end-point alignment strategy, performing data fusion processing based on visual marker data, image acquisition data, and inertial measurement unit data to obtain relative pose information; when the current fusion strategy is an anti-disturbance landing strategy, performing data fusion processing based on inertial measurement unit data and vehicle motion state data to obtain relative pose information.

[0048] In some examples, the ultra-wideband (UWB) base station array data consists of the raw ranging communication data collected by the UWB array deployed on the vehicle-mounted platform and the target UAV's onboard UWB tag, as well as the base station's own position calibration data. The vehicle-mounted platform can deploy at least three UWB base stations according to preset rules, using a high-frequency acquisition frequency to conduct ranging communication with the UAV's onboard UWB tag, collecting real-time ranging data. Simultaneously, the position data of each base station in the vehicle coordinate system, calibrated during system initialization, is used as the basic parameters. After the vehicle-mounted computing and communication unit completes data format standardization, validity verification, and timestamp marking, the UWB base station array data is formed. This data is stored in real-time and broadcast to the UAV. For example, the UWB base station array data includes the real-time ranging value of 40 meters between base station 1 and the onboard tag, the real-time ranging value of 41 meters between base station 2 and the onboard tag, and the three-dimensional coordinate calibration data of each base station in the vehicle coordinate system.

[0049] Visual marker data consists of the inherent feature data and position calibration data of cooperative visual markers deployed at the center of the pre-set landing area on the vehicle-mounted platform. It serves as the vehicle-mounted reference data for visual recognition and relative pose calculation of the target UAV. The system can collect features and calibrate the spatial position of cooperative visual markers in the landing area of ​​the vehicle-mounted platform. The unique coding information, geometric dimensions, and three-dimensional position information of the markers in the vehicle coordinate system are entered into the vehicle's computing and communication unit. After data solidification and verification, visual marker data is formed. This data is stored on the vehicle and can be retrieved in real time during the UAV's landing process. For example, visual marker data includes the exclusive coding data of the AprilTag marker, the geometric dimension data of an 80-centimeter side length, and the three-dimensional coordinate data of the vehicle coordinate system at the center of the landing area.

[0050] Vehicle motion state data is parameter data reflecting the real-time kinematic characteristics of the onboard platform during its operation. It serves as reference data for compensating for dynamic disturbances in the vehicle body during the anti-disturbance landing phase of the target UAV. The onboard platform can collect motion parameters such as vehicle speed, angular velocity, and attitude angle in real time from the vehicle control system via the vehicle controller LAN bus. After data filtering, zero-bias correction, and timestamp marking by the onboard computing and communication unit, the vehicle motion state data is generated. This data is updated at a preset frequency and broadcast to the UAV. For example, the vehicle motion state data includes motion parameters such as the onboard platform's longitudinal speed of 20 kilometers per hour, the vehicle's yaw rate of 0.1 radians per second, and the vehicle's pitch angle of 0 degrees Celsius.

[0051] The first global navigation positioning data is the real-time global positioning and speed data collected by the vehicle platform through the Global Navigation Satellite System. The global navigation satellite system receiver on the vehicle platform receives satellite navigation signals at a preset frequency, calculates the three-dimensional coordinates, real-time speed, and heading angle of the vehicle platform in the geographic coordinate system, and after the vehicle computing and communication unit completes the data validity verification, the first global navigation positioning data is formed. This data is broadcast in real time and stored locally. For example, the first global navigation positioning data includes the longitude, latitude, and altitude geographic coordinates of the vehicle platform, as well as the calculated data such as the vehicle's real-time driving speed and heading angle.

[0052] The second global navigation and positioning data is the real-time global positioning and flight speed data collected by the target UAV through the Global Navigation Satellite System. The receiver of the Global Navigation Satellite System on the target UAV can receive satellite navigation signals at a preset frequency and calculate the UAV's three-dimensional coordinates, flight speed, and heading angle in the geographic coordinate system. After the airborne sensor fusion center module completes the data denoising and time synchronization processing, the second global navigation and positioning data is formed. This data is cached locally and can be interacted with the vehicle platform. For example, the second global navigation and positioning data includes the UAV's longitude, latitude, and altitude geographic coordinates, as well as the calculated data such as the UAV's flight speed and heading angle.

[0053] Ultra-wideband (UWB) tag data is the raw data collected by the UWB tag on the target UAV and the UWB base station array on the vehicle platform for ranging communication. It is the basic UWB data for calculating the position of the target UAV relative to the vehicle platform. The UWB tag on the target UAV can conduct bidirectional ranging communication with the UWB base station array on the vehicle platform at a high frequency to collect real-time ranging data with each base station. After the airborne sensor fusion center module completes the data validity screening and time synchronization processing, the UWB tag data is formed. This data is cached locally on the UAV in real time. For example, the UWB tag data includes ranging communication data such as the real-time ranging value of 35 meters between the airborne tag and vehicle base station 1, and the real-time ranging value of 36 meters between the airborne tag and vehicle base station 2.

[0054] Image acquisition data consists of real-time images and video frames of the vehicle platform and its surrounding environment captured by the onboard visual camera of the target UAV. It serves as the basic visual data for UAV visual recognition and relative pose calculation. The onboard visual camera of the target UAV continuously acquires real-time image data from the front and below during flight at a preset frame rate. After image denoising, format conversion, and feature enhancement processing are completed by the onboard sensor fusion center module, the image acquisition data is generated and cached locally on the UAV. For example, the image acquisition data includes visual marker images of the landing area of ​​the vehicle platform, images of the vehicle platform body, and continuous video frame image data captured by the UAV.

[0055] Inertial measurement unit (IMU) data consists of real-time motion inertial parameters collected by the onboard IMU of the target UAV. It serves as onboard data for UAV attitude recursion, attitude maintenance, and compensation for various positioning errors. The IMU collects inertial parameters such as three-axis angular velocity and three-axis acceleration at a high frequency of 100 Hz. After zero-bias compensation, data filtering, and preliminary attitude angle recursion are performed by the onboard sensor fusion center module, IMU data is generated. This data is cached locally on the UAV in real time and participates in fusion processing. For example, IMU data includes inertial parameters such as the UAV's three-axis angular velocity of 0.02 radians per second and three-axis acceleration of 0.1 meters per second squared, as well as preliminary processed attitude angle recursion data.

[0056] Under the current fusion strategy of long-range return-to-home, data fusion processing is performed based on the first global navigation positioning (GNSS) data, inertial measurement unit (INS) data, and second GNSS data to obtain relative pose information. This is a multi-source sensor data fusion execution logic specific to the long-range return-to-home phase. It uses dual-end global positioning data from the global navigation satellite system as the core and INS data as an auxiliary, adapting to the navigation requirements of UAVs for long-range coarse positioning. The target UAV's sensor fusion center module can parse the long-range return-to-home strategy, then retrieve the first GNSS data from the vehicle broadcast data and the second GNSS data from the local cache. The data and inertial measurement unit (IMU) data are first registered in the geographic coordinate system and synchronized in time with the dual-end global navigation positioning (GNPS) data. Then, a preset fusion algorithm is called to integrate, filter, optimize, and correct the errors of the three types of data. Finally, the relative pose information is calculated through a spatial pose calculation model. For example, the vehicle-mounted first GNPS data and the UAV's second GNPS data are retrieved, and their initial relative positions in the geographic coordinate system are calculated. Then, the UAV's IMU data is fused to correct the update delay error of the GNPS data, and the remote coarse positioning relative pose information of the UAV relative to the vehicle platform is calculated.

[0057] Under the current fusion strategy of mid-range approach, data fusion processing is performed based on ultra-wideband tag data, ultra-wideband base station array data, image acquisition data, and inertial measurement unit data to obtain relative pose information. This is a multi-source sensor data fusion execution logic specifically for the mid-range approach phase. It uses the relative positioning of ultra-wideband dual-end ranging data as its core, fusing visual image acquisition data and inertial measurement unit data to meet the positioning requirements of UAVs for precise mid-range approach to vehicle-mounted platforms. The target UAV's sensor fusion center module can parse the mid-range approach strategy and then selectively retrieve ultra-wideband base station array data from the vehicle broadcast data and retrieve data from its local cache. The process involves first performing ranging calculations on the ultra-wideband (UWB) dual-end data, image acquisition data, and inertial measurement unit (IMU) data, and then performing collaborative preprocessing on the image acquisition data and IMU data. Next, a pre-defined fusion algorithm is invoked to integrate, optimize, and correct the deviations of the four types of data. Finally, a spatial pose calculation model is used to calculate the relative pose information. For example, by fusing UWB base station array data and UWB tag data, the centimeter-level horizontal position of the UAV relative to the vehicle platform is calculated. Then, by fusing the UAV's image acquisition data and IMU data to correct altitude and attitude deviations, the mid-range precise relative pose information of the UAV relative to the vehicle platform is calculated.

[0058] With the current fusion strategy being an end-of-pipe alignment strategy, data fusion processing is performed based on visual marker data, image acquisition data, and inertial measurement unit (IMU) data to obtain relative pose information. This is a dedicated multi-source sensor data fusion execution logic for the end-of-pipe alignment stage. Using vehicle-mounted visual marker data as a reference and UAV image acquisition data as the recognition basis, it fuses IMU data to meet the high-precision positioning requirements of UAVs hovering and aligning precisely above the vehicle platform's landing area. The target UAV's sensor fusion center module can parse the end-of-pipe alignment strategy, then directionally retrieve visual marker data from the vehicle broadcast data and image acquisition data and IMU data from the local cache. The unit data is first processed by a visual recognition algorithm to match the cooperative visual markers and vehicle-mounted visual markers in the image acquisition data to calculate the initial visual pose. Then, inertial measurement unit data is fused to correct the instantaneous visual recognition errors caused by vehicle bumps and changes in UAV attitude. Finally, a preset fusion algorithm is called to complete the data optimization and calculate the relative pose information. For example, the feature and position data of the vehicle-mounted AprilTag are retrieved, matched with the image of the marker in the UAV image acquisition data, and the initial pose is calculated. Then, inertial measurement unit data is fused to correct the recognition deviation and calculate the sub-decimeter-level high-precision relative pose information of the UAV relative to the center of the landing area of ​​the vehicle platform.

[0059] Under the current fusion strategy of anti-disturbance landing, data fusion processing based on inertial measurement unit (IMU) data and vehicle motion state data is performed to obtain relative pose information. This is a multi-source sensor data fusion execution logic specific to the anti-disturbance landing phase. Using UAV high-frequency IMU data as the core, it integrates vehicle motion state data for dynamic disturbance compensation, adapting to the positioning requirements of synchronous and smooth landing of the UAV and the vehicle platform. The target UAV's sensor fusion center module can parse the anti-disturbance landing strategy, then directionally retrieve vehicle motion state data from the vehicle broadcast data and retrieve inertial measurement data from the local cache. For unit data, first, time synchronization and spatial coordinate system registration of the two types of data are completed. Then, vehicle motion state data is used as dynamic disturbance compensation. A preset fusion algorithm is called to fuse inertial measurement unit data and compensation data to correct the predicted motion trajectory of the UAV. Finally, the disturbance-resistant relative pose information is calculated. For example, the high-frequency inertial measurement unit data of the UAV is retrieved to recursively predict the motion trajectory. Then, vehicle motion state data such as vehicle speed and yaw rate are fused to dynamically compensate the predicted trajectory to offset vehicle turbulence and motion disturbances. The relative pose information of the UAV relative to the vehicle platform during the touchdown phase is calculated.

[0060] By implementing the above embodiments, global navigation positioning data, ultra-wideband positioning data, visual information, and inertial measurement unit data are integrated at different stages and processed in conjunction with vehicle motion state information. This allows for full utilization of the advantages of various sensors under different distance conditions, improving the pose estimation accuracy of the UAV relative to the vehicle platform. For example, coarse positioning is achieved using navigation positioning data at long range, relative positioning is achieved using ultra-wideband and visual information at mid-range, precise alignment is achieved using visual markers at the terminal stage, and stable control is achieved by combining vehicle motion state at the touchdown stage. This comprehensively improves the positioning reliability and control stability of the UAV during the landing process on the mobile platform.

[0061] In some embodiments, the aforementioned data fusion processing based on the first global navigation positioning data, inertial measurement unit data, and second global navigation positioning data to obtain relative pose information may include: determining the first relative position of the target UAV relative to the vehicle platform based on the first global navigation positioning data and the second global navigation positioning data; and fusing the first relative position and the inertial measurement unit data based on a preset filtering algorithm to obtain relative pose information.

[0062] In some examples, the first relative position is calculated by the sensor fusion center module of the target UAV based on the first global navigation positioning data of the vehicle platform and the second global navigation positioning data of the UAV. The spatial basic position information of the target UAV relative to the vehicle platform changes in real time with the update of the global navigation positioning data of both ends. The sensor fusion center module of the target UAV can retrieve the first global navigation positioning data from the vehicle broadcast data and retrieve the second global navigation positioning data from the local cache. First, the two types of data are unified under the same geographic coordinate system. Then, the spatial position calculation algorithm is used to perform difference calculation on the three-dimensional coordinate data of both ends to calculate the basic position parameters of the UAV in three-dimensional space relative to the vehicle platform. After data validity verification and outlier removal, it is determined as the first relative position. For example, the first relative position can be presented as the target UAV's relative horizontal coordinate of 10 meters, relative vertical coordinate of 8 meters, and relative height of 30 meters relative to the vehicle platform, or it can be presented as the target UAV's spatial position 50 meters behind the vehicle platform and at a height of 25 meters.

[0063] The preset filtering algorithm is a dedicated data filtering optimization algorithm pre-configured in the target UAV's sensor fusion center module during the system initialization phase. It is adapted to the fusion processing of the first relative position and inertial measurement unit data during the remote return phase. The algorithm's function is to eliminate measurement errors from a single data source, compensate for the performance defects of different data, achieve complementary advantages of multi-source data, and improve the accuracy and continuity of relative pose information. Based on the positioning accuracy requirements of the vehicle-mounted UAV during the remote return phase and the performance characteristics of the GNSS data and inertial measurement unit data, a suitable algorithm type can be selected from mainstream filtering algorithms. The algorithm is written as an executable program and embedded in the target UAV's sensor fusion center module. At the same time, an algorithm parameter adjustment interface is reserved so that the filtering parameters can be fine-tuned according to the actual application scenario. The algorithm program is stored in the onboard storage unit for the system to call in real time. For example, the preset filtering algorithm can be an Extended Kalman Filter (EKF) or an Error State Kalman Filter (ESKF). Both types of algorithms can achieve efficient fusion and error correction of position and inertial data.

[0064] The sensor fusion center module of the target UAV can first perform time synchronization calibration between the calculated first relative position and the locally cached inertial measurement unit (IMU) data to ensure the consistency of the two types of data in the time dimension. Then, a preset filtering algorithm is loaded into the module, using the first relative position as the algorithm's observation value and the IMU data as the algorithm's motion state recursion basis. The algorithm is then substituted to complete the data filtering optimization, error correction, and state estimation. Finally, through the pose calculation model in the vehicle coordinate system, the complete relative pose information of the target UAV relative to the vehicle platform is obtained from the fused data analysis. For example, the calculated first relative position is input into the extended Kalman filter algorithm as the observation value, and the three-axis angular velocity and three-axis acceleration data collected by the IMU are used as the recursive data of the UAV's motion state. After filtering and fusion by the algorithm, the position deviation caused by the lag in the update of the global navigation satellite system data is corrected, and the real-time attitude angle of the UAV is calculated. Finally, the relative pose information containing three-dimensional position and three-axis attitude angle is obtained.

[0065] By implementing the above embodiments, the first relative position of the UAV relative to the vehicle platform is determined using the global navigation and positioning data of the vehicle platform and the UAV. Combined with the inertial measurement unit data, the relative pose estimation result can be obtained relatively stably at long distances. Effective initial positioning can be achieved when the UAV is far away from the vehicle platform. The positioning result is compensated and smoothed using inertial information, thereby improving the positioning stability of the UAV during its return to the vehicle platform at long distances and providing a reliable pose basis for the subsequent landing phase.

[0066] In some embodiments, the aforementioned data fusion processing based on UWTA tag data, UWTA base station array data, image acquisition data, and inertial measurement unit data to obtain relative pose information may include: determining a second relative position of the target UAV relative to the vehicle-mounted platform based on UWTA tag data and UWTA base station array data; performing visual inertial odometry assessment based on image acquisition data and inertial measurement unit data to obtain visual inertial pose information of the target UAV; and fusing the second relative position and visual inertial pose information based on a preset filtering algorithm to obtain relative pose information.

[0067] In some examples, the second relative position is calculated by the target UAV's sensor fusion center module based on airborne UWB tag data and vehicle-mounted UWB base station array data. This high-precision spatial position information of the target UAV relative to the vehicle platform relies on the ranging characteristics of UWB technology to achieve centimeter-level horizontal positioning accuracy. It serves as the positional basis for the UAV's relative pose calculation during the mid-range approach phase, and its calculation results change dynamically in real time as the UWB dual-end ranging data is updated. The target UAV's sensor fusion center module can selectively retrieve UWB base station array data from the vehicle-mounted broadcast data and retrieve UWB data from its local cache. Broadband tag data is first unified into the vehicle coordinate system preset by the vehicle platform. Then, spatial position is calculated by combining the position calibration data of the base station array with the real-time ranging data from both ends through a polygonal positioning and other ranging algorithms. After data validity verification and outlier removal, the second relative position is determined. For example, the second relative position can be presented as the relative horizontal coordinate of the target UAV relative to the center of the landing area of ​​the vehicle platform is 0.5 meters, the relative vertical coordinate is 0.3 meters, and the relative height is 8 meters. It can also be presented as the centimeter-level spatial position of the UAV 20 meters behind the vehicle platform, with a horizontal offset of 0.2 meters and a height of 6 meters.

[0068] Visual-inertial pose information (VIS) is the complete pose information of the target UAV, obtained by the sensor fusion center module of the target UAV based on airborne image acquisition data and inertial measurement unit (IMU) data through visual-inertial odometry (VIO). This information integrates the absolute pose perception advantage of visual data and the high-frequency motion recursion advantage of IMU data, including parameters such as the UAV's three-dimensional position, real-time velocity, and spatial attitude angles. It is key data for compensating for the shortcomings of ultra-wideband positioning (UWB) in altitude perception and attitude calculation during the mid-range approach phase. The target UAV's sensor fusion center module can synchronously retrieve image acquisition data and inertial measurement unit (IMU) data from its local cache. For unit data, the time synchronization of the two types of data and the spatial calibration of the camera-inertial measurement unit are first completed. Then, the visual inertial odometry method is run to extract, track and match environmental feature points in the image acquisition data. Combined with the three-axis angular velocity and three-axis acceleration of the inertial measurement unit data, motion state recursion and pose calculation are performed. After error correction, visual inertial pose information is obtained. For example, visual inertial pose information may include the three-dimensional position coordinates of the UAV, the flight speed of 0.8 m / s, and the spatial attitude angles of roll angle 0°, pitch angle 0.5°, and yaw angle 1°, along with real-time trend data of velocity and attitude changes.

[0069] The sensor fusion center module of the target UAV can first synchronize the calculated second relative position with the visual-inertial pose information in time and perform coordinate system one calibration to ensure the consistency of the two types of data in the time and space dimensions. Then, a preset filtering algorithm is loaded into the module, using the second relative position as the position observation value of the algorithm and the visual-inertial pose information as the basis for the algorithm's motion state recursion and attitude observation. Substitute these values ​​into the algorithm to complete the data filtering optimization, error correction, and optimal state estimation. Finally, through the pose calculation model in the vehicle coordinate system, the complete relative pose information of the target UAV relative to the vehicle platform is obtained from the fused data analysis. For example, the second relative position input error state Kalman filter algorithm of the ultra-wideband solution can be used as the horizontal position observation value. At the same time, the height, velocity, and attitude angle data of the visual-inertial pose information can be used as the supplementary observation and state recursion basis of the algorithm. After the algorithm filters and fuses, the deviation of ultra-wideband in altitude positioning and the cumulative error of visual-inertial odometry are corrected, and finally, relative pose information containing centimeter-level horizontal position, accurate height, and real-time attitude is obtained.

[0070] Through the implementation of the above embodiments, the second relative position of the UAV relative to the vehicle platform is determined based on ultra-wideband tag data and ultra-wideband base station array data. Visual inertial odometry is then evaluated by combining image acquisition data and inertial measurement unit data. After fusion processing through filtering algorithms, more accurate relative pose information can be obtained in the mid-range stage when the UAV gradually approaches the vehicle platform. By utilizing the complementary characteristics of ultra-wideband positioning and visual inertial information, the relative positioning accuracy in the mid-range stage can be improved, which is conducive to the stable approach of the UAV to the moving vehicle platform.

[0071] In some embodiments, the aforementioned data fusion processing based on visual marker data, image acquisition data, and inertial measurement unit data to obtain relative pose information may include: identifying cooperative markers corresponding to the visual marker data based on the image acquisition data; calculating the cooperative markers using a perspective n-point algorithm to obtain the six-degree-of-freedom relative pose of the target UAV relative to the cooperative markers; determining the inertial pose information of the target UAV based on the inertial measurement unit data; and fusing the six-degree-of-freedom relative pose and inertial pose information based on a preset filtering algorithm to obtain the relative pose information.

[0072] In some examples, the cooperative marker is a dedicated visual positioning marker uniquely corresponding to the visual marker data, deployed at the center of the pre-set landing area of ​​the vehicle-mounted platform. This marker has a unique coded pattern and fixed geometric dimensions, serving as a physical reference for achieving high-precision visual pose calculation during the terminal alignment phase of the target UAV. The marker can be deployed at the center of the landing area of ​​the vehicle-mounted platform according to a pre-set specification. At the same time, the coded features, geometric dimensions, and three-dimensional position of the marker in the vehicle coordinate system are collected and calibrated. The relevant information is entered into the vehicle-mounted computing and communication unit to form visual marker data. This marker is a physical entity that maintains a fixed position during landing for identification and collection by the UAV's onboard camera. For example, the cooperative marker can be an AprilTag or a QR code with a unique code. A commonly used deployment specification is a square April tag with a side length of 80 centimeters, fixed at the center of the landing area of ​​the vehicle-mounted platform.

[0073] The perspective n-point algorithm is a dedicated visual pose calculation algorithm pre-configured in the target UAV sensor fusion center module. It is used to calculate the three-dimensional spatial pose through two-dimensional image feature points. This algorithm establishes a perspective projection relationship by matching the coordinates of the two-dimensional feature points of the cooperative marker in the image with the actual three-dimensional feature point coordinates of the cooperative marker, and calculates the spatial pose of the camera relative to the cooperative marker. It is an algorithm for achieving high-precision visual pose calculation in the end-effector alignment stage. For example, the perspective n-point algorithm can adapt to the calculation requirements of different numbers of feature points. When the four corner features of the cooperative marker are identified, the spatial pose calculation of the UAV relative to the cooperative marker can be completed. It is a classic pose calculation algorithm in the field of visual positioning.

[0074] The six-degree-of-freedom relative pose is the complete spatial pose information of the target UAV relative to the cooperative marker on the vehicle platform, calculated by the perspective n-point algorithm. It includes three-dimensional positional degrees of freedom and three-dimensional attitude degrees of freedom. The three-dimensional positional degrees of freedom reflect the UAV's lateral, longitudinal, and vertical relative position to the cooperative marker in the vehicle coordinate system, while the three-dimensional attitude degrees of freedom reflect the UAV's roll, pitch, and yaw angles relative to the cooperative marker. This pose information has sub-decimeter-level positioning accuracy and serves as the visual observation basis for UAV pose calculation during the end-of-line alignment stage. It can be matched by the target UAV's sensor fusion center module using the perspective n-point algorithm. The 2D image feature points of the cooperative markers collected by the airborne camera and the actual 3D feature points of the cooperative markers in the visual marker data are used by the algorithm to calculate the 3D position parameters and 3D attitude parameters of the UAV relative to the cooperative markers. The combination of these two types of parameters constitutes the six-degree-of-freedom relative pose. The calculation results are cached for later use after validity verification. For example, the six-degree-of-freedom relative pose can be presented as position parameters with a relative horizontal coordinate of 0.1 meters, a relative vertical coordinate of 0.05 meters, and a relative altitude of 1 meter, as well as attitude parameters with a roll angle of 0 degrees, a pitch angle of 0 degrees, and a yaw angle of 0 degrees. This combination of parameters realizes the accurate representation of the UAV's full-dimensional pose relative to the cooperative markers.

[0075] Inertial pose information is the real-time motion pose information of the target UAV, calculated by the sensor fusion center module based on data from the onboard inertial measurement unit. This information relies on the high-frequency acquisition characteristics of the inertial measurement unit (IFU) at the 100Hz level to achieve continuous updates of pose data. It includes parameters such as the UAV's real-time attitude angles, motion acceleration, and position recursion trend. This data is crucial for compensating for instantaneous errors in visual pose calculation and achieving continuous pose perception. The sensor fusion center module of the target UAV can retrieve the three-axis data acquired by the IFU from its local cache. The raw data of angular velocity and three-axis acceleration are preprocessed by zero bias compensation, data filtering, and kinematic recursion to obtain the real-time attitude angle and position recursion information of the UAV. The combination of the two types of information is the inertial attitude information, which is updated in real time as the inertial measurement unit collects data. For example, the inertial attitude information can be presented as motion parameters of three-axis angular velocity of 0.01 radians per second and three-axis acceleration of 0.05 meters per square second, as well as attitude parameters of roll angle of 0.2 degrees and pitch angle of 0.1 degrees, along with short-term position recursion offset data of the UAV.

[0076] The sensor fusion center module of the target UAV can first synchronize the calculated six-DOF relative pose and inertial pose information in time and perform coordinate system one calibration to ensure the consistency of the two types of data in the time and space dimensions. Then, a preset filtering algorithm is loaded into the module, using the six-DOF relative pose as the high-weight observation value of the algorithm and the inertial pose information as the basis for the algorithm's motion state recursion. Substitute these values ​​into the algorithm to complete the data filtering optimization, error correction, and optimal state estimation. Finally, through the pose calculation model in the vehicle coordinate system, the complete relative pose information of the target UAV relative to the vehicle platform is calculated from the fused data. For example, the six-DOF relative pose input error state Kalman filter algorithm is used as the core observation value, and the inertial pose information is used as the state recursion data of the algorithm. After filtering and fusion by the algorithm, the instantaneous visual pose recognition error caused by vehicle bumps and UAV attitude changes, as well as the cumulative recursion error of the inertial measurement unit data, are corrected, and finally, the relative pose information containing sub-decimeter position and high-precision attitude is obtained.

[0077] By implementing the above embodiments, visual markers in the image acquisition data are identified, and the six-degree-of-freedom relative pose of the UAV relative to the cooperative markers is calculated using the perspective n-point algorithm. At the same time, combined with the inertial measurement unit data for fusion processing, high-precision pose alignment information can be obtained in the terminal stage when the UAV approaches the vehicle platform. This can improve the spatial positioning accuracy of the UAV in the pre-landing stage, enabling the UAV to more accurately align with the landing area of ​​the vehicle platform, thereby improving the accuracy of the landing operation.

[0078] In some embodiments, the aforementioned data fusion processing based on inertial measurement unit data and vehicle motion state data to obtain relative pose information may include: performing motion recursion based on inertial measurement unit data to obtain the predicted motion trajectory of the target UAV; using vehicle motion state data as a feedforward compensation amount to dynamically correct the predicted motion trajectory to obtain the compensated motion trajectory; and updating the state of the compensated motion trajectory based on a preset filtering algorithm to obtain relative pose information.

[0079] In some examples, motion recursion is a process in which the sensor fusion center module of the target UAV continuously calculates the UAV's motion state over a short period of time based on data from the onboard inertial measurement unit and a kinematic model. This process fully utilizes the high-frequency acquisition characteristics of the inertial measurement unit (IFU) at the 100 Hz level to achieve continuous recursion of the UAV's motion state, and is a method for calculating the UAV's trajectory during the touchdown phase. The sensor fusion center module of the target UAV can retrieve the raw three-axis angular velocity and three-axis acceleration data collected by the IFU from its local cache, and first perform preprocessing such as zero-bias compensation, data filtering, and attitude calculation. Then, the preset kinematic recursion algorithm is invoked, taking the current pose and motion state of the UAV as the initial value. Based on the real-time changes in the inertial measurement unit data, the position, velocity, attitude and other motion states of the UAV at the next moment are continuously calculated. The entire process is automatically executed at high frequency by the airborne computing unit. For example, the motion recursion is based on the three-axis acceleration of 0.1 m / s² and the vertical descent angular velocity of 0.02 radians per second collected by the inertial measurement unit. Combined with the initial state of the UAV currently hovering 3 meters above the vehicle platform, the position and attitude change trends of the UAV in the next 0.1 seconds and 0.2 seconds are recursively obtained.

[0080] The predicted trajectory is the expected trajectory of the UAV under short-term conditions without external disturbances, obtained by the sensor fusion center module of the target UAV through motion recursion. This trajectory includes spatiotemporal parameters such as the UAV's three-dimensional position, flight speed, and spatial attitude at different points in time. It serves as the basic reference for UAV trajectory planning during the anti-disturbance landing phase, and its trajectory trend is dynamically adjusted in real time according to changes in inertial measurement unit data. The sensor fusion center module can integrate and fit the motion state parameters of the UAV at each moment obtained by motion recursion according to the time series to form continuous trajectory data. After data validity verification and outlier removal, it is determined as the predicted trajectory. For example, the predicted trajectory is a continuous spatial trajectory of the UAV descending vertically at a rate of 0.2 m / s and a forward flight speed of 0.3 m / s, and then descending from 3 meters above the vehicle platform to 2 meters and moving forward 0.3 meters in the following second. The trajectory includes the three-dimensional position and attitude parameters corresponding to every 0.01 seconds.

[0081] The feedforward compensation amount is a quantitative compensation parameter obtained by calibrating the vehicle motion state data of the vehicle platform according to the vehicle coordinate system and the kinematic characteristics of the UAV. This parameter is used to offset the turbulence, vibration and dynamic disturbances caused by the vehicle motion generated during the vehicle platform's operation. It is the quantitative basis for correcting the predicted trajectory of the UAV, and its value changes in real time with the update of the vehicle motion state data. The sensor fusion center module of the target UAV can retrieve the vehicle motion state data from the vehicle broadcast data, first unify the data to the motion reference coordinate system of the UAV, and then use a preset compensation amount conversion algorithm to convert the vehicle's motion parameters such as speed, angular velocity and attitude angle into quantitative compensation parameters adapted to the UAV trajectory correction. The calibrated parameters are the feedforward compensation amount. For example, the feedforward compensation amount is to convert the longitudinal driving speed of the vehicle platform of 20 kilometers per hour into the forward flight speed compensation amount of 5.6 meters per second for the UAV, and the yaw angular velocity of the vehicle of 0.1 radians per second into the yaw angular velocity compensation amount of 0.1 radians per second for the UAV, so as to match the real-time motion state of the vehicle platform.

[0082] The compensated trajectory is the actual planned trajectory of the UAV obtained by the sensor fusion center module of the target UAV after substituting the feedforward compensation amount into the trajectory correction model and dynamically correcting the predicted trajectory. This trajectory fits the real-time driving state of the vehicle platform and can effectively offset the impact of vehicle dynamic disturbances on the UAV's touchdown. It serves as the basis for the UAV's flight trajectory during the anti-disturbance touchdown phase. The sensor fusion center module can call a preset trajectory dynamic correction model, using the feedforward compensation amount as a correction factor, to correct the position, velocity, attitude, and other spatiotemporal parameters of the predicted trajectory point by point. After correction, the trajectory data is smoothed and its validity is verified to finally obtain the compensated trajectory. For example, the compensated trajectory is based on the original predicted trajectory, with the addition of a forward velocity compensation of 5.6 m / s and a yaw angular velocity compensation of 0.1 radians per second. After correction, the forward flight speed of the UAV is consistent with the driving speed of the vehicle platform, and the yaw motion is synchronized with the lateral sway of the vehicle body. During the vertical descent from 3 meters to 2 meters, there is no relative displacement between the UAV and the vehicle platform.

[0083] The sensor fusion center module of the target UAV can first perform timestamp calibration and coordinate system registration on the compensated motion trajectory data to ensure that the data is consistent with the vehicle coordinate system. Then, a preset filtering algorithm is loaded into the module, and the compensated motion trajectory is used as the state prediction value of the algorithm. The algorithm is substituted into the algorithm to complete the real-time update of the motion state, error correction and optimal state estimation. Finally, the pose calculation model in the vehicle coordinate system is used to calculate the complete relative pose information of the target UAV relative to the vehicle platform from the updated state data. For example, the state Kalman filter algorithm with the compensated motion trajectory input error is used as the state prediction value. After the algorithm completes the real-time update of the motion state and the correction of the cumulative error of the inertial measurement unit, the high-precision relative pose information of the UAV relative to the center of the landing area of ​​the vehicle platform is calculated, with a relative horizontal coordinate of 0.05 meters, a relative vertical coordinate of 0.03 meters, a relative height of 1 meter, and roll, pitch and yaw angles of 0°.

[0084] By implementing the above embodiments, motion recursion is performed using inertial measurement unit data to predict the trajectory of the UAV, and vehicle motion state data is used as feedforward compensation to dynamically correct the predicted trajectory. Then, state updates are performed through filtering algorithms. This can effectively compensate for the impact of vehicle motion during the UAV's touchdown phase, enabling the UAV to maintain a stable relative motion relationship while the vehicle is in continuous motion. This improves the landing stability of the UAV on the mobile platform and reduces attitude disturbances during landing.

[0085] In some embodiments, to achieve the multi-stage data fusion positioning of the above-mentioned vehicle-mounted UAV dynamic landing, a corresponding vehicle-mounted UAV dynamic landing hierarchical fusion positioning system can also be built. This system may include a vehicle platform subsystem and a UAV onboard subsystem. The two establish a bidirectional data interaction connection through a low-latency communication link. Each subsystem is divided according to a preset functional module and works collaboratively to provide full-dimensional support for the entire process of vehicle-mounted UAV dynamic landing, including perception, computing, communication, and positioning fusion. The vehicle platform subsystem, as the ground perception and data interaction terminal, provides the UAV with a relative positioning reference and real-time vehicle motion status information. The UAV onboard subsystem, as the flight perception and fusion computing terminal, completes multi-source sensor data acquisition, hierarchical fusion calculation, and positioning strategy execution. The collaborative cooperation of the two subsystems achieves high-precision and high-robust positioning for vehicle-mounted UAV dynamic landing.

[0086] In some examples, the vehicle-mounted platform subsystem may include an ultra-wideband (UWB) base station array, visual markers, a vehicle state perception unit, and an onboard computing and communication unit. The UWB base station array is deployed at at least three known locations on the vehicle platform. By precisely calibrating the relative positions of each base station, a relative coordinate system for the vehicle is formed, providing a spatial reference for UAV ranging. For example, five UWB base stations are deployed around the landing area of ​​the vehicle platform, and a full-dimensional ranging network is constructed after position calibration. Visual markers are set at the center of the predetermined landing area of ​​the vehicle platform and have a unique coded pattern, providing a physical reference for UAV visual positioning. For example, an AprilTag marker with a side length of approximately 80 centimeters is deployed at the center of the landing area; its unique code enables accurate visual recognition. The vehicle state perception unit is used to acquire and upload motion state information such as vehicle speed, angular velocity, and attitude angle in real time. It can collect relevant data through the vehicle's built-in bus system to ensure the real-time performance and accuracy of the data. The onboard computing and communication unit, as the core of the vehicle-mounted platform subsystem, is used to fuse and process information collected by various onboard sensors. Simultaneously, it completes bidirectional data broadcasting and reception with the UAV through a low-latency communication link, achieving real-time synchronization of data between the two ends.

[0087] The UAV's onboard subsystem may include a multi-source sensor module and a sensor hierarchical fusion center module. The multi-source sensor module provides comprehensive perception data for positioning fusion, while the sensor hierarchical fusion center module is the core of the system, responsible for executing data fusion algorithms and phased positioning strategies. The multi-source sensor module integrates a global navigation satellite system receiver, an ultra-wideband tag, a visual camera, an inertial measurement unit, and an optional lidar. Each sensor synchronously collects data at a preset frequency, providing multi-dimensional raw perception information for positioning fusion. The sensor hierarchical fusion center module runs an adaptive fusion algorithm based on extended Kalman filtering or error state Kalman filtering. This module is configured to execute a phased positioning strategy matched to the UAV's landing distance. The first phase is the long-range return phase, which is dominated by the absolute position information provided by the global navigation satellite system. The first stage involves integrating inertial measurement unit (IMU) data to guide the UAV to the approximate airspace of the vehicle platform. The second stage is the mid-range approach stage, where the UAV enters the effective range of the ultra-wideband (UWB) system. The centimeter-level relative horizontal position provided by the UWB system is tightly coupled and fused with the position and velocity estimated by the visual inertial odometry (VIO) as the core positioning information. The third stage is the end-point precision alignment stage, where the cooperative visual marker enters and stabilizes in the camera's field of view. The high-precision six-DOF relative pose calculated based on the marker is used as the optimal observation value input to the fusion filter, guiding the UAV to hover precisely above the landing point. The fourth stage is the disturbance-resistant touchdown stage, where the role of high-frequency IMU data in attitude control is strengthened during the final touchdown stage. The real-time motion state of the vehicle obtained from the communication link is tightly integrated to provide feedforward compensation for the landing trajectory, thereby counteracting disturbances caused by vehicle turbulence and motion.

[0088] In some embodiments, the vehicle-mounted UAV dynamic landing positioning method based on the above-mentioned vehicle-mounted UAV dynamic landing hierarchical fusion positioning system may include the following steps: S1 is the system initialization step, which completes the time synchronization and spatial calibration of airborne multi-sensors, including visual cameras, inertial measurement units, and ultra-wideband tags, to ensure the consistency of data from each sensor in time and space dimensions, laying the foundation for subsequent fusion processing; S2 is the remote return step, in which the UAV enters the first stage after receiving the landing command, plans the initial return path based on its own global navigation satellite system position and the global navigation satellite system position broadcast by the vehicle platform, and simultaneously fuses the global navigation satellite system position, velocity, and inertial measurement unit data to complete the initial navigation; S3 is the mid-range approach step, in which the second stage is triggered after the UAV enters the ultra-wideband network coverage area, and the fusion filter uses ultra-wideband relative position and visual inertial odometry solution information as the main observations to continuously estimate the precise position and velocity of the UAV relative to the vehicle platform, guiding the UAV to gradually approach the vehicle platform; S4 is the terminal precision alignment step. In the first stage, after the airborne camera recognizes the cooperative visual marker, the system enters the third stage. The visual processing module calculates the relative pose in real time using the perspective n-point algorithm and uses it as a high-weighted observation input to the fusion filter to achieve sub-decimeter-level position and attitude alignment. In the second stage, S5 is the anti-disturbance landing step. After the UAV descends to a predetermined low altitude, such as below five meters, it enters the fourth stage. The system uses high-frequency inertial measurement unit data as the core of attitude control, and simultaneously receives real-time speed and angular velocity information uploaded by the vehicle. In the fusion algorithm, this information is used as known control input to dynamically compensate for the predicted trajectory of the UAV, achieving a synchronous and smooth landing with the mobile platform. In the third stage, S6 is the adaptive noise adjustment step. Throughout the landing process, the fusion algorithm continuously performs adaptive noise adjustment, dynamically adjusting the observation noise matrix weights in the filtering based on the signal quality of each sensor, including the signal-to-noise ratio of the global navigation satellite system, image clarity, and ultra-wideband residuals. In the event of a short-term failure of a sensor, the system smoothly transitions to a backup mode dominated by other sensors to ensure the continuity and stability of positioning fusion.

[0089] In specific examples, taking the dynamic landing of a quadcopter drone on the roof platform of a moving electric van as an example, the hardware deployment, software and algorithm implementation, and actual landing process of the above system are explained in detail. Regarding hardware deployment, five ultra-wideband base stations are installed at the four corners and center of the van's roof platform to accurately measure their relative positions and construct a vehicle coordinate system. An AprilTag marker with a side length of approximately 80 centimeters is placed at the center of the platform. The van obtains information such as vehicle speed and yaw rate through the onboard controller's local area network bus and broadcasts it externally through the onboard computer and fifth-generation vehicle wireless communication module. The drone is equipped with a real-time dynamic global navigation satellite system module, an ultra-wideband tag, a global shutter binocular camera, an industrial-grade inertial measurement unit (IMU) i.e., an attitude reference system, and a Jetson AGX Orin as a fusion computing unit to provide computing power support for data fusion and algorithm operation. Regarding software and algorithm implementation, a robot operating system is deployed on the Jetson AGX Orin. The system (ROS) runs various sensor drivers, visual inertial odometry calculation methods such as VINS-Fusion and AprilTag detection programs, and core fusion algorithms. The core fusion algorithm uses error-state Kalman filtering, and sets the state variables as the three-dimensional position, velocity, attitude of the UAV relative to the vehicle coordinate system, as well as the zero bias error of the inertial measurement unit. During system initialization, offline calibration and online time synchronization of the camera-inertial measurement unit are completed. At the same time, an adaptive weight strategy is set, setting the noise covariance of the ultra-wideband position observation to be positively correlated with the magnitude of its calculated residual, and setting the noise covariance of visual observation to be negatively correlated with the number of feature points tracked in the image and the confidence of the marker detection, so as to realize the dynamic adjustment of sensor weights.In the actual landing process, the truck travels at a constant speed of 20 kilometers per hour in a straight line. The drone initiates the landing procedure 100 meters behind the truck. Phase one is the long-range return phase, where the relative distance is greater than 50 meters. The drone flies towards the truck based on its own real-time dynamic GNSS position and the real-time dynamic GNSS position broadcast by the truck. The fusion filter mainly uses GNSS observations. Phase two is the mid-range approach phase, where the relative distance is 10 to 50 meters. The drone's UWB tag receives sufficient base station signals. The filter uses the UWB relative position as the core horizontal position observation, while fusing the altitude and speed estimated by visual inertial odometry. At this time, if the GNSS... If the signal is lost due to bridge obstruction or other reasons, the system can achieve seamless switching. Phase three is the precise alignment phase at a relative distance of three to ten meters. The camera clearly identifies the AprilTag marker, and the visual pose calculation result is input into the filter at a high update rate. The drone precisely adjusts to hover one meter directly above the marker. Phase four is the anti-disturbance touchdown phase at a relative distance of less than three meters until ground contact. The drone begins to descend, and the fusion algorithm uses the real-time longitudinal velocity uploaded by the truck as a feedforward quantity, superimposed on the forward velocity control command of the drone, keeping the drone and the truck relatively stationary. At the same time, it relies on high-frequency inertial measurement unit data to ensure attitude stability until the landing lock mechanism is triggered, completing the entire dynamic landing process.

[0090] Through the implementation of the above embodiments, a hierarchical fusion positioning system with dual-end collaboration between the vehicle-mounted platform and the UAV is established. Based on this system, a phased dynamic landing positioning method is executed. Combined with specific scenarios, hardware deployment and software algorithm implementation are completed. This enables accurate fusion of multi-source sensor data and adaptive execution of phased positioning strategies throughout the entire dynamic landing process of the vehicle-mounted UAV. By utilizing adaptive noise adjustment and sensor seamless switching mechanisms, the robustness of positioning fusion is effectively improved, which can cope with sensor failure in complex environments. At the same time, by using feedforward compensation to offset vehicle movement and turbulent disturbances, accurate and stable landing of the UAV on the vehicle platform is achieved, significantly improving the success rate and stability of dynamic landing of the vehicle-mounted UAV.

[0091] Furthermore, as an implementation of the aforementioned method embodiments, this application also provides a vehicle-mounted unmanned aerial vehicle (UAV) landing control device for implementing the aforementioned method embodiments. This device embodiment corresponds to the aforementioned method embodiments. For ease of reading, this vehicle-mounted UAV landing control device embodiment will not repeat the details of the aforementioned method embodiments one by one, but it should be understood that the device in this application embodiment can correspondingly implement all the contents of the aforementioned method embodiments. For example... Figure 2As shown, the vehicle-mounted drone landing control device 20 includes: a data acquisition unit 201, a strategy determination unit 202, a pose determination unit 203, and a landing control unit 204. The data acquisition unit 201 acquires first sensor data from the vehicle platform and second sensor data from the target drone. The strategy determination unit 202 determines the current fusion strategy from a multi-stage fusion strategy based on the relative distance between the target drone and the vehicle platform. The pose determination unit 203 performs fusion processing on the first and second sensor data based on the current fusion strategy to determine the relative pose information of the target drone relative to the vehicle platform. The landing control unit 204 generates landing control commands for the target drone based on the relative pose information to guide the target drone to land on the vehicle platform while the vehicle is in motion.

[0092] In some embodiments, the multi-stage fusion strategy includes a long-range return-to-home strategy, a mid-range approach strategy, a terminal alignment strategy, and a disturbance-resistant landing strategy. The strategy determination unit 202 is further configured to determine the current fusion strategy as a long-range return-to-home strategy when the relative distance is greater than a first preset distance; determine the current fusion strategy as a mid-range approach strategy when the relative distance is less than or equal to the first preset distance and greater than a second preset distance; determine the current fusion strategy as a terminal alignment strategy when the relative distance is less than or equal to the second preset distance and greater than a third preset distance; and determine the current fusion strategy as a disturbance-resistant landing strategy when the relative distance is less than or equal to the third preset distance.

[0093] In some embodiments, the first sensor data includes ultra-wideband base station array data, visual marker data, vehicle motion state data, and first global navigation positioning data; the second sensor data includes second global navigation positioning data, ultra-wideband tag data, image acquisition data, and inertial measurement unit data. The pose determination unit 203 is further configured to: 1) perform data fusion processing based on the first global navigation positioning data, inertial measurement unit data, and second global navigation positioning data to obtain relative pose information when the current fusion strategy is a long-range return-to-home strategy; 2) perform data fusion processing based on ultra-wideband tag data, ultra-wideband base station array data, image acquisition data, and inertial measurement unit data to obtain relative pose information when the current fusion strategy is an end-point alignment strategy; 3) perform data fusion processing based on visual marker data, image acquisition data, and inertial measurement unit data to obtain relative pose information when the current fusion strategy is an anti-disturbance landing strategy; and 4) perform data fusion processing based on inertial measurement unit data and vehicle motion state data to obtain relative pose information when the current fusion strategy is an anti-disturbance landing strategy.

[0094] In some embodiments, the pose determination unit 203 is further configured to determine the first relative position of the target UAV relative to the vehicle platform based on the first global navigation positioning data and the second global navigation positioning data; and to perform fusion processing on the first relative position and the inertial measurement unit data based on a preset filtering algorithm to obtain relative pose information.

[0095] In some embodiments, the pose determination unit 203 is further configured to determine the second relative position of the target UAV relative to the vehicle platform based on UAV tag data and UAV base station array data; perform visual inertial odometry assessment based on image acquisition data and inertial measurement unit data to obtain visual inertial pose information of the target UAV; and perform fusion processing on the second relative position and visual inertial pose information based on a preset filtering algorithm to obtain relative pose information.

[0096] In some embodiments, the pose determination unit 203 is further configured to: identify cooperative markers corresponding to visual marker data based on image acquisition data; calculate the cooperative markers using a perspective n-point algorithm to obtain the six-degree-of-freedom relative pose of the target UAV relative to the cooperative markers; determine the inertial pose information of the target UAV based on inertial measurement unit data; and fuse the six-degree-of-freedom relative pose and inertial pose information based on a preset filtering algorithm to obtain relative pose information.

[0097] In some embodiments, the pose determination unit 203 is further configured to perform motion recursion based on inertial measurement unit data to obtain the predicted motion trajectory of the target UAV; use vehicle motion state data as feedforward compensation to dynamically correct the predicted motion trajectory to obtain the compensated motion trajectory; and update the state of the compensated motion trajectory based on a preset filtering algorithm to obtain relative pose information.

[0098] This application also provides a computer-readable storage medium storing computer-executable instructions or computer programs, which, when executed by a processor, will cause the processor to perform any step of the vehicle-mounted UAV landing control method provided in this application.

[0099] In some embodiments, the computer-readable storage medium may be a random access memory (RAM), a read-only memory (ROM), flash memory, a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM); or it may be a variety of devices that include one or any combination of the above-mentioned memories.

[0100] In some embodiments, computer-executable instructions may take the form of programs, software, software modules, scripts, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as stand-alone programs or as modules, components, subroutines, or other units suitable for use in a computing environment.

[0101] In some embodiments, computer-executable instructions may, but do not necessarily, correspond to files in a file system, and may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a HyperText Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple co-located files (e.g., files that store one or more modules, subroutines, or code sections).

[0102] In some embodiments, computer-executable instructions may be deployed to execute on an electronic device, or on multiple electronic devices located at one location, or on multiple electronic devices distributed across multiple locations and interconnected via a communication network.

[0103] like Figure 3 As shown, this application also provides a vehicle-mounted drone 30, including a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and executable on the processor. When the processor 320 executes the computer program 311, it implements any step of the above-described vehicle-mounted drone landing control method.

[0104] This application also provides a computer program product comprising a computer program or computer-executable instructions stored in a computer-readable storage medium. The processor of the vehicle-mounted drone reads the computer program or computer-executable instructions from the computer-readable storage medium and executes the computer program or computer-executable instructions, causing the vehicle-mounted drone to perform any step of the vehicle-mounted drone landing control method described above.

[0105] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application 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 do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for controlling the landing of a vehicle-mounted unmanned aerial vehicle, characterized in that, include: Acquire the first sensor data from the vehicle-mounted platform and the second sensor data from the target drone; Based on the relative distance between the target UAV and the vehicle platform, the current fusion strategy is determined from the multi-stage fusion strategy; Based on the current fusion strategy, the first sensor data and the second sensor data are fused to determine the relative pose information of the target UAV relative to the vehicle platform. Based on the relative pose information, landing control commands are generated for the target UAV to guide the target UAV to land on the vehicle platform while the vehicle is in motion.

2. The vehicle-mounted unmanned aerial vehicle landing control method according to claim 1, characterized in that, The multi-stage fusion strategy includes a long-range return-to-home strategy, a mid-range approach strategy, a terminal alignment strategy, and an anti-disturbance landing strategy. The step of determining the current fusion strategy from the multi-stage fusion strategy based on the relative distance between the target UAV and the vehicle platform includes: If the relative distance is greater than a first preset distance, the current fusion strategy is determined to be the remote return-to-home strategy; If the relative distance is less than or equal to the first preset distance and greater than the second preset distance, the current fusion strategy is determined to be the mid-range approach strategy. If the relative distance is less than or equal to the second preset distance and greater than the third preset distance, the current fusion strategy is determined to be the end alignment strategy; If the relative distance is less than or equal to the third preset distance, the current fusion strategy is determined to be the anti-disturbance slamming strategy.

3. The vehicle-mounted unmanned aerial vehicle landing control method according to claim 2, characterized in that, The first sensor data includes ultra-wideband base station array data, visual marker data, vehicle motion state data, and first global navigation positioning data; the second sensor data includes second global navigation positioning data, ultra-wideband tag data, image acquisition data, and inertial measurement unit data. The step of fusing the first sensor data and the second sensor data based on the current fusion strategy to determine the relative pose information of the target UAV relative to the vehicle platform includes: When the current fusion strategy is the remote return-to-home strategy, data fusion processing is performed based on the first global navigation positioning data, the inertial measurement unit data, and the second global navigation positioning data to obtain the relative pose information; When the current fusion strategy is the mid-range approach strategy, the relative pose information is obtained by performing data fusion processing based on the ultra-wideband tag data, the ultra-wideband base station array data, the image acquisition data, and the inertial measurement unit data. When the current fusion strategy is the end alignment strategy, data fusion processing is performed based on the visual marker data, the image acquisition data, and the inertial measurement unit data to obtain the relative pose information; When the current fusion strategy is the anti-disturbance landing strategy, the relative pose information is obtained by performing data fusion processing based on the inertial measurement unit data and the vehicle motion state data.

4. The vehicle-mounted unmanned aerial vehicle landing control method according to claim 3, characterized in that, The process of fusing data based on the first global navigation positioning data, the inertial measurement unit data, and the second global navigation positioning data to obtain the relative pose information includes: Based on the first global navigation positioning data and the second global navigation positioning data, the first relative position of the target UAV relative to the vehicle platform is determined; The relative pose information is obtained by fusing the first relative position and the inertial measurement unit data based on a preset filtering algorithm.

5. The vehicle-mounted unmanned aerial vehicle landing control method according to claim 3, characterized in that, The process of fusing data based on the ultra-wideband tag data, the ultra-wideband base station array data, the image acquisition data, and the inertial measurement unit data to obtain the relative pose information includes: Based on the ultra-wideband tag data and the ultra-wideband base station array data, the second relative position of the target UAV relative to the vehicle platform is determined; Visual inertial odometry (VIO) is performed based on the image acquisition data and the inertial measurement unit (IMU) data to obtain the visual inertial pose information of the target UAV. The relative pose information is obtained by fusing the second relative position and the visual-inertial pose information based on a preset filtering algorithm.

6. The vehicle-mounted unmanned aerial vehicle landing control method according to claim 3, characterized in that, The process of fusing data based on the visual marker data, the image acquisition data, and the inertial measurement unit data to obtain the relative pose information includes: Based on the image acquisition data, the cooperative markers corresponding to the visual marker data are identified; The cooperative marker is solved by the perspective n-point algorithm to obtain the six-degree-of-freedom relative pose of the target UAV relative to the cooperative marker; Based on the data from the inertial measurement unit, the inertial pose information of the target UAV is determined; The relative pose information of the six degrees of freedom and the inertial pose information are fused based on a preset filtering algorithm to obtain the relative pose information.

7. The vehicle-mounted unmanned aerial vehicle landing control method according to claim 3, characterized in that, The process of fusing data based on the inertial measurement unit data and the vehicle motion state data to obtain the relative pose information includes: Based on the data from the inertial measurement unit, motion recursion is performed to obtain the predicted motion trajectory of the target UAV; The vehicle motion state data is used as a feedforward compensation amount to dynamically correct the predicted motion trajectory, resulting in a compensated motion trajectory. The state of the compensated motion trajectory is updated based on a preset filtering algorithm to obtain the relative pose information.

8. A vehicle-mounted unmanned aerial vehicle (UAV) landing control device, characterized in that, include: The data acquisition unit is used to acquire the first sensor data of the vehicle platform and the second sensor data of the target drone. The strategy determination unit is used to determine the current fusion strategy from the multi-stage fusion strategies based on the relative distance between the target UAV and the vehicle platform; The pose determination unit is used to perform fusion processing on the first sensor data and the second sensor data based on the current fusion strategy to determine the relative pose information of the target UAV relative to the vehicle platform. The landing control unit is used to generate landing control commands for the target UAV based on the relative pose information, so as to guide the target UAV to land on the vehicle platform while the vehicle is in motion.

9. A vehicle-mounted unmanned aerial vehicle, comprising: The memory and processor are characterized in that the processor is used to implement the steps of the vehicle-mounted unmanned aerial vehicle landing control method as described in any one of claims 1 to 7 when executing a computer program stored in the memory.

10. A computer-readable storage medium having stored thereon computer-executable instructions or a computer program, characterized in that, When the computer-executable instructions or the computer program are executed by a processor, the steps of the vehicle-mounted unmanned aerial vehicle landing control method as described in any one of claims 1 to 7 are implemented.

Images