Fixed-wing uav swarm dense recovery method and system

Abstract

The application relates to the technical field of unmanned aerial vehicle recovery, and discloses a fixed-wing unmanned aerial vehicle group intensive recovery method and system. The method comprises the following steps: a vehicle-mounted control center receives a recovery application of a group of unmanned aerial vehicles, and distributes a guide channel and an initial access point according to the unmanned aerial vehicle state and real-time vehicle motion information; a vehicle-mounted multi-channel guide system sends a high-precision positioning signal by using beam forming, and generates a dynamically adjusted recovery path in combination with a vehicle moving track; the vehicle-mounted control center executes an intensive recovery scheduling algorithm, determines a recovery sequence and a time window based on a battery residual capacity and a space-time conflict detection result; the unmanned aerial vehicles fly along the path to a recovery platform, land by visual positioning assistance, and trigger mechanical locking and interface docking; and a vehicle-mounted energy management system implements dynamic charging power distribution according to the battery state and a vehicle load. The application realizes high concurrency, safety, efficient intensive recovery and rapid energy supplement of the group of unmanned aerial vehicles on a moving vehicle.

Description

Technical Field

[0001] This invention relates to the field of drone recovery technology, and in particular to a method and system for the intensive recovery of fixed-wing drone swarms. Background Technology

[0002] With the rapid development of drone technology, vertical takeoff and landing (VTOL) fixed-wing drones have shown great application potential in fields such as material transportation, intelligence reconnaissance, and communication relay due to their combination of the long endurance and high speed of fixed-wing aircraft with the vertical takeoff and landing and runway-free advantages of multi-rotor drones. In practical applications, especially in complex environments such as emergency rescue, drones often need to operate collaboratively in swarms. To improve mission continuity and coverage, ground mobile vehicles are typically used as mobile base stations to transport, recover, and resupply drone swarms.

[0003] However, existing vehicle-mounted drone recovery methods mainly rely on manual operation or single-channel automatic guidance, which is insufficient to meet the needs of dense drone swarm recovery. Specifically, existing technologies have the following problems: First, the vehicle environment is complex. During vehicle operation, vibrations, attitude changes, and positional shifts occur. Traditional single-positioning guidance methods struggle to achieve precise synchronous guidance for multiple drones, easily leading to recovery collisions or failures. Second, there is a lack of efficient scheduling algorithms. When multiple drones simultaneously request recovery, air congestion and path conflicts can easily occur, resulting in low recovery efficiency or even crashes. Third, charging management is inefficient. The vehicle's power supply is limited and cannot dynamically allocate power based on the battery status of multiple drones and the vehicle's operating conditions, leading to slow charging speeds or affecting the normal operation of other vehicle equipment.

[0004] Therefore, there is an urgent need for a method that can solve the above-mentioned technical problems, realize the safe, dense and efficient recovery of drone swarms on mobile vehicle platforms, and enable intelligent dynamic charging scheduling. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for the intensive recovery of fixed-wing UAV swarms, which solves the problems of difficult recovery, low efficiency and extensive charging management of UAV swarms on mobile platforms through multi-channel guidance, intensive scheduling algorithms and dynamic charging management.

[0006] To achieve the above objectives, the first aspect of the present invention provides a method for dense recovery of a swarm of fixed-wing unmanned aerial vehicles, comprising the following steps: The vehicle control center receives recovery request information from the drone swarm and assigns a unique guidance channel and initial recovery access point to each drone requesting recovery based on the current status information of the drone swarm and the real-time movement information of the vehicle. The vehicle-mounted multi-channel guidance system sends high-precision relative positioning signals to the UAV through the allocated guidance channels. After receiving the signals, the UAV calculates its real-time spatial position relative to the vehicle. The vehicle-mounted control center generates a dynamically adjusted recovery path for each UAV based on the real-time spatial position and the vehicle's movement trajectory. The vehicle-mounted control center executes a dense recovery scheduling algorithm. Based on the real-time spatial location of each drone, the remaining battery power, and the conflict detection results of the recovery path, it determines the recovery order and time window of the drone swarm and sends the recovery order and time window to the corresponding drones. The drone flies to the vehicle-mounted recovery platform along a dynamically adjusted recovery path according to the recovery order and time window. The vehicle-mounted recovery platform controls the mechanical structure to physically lock and interface with the drone based on the landing signal triggered by the drone. The vehicle energy management system dynamically allocates charging power based on the battery status of the docked drone and the overall energy load of the vehicle. After charging is completed, it updates the drone status according to mission requirements and either releases the drone or keeps it parked.

[0007] A second aspect of the present invention provides a dense recovery system for a swarm of fixed-wing unmanned aerial vehicles, applicable to mobile vehicles, comprising: The vehicle-mounted control center is used to receive recovery request information sent by the drone swarm, and to allocate a unique guidance channel and initial recovery access point to each drone requesting recovery based on the current status information of the drone swarm and the real-time movement information of the vehicle; it is also used to execute a dense recovery scheduling algorithm to determine the recovery order and time window of the drone swarm based on the real-time spatial location of each drone, the remaining battery power, and the conflict detection results of the recovery path. The vehicle-mounted multi-channel guidance system is used to send high-precision relative positioning signals to the UAV through the allocated guidance channels and receive the relative position information transmitted back by the UAV. Combined with the vehicle's movement trajectory, it generates a dynamically adjusted recovery path for each UAV. The vehicle-mounted recovery platform is used to control the mechanical structure to physically lock and interface with the drone based on the landing signal triggered by the drone. The vehicle-mounted energy management system is used to dynamically allocate charging power based on the battery status of the docked drone and the overall energy load of the vehicle, in conjunction with the vehicle-mounted control center.

[0008] Compared with the prior art, the present invention has the following significant beneficial technical effects: (1) Significantly improved the efficiency and throughput of drone swarm recovery.

[0009] Existing drone recovery technologies typically employ a single-channel, one-by-one guidance approach or require large safety intervals, resulting in long recovery times and low efficiency when recovering multiple drones. This invention, through a vehicle-mounted multi-channel guidance system and a dense recovery scheduling algorithm, can simultaneously allocate independent guidance channels and spatiotemporally conflict-free recovery paths to multiple drones. The system supports parallel entry of multiple drones at different airspace levels near the vehicle, breaking through the bottleneck of traditional serial recovery and significantly improving the drone recovery density per unit time and the turnaround efficiency of the entire swarm mission.

[0010] (2) High-precision recovery was achieved under complex dynamic mobile platforms.

[0011] Existing technologies mostly rely on fixed ground stations or large, stable ships for recovery. However, for wheeled / tracked vehicles maneuvering at high speeds on rough terrain, it's difficult to overcome the positioning errors and landing risks caused by vehicle vibrations and drastic attitude changes. This invention introduces a vehicle trajectory envelope prediction and real-time motion compensation algorithm, combined with spline interpolation dynamic path generation technology, to predict the vehicle's instantaneous motion state and adjust the recovery path in real time. This enables the UAV to accurately track moving targets in a non-inertial frame of reference. Even when the vehicle is making sharp turns or experiencing severe vibrations, the UAV can still accurately align and land on the recovery platform, overcoming the geographical limitations of fixed recovery sites.

[0012] (3) It solved the aerial conflicts and safety hazards during the intensive recovery process.

[0013] During the dense recovery of drone swarms, the simultaneous operation of multiple drones in confined airspace greatly increases the risk of collisions or wake interference. Existing technologies often rely on manual avoidance or simple distance thresholds, lacking a system-level conflict resolution mechanism. This invention utilizes a spatiotemporal conflict detection algorithm to discretize the recovery path in four-dimensional spatiotemporal space (three-dimensional space + time), accurately identifying the risk of path overlap. Simultaneously, combined with a battery remaining power priority strategy, it automatically replans conflicting paths or shifts time windows, eliminating collision risks at the algorithmic level and achieving unattended, automated, and safe recovery.

[0014] (4) Enhanced locking reliability in harsh physical environments.

[0015] On moving vehicles, drones relying solely on their own weight or simple electromagnetic adsorption are prone to detachment due to vehicle vibrations. This invention designs a dual mechanical and electromagnetic locking mechanism, combined with visual-assisted positioning and attitude synchronization technology. The system instantly achieves rigid physical locking the moment the drone touches the ground, effectively resisting the impact of vehicle acceleration, braking, and off-road driving. Furthermore, the floating blind-plug interface allows for greater positional tolerance, ensuring reliable electrical connection even under dynamic shaking, significantly improving the system's environmental adaptability and robustness.

[0016] (5) Optimized vehicle energy utilization efficiency and battery life.

[0017] Existing vehicle-mounted charging systems often employ a simple constant voltage and constant current mode, which not only results in slow charging speeds but also makes them prone to grid overload or battery damage when vehicle load changes. This invention uses a dynamic charging power allocation algorithm to balance vehicle driving load and charging demand in real time, ensuring maximum charging efficiency without affecting vehicle maneuverability. Furthermore, through a pulse depolarization charging strategy and a temperature-efficiency correlation model, the system can assess battery internal resistance and health status in real time. This prevents battery overheating, extends battery cycle life, and enables high-current rapid charging, significantly shortening the preparation time for drone swarms to re-deploy. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of one embodiment of the dense recovery method for a fixed-wing UAV swarm in this invention. Detailed Implementation

[0019] The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0020] Example 1 This embodiment provides an overall implementation process for a method of dense recovery and dynamic charging scheduling of VTOL fixed-wing UAV swarms based on vehicle-mounted multi-channel guidance. This method is particularly suitable for special vehicles traveling on rough terrain, enabling rapid recovery and charging of ten VTOL fixed-wing UAVs after completing their missions. Please refer to [link to relevant documentation]. Figure 1 , Figure 1 This is a schematic diagram of an embodiment of the dense recovery method for a fixed-wing UAV swarm in this invention, including: 101. The vehicle control center receives the recovery request information sent by the drone swarm, and assigns a unique guidance channel and initial recovery access point to each drone requesting recovery based on the current status information of the drone swarm and the real-time movement information of the vehicle. When a drone in a swarm completes its designated patrol or transport mission, or when its onboard battery management system detects that its remaining battery power is below the 20% safety threshold, the drone will automatically generate a recovery request data packet. This data packet contains the drone's unique identification code, its current absolute position coordinates provided by the Global Navigation Satellite System, its remaining battery power percentage, and its current flight speed vector.

[0021] Upon receiving the data packet, the vehicle control center does not immediately instruct the vehicle to return. Instead, it first acquires the vehicle's real-time status through the onboard inertial measurement unit and wheel speed sensors. This includes the vehicle's speed, heading, pitch angle, roll angle, and the frequency of vertical bumps caused by uneven road surfaces. Based on this data, the vehicle control center uses a kinematic prediction model to calculate the vehicle's three-dimensional trajectory envelope over the next sixty seconds. This trajectory envelope is a dynamically changing virtual spatial region, representing the possible spatial range of the recovery platform.

[0022] Subsequently, based on the trajectory envelope and the drone's current location, the system estimates the drone's arrival time using the shortest time principle.

[0023] To prevent communication congestion or mutual interference among multiple drones while they are waiting in the air, the system assigns a unique guidance channel number to each approved drone and sets an initial recovery access point upstream of the trajectory envelope. This access point is not a fixed coordinate, but a relative spatial coordinate point that is updated in real time as the vehicle moves, serving as the starting landmark for the drone to enter the recovery guidance procedure.

[0024] For example, suppose a drone codenamed "001" completes a border reconnaissance mission and its remaining battery level drops to 18%, so it sends a recovery request to the recovery vehicle. The request includes its current location as 118 degrees East longitude and 40 degrees North latitude, with a flight speed of 20 meters per second. Simultaneously, the recovery vehicle is traveling north across undulating terrain at 15 meters per second, with the vehicle vibrating vertically at a frequency of two hertz due to road bumps. Upon receiving the request from "001," the onboard control center predicts that the recovery platform behind the vehicle will be in a northward-moving, vertically fluctuating area in the future. The system calculates that if "001" flies at its current full speed, it will take approximately 120 seconds to reach the vehicle. Therefore, the system assigns "001" guidance channel 01 and sets a virtual point 500 meters behind and 100 meters above the vehicle as the initial access point, instructing "001" to fly to that point and wait.

[0025] 102. The vehicle-mounted multi-channel guidance system sends high-precision relative positioning signals to the UAV through the allocated guidance channels. After receiving the signals, the UAV calculates its real-time spatial position relative to the vehicle. The vehicle-mounted control center generates a dynamically adjusted recovery path for each UAV based on the real-time spatial position and the vehicle's movement trajectory. This step utilizes the assigned guidance channels to establish a high-precision relative positioning relationship, and based on this, generates a recovery trajectory that can adapt to the dynamic movement of the vehicle.

[0026] After step 101 is completed, the UAV establishes an initial connection with the recovery vehicle. However, the positioning accuracy of global navigation satellite systems is typically at the meter level, which is insufficient for precise landing. Therefore, the vehicle-mounted multi-channel guidance system begins operation. This system integrates a phased array antenna and an ultra-wideband positioning base station, and can concentrate the energy of the directional signal beam in the corresponding direction of the UAV using beamforming technology, based on the guidance channel number assigned in step 101.

[0027] The system sends a guidance pulse signal to the drone, containing microsecond-level timestamps and precise angle information. Upon receiving the signal, the drone first measures the time difference between transmission and reception, multiplies it by the speed of light to obtain a rough distance, and then calculates the extremely precise relative angle using the phase difference between multiple antennas. Combining this with acceleration and angular velocity data provided by the drone's own inertial navigation system, the system performs data fusion using an extended Kalman filter algorithm to calculate the drone's real-time spatial position relative to the vehicle coordinate system, achieving centimeter-level accuracy.

[0028] The drone then transmits this high-precision relative position data back to the vehicle-mounted control center. The control center then processes this data in conjunction with the real-time movement trajectory of the recovery vehicle to compensate for displacement deviations caused by the vehicle's movement and bumps.

[0029] Finally, a smooth, dynamic recovery path is generated using a cubic spline interpolation algorithm. This path starts at the initial access point in step 101 and ends at the center locking point of the vehicle recovery platform. The curvature of the path is dynamically adjusted according to the drone's flight performance and the vehicle's real-time speed to ensure that the drone can approach the platform in the safest possible manner.

[0030] For example, after UAV "001" arrives at the initial access point, the vehicle-mounted guidance system sends a high-frequency positioning signal to it via channel 01. Because the vehicle is traveling north and experiencing bumps, the guidance system adjusts its beam direction in real time, always keeping it aligned with "001". The onboard computer of "001" calculates its position as 495 meters behind and 99 meters above the vehicle in the vehicle's coordinate system based on the signal and transmits this data back to the vehicle. The vehicle control center detects that the vehicle has moved forward 15 meters in the previous second, and the platform has descended 0.1 meters due to the bumps, so it immediately corrects the target position. The system-generated path instruction for "001" is to first dive at a 30-degree angle to reduce altitude, then level off as it approaches the rear of the vehicle, entering a following mode synchronized with the vehicle's speed, and finally smoothly transitioning to hover 0.5 meters directly above the recovery platform.

[0031] 103. The vehicle-mounted control center executes a dense recovery scheduling algorithm to determine the recovery order and time window of the drone swarm based on the real-time spatial location of each drone, the remaining battery power, and the conflict detection results of the recovery path, and sends the recovery order and time window to the corresponding drones. This step is the initiation phase of the entire recovery process. Its core task is to establish the initial connection between the drone and the mobile vehicle and to allocate resources based on the current dynamic environment. Specifically, the onboard control center needs to constantly monitor the communication frequency bands in the air and wait for drones to initiate recovery requests. This step is crucial to ensuring the safety of dense drone swarm recovery, mainly addressing the spatiotemporal conflicts when multiple drones are recovered simultaneously, and determining the recovery order and precise time window.

[0032] When multiple drones simultaneously request recovery, the vehicle-mounted control center initiates a dense recovery scheduling algorithm. This algorithm first constructs a multi-objective optimization model, with objective functions including minimizing the total recovery time and minimizing the total battery consumption of all drones. Input parameters include the real-time spatial location, remaining battery power, and the dynamic recovery path generated in step 102 for all drones to be recovered. The core of the algorithm lies in conflict detection. The system discretizes the recovery path of each drone in the time dimension, sampling once every 0.1 seconds to form a four-dimensional trajectory point sequence containing position coordinates and timestamps.

[0033] Next, the system calculates the Euclidean distance between the trajectory point sequences of any two different drones. If it is found that the Euclidean distance between two drones at a certain moment is less than a preset safety threshold (e.g., 5 meters), and the time difference between them is less than a preset time buffer threshold (e.g., 2 seconds), then it is determined that the two drones have a potential risk of mid-air collision or wake interference, i.e., a conflict is identified.

[0034] For detected conflicts, the algorithm prioritizes drones based on their remaining battery power. Drones with lower remaining battery power have higher priority. The algorithm then replans paths to eliminate spatial overlap by adjusting the flight speed of lower-priority drones (e.g., instructing them to circle or decelerate) to shift their time windows or fine-tuning their altitude. Finally, the system generates a detailed recovery sequence table, specifying the precise arrival time, flight speed profile, and altitude adjustment instructions for each drone, and sends this information to the corresponding drone.

[0035] For example, besides "001", "002" and "003" also applied for recovery almost simultaneously. The scheduling algorithm detected that the recovery paths generated by "002" and "003" highly overlapped in the airspace 200 meters from the rear of the vehicle, and their arrival times differed by only 0.5 seconds, posing a significant risk of collision. The system further checked their battery levels and found that "002" had 15% remaining, while "003" had 25%. Therefore, the system determined that "002" had a higher priority. The scheduling algorithm kept "002's" original path unchanged and instructed "003" to add a 50-meter diameter hovering command to its path, delaying its arrival time in the conflict area by 10 seconds. Subsequently, the system sent updated scheduling instructions to the three drones: "001" to proceed immediately, "002" to proceed after a 15-second interval, and "003" to proceed after a 25-second interval.

[0036] 104. The UAV flies to the vehicle-mounted recovery platform along a dynamically adjusted recovery path according to the recovery sequence and time window. The vehicle-mounted recovery platform controls the mechanical structure to physically lock and interface with the UAV based on the landing signal triggered by the UAV. This step involves switching the drone from flight mode to physical lock mode, requiring a quick and secure mechanical and electrical connection while the vehicle is in constant motion.

[0037] After receiving the approach command generated in step 103, the UAV flies along the dynamically adjusted recovery path. When the UAV reaches directly above the recovery platform, it enters the final landing phase. The UAV activates its downward-facing vision sensor to capture images of the recovery platform's surface and uses image recognition algorithms to locate specific markings (such as concentric circles or QR codes) pre-painted on the platform. Based on the deformation and offset of the markings in the image, the UAV calculates its lateral and attitude angular deviations relative to the platform plane and corrects its attitude by adjusting the motor speed to ensure that its belly remains strictly parallel to the recovery platform plane.

[0038] The drone then descends slowly until the pressure sensor on its landing gear touches the ground, triggering a landing signal. Upon receiving this signal, the vehicle-mounted recovery platform immediately activates its rapid locking mechanism. This mechanism employs a dual locking system of electromagnetic attraction and mechanical grippers. The energized electromagnet generates a strong suction force, instantly attaching the drone's landing gear to the platform surface; simultaneously, the mechanical grippers, driven by a motor, quickly close, locking the drone's wing roots or landing gear struts, creating a physical anti-detachment lock to prevent the drone from slipping or tipping over due to severe vehicle jolting.

[0039] After locking is confirmed, the telescopic interface probe on the platform rises upward under the push of the servo motor. The probe tip has a flexible contact that can automatically align and insert into the charging interface groove on the bottom of the drone. The interface detection circuit monitors the continuity of the circuit in real time. Once the voltage and resistance values ​​are detected to be within the normal range, the electrical connection is confirmed to be successful.

[0040] For example, "001" flew to the recovery platform according to dispatch instructions. Its visual sensors detected an "H"-shaped mark on the platform and found that its nose was 5 centimeters to the left. It then adjusted the rotor thrust difference to correct the deviation. The landing gear wheels gently touched the platform surface, and the landing signal was issued. The recovery platform immediately activated, and a 3,000-watt powerful electromagnet attracted "001's" metal landing gear. At the same time, the four mechanical claws on both sides quickly closed like hands, firmly gripping the fuselage. At this moment, the vehicle ran over a rock and violently bumped, but "001" remained motionless. Immediately afterwards, three gold-plated probes in the center of the platform rose and precisely inserted into the charging port on the bottom of "001". The feedback resistance of the detection circuit was 0.1 ohms, indicating a good connection and that charging could begin.

[0041] 105. The vehicle energy management system dynamically allocates charging power based on the battery status of the docked drone and the overall energy load of the vehicle. After charging is completed, it updates the drone status according to mission requirements and either releases the drone or keeps it parked.

[0042] This step is the post-docking energy management phase, which aims to optimize the charging efficiency of multiple drones and prepare them for re-deployment when the vehicle's power supply is limited.

[0043] After the drones lock onto and dock, the vehicle's energy management system takes over charging control. The system first reads real-time data from the battery management system of each locked drone via the charging interface's data bus, including the battery's current cell temperature, state of charge, health status, and maximum allowable charging current. Simultaneously, the system monitors the vehicle's own energy status, including the engine's output power, the charge level of the auxiliary battery, and the load on other electrical devices within the vehicle.

[0044] Based on the above data, the system calculates the maximum total power available for drone charging. This maximum is dynamically adjusted; for example, when a vehicle accelerates uphill, the power allocated to charging automatically decreases. Subsequently, the system uses a power allocation algorithm to dynamically adjust the charging current for each drone within the constraint of the maximum total power. The core strategy of the algorithm is state-of-charge priority, meaning that a larger charging current is allocated to the drone with the lowest battery level to help it reach takeoff charge as quickly as possible; for drones with relatively higher battery levels, a smaller current is allocated for trickle charging or heat preservation.

[0045] In addition, the system incorporates a battery temperature protection mechanism. If the battery temperature of a drone becomes too high, the system will forcibly reduce or stop its charging current until the temperature returns to normal. When a drone's battery level reaches a preset mission requirement (e.g., 90%), the system will stop charging and update the drone's status to ready. At this point, based on mission instructions from the vehicle control center, the system can control the unlocking mechanism to release the drone, allowing it to execute the next mission, or maintain its locked parking state while the vehicle is stationary.

[0046] For example, a recovery vehicle retrieves three drones. While climbing a hill, the engine is under heavy load, and the system calculates that it can only provide 5000 watts of power to charge the drones. Drone "001" has 18% battery and a battery temperature of 20 degrees Celsius; Drone "002" has 15% battery and a battery temperature of 45 degrees Celsius; and Drone "003" has 25% battery and a battery temperature of 25 degrees Celsius. The system determines that Drone "002" has the lowest battery level but the highest temperature, allocating it 1000 watts of power for current-limited charging to prevent overheating; Drone "001" is allocated 2500 watts of fast charging power; and Drone "003" is allocated 1500 watts. As the vehicle enters a flat section of road, the engine load decreases, and the system increases the total charging power limit to 8000 watts, automatically increasing the charging power for Drone "002". Ten minutes later, Drone "001" reaches 95% battery, the system stops charging it, and marks its status as mission ready. At this moment, a new mission command was issued, the system controlled "001" to release its mechanical claw, the probe retracted, and "001" took off again.

[0047] Example 2 This embodiment further illustrates step 101 in Embodiment 1. It details how the vehicle-mounted control center receives and processes the drone swarm recovery request, and how, in a complex and dynamic vehicle environment, it intelligently allocates a dedicated guidance channel and initial access point to each drone. This embodiment is the starting point of the entire intensive recovery process, and the rationality of its allocation directly determines the efficiency and safety of subsequent guidance and recovery.

[0048] In this embodiment, step 101 further includes: 1011. When each drone in the drone swarm detects that its own battery level is below a preset threshold or receives a ground command, it sends a recovery request to the vehicle control center, which includes its identity, current location coordinates, remaining battery level, and current flight speed.

[0049] In the wide airspace where drone swarms perform missions, each drone is a distributed, independent node. They need to autonomously determine when to initiate a recovery request based on their own energy status and mission status.

[0050] Each VTOL fixed-wing UAV integrates an intelligent mission management and energy monitoring module. This module reads voltage, current, and cell temperature data from the battery management system in real time, and uses the ampere-hour integration method and open-circuit voltage method to accurately calculate the remaining battery percentage. The system has multiple preset thresholds, including a "Level 1 warning threshold" (e.g., 30%) and a "forced return threshold" (e.g., 15%). When the remaining battery level drops to the Level 1 warning threshold, the UAV will enter a low-power standby mode; once it drops to the forced return threshold, regardless of whether the current mission is completed, the system will automatically abort the current mission, lock the flight path, and prepare to initiate a recovery request.

[0051] In addition to being triggered by battery power, the drone also listens for broadcast commands from the vehicle's control center via an encrypted data link. If the vehicle's control center needs to recall the drone due to tactical changes, sudden weather changes, or the vehicle about to enter a dangerous area, it will send a mandatory recall command code with the highest priority. Upon receiving this command, the drone will immediately cease its current actions and switch to the recovery process.

[0052] When constructing the recovery request information data packet, the drone not only includes the basic status information mentioned above, but also packages the following key data to ensure the accuracy of subsequent scheduling: Identification code: This is a globally unique identifier used to distinguish individuals within a cluster and prevent communication conflicts.

[0053] Current location coordinates: based on longitude, latitude, and altitude obtained from the Global Navigation Satellite System module, with an accuracy of sub-meter level.

[0054] Remaining battery power: Provided in both percentage and remaining flight time (minutes) format, allowing dispatch centers to intuitively assess urgency.

[0055] Current flight speed: including ground speed and air speed, used to calculate the time required to reach the recovery vehicle.

[0056] Airborne health status: including whether the power system is malfunctioning, communication link quality index, etc., is used to assess whether the drone has the ability to land normally or requires emergency intervention.

[0057] After being compressed and encrypted, this data is transmitted to the ground-based vehicle control center via a time-division multiple access method through an interference-resistant VHF or UHF data link.

[0058] For example, a drone codenamed "Alpha" was performing a long-endurance reconnaissance mission when its battery management system detected that the voltage of a single battery cell had dropped to 3.6 volts, with an estimated remaining charge of 14%. At this point, "Alpha's" flight control computer determined that it had reached the mandatory return-to-home threshold. It immediately stopped circling, automatically adjusted its nose to point towards the designated recovery area, and simultaneously assembled a data packet. The data packet contained: "Alpha's" ID code, its current location (116.4 degrees east longitude, 39.9 degrees north latitude), altitude (500 meters), remaining charge (14%), and current ground speed (22 meters per second). Furthermore, because "Alpha" had encountered strong crosswinds during flight, its wing sensors recorded slight structural stress data, which was also packaged into the data packet. This data packet was then transmitted to the ground via frequency hopping signals.

[0059] 1012. After receiving the recovery request information, the vehicle control center, in combination with the vehicle's current driving speed, driving direction and vehicle body posture and bump status, predicts the vehicle's motion trajectory envelope within a preset time period in the future.

[0060] Unlike fixed ground-based recovery points, vehicle-mounted recovery platforms operate in a non-inertial frame of motion. Changes in vehicle speed, centrifugal force during steering, and road surface disturbances can all cause unpredictable and drastic changes in the platform's position and attitude in space. Therefore, a dynamic capture window must be established.

[0061] The onboard control center is integrated into the vehicle's central control computer and is equipped with a high-performance real-time operating system. Upon receiving a drone recovery request, the control center immediately activates the motion prediction algorithm module. This module collects the following real-time motion parameters through the onboard sensor network: (1) Motion parameters: The absolute ground speed of the vehicle is obtained by wheel speed sensor and Doppler radar; the heading angle, pitch angle and roll angle of the vehicle are obtained by fiber optic gyroscope and electronic compass; the geodetic coordinates of the vehicle are obtained by global navigation satellite system.

[0062] (2) Disturbance parameters: The frequency and amplitude of vertical bouncing caused by uneven road surface are obtained by the displacement sensor on the suspension system and the vehicle body accelerometer, i.e. the bumpy state.

[0063] The motion prediction algorithm, based on a vehicle dynamics model, takes current motion and disturbance parameters as input and uses Monte Carlo simulation or Kalman filtering to predict the vehicle's trajectory over a preset time period (e.g., 60 to 120 seconds). This trajectory is not a single line, but a geometric shape that continuously expands, contracts, and rotates in three-dimensional space—the motion trajectory envelope. This envelope defines the maximum spatial range where the center point of the recovery platform may appear. For example, when the vehicle is traveling at high speed in a straight line, the envelope is a long, thin tube; when the vehicle is traveling at low speed on rough terrain, the envelope is an irregular, larger mass.

[0064] The prediction process also considers the vehicle's handling characteristics, such as maximum steering angle, maximum braking acceleration, and maximum climbing ability. If the vehicle is performing a maneuver to avoid an obstacle, the prediction algorithm will correspondingly increase the volume of the envelope to accommodate possible extreme position changes, ensuring that the recovery command has sufficient safety margin.

[0065] The ground-based recovery vehicle is traveling at 60 km / h on an undulating hilly dirt road. The onboard control center receives a recovery request from "Alpha". At this time, sensors show the vehicle's heading is due north, pitch angle is ±5 degrees (uphill), and roll angle is zero. Simultaneously, the accelerometer detects the vehicle experiencing a vertical bump with a frequency of 1.5 Hz and an amplitude of ±10 cm. The control center's motion prediction algorithm immediately runs. Considering the vehicle's current uphill position and rough road surface, the algorithm predicts that within the next 90 seconds (i.e., the estimated arrival time of "Alpha"), the vehicle will move 1500 meters in its forward direction. Due to the bumps and uphill climb, the recovery platform's height will fluctuate by 20 cm above and below the baseline height, and the heading may sway left and right by 5 degrees to avoid potholes. Based on this, the system generates an ellipsoidal motion trajectory envelope centered on the rear recovery platform, with dimensions of 10 meters, 5 meters, and 3 meters, and uses this envelope as the benchmark for subsequent planning.

[0066] 1013. Based on the motion trajectory envelope and the current position coordinates of the UAV swarm, calculate the estimated arrival time of each UAV to the vehicle recovery area, and assign each UAV a non-interfering guidance channel number and corresponding initial access point spatial coordinates according to the order of the estimated arrival times or the preset priority rules.

[0067] The purpose of this step is to decouple multiple drones in the spatiotemporal dimension, preventing them from occupying the same airspace resources at the same time, and at the same time designating a gate for each drone to enter the recovery guidance phase.

[0068] The control center has multiple available guidance channels (e.g., eight independent channels), each corresponding to a specific set of communication frequencies and signal codes. To ensure the purity and directionality of the guidance signals, any two UAVs that are close to each other in the recovery space-time domain cannot use the same channel to avoid co-channel interference or multipath effect confusion.

[0069] The allocation process begins with spatiotemporal calculations. The system performs vector calculations on the current position coordinates of each requesting drone and the predicted trajectory envelope of the vehicle. Considering that the vehicle is a moving target, the drone cannot fly directly to the vehicle's current coordinates but must fly to the vehicle's coordinates at some future point in time. This involves calculating an interception point. The system assumes that the drone maintains its current speed and the vehicle maintains its predicted motion, and uses an iterative approximation method to calculate the moment when the two are closest or reach the preset recovery distance, i.e., the estimated arrival time.

[0070] Based on the estimated arrival time, the system initially prioritizes the drones. If two drones arrive very close together (e.g., the time difference is less than five seconds), the system will further assess their priorities. The priority rules are determined by a weighted average of several factors, including: remaining battery power (lower battery power, higher priority), mission importance (e.g., drones carrying critical data have higher priority), and onboard health status (faulty drones have the highest priority and require emergency recovery).

[0071] After the recovery sequence is determined, the system begins to allocate guidance channels and initial access points.

[0072] Guidance channel allocation: A graph coloring algorithm or a simple interval allocation method is used. For example, the first drone in sequence is assigned channel one, and the second drone in sequence is assigned channel two; if the first and third drones in sequence are spatially far apart, the third drone in sequence can reuse channel one.

[0073] Initial Access Point Allocation: The initial access point is the starting point from which the drone leaves cruise mode and enters the coverage area of ​​the vehicle-mounted high-precision guidance system. This point cannot be set directly above the vehicle (to prevent collision) or in the turbulence area behind the vehicle (to prevent loss of control). It is usually set at a virtual location above or to the side of the vehicle. This location is a dynamic coordinate relative to the vehicle's coordinate system. The system will assign a different access point to each drone; for example, the first drone's access point might be set 100 meters above and behind the vehicle, and the second 120 meters above and behind the vehicle, thus creating a vertical hierarchy and maintaining a safe distance.

[0074] Finally, the system packages the allocation results into a confirmation command and sends it to the corresponding drone via a wide area communication link.

[0075] For example, while the control center was processing the application for "Alpha," it discovered that another drone, "Beta," had also submitted an application.

[0076] Time calculations: Calculations show that "Alpha" is 20 kilometers away from the recovery vehicle, flying at a speed of 22 meters per second, and will take approximately 15 minutes to arrive. "Beta" is 15 kilometers away, and will take approximately 11 minutes to arrive.

[0077] Sorting and Priority: Although "Beta" arrived first, the system detected that "Alpha" had only 14% battery remaining, while "Beta" had 30%. Based on the principle of prioritizing low battery, the system adjusted the recycling sequence of "Alpha" to first, and "Beta" to second.

[0078] Channel allocation: The system allocates channel 01 (frequency of a certain hertz, code A) to the “Alpha” guidance channel and channel 02 (frequency of a certain hertz, code B) to the “Beta” guidance channel to avoid interference.

[0079] Access point settings: Considering that "Alpha" has priority and low battery, the system sets the nearest access point for it at 80 meters above and behind the vehicle. To prevent "Beta" from following too closely and causing danger, the system sets the access point for "Beta" at 150 meters above and behind the vehicle and instructs it to hover at the access point and wait for "Alpha" to land.

[0080] Sending instructions: The vehicle center sends an instruction to "Alpha": "Approved for recovery, assigned channel 01, target access point relative coordinates (80 meters behind, 80 meters above), full speed ahead." At the same time, it sends an instruction to "Beta": "Wait in line, assigned channel 02, target access point relative coordinates (150 meters behind, 150 meters above), maintain circling upon arrival." Through the precise collaboration of the above three sub-steps, Example 2 achieves accurate response and orderly resource allocation to multiple drone recovery requests in a mobile vehicle platform environment, laying a solid spatiotemporal foundation for subsequent high-precision guidance and physical recovery.

[0081] Example 3 This embodiment provides a detailed explanation of step 102 in Embodiment 1. It elaborates on how, after completing channel allocation and access point setting in Embodiment 2, a vehicle-mounted multi-channel guidance system is used to achieve high-precision relative positioning of UAVs in the airspace surrounding the moving vehicle, and how a recovery path that adapts to dynamic environmental changes is generated in real time based on the vehicle's nonlinear motion. This step is crucial for solving the problems of "difficult tracking and aiming, accurate positioning, and optimal path" in mobile platform recovery.

[0082] In this embodiment, step 102 further includes: 1021. The vehicle-mounted multi-channel guidance system uses beamforming technology to send ultra-wideband pulse signals or laser guidance signals with timestamps and angle information to the corresponding UAV on each guidance channel.

[0083] The core of this invention lies in establishing a highly interference-resistant and highly directional communication and measurement link. Because vehicles operate in complex electromagnetic environments and are in motion, traditional omnidirectional antennas cannot meet the signal isolation requirements for simultaneous recovery of multiple UAVs. Therefore, this invention employs advanced phased array beamforming technology.

[0084] The vehicle-mounted multi-channel guidance system is installed inside the fairing on the top of the vehicle and consists of a multi-element phased array antenna array, an RF transceiver module, and a baseband signal processor. After receiving the guidance channel number assigned in Embodiment 2, the system forms an independent narrow beam on the array for each assigned UAV.

[0085] Beam pointing control: The baseband processor calculates the phase weighting value of each element in the antenna array based on the approximate coordinates of the Global Navigation Satellite System transmitted back by the UAV. By adjusting the phase difference of the signals of each element, coherent interference of the transmitted signals in space is achieved, thereby concentrating the energy within an extremely narrow main lobe and precisely pointing it at the target UAV. At the same time, nulls are formed in the directions of adjacent UAVs, minimizing signal crosstalk between channels.

[0086] Signal Modulation and Transmission: The system selects ultra-wideband pulse signals as the carrier due to their extremely high time resolution (nanosecond level), ability to penetrate rain and fog, and strong resistance to multipath effects. Each pulse signal is loaded with a precise atomic clock timestamp and information indicating the current antenna beam pointing angle. For the laser guidance channel (as a backup or high-precision mode), the system modulates a laser-coded beam, which also contains angle and synchronization information.

[0087] Continuous tracking: The beam is not fixed but controlled in a closed loop. The system monitors the signal strength indication transmitted back by the drone in real time and fine-tunes the beam direction through iterative algorithms, always locking the beam center on the drone's body. Even if the drone is maneuvering at high speed or the vehicle is violently bumpy, the link will not be interrupted.

[0088] This design is equivalent to laying an invisible fiber optic cable for each drone, creating a multi-layered proprietary guidance network above the vehicle.

[0089] For example, suppose the vehicle control center has already assigned channel 01 to "Alpha". The phased array controller of the guidance system immediately calculates the array parameters and precisely points the main lobe beam at "Alpha", which is 80 meters above and behind the vehicle. At this time, another drone, "Beta", is hovering at 150 meters. While pointing at "Alpha", the system automatically generates a signal null in the direction of "Beta". Subsequently, the system sends ultra-wideband pulses to "Alpha" at a rate of 10 trillion times per second. Each pulse is marked with the information "sent at 12:00:00:00:00 milliseconds" and "beam elevation angle 30 degrees, azimuth angle 0 degrees". Even if the vehicle suddenly makes a sharp left turn while driving, the inertial navigation equipment of the guidance system senses the turning angular velocity and immediately drives the beam to deflect five degrees to the left, always keeping the signal closely following "Alpha" like a spotlight.

[0090] 1022. After receiving the guidance signal, the UAV calculates its three-dimensional relative coordinates with respect to the vehicle coordinate system by using the signal propagation delay and phase difference, and combines it with its own inertial navigation data for filtering and fusion to obtain a high-precision relative position estimate.

[0091] In mobile recovery scenarios, absolute coordinates (such as latitude and longitude) cannot meet centimeter-level docking requirements due to error accumulation and communication latency. Therefore, it is essential to obtain the real-time relative position with respect to the vehicle's coordinate system. The UAV's onboard unit is equipped with a matching microstrip array antenna and a high-speed signal processing unit. Upon receiving a pulse signal from the vehicle-mounted guidance system, the following processing flow is executed: (1) Time Delay Ranging: The UAV reads its own clock and records the moment the pulse signal is received. By comparing the difference between the received moment and the transmission moment carried in the pulse, the flight time of the signal in the air is calculated. Since the speed of light is known, the straight-line slant distance between the UAV and the vehicle-mounted antenna can be obtained by multiplying the flight time by the speed of light.

[0092] (2) Phase difference angle measurement: The UAV uses multiple receiving antennas (at least three non-collinear) distributed on the fuselage to measure the phase difference of the same pulse signal reaching different antennas. The phase difference reflects the path difference of the signal to each antenna. Based on the path difference and the known antenna spacing, the signal incident direction, i.e., the pitch angle and azimuth angle of the UAV relative to the vehicle, can be calculated using geometric relationships.

[0093] (3) Coordinate transformation and fusion: Combining the calculated slant range, pitch angle and azimuth angle, the position of the UAV is determined in the spherical coordinate system. Then, it is transformed into Cartesian relative coordinates with the vehicle center as the origin, the vehicle longitudinal axis as the X-axis, the horizontal axis as the Y-axis and the vertical axis as the Z-axis through trigonometric function transformation.

[0094] (4) Filtering Optimization: To eliminate signal noise and non-line-of-sight errors, the system introduces an extended Kalman filter algorithm. The calculated relative coordinates are used as observations, and the acceleration and angular velocity measured by the UAV's own inertial navigation system are used as control variables to predict and correct the UAV's motion state. The filtering algorithm can effectively smooth data jumps caused by vehicle bumps and output a continuous, smooth, and highly confident relative position trajectory.

[0095] Through this process, the drone no longer knows its exact location on Earth, but rather precisely knows its location "how many meters to the left rear and how many meters above" the vehicle.

[0096] For example, the UAV "Alpha" received an ultra-wideband pulse, measuring the signal propagation time as 330 nanoseconds, and calculated a slant range of approximately 99 meters. Simultaneously, the phase difference measured by the antennas at both ends of its wings corresponded to an incident azimuth angle of -5 degrees (i.e., to the left rear of the vehicle), while the phase difference measured by the tail antenna and the fuselage antenna corresponded to an incident elevation angle of 30 degrees. The flight control computer quickly performed a trigonometric transformation, determining that "Alpha" was located at X-axis -85 meters, Y-axis -8 meters, and Z-axis +49 meters in the vehicle coordinate system. At this moment, due to a crosswind, "Alpha" sideslipped, and the inertial navigation system detected the sideslip acceleration. The extended Kalman filter then intervened, recognizing that the position calculated solely based on the signal was lagging behind the actual motion. Therefore, it combined the inertial data to perform forward compensation on the calculated position, ultimately outputting an optimized relative position correction of X-axis -84.5 meters, Y-axis -9.1 meters, and Z-axis +49 meters, improving accuracy to the centimeter level.

[0097] 1023. The vehicle control center receives the high-precision relative position estimate transmitted back by the UAV, combines it with the position compensation amount generated by the real-time movement of the vehicle, and uses the spline interpolation algorithm to generate a dynamic recovery path with the initial access point as the starting point and the vehicle recovery platform locking point as the ending point.

[0098] The challenge in this step lies in the fact that the endpoint of the path (the recovery platform) changes constantly with the vehicle's movement, and the starting point (the initial access point) also needs to be updated based on the vehicle's movement. This invention employs dynamic spline interpolation technology to generate a recovery trajectory that satisfies the kinematic constraints of the UAV while smoothly adapting to vehicle disturbances.

[0099] After receiving the high-precision relative position transmitted back by the drone, the vehicle control center activates the path planner. The path planner not only calculates the path but also calculates the vehicle's motion compensation in real time.

[0100] Motion compensation calculation: Due to signal propagation delays, by the time the UAV receives the signal and transmits its position, the vehicle has already moved a certain distance. The control center uses the vehicle's high-frequency inertial data to calculate the vehicle's displacement and rotation increments from the moment the signal was sent to the present moment. This is the position compensation amount, representing the distance the vehicle has traveled. The control center adds this compensation amount to the relative position transmitted by the UAV to obtain the absolute relative position of the UAV with respect to the vehicle's current real-time attitude.

[0101] Critical point selection: Between the initial access point and the recovery platform lock point (usually the center of the vehicle's rear deck), the planner selects several critical path points, either uniformly or non-uniformly, based on flight altitude and distance. These points should avoid the vehicle's wake turbulence zone (usually 10 to 30 meters above and behind the vehicle) and the exhaust pipe's hot airflow zone.

[0102] Spline interpolation generation: Using cubic B-spline or quintic polynomial interpolation algorithms, the key points mentioned above are connected. The advantage of spline interpolation is that the generated curve has continuous first derivatives (velocity) and second derivatives (acceleration), which means that the UAV does not need to make frequent sharp turns or sudden accelerations and decelerations when tracking the path, and the flight trajectory is very smooth.

[0103] Dynamic refresh mechanism: The generated path is not a one-time event. Every time the vehicle moves or the drone's position is updated, the path is recalculated. As the vehicle moves forward and turns, the entire recovery path translates and rotates in space like a conveyor belt, but the shape of the path remains stable relative to the vehicle's coordinate system, guiding the drone to land smoothly on the moving platform.

[0104] For example, the onboard center calculates that due to communication delays and vehicle movement, Alpha's actual relative position needs to be compensated by 2 meters in the positive X-axis direction. The planner then selects the starting point in the relative coordinate system: 80 meters behind the vehicle, 50 meters high (i.e., the access point in Example 2); the midpoint: 30 meters behind the vehicle, 15 meters high (to avoid wake turbulence); and the ending point: 0 meters behind the vehicle, 0 meters high (deck center). Using fifth-order polynomial spline interpolation, the system generates a smooth descent curve. Suddenly, the vehicle swerves sharply to the right to avoid a pothole, causing the rear of the vehicle to swing to the left. The path planner detects this change within milliseconds and immediately rotates all path points ten degrees to the left around the vehicle's center. At this moment, from the perspective of an observer outside the vehicle, Alpha's predetermined trajectory swings instantly to the left like a dancing ribbon, accurately predicting the vehicle's turn and ensuring that the drone always aligns with the rotated deck center. The system sends this dynamically updated path parameter to Alpha, instructing it to fly along the dynamically updated path parameter.

[0105] Example 4 This embodiment is a further detailed description of step 103 in Embodiment 1. This embodiment elaborates on how the vehicle-mounted control center solves air traffic control problems in complex scenarios where multiple drones simultaneously request recovery, through intelligent scheduling algorithms. Its core lies in addressing the challenges brought about by dense recovery: that is, how to coordinate multiple dynamically flying drones within a limited time and space window to ensure their safe, orderly, and efficient landing on the mobile vehicle platform.

[0106] In this embodiment, step 103 further includes: 1031. The vehicle-mounted control center constructs a scheduling optimization model with the objective functions of minimizing recovery time and balancing battery consumption. The input parameters include the real-time spatial location of all UAVs to be recovered, the remaining battery power, and the generated dynamic recovery path.

[0107] The core of this sub-step lies in establishing a multi-objective optimization decision-making model. Traditional single-objective scheduling often only pursues speed, which may cause some drones with low battery levels to run out of power and crash due to waiting in the queue for too long. Therefore, this embodiment adopts a time-energy dual optimization model.

[0108] The scheduling processor in the vehicle control center constructs a scheduling model based on a mixed-integer linear programming or genetic algorithm framework. This model defines two core objective functions: (1) Minimize total recovery time: that is, minimize the total time from the first UAV entering the recovery process to the last UAV completing the lock-on. This requires the system to allow UAVs to process in parallel as much as possible and make full use of the resources of multiple guidance channels and recovery positions.

[0109] (2) Battery consumption equalization and minimization: Here, equalization means avoiding excessive waiting for a certain drone; minimization means minimizing the high energy consumption of the drone during the waiting and circling period.

[0110] In this embodiment, the input parameters of the model constitute the boundary conditions for decision-making: Real-time spatial location: Determines the current distance and situation of the drone.

[0111] Remaining battery power: This is a key indicator for calculating urgency. The lower the battery level, the higher its weighting penalty in the model, meaning the system will tend to allow the drone to cut in line or recover it via a better path to reduce the risk of it running out of power.

[0112] Dynamic recovery path: Each path includes a preset flight distance, turning radius, and estimated time.

[0113] When building the model, the system also introduces constraints, such as: only one drone can be served by the same guidance channel at a time; the recovery platform can only perform one locking operation at a time; and a legally mandated safe distance must be maintained between any two drones. Under these constraints, the scheduling model will iteratively search for an optimal or suboptimal solution set, which defines the arrival sequence of each drone.

[0114] For example, the vehicle-mounted control center simultaneously faces three drones requesting recovery: "Alpha" has 15% remaining battery power and is 5 kilometers away; "Beta" has 40% remaining battery power and is 10 kilometers away; and "Gamma" has 20% remaining battery power and is 5.5 kilometers away. If sorted solely by distance, "Alpha" and "Gamma" would arrive almost simultaneously, leading to significant conflict. If sorted solely by battery power, "Alpha" would have the highest priority, followed by "Gamma," and finally "Beta." After the scheduling model runs, comprehensive calculations show that if "Gamma" is allowed to hover and wait for "Alpha," although "Alpha" can recover the fastest, "Gamma" also has limited battery power, and hovering could increase its risk. The model's final solution is to leverage the multi-channel advantage, instructing "Alpha" and "Gamma" to use two different guidance channels (left and right) with a small time difference (e.g., three seconds) to stagger their arrival, while allowing the well-charged "Beta" to hover in a large circle on the outer perimeter. This approach ensures a short total recovery time (with both aircraft recovering simultaneously) while also taking into account the battery status of "Gamma," preventing it from running out of power due to prolonged hovering.

[0115] 1032. Using this scheduling optimization model, spatiotemporal conflict detection is performed on the dynamic recovery paths of all UAVs to be recovered, and path segments that overlap in three-dimensional space and time are identified.

[0116] Even with a preliminary scheduling plan, the recovery paths change dynamically as vehicles move, and originally separate paths may overlap at some point due to special vehicle maneuvers (such as sharp turns causing paths to converge). Therefore, refined spatiotemporal conflict detection is necessary.

[0117] Spatiotemporal conflict detection is a four-dimensional (three-dimensional space + time) geometric calculation process. The system treats the dynamic recovery path of each UAV as a pipe extending in four-dimensional space. The cross-section of the pipe is determined by the safety distance (e.g., a cylinder with a radius of 10 meters), representing the absolutely safe airspace that the UAV must maintain during flight.

[0118] The detection process in this embodiment adopts a discrete sampling strategy: (1) Path discretization: The system slices the recovery path of each UAV along the time axis, for example, sampling a coordinate point every zero point one second. This results in a series of trajectory point sequences with timestamps.

[0119] (2) Pairwise comparison: The system traverses and compares the trajectory point sequences of any two UAVs. For point P_a at time T_a of UAV A and point P_b at time T_b of UAV B, the spatial distance between them is calculated.

[0120] (3) Conflict determination: If the spatial distance is less than the preset safety threshold (e.g., 5 meters) and |T_a - T_b| is less than the preset time threshold (e.g., 2 seconds), then the two drones are determined to have a spatiotemporal conflict. This means that if they fly along the current path, the two drones will be very close at some point, and may even collide or be severely interfered with by the wake turbulence.

[0121] (4) Dynamic prediction: Considering that the vehicle is moving and the path itself is changing, the system will combine the real-time motion prediction of the vehicle to dynamically push forward the future path points to ensure that the detected conflict is a real future risk, rather than just an overlap of historical data.

[0122] For example, the system is detecting the paths of "Alpha" and "Gamma". Although the initial scheduling scheme allows them to enter the field 3 seconds apart, the vehicle travels to a sharp turn. The path point sequence of "Alpha" shows that at the 12th second, it will be 10 meters behind and to the left of the vehicle, 5 meters high. The path point sequence of "Gamma" shows that at the 13th second, it will be 12 meters behind and to the left of the vehicle, 5 meters high. The system calculates that the spatial distance between the two is 2 meters, less than the 5-meter safety threshold; the time difference is 1 second, less than the 2-second time threshold. The system immediately issues a red alert: a spatiotemporal conflict has been detected! Specifically, during the vehicle's left turn, the two recovery paths overlap at the point where they meet on the left rear of the vehicle. Without intervention, the two drones will nearly collide in mid-air.

[0123] 1033. Based on the collision detection results and the priority of remaining battery power, perform path replanning or time window offset for drones with overlapping paths, and generate a recovery sequence table containing the precise recovery time and flight speed adjustment instructions for each drone.

[0124] This step is the final stage of conflict resolution and command generation. Once a conflict is detected, the system must immediately execute a corrective strategy. The principle of correction is to avoid prioritizing low battery performance while also considering recovery efficiency; that is, to prioritize protecting low-battery drones and minimize their impact on overall recovery efficiency.

[0125] Based on the severity of the conflict and the status of the drone, the system selects one or more of the following remedial strategies: (1) Time window offset (aerial waiting): The commanded UAVs with lower priority hover or circle at a safe altitude before the conflict point, allowing the UAVs with higher priority to pass through the conflict area first. The offset is accurate to the second to ensure sufficient safe interval.

[0126] (2) Path replanning (spatial avoidance): If the time offset would prevent the low-battery drone from being recovered in time, the system will modify the drone's flight path. For example, it can adjust the angle at which the drone enters the recovery platform, or instruct it to enter from the other side of the vehicle (e.g., from the left to the right), thereby completely avoiding the conflict in space.

[0127] (3) Speed ​​adjustment (compression / stretch): Instruct the UAV to fine-tune its flight speed in non-conflict zones. For example, slightly accelerate the leading UAV to fly away from the conflict zone, or slightly decelerate the trailing UAV to lag behind, thereby increasing the distance.

[0128] After the corrections are complete, the system generates the final recovery sequence table. This table is not just a simple queue list, but a flight schedule for each drone. The table includes: Precise recovery timing: For example, "Alpha" must make contact with the deck at 12 minutes and 30 seconds.

[0129] Critical node speed: For example, the speed drops to 15 meters per second when passing through a conflict zone, and accelerates to 25 meters per second on straight sections.

[0130] Altitude layer instructions: For example, “Gamma” maintains an altitude of 50 meters while waiting and descends to 30 meters upon approach.

[0131] The instruction list is encapsulated into a dedicated navigation data packet and sent to the corresponding drone in real time.

[0132] For example, regarding the conflict between "Alpha" and "Gamma" detected in the previous step on the left rear of the vehicle, the system checks the battery level: "Alpha" is 15%, and "Gamma" is 20%. Although "Alpha" is lower, "Gamma" is also at the low battery warning threshold, making prolonged hovering unsuitable. The system decides to employ a hybrid strategy combining spatial avoidance with a slight time shift: The system replanned the recovery path for "Gamma," changing its entry point from the left rear side of the vehicle to the right rear side, bypassing the conflict area. "Alpha" maintained its original left rear side path but was instructed to increase its speed by 10% as it approached the rendezvous point for a faster passage.

[0133] The final recycling sequence table is generated as follows: “Alpha”: Estimated contact time 12 minutes and 35 seconds, path left-hand, normal speed throughout.

[0134] “Gamma”: Expected contact time is 12 minutes and 40 seconds. The path is to the right. At the 10-minute mark, it will perform a large-radius rightward maneuver to avoid the airflow on the left.

[0135] The final command was issued, and the two drones were guided back, one to the left and one to the right, one in front and one behind, safely and smoothly flying towards the mobile recovery platform.

[0136] Example 5 This embodiment provides a further detailed description of step 104 in embodiment 1. This embodiment details the process by which the UAV, after completing aerial guidance, finally contacts and achieves stable parking and electrical connection on the mobile vehicle platform. This is the stage with the lowest tolerance for error during recovery, requiring millisecond-level response and millimeter-level docking accuracy under harsh physical conditions such as vehicle vibration and airflow disturbance.

[0137] In this embodiment, step 104 further includes the following sub-steps: 1041. The UAV flies to a hovering point above the recovery platform, uses visual sensors to identify the marked graphics on the recovery platform, adjusts its attitude to make the belly of the UAV parallel to the plane of the recovery platform, and slowly descends until the landing gear touches the ground and triggers the landing signal.

[0138] This step focuses on using visual corrections to achieve the final soft landing alignment. Although radio guidance has been used previously, at extremely close range (the last meter), radio signals may be subject to multipath interference and cannot detect instantaneous tilting of the recovery platform (such as one side tilting up due to a vehicle running over a rock). Therefore, visual sensors are crucial.

[0139] When the drone flies along the dynamic path to a hovering point ±0.5 meters above the recovery platform, the flight control system switches to precision vision landing mode.

[0140] Visual Recognition: A high-definition wide-angle camera or infrared camera mounted on the underside of the drone is activated to capture real-time images of the recovery platform surface. The recovery platform surface is coated with high-contrast feature markings, such as QR codes, ArUco markers, or concentric circle arrays. The onboard image processing unit analyzes the video stream in real time, extracts the corner information of the markings, and calculates the lateral deviation (X, Y axis deviation) and altitude deviation (Z axis deviation) of the drone relative to the center of the markings using a perspective transformation algorithm.

[0141] Attitude synchronization: The system further analyzes the geometric deformation of the marked graphic (e.g., a square becomes a trapezoid) to deduce the current pitch and roll angles of the recovery platform. The UAV no longer maintains a horizon attitude but dynamically adjusts its motor speeds to force the fuselage attitude to remain perfectly parallel to the actual tilt angle of the recovery platform. This step is to prevent the UAV from landing in the center but then tipping over due to uneven wings causing one side of the landing gear to bear the load first.

[0142] Ground contact trigger: after completing alignment and synchronization, the unmanned aerial vehicle (UAV) descends vertically at an extremely low speed (e.g., 0.3 meters per second). When the buffer strut on the landing gear is compressed, or a pressure contact switch installed at the bottom of a landing skid detects physical pressure, the circuit is instantly conducted to generate a high-level pulse signal, that is, the landing signal. This signal not only turns off the motor through the airborne internal bus, but also is urgently sent to the vehicle-mounted recovery platform controller via a wireless link.

[0143] For example, the UAV "Alpha" hovers 1 meter above the recovery platform. At this time, the right rear wheel of the vehicle rolls over a raised rock, causing the recovery platform to instantly tilt 5 degrees to the right. The UAV's downward-looking camera captures the "hui"-shaped marker on the platform and finds that the shape of the marker is distorted. The flight control computer immediately calculates the correction amount, controls the right rotor to reduce pulling force and the left rotor to increase pulling force, tilts the fuselage 5 degrees to the right, and perfectly fits the inclined surface of the platform. Then, "Alpha" descends slowly. In an instant of 0.01 seconds, the micro switch at the bottom of the left landing gear skid touches the platform deck, the landing signal is generated immediately, the motor and rotors are locked and stopped immediately, and the UAV lands silently on the platform by virtue of inertia.

[0144] 1042. After receiving the landing signal, the vehicle-mounted recovery platform immediately activates the electromagnetic locking device or mechanical clamping claws on the surface of the platform, and quickly physically fixes the UAV under the bumpy environment of the vehicle to prevent the UAV from slipping or overturning.

[0145] It is extremely unsafe to stay on a moving vehicle only relying on the UAV's own gravity, and the inertia force generated by vehicle acceleration, braking or turning is sufficient to throw the UAV out. Therefore, an active strong locking mechanism must be adopted.

[0146] An embedded controller and a high-power execution unit are integrated inside the vehicle-mounted recovery platform. When the sensor of the platform (e.g., a pressure sensor array) directly detects the impact of a falling object, or receives the radio landing signal sent by the UAV, the controller triggers the locking procedure.

[0147] Electromagnetic adsorption: a high-strength electro-permanent magnet array is pre-embedded on the surface of the platform. Within millisecond-level time after receiving the signal, the electromagnet is energized to generate a strong adsorption magnetic field. Since the landing gear or the bottom of the fuselage of the UAV contains magnetic metal material (or a preset ferromagnetic induction plate), the UAV can be tightly adsorbed on the deck to overcome the lateral slip force.

[0148] Mechanical locking: To prevent the drone from bouncing and demagnetizing due to severe vehicle vibrations, the system synchronously activates mechanical grippers. These grippers typically employ a four-bar linkage mechanism, mounted at the four corners or sides of the platform. Servo motors drive the linkages, causing the rubber-coated grippers to rapidly retract inward, locking onto the drone's landing gear struts, wing roots, or a dedicated locking handle. The mechanical lock has a self-locking function; once closed, it will not release even if power is cut off.

[0149] Rigid connection: The dual locking of electromagnetic and mechanical components turns the drone and the vehicle into a rigid whole, ensuring that the drone moves with the vehicle without relative displacement.

[0150] For example, as soon as "Alpha" issued its landing signal, the onboard platform controller detected the command. Despite the vehicle being subjected to severe turbulence, the electromagnets on the platform surface reached peak attraction within 0.05 seconds, instantly and firmly holding "Alpha's" aluminum alloy landing gear in place. Immediately afterwards, the electric push rods on both sides of the platform propelled powerful mechanical claws, like two robotic hands, precisely locking onto the locking slots beneath "Alpha's" fuselage. The very next second after the lock was completed, the vehicle lurched across a deep ditch, causing violent shaking, but thanks to the double locking mechanism, the "Alpha" drone did not bounce or shift in the slightest.

[0151] 1043. After the physical locking is completed, the vehicle-mounted recovery platform controls the telescopic interface probe to rise upwards and blindly connect with the charging interface on the bottom of the drone, and confirms the reliability of the electrical connection through the interface detection circuit.

[0152] This step represents a key technical challenge in achieving energy replenishment: dynamic blind insertion. Because the drone's landing position varies by a tiny millimeter each time, and the vehicle is constantly vibrating, the interface must possess high fault tolerance and automatic alignment capabilities.

[0153] The docking process will only begin after the positioning sensor of the mechanical locking device sends a signal indicating that the locking is complete. A platform-active, drone-passive strategy is employed: the probe is mounted on the recovery platform and lifted upwards to connect with the drone.

[0154] (1) Blind insertion structure: The telescopic interface probe adopts a conical or funnel-shaped guide head design, and the probe itself has a certain elastic floating mechanism (such as a spring reset structure). The charging interface at the bottom of the drone is a large-diameter socket with elastic contacts inside.

[0155] (2) Lifting and docking: The linear motor or cylinder inside the platform drives the probe to move upward. Even if there is a horizontal deviation of two or three centimeters in the landing position of the UAV, the conical guide head will automatically slide straight through the lateral component force when it contacts the socket, guiding the probe to insert into the center of the socket.

[0156] (3) Electrical Confirmation: When the probe is fully inserted, the limit switch is triggered. The interface detection circuit immediately sends a weak detection current (e.g., 10 mA) to measure the voltage drop and resistance value of the circuit. If the detected resistance value is within a preset reasonable range (e.g., less than 50 milliohms) and the on / off logic of each pin (power positive, power negative, communication line) is correct, the connection is considered successful. If a short circuit or open circuit is detected, the system will stop outputting high-power charging current and issue an alarm.

[0157] For example, after the "Alpha" drone was locked by the mechanical gripper, the three spring-loaded interfaces in the center of the platform slowly rose under the drive of servo motors. Because the "Alpha" landed slightly to the left (5 millimeters), the guide cones at the top of the probes touched the edge of the socket. However, under the continuous thrust, the guide cones slid in smoothly, automatically correcting their centerline using their own elastic floating mechanism. With a soft "click," the three gold-plated probes were fully inserted into the grooves at the bottom of the "Alpha." The interface detection circuit quickly started working, detecting a power circuit resistance of 0.05 ohms, indicating a normal communication handshake signal. The vehicle control center then illuminated the green indicator light on the dashboard, signifying that the "Alpha" had successfully connected to the vehicle's energy and information network and was ready to begin charging.

[0158] Example 6 This embodiment provides a further detailed explanation of step 105 in Embodiment 1. It describes in detail how, after a successful drone lock-on and docking, the vehicle-mounted energy management system implements intelligent dynamic charging power allocation under complex constraints, including limited power of the vehicle's mobile power supply, significant fluctuations in the vehicle's own load, and varying drone battery states. This process ensures that the drone swarm can quickly replenish its power while simultaneously guaranteeing the electrical safety of the vehicle and other critical equipment.

[0159] In this embodiment, step 105 further includes the following sub-steps: 1051. The vehicle-mounted energy management system reads the battery management system data of each locked drone to obtain the current battery temperature, remaining power, and maximum allowable charging current.

[0160] Before any power transfer can be performed, the physical characteristics and safety boundaries of the load (i.e., the drone battery) must be accurately understood, because the battery’s ability to withstand charging current varies greatly depending on its condition.

[0161] After completing the interface docking in Example 5, the onboard energy management system establishes a handshake connection with the battery management system of each drone through the communication bus (such as CAN bus or RS-485 bus) in the charging interface. The system does not simply output full power, but first performs a comprehensive data polling.

[0162] (1) Data Acquisition: The system requests to read the key health parameters of each UAV, specifically including: Battery temperature: especially the highest and lowest individual cell temperatures. Temperature directly affects charging safety; low temperatures require preheating, and high temperatures require cooling.

[0163] Remaining charge: This refers to the state of charge. Low-charge batteries typically have a strong ability to accept high current, but as the charge increases, the current must be reduced to prevent overcharging.

[0164] Maximum permissible charging current: This is the safe upper limit calculated in real time by the vehicle energy management system (BMS) based on the battery aging level, temperature, and voltage.

[0165] Battery internal resistance and voltage: used to evaluate charging efficiency.

[0166] (2) Status classification: The vehicle energy management system classifies the drones to be charged into different categories based on the data read, such as: "cold drone" (temperature below 5 degrees, requires trickle preheating), "hot drone" (temperature above 45 degrees, requires cooling or stopping charging), "starved drone" (battery below 20%, requires fast charging) and "normal drone" (good status, allocated as needed).

[0167] For example, after the "Alpha" drone is locked, the onboard energy management system immediately reads its data: "Alpha's" current battery temperature is 25 degrees Celsius, within the optimal operating range; remaining battery power is 14%, indicating a "starved" state; the BMS reports the current maximum allowable charging current as 40 amps. Meanwhile, the locked "Beta" data shows: battery temperature 48 degrees Celsius, indicating a "warmed" state; remaining battery power 60%; due to excessive temperature, the BMS proactively limits the maximum allowable charging current to 5 amps. The system marks this data as the charging requirement profile for each drone.

[0168] 1052. The system monitors the output power margin of the vehicle's engine or generator set and the remaining power of the on-board auxiliary energy storage unit in real time, and calculates the upper limit of the total power that can be used to charge the drone.

[0169] The electrical energy on a vehicle is not unlimited; the vehicle's operation and communication equipment both require electrical energy. The charging system should not compromise the vehicle's core functions in order to charge the drone.

[0170] The vehicle energy management system is connected to the vehicle's main DC bus and monitors the energy flow of the entire vehicle's power grid in real time through high-precision voltage and current sensors.

[0171] Source monitoring: The system reads the real-time output power of the generator set driven by the vehicle's engine and obtains the remaining torque information fed back by the engine control unit to assess how much room for improvement the generator has. At the same time, it monitors the state of charge of the on-board auxiliary energy storage unit (such as a supercapacitor or auxiliary lithium battery).

[0172] Load statistics: The system summarizes the actual power consumption of all currently running high-energy-consuming devices in the vehicle (such as drive motors, communication equipment, air conditioning, and task computers).

[0173] Upper limit calculation: Maximum available total power = Current maximum stable output power of generator - Current load power of other equipment in the vehicle + Maximum instantaneous discharge power of auxiliary energy storage unit.

[0174] This upper limit is a dynamically changing physical quantity. For example, when the vehicle is stationary and high-energy-consuming equipment is off, the upper limit is very high; when the vehicle accelerates rapidly uphill and high-energy-consuming equipment is operating at full power, the upper limit may drop very low or even to zero. The system will set the total power valve of the drone swarm charging module based on this real-time upper limit.

[0175] For example, the recovery vehicle is stationary, the engine is idling, the generator has ample output power, and the auxiliary energy storage unit has sufficient charge. The system calculates that the maximum total power available for charging the drone is 10 kilowatts. Suddenly, the driver starts the vehicle and begins high-speed maneuvering, activating high-energy-consuming equipment, instantly increasing the vehicle's load. The system detects a slight drop in the bus voltage, immediately recalculates, and lowers the maximum available total power to 4 kilowatts to prioritize power supply for vehicle maneuvering and high-energy-consuming equipment.

[0176] 1053. Based on the battery status of each drone and the calculated total power limit, the charging current of each drone is dynamically adjusted using a power allocation algorithm. Under the premise of ensuring that the total power limit is not exceeded, the drone with the lowest battery level is given priority to be allocated the maximum charging power.

[0177] This step is a multi-objective optimization process that seeks the optimal charging efficiency under the conditions of satisfying both "safety constraints" and "total quantity constraints".

[0178] The system runs a real-time power allocation algorithm that runs every 100 milliseconds and adjusts based on the latest data.

[0179] Priority sorting: The algorithm first prioritizes drones based on their battery status. The typical weighting formula is: Priority = Base Weight × (1 - Remaining Battery Power) × Temperature Health Factor. This means that the lower the battery power, the higher the priority; abnormal temperatures (too high or too low) will lower the priority or limit power.

[0180] Allocation strategy: Phase 1 (Safety Satisfaction): Allocate the minimum current (such as trickle) required to maintain the battery activity of all drones.

[0181] Phase Two (Urgency Satisfaction): If there is still a remaining total power after allocating the minimum current, the system will prioritize increasing the charging current for the drone with the lowest remaining power until the maximum allowable charging current of that drone or the limit allowed by its battery temperature is reached.

[0182] Phase 3 (Balanced Allocation): If all low-battery drones are at full capacity and still have surplus power, it is allocated proportionally to other drones with higher battery levels.

[0183] Dynamic adjustment: If the total power limit in step 1052 suddenly decreases (e.g., vehicle acceleration), the algorithm will, in reverse priority, prioritize reducing the charging current of high-battery drones, or even temporarily interrupt the charging of high-battery drones, to fully protect the charging process of low-battery drones and the stability of the vehicle's power grid. The system precisely controls the output current by adjusting the duty cycle of the DC-DC converter.

[0184] For example, the system currently calculates a total power limit of 5 kilowatts.

[0185] "Alpha" drone: 14% battery, maximum allowable current 40A (approximately 1.5kW), highest priority.

[0186] "Beta" drone: 60% battery, maximum allowable overheating 5A (approximately 0.2kW), lowest priority.

[0187] "Gamma" drone: 25% battery, maximum allowable current 30A (approximately 1.1kW), medium priority.

[0188] The algorithm runs as follows: First, ensure the basic power supply. Second, fully support the "Alpha" UAV, allocating 40A (1.5kW).

[0189] Secondly, the "Gamma" drone is allocated 30A (1.1kW). Due to temperature limitations, the "Beta" drone can only be allocated 5A (0.2kW).

[0190] The current total power consumption is 2.8kW, with 2.2kW remaining. At this point, although "Gamma" has not yet reached its limit, in order to accelerate its recovery speed, the system decides to allocate all the remaining power to it (assuming its thermal design allows for short-term overcharging), increasing its current to 60A (2.2kW).

[0191] At this point, the total power is 2.8 + 2.2 = 5.0kW, which is just the upper limit.

[0192] Five seconds later, the vehicle began to accelerate, and the maximum total power output instantly dropped to 2kW. The algorithm responded immediately. Keep the current of UAV "Alpha" unchanged at 40A (1.5kW) (highest priority). Reduce the current of UAV "Gamma" from 60A to 12A (approximately 0.4kW). Keep the current of UAV "Beta" unchanged at 5A.

[0193] Total power consumption is 1.5 + 0.4 + 0.2 = 2.1kW (slightly exceeding the limit; the system will fine-tune the generator excitation to increase the output slightly, or continue to slightly reduce the Gamma current).

[0194] This millisecond-level dynamic adjustment ensured that the most urgently needed "Alpha" drone could regain its takeoff capability as quickly as possible, while also preventing the vehicle's power grid from overloading and tripping.

[0195] Example 7 This embodiment provides a further detailed explanation of step 1032 in Embodiment 4. It describes in detail how, in a vehicle-mounted mobile recovery scenario, a high-precision spatiotemporal conflict detection algorithm can be used to identify potential collision risks among multiple drones on a dynamic recovery path. Unlike static detection, this embodiment introduces a vehicle motion disturbance compensation mechanism to ensure the accuracy and reliability of the detection results under vehicle vibration and maneuvering conditions, providing a solid foundation for subsequent scheduling decisions.

[0196] In this embodiment, step 1032 further includes the following sub-steps: 10321. Discretize the dynamic recovery path of each UAV on the time axis to generate a four-dimensional path point sequence containing position coordinates and timestamps.

[0197] The core of this step lies in transforming the continuous, nonlinear dynamic path into a discrete data structure that can be processed by a computer. Since the motion of vehicles and drones is continuous, directly comparing two infinitely thin curves is computationally extremely complex; therefore, a discretization strategy must be adopted.

[0198] The scheduling module in the vehicle-mounted control center obtains the dynamic recovery path for each UAV from the path planner. These paths are typically high-order polynomials or spline curves. For collision detection, the system first sets a sampling time interval, usually 0.1 seconds or less, depending on the UAV's flight speed and the system's computing power. Shorter sampling intervals result in higher detection accuracy, but the computational load increases exponentially.

[0199] Timeline expansion: The system extends backward along the predicted recovery timeline, with the current moment as zero. For example, for a drone expected to land in 60 seconds, the system will mark a time point every 0.1 seconds within the time range of 0 to 60 seconds.

[0200] Coordinate mapping: For each time point, the system substitutes the UAV's path equation into the time variable to calculate the UAV's absolute coordinates in three-dimensional space at that moment. These coordinates are the absolute position in the Global Navigation Satellite System coordinate system.

[0201] Four-dimensional sequence generation: The system combines a timestamp with its corresponding three-dimensional coordinates into a quadruple, representing the UAV's location at position P_i at time t_i. All these quadruples, arranged in chronological order, constitute the UAV's four-dimensional waypoint sequence.

[0202] Bounding box construction: To further simplify calculations, the system also constructs a safe bounding box around each waypoint. Considering the size of the drone and sensor errors, this bounding box is typically a cuboid or sphere with a side length or radius larger than the physical size of the drone itself.

[0203] For example, the dynamic recovery path of UAV "001" is determined. The system sets the sampling interval to 0.1 seconds. At time T=0 seconds (current), its coordinates are calculated as (X1, Y1); at time T=0.1 seconds, its coordinates are calculated as (X2, Y2); at time T=0.2 seconds, its coordinates are calculated as (X3, Y4), and so on, until T=50 seconds (landing time). The system generates a sequence containing 500 points. At the same time, the system adds a virtual sphere with a radius of 3 meters around each coordinate point, representing the safe space occupied by UAV "001". In this way, the flight trajectory of UAV "001" is abstracted as a chain composed of 500 spatiotemporal spheres.

[0204] 10322. Calculate the Euclidean distance between any two drone path point sequences. When the Euclidean distance is less than a preset safety threshold and the time difference is less than a preset time threshold, it is determined that there is a spatial conflict between the two drones.

[0205] This embodiment employs a dual constraint of "space + time," meaning that a conflict is only determined when two drones are too close in both space and time. This avoids misjudging situations where drones "safely pass through the same airspace one after the other" as conflicts.

[0206] The system employs a double-loop traversal algorithm to compare drones in pairs. For any two drones A and B, the system obtains their path point sequences Seq_A and Seq_B, respectively.

[0207] Distance calculation: The system selects a point P_A(t_a) in Seq_A and a point P_B(t_b) in Seq_B, and calculates the straight-line distance between the two points using the Euclidean distance formula. This distance represents the distance between the centers of the two spacetime spheres.

[0208] Threshold determination: The system presets two key thresholds: Space safety threshold: typically the sum of the safety radii of two drones plus an additional buffer distance (e.g., 10 meters).

[0209] Time safety threshold: usually a few seconds (e.g., 3 seconds), representing the minimum time interval that the front and rear machines must maintain in order to prevent wake interference or insufficient response.

[0210] Logical judgment: If the calculated Euclidean distance is less than the spatial safety threshold, and |t_a - t_b| (the absolute value of the difference between the two timestamps) is less than the time safety threshold, a collision alarm is triggered. This means that two drones will appear in very close spatial locations at almost the same time, posing a significant risk of collision.

[0211] Conflict point marking: Once a conflict is identified, the system will record the time of the conflict, the spatial coordinates, and the IDs of the two drones involved, forming a conflict report.

[0212] For example, the system is comparing drones "001" and "002". During processing, it is found that the coordinates of drone "001" at T=30 seconds are (X3, Y3), while the coordinates of drone "002" at T=30.1 seconds are (X6, Y6).

[0213] The system calculates the Euclidean distance: the straight-line distance between the two points is 8 meters.

[0214] System check thresholds: The preset space safety threshold is 10 meters, and the time safety threshold is 3 seconds.

[0215] Judgment result: 8 meters is less than 10 meters, and 0.1 seconds (i.e., 30.1 - 30) is less than 3 seconds. Both conditions are met, and the system determines that a serious conflict exists. That is, "001" and "002" will collide in mid-air or have a dangerous approach at 30 seconds. The system immediately highlights this conflict area on the dispatch interface.

[0216] 10323. Based on the disturbance effect of vehicle movement on the spatial position of the recovery platform, the path point sequence is dynamically corrected, and conflict detection is re-performed on the corrected path until it is ensured that all paths do not overlap in time and space.

[0217] This step introduces a dynamic disturbance compensation mechanism, which is the key difference between this invention and static recovery technology. Because the vehicle is moving, especially during off-road driving with its bumps and pitches, it changes the actual position of the recovery platform in space, thus affecting the actual flight trajectory of the UAV relative to the ground. Ignoring these disturbances during detection could lead to serious misjudgments.

[0218] After the initial test is completed and preliminary scheduling results are generated, the system does not immediately make a decision, but instead enters a closed-loop correction process.

[0219] Disturbance data injection: The system obtains disturbance prediction data for future time periods from the vehicle dynamics model. This includes vertical displacement caused by road surface unevenness, roll angle changes caused by steering, and pitch angle changes caused by acceleration and deceleration. These disturbances are treated as high-frequency noise superimposed on the vehicle's trajectory.

[0220] Dynamic waypoint correction: The system maps these disturbance data onto the waypoint sequence of each UAV. Specifically, for each point in the path sequence, the system queries the predicted disturbance amount of the vehicle at that moment based on the timestamp corresponding to that point. If the vehicle makes an upward jump at that moment, then the waypoints of all UAVs that use that vehicle as a reference target will also be shifted upward in the absolute coordinate system accordingly. This correction simulates a scenario in the real physical world where the platform is moving and the guidance point moves accordingly.

[0221] Secondary detection and iteration: After jitter correction based on vehicle disturbances at all path points, the system re-runs the conflict detection algorithm in step 10322. This may uncover new, previously undetected conflicts. For example, two drones that were originally 5 meters apart in height may have their guidance paths suddenly lifted and overlapped with the other's due to a severe jolt from the vehicle.

[0222] Iterative Loop: If a conflict is detected, the system will trigger the scheduling algorithm to fine-tune the path or time window (as described in Example 4), generate a new path point sequence, inject perturbation data again, and detect again. This process continues until all path point sequences still meet the spatiotemporal safety threshold even with the maximum predicted perturbation.

[0223] For example, the initial detection showed that the paths of drones "001" and "002" were 5 meters apart vertically in space, which seemed safe. However, the system was injected with a prediction of vehicle disturbance 10 seconds in the future: the vehicle would run over a large rock in the 10th second, causing the rear of the vehicle (recovery platform) to bounce up 0.5 meters instantly, accompanied by a roll to the left.

[0224] The system then corrected the waypoints of all drones based on this information. After the correction, it was found that the vertical distance between the originally lower path of drone "001" and the originally higher path of drone "002" was shortened to 3 meters due to the vehicle's upward jump. At the same time, due to the vehicle's leftward tilt, the two paths also converged horizontally.

[0225] A secondary detection triggered an alarm: the corrected distance was less than the 10-meter threshold. The system recognized this as a potential dynamic conflict and decided to postpone the approach time of UAV "002" by 5 seconds. The system generated a new waypoint sequence and injected disturbances again for a third detection. This time, even with vehicle jolting, the safe distance between the two UAVs remained above 10 meters. The system confirmed that this scheme was safe and reliable in a dynamic environment and ultimately locked in this recovery and scheduling scheme.

[0226] Example 8 This embodiment provides a further detailed explanation of step 1053 in Embodiment 6. It describes in detail how, based on dynamic power allocation, the charging process of a single drone can be further refined through real-time monitoring of internal battery characteristics (such as internal resistance and temperature) and an advanced pulse charging strategy. This method not only ensures charging safety but also significantly improves charging efficiency and battery life, solving the problem of poor adaptability of traditional constant current and constant voltage charging methods in mobile vehicle environments.

[0227] In this embodiment, step 1053 further includes the following sub-steps: 10531. Set the battery temperature protection threshold. When the battery temperature of a drone is detected to exceed the protection threshold, set the charging power of the drone to zero or a very low trickle charging mode.

[0228] Lithium batteries are extremely sensitive to temperature. High temperatures can lead to thermal runaway or even fire, while low temperatures can cause lithium plating and damage the battery. Therefore, temperature protection is a prerequisite that takes precedence over all other charging strategies.

[0229] The onboard energy management system stores the safety temperature profiles for different types of lithium batteries. During charging, the system reads temperature sensor data from each drone's battery at a frequency of at least 10 Hz.

[0230] Threshold setting: Set two levels of protection thresholds. Level 1 is the warning threshold (e.g., 45°C), and Level 2 is the cut-off threshold (e.g., 55°C). Also set a low temperature threshold (e.g., 0°C).

[0231] When the temperature falls between the warning threshold and the cut-off threshold, the system considers the battery to be overheated. At this point, the charging power will be forcibly limited to an extremely low level (such as 0.1C, which is one-tenth of the capacity current), or charging will be stopped and the battery's built-in fan or the vehicle's liquid cooling system will be activated to dissipate heat until the temperature returns to a safe range.

[0232] When the temperature exceeds the cutoff threshold, the system determines it to be in serious danger, immediately sets the charging power to zero, completely cuts off the current circuit, and sends a high temperature alarm to the operator via the vehicle network.

[0233] When the temperature is below the low temperature threshold, the system will start a preheating program, using a small current pulse to heat the inside of the battery, and then enter the normal charging mode after the temperature rises.

[0234] For example, the "Beta" drone was undergoing high-power fast charging. Due to the limited heat dissipation conditions inside the vehicle's sealed compartment, coupled with the residual heat from the drone's battery during flight, its temperature rapidly increased. When the system detected that the drone's battery temperature reached 46 degrees Celsius, exceeding the first-level warning threshold, the system immediately triggered protection logic, instantly cutting off the original 30-amp charging current and switching to a 0.5-amp trickle charge. At this time, the vehicle's cooling fan powerfully blew air onto the drone's battery compartment. Two minutes later, the temperature dropped to 40 degrees Celsius, the system determined it was safe, and only then slowly restored the charging current.

[0235] 10532. A pulse charging strategy is adopted, in which a short discharge depolarization pulse is inserted during the charging cycle. The internal impedance of the battery is evaluated in real time by monitoring the voltage drop rate after the depolarization pulse, and the charging current amplitude of the next cycle is adjusted accordingly.

[0236] Traditional constant-current charging can cause concentration polarization and ohmic polarization inside the battery, resulting in a rapid increase in terminal voltage, which may mistakenly indicate that the battery is fully charged and thus limit the current. Pulse charging eliminates this polarization by pausing and reversing the discharge, allowing the battery to accept a larger current.

[0237] The vehicle charger adopts a cyclic waveform of "positive charging - charging stop - negative depolarization".

[0238] Depolarization pulse insertion: After a 10-second positive high-current charging cycle, the system pauses charging for 0.5 seconds, followed by a 0.1-second negative low-current pulse (discharge pulse). This negative pulse neutralizes the accumulated charge on the battery electrode surface, eliminates the polarization effect, and causes the battery terminal voltage to drop rapidly, restoring its true potential.

[0239] Internal resistance assessment: At the instant the negative pulse ends, the system captures the battery voltage change curve at a microsecond-level sampling rate. The battery's internal impedance (internal resistance) can be estimated using the formula R = ΔV / ΔI, where ΔV is the voltage drop and ΔI is the current value of the negative pulse.

[0240] Dynamic adjustment: Battery internal resistance is the best indicator of battery health and charge acceptance. If the measured internal resistance is low, it indicates that the battery is in good condition and can accept a large current. The system will appropriately increase the amplitude of the forward charging current in the next cycle. If the measured internal resistance increases, it indicates that the resistance to the internal chemical reaction of the battery has increased or that it is close to being fully charged. The system will then reduce the charging current to avoid generating excessive Joule heat.

[0241] For example, the "Alpha" drone is currently in pulse charging mode. The charger first outputs 40 amps of current for 10 seconds, then stops. Immediately afterward, the system outputs 5 amps of current in reverse for 0.1 seconds (depolarization pulse). At this instant, the system's voltage monitoring circuit records a drop in battery voltage from 16.5 volts to 16.3 volts (a drop of 0.2 volts). The system calculates: internal resistance R = 0.2V / 5A = 0.04 ohms. This internal resistance is very low, indicating that the "Alpha" drone's battery has excellent internal activity. Therefore, in the next 10-second charging phase, the system boldly increases the current to 45 amps. As charging progresses, the subsequently detected internal resistance gradually rises to 0.1 ohms, and the system then gradually reduces the current to 20 amps, 10 amps, until fully charged.

[0242] 10533. Establish a correlation model between charging efficiency and battery temperature. During the process of dynamically adjusting the charging current, maintain the charging efficiency within the preset high-efficiency range, while ensuring that the battery temperature rise rate is controlled within a safe range.

[0243] Simply fast charging may result in low current utilization (which turns into heat), while being overly conservative wastes time. This embodiment seeks the optimal balance between speed and safety / efficiency by establishing a correlation model.

[0244] The vehicle energy management system incorporates a three-dimensional mapping table or neural network model of current-temperature-efficiency. This model is trained based on a large amount of battery experimental data.

[0245] Model inputs: real-time battery temperature, current charging current, ambient temperature.

[0246] Model prediction: The model outputs the estimated charging efficiency (how much electrical energy is converted into chemical energy and how much into heat energy) and the estimated temperature rise rate (how many degrees the temperature rises per minute) under the current operating conditions.

[0247] Closed-loop control: System set target constraints: Constraint A: Charging efficiency must be higher than 90% (to avoid energy waste).

[0248] Constraint B: The rate of temperature rise must be less than one degree Celsius per minute (to prevent thermal runaway).

[0249] Constraint C: Total power shall not exceed the vehicle’s available upper limit.

[0250] Decision execution: If the current causes the predicted efficiency to fall below 90% (meaning excessive heat generation), the system will proactively reduce the current, even if the battery is not fully charged, to bring the operating point back to the high-efficiency zone. Conversely, if the temperature rise is slow and efficiency is high, the system will attempt to slightly increase the current to tap into its potential.

[0251] For example, the "Gamma" drone is charging, and the current ambient temperature is 35 degrees Celsius, which is considered high. The system calls the relevant model for evaluation. If the current is maintained at 30 amps, the model predicts: charging efficiency is 85% (due to significant heat loss from internal resistance), and the temperature rise rate is 1.5 degrees Celsius per minute (too fast, risky). The system determines this operating condition is unacceptable.

[0252] The system attempted to reduce the current to 20 amps. The model predicted that charging efficiency would increase to 95% and the temperature rise rate would decrease to 0.8 degrees Celsius per minute. The system decided to lock 20 amps as the optimal charging current. Five minutes later, the battery cooling system of the "Gamma" drone took effect, the temperature dropped, and the model recalculated, showing that efficiency could still be maintained at 92% at 25 amps. The system then automatically increased the current to 25 amps, achieving precise management of the charging process that was both fast and safe.

[0253] Example 9 This embodiment details the specific hardware architecture, software module division, and collaborative working mechanism among the subsystems of a dense recovery system for a fixed-wing UAV swarm based on a vehicle-mounted platform. This system is applied to the dense recovery methods for fixed-wing UAV swarms described in Implementations 1-8, aiming to solve the engineering challenges of simultaneously receiving, precisely guiding, reliably locking, and efficiently replenishing energy for multiple fixed-wing UAVs on a mobile, highly dynamic vehicle platform.

[0254] (1) System overall architecture design The fixed-wing UAV swarm dense recovery system provided in this embodiment is integrated onto a highly mobile wheeled or tracked special vehicle chassis. This vehicle serves not only as a transportation tool but also as an aerial mobile command and refueling station. The system is physically divided into four main functional areas: a driving and control area, a central data processing area, a recovery operation area, and an energy and power area.

[0255] Driving control area: Includes the vehicle driving control terminal and the monitoring instrument panel of the recycling system, allowing operators to intuitively view the recycling status.

[0256] Central Data Processing Area: Equipped with a ruggedized vehicle-mounted control center cabinet, containing high-performance computing servers, data loggers, and communication switches.

[0257] Recovery operation area: Located at the rear or top of the vehicle, it is equipped with an antenna array of the vehicle-mounted multi-channel guidance system and a physical deck of the vehicle-mounted recovery platform.

[0258] Energy and Power Zone: Includes the power distribution cabinet for the vehicle's main engine, generator set, on-board auxiliary energy storage unit, and on-board energy management system.

[0259] The entire system adopts a distributed bus architecture, using gigabit Ethernet and vehicle CAN bus to connect various subsystems, ensuring low latency and high reliability of data transmission.

[0260] (2) Vehicle control center The vehicle-mounted control center is used to receive recovery request information sent by the drone swarm, and to allocate a unique guidance channel and initial recovery access point to each drone requesting recovery based on the current status information of the drone swarm and the real-time movement information of the vehicle; it is also used to execute a dense recovery scheduling algorithm to determine the recovery order and time window of the drone swarm based on the real-time spatial location of each drone, the remaining battery power, and the conflict detection results of the recovery path.

[0261] In this embodiment, the vehicle control center is responsible for overall decision-making and task scheduling.

[0262] Hardware Architecture: The core utilizes an industrial-grade server with a multi-core processor architecture, possessing powerful floating-point computing capabilities and supporting parallel processing of multiple drone data streams. The server is equipped with a ruggedized hard disk array for recording flight data and recovery logs. Communication interface cards support shortwave, VHF, data link, and 5G / 4G cellular network communication, ensuring beyond-line-of-sight connectivity with the drone swarm.

[0263] Software modules: running on a real-time operating system, mainly including a task management module, a communication protocol stack, a situational awareness display module, and a core intensive recycling scheduling engine.

[0264] Task Management Module: Responsible for parsing the recovery requests sent by the drone, and parsing information such as identification, location, battery level, and payload status.

[0265] Dense Recovery Scheduling Engine: This is the core software component of the control center, corresponding to the scheduling function in the dense recovery method for fixed-wing UAV swarms. It incorporates an optimization model based on genetic algorithms or ant colony algorithms. When multiple UAV requests are received, the engine comprehensively considers the real-time motion information of the vehicle (such as speed, acceleration, heading, and turbulence) and the UAV status to calculate the optimal recovery sequence. For example, it determines which UAV has the most urgent battery charge, which UAV is closest, and, combined with the vehicle's future trajectory, allocates a time window to each UAV that will not cause spatiotemporal conflicts.

[0266] The workflow of the vehicle control center is as follows: The control center first monitors the airspace and immediately decodes any received recovery request data packets. Then, the scheduling engine activates, using vehicle dynamics models to predict the vehicle's position envelope over the next 30 to 60 seconds. Based on this envelope and the drone's current position, the system assigns each drone a unique logical channel number (e.g., channel 01 to channel 08) and virtual initial access point coordinates. Finally, the control center sends the assignment results to the corresponding drone via uplink and simultaneously forwards relevant parameters to the onboard multi-channel guidance system and the onboard recovery platform, preparing them for operation.

[0267] (3) Vehicle-mounted multi-channel guidance system The vehicle-mounted multi-channel guidance system is used to send high-precision relative positioning signals to the UAV through the allocated guidance channels, and receive the relative position information transmitted back by the UAV. Combined with the vehicle's movement trajectory, it generates a dynamically adjusted recovery path for each UAV.

[0268] In this embodiment, the vehicle-mounted multi-channel guidance system is responsible for providing centimeter-level relative positioning services during the approach phase of the UAV.

[0269] Hardware Configuration: The system mainly consists of a phased array antenna array, a radio frequency transceiver unit, a baseband signal processing unit, and a time synchronization unit. The phased array antenna array is typically installed on the top or rear of the vehicle near the recovery platform. It employs a flat panel array design, providing excellent aerodynamic characteristics and reducing wind resistance at high speeds. The time synchronization unit uses an atomic clock or a GPS / BeiDou timing module to ensure the entire system's time accuracy reaches the nanosecond level.

[0270] The working principle of the vehicle-mounted multi-channel guidance system is as follows: Signal Transmission: After the control center allocates channels, the baseband processing unit of the guidance system generates a high-precision ranging signal with a specific pseudo-random code based on the channel number. The phased array antenna forms a directional beam by electronically controlling the phase of each array element, precisely pointing it at the target UAV. This beamforming technology allows the system to serve multiple UAVs simultaneously without interference between channels.

[0271] Relative positioning: After receiving the signal, the UAV measures the propagation delay and phase difference to calculate its position relative to the vehicle, and then sends it back to the vehicle via downlink. The onboard guidance system receives this transmitted data and, combined with the vehicle's own inertial navigation data, calculates the UAV's high-precision relative position in the vehicle's coordinate system.

[0272] Path guidance: Based on high-precision relative position, the guidance system generates a dynamic recovery path and converts it into guidance commands that are continuously sent to the UAV to guide it to the recovery platform.

[0273] (4) Vehicle-mounted recycling platform The vehicle-mounted recovery platform is used to control the mechanical structure to physically lock and interface with the drone based on the landing signal triggered by the drone. In this embodiment, the vehicle-mounted recovery platform is the physical location where the drone finally lands and is also the key mechanical structure for achieving secure locking.

[0274] Mechanical Structure: The platform's main body is constructed from high-strength aluminum alloy or carbon fiber composite materials, featuring lightweight and high rigidity. Visual aids (such as high-contrast ArUco QR codes) are mounted on the platform surface for final visual alignment of the drone. An integrated active damping system uses hydraulic or electromagnetic active suspension to counteract vertical bumps generated by vehicle movement, providing the drone with a relatively stable virtual horizontal plane.

[0275] The locking and docking mechanism includes the following components: Mechanical locking: Electromagnets and mechanical grippers are arranged around the platform. The electromagnets provide the initial attraction force to prevent the drone from slipping; the mechanical grippers use a linkage mechanism driven by a servo motor, which can close within milliseconds after the drone touches the ground, locking the drone's landing gear or fuselage reinforcement to achieve a rigid connection.

[0276] Electrical Connection: A telescopic probe interface is located in the center of the platform. The probe adopts a floating design, allowing for a certain degree of horizontal alignment error. After mechanical locking is confirmed, the probe is lifted by a cylinder and inserted into the charging socket on the bottom of the drone to achieve electrical connection.

[0277] Sensor integration: The platform surface is embedded with a pressure sensor array and an infrared proximity sensor to sense the drone's landing position and attitude in real time. Once a landing signal is detected, the locking procedure is immediately triggered.

[0278] (5) On-board energy management system The vehicle-mounted energy management system is used to dynamically allocate charging power based on the battery status of the docked drone and the overall energy load of the vehicle, in conjunction with the vehicle-mounted control center.

[0279] In this embodiment, the on-board energy management system is responsible for providing safe and efficient charging services for the drone swarm under dynamic vehicle operating conditions.

[0280] Circuit Architecture: The system adopts a high-voltage DC bus power supply architecture. The main power is provided by a generator driven by the vehicle engine, supplemented by a high-power supercapacitor or lithium battery pack as energy storage buffer. Each charging channel is equipped with an independent AC / DC or DC / DC converter.

[0281] The management logic of the vehicle energy management system is as follows: The energy management system communicates with the control center via a CAN bus. It monitors the battery status (such as voltage, current, and temperature) of each docked drone in real time, as well as the vehicle's own electrical load (such as drive motors, communication equipment, and air conditioning). Based on the control center's scheduling instructions, the energy management system dynamically adjusts the output power of each channel. For example, when the vehicle accelerates rapidly and the generator power is strained, the system will automatically reduce the charging power of non-emergency drones, prioritizing the charging needs of vehicle operation and critical drones.

[0282] (6) Examples of collaborative work scenarios in the system Suppose a recovery vehicle equipped with this system is on a mission in the field, and three fixed-wing drones, "001", "002", and "003", apply for recovery.

[0283] Step 1: The vehicle control center receives the request and calculates that "001" has the lowest battery level and is the closest. The dispatch engine decides that "001" has priority and is assigned channel 01; "002" is second and is assigned channel 02; "003" hovers and waits.

[0284] Step 2: The multi-channel guidance system activates the beams of channels 01 and 02, pointing them towards "001" and "002" respectively. Following the guidance instructions, the two drones enter from the left and right sides of the vehicle respectively, forming a dual-path parallel approach.

[0285] Step 3: When "001" flies above the platform, the vision system identifies the marker, adjusts its attitude, and lands. The platform sensors detect the ground contact signal, the robotic gripper locks instantly, and the probe rises to dock.

[0286] Step 4: The energy management system detects that "001" has successfully connected, reads its battery level as 10%, and its temperature as 20°C. At this time, the vehicle is traveling at a constant speed with ample power, and the system allocates the maximum charging current to "001".

[0287] Step 5: Subsequently, "002" also successfully landed at another predetermined location (or at another time at the same location), and the system completed the recovery and charging of all drones in sequence as planned.

[0288] As can be seen from the above embodiments, this system, through its highly integrated hardware architecture and intelligent software scheduling, perfectly realizes the dense, safe, and efficient recovery and resupply of multiple drones on mobile vehicles.

[0289] Example 10 This embodiment focuses on the detailed technical implementation of the core components of the vehicle-mounted multi-channel guidance system—the phased array antenna array and the path generation module. This is a key subsystem that ensures that multiple UAVs can achieve mutual non-interference, accurate positioning, and dynamic tracking in complex electromagnetic environments and under high vehicle maneuverability.

[0290] (1) Hardware implementation of phased array antenna array In this embodiment, the core component of the vehicle-mounted multi-channel guidance system is an active phased array antenna array that operates in the microwave frequency band (e.g., C-band or X-band).

[0291] Array Layout: The antenna array is designed as a flat rectangular structure, approximately 600mm × 400mm in size, and is installed inside the roof fairing in front of the vehicle recovery platform. The array consists of hundreds of microstrip patch antenna elements arranged in a rectangular grid. To achieve signal transmission and reception isolation, the array employs a separate transmission and reception design or a time-division multiplexing design, with a low-noise amplifier and a power amplifier connected to the rear end of each element.

[0292] Beam control network: The core of the array is the beam control network, which contains multiple digital phase shifters and attenuators. By controlling the feed phase of each antenna element through an FPGA (Field-Programmable Gate Array), a beam pointing in a specific direction can be synthesized in space. For example, to point the beam at a 30-degree elevation angle to the left front, the system calculates the required phase difference for each element and controls the phase shifters to perform the corresponding phase shift. This electronic scanning method is significantly superior to traditional mechanically rotating antennas, achieving scanning speeds in the microsecond range and enabling the simultaneous tracking of dozens of high-speed moving targets.

[0293] Multi-channel isolation technology: To support the simultaneous recovery of multiple UAVs, the system employs spatial division multiple access (SDMA) technology. Through beamforming algorithms, the system can form multiple independent narrow beams (e.g., beamwidth less than 5 degrees) in space, each pointing towards a different UAV. Simultaneously, nulls are formed in non-target directions, significantly reducing signal interference between channels. This design ensures that even when "001" and "002" are very close together, the guidance signals they receive remain clear and independent, without crosstalk.

[0294] (2) Beamforming and signal transmission mechanism Guidance signal generation: The baseband signal processing unit is responsible for generating the ranging signal. This system uses an ultra-wideband (UWB) pulse signal as the carrier. UWB signals have extremely narrow pulse widths (nanosecond level) and extremely high time resolution, enabling centimeter-level ranging accuracy. Each guidance channel is assigned a unique pseudo-random noise (PN) code sequence. Before transmission, information such as the channel number, timestamp (accurate to nanoseconds), and vehicle attitude angle are modulated onto the carrier.

[0295] Adaptive beam tracking: The system is not statically pointing, but possesses closed-loop tracking capability. During its approach, the UAV continuously transmits response signals to the vehicle. Upon receiving these signals, the onboard array uses digital beamforming technology to precisely calculate the UAV's angular position based on the phase difference of the signal reaching different array elements. The controller updates the phase shifter's control word in real time based on this angular position, ensuring the beam remains locked onto the UAV. Even if the vehicle experiences severe swaying due to rough terrain, the roof-mounted array uses data from the inertial measurement unit for attitude compensation, maintaining absolute stability of the beam pointing in space.

[0296] (3) Internal logic and algorithm of the path generation module The path generation module is the bridge connecting positioning perception and flight control. Its core task is to transform discrete positioning points into smooth, flyable, and dynamic trajectories.

[0297] Input data preprocessing: The module receives raw relative positioning data (NED coordinate system: NE-G) from the guidance system. This data typically contains high-frequency noise. The module first uses a Kalman filter algorithm, combined with the vehicle's own motion model (such as a uniform speed or uniform acceleration model) and the UAV's kinematic constraints (maximum overload, maximum climb rate), to smooth the position data, filter out outliers, and output the optimal state estimation vector.

[0298] Vehicle Motion Compensation: Since the recovery platform moves with the vehicle, directly planning the path using absolute coordinates could lead to the drone rear-ending or colliding with the vehicle. The path generation module introduces a relative motion compensation algorithm. It reads wheel speedometer and IMU data from the vehicle chassis to calculate the vehicle's displacement vector within the predicted future timeframe. Then, it inversely transforms the predicted position of the drone in the absolute coordinate system to a volume coordinate system with the vehicle's center as the origin. In this way, the planned path is a virtual pipeline that is stationary (or slowly changing) relative to the vehicle, greatly simplifying drone tracking.

[0299] Spline interpolation path generation: In volume coordinates, the module determines the starting point (the UAV's current position), intermediate points (waypoints used for altitude adjustment and wake avoidance), and the ending point (the center of the recovery platform). A smooth curve connecting these points is generated using either cubic B-spline (B-Spline) or fifth-order polynomial interpolation algorithms. The B-spline curve has a continuous second derivative, meaning the generated path curvature changes continuously. This eliminates the need for frequent rudder maneuvers when the UAV tracks the path, resulting in smoother flight and effectively reducing aerodynamic wear.

[0300] Dynamic refresh mechanism: Path generation is a real-time, cyclical process with a refresh rate as high as 50Hz. This means the system replans the path every 0.02 seconds. This high-frequency refresh makes the path highly flexible. For example, when the vehicle suddenly makes a sharp left turn, the body coordinate system rotates accordingly, and the path planner immediately calculates a new curve. From an external observer's perspective, the recovery path guides the drone to naturally follow the vehicle's turn, always accurately pointing towards the platform's center.

[0301] For example, suppose a vehicle is traveling at 60 km / h and is preparing to recover two drones.

[0302] Initialization: The path generation module receives the scheduling instruction and establishes task threads for "001" of channel 01 and "002" of channel 02.

[0303] Guidance and direction setting: The phased array simultaneously forms two beams. Beam A points to "001" and transmits a UWB pulse with PN code A; beam B points to "002" and transmits a UWB pulse with PN code B.

[0304] Solution and Planning: The UAV calculates the relative position and transmits it back. The path generation module uses Kalman filtering to detect that position "001" has deviated 2 meters to the left from the planned route. At the same time, the vehicle's IMU detects that the front of the vehicle is turning to the right.

[0305] Dynamic Correction: The path generation module immediately runs the compensation algorithm. First, it calculates the angle of rotation of the body coordinate system caused by the vehicle's right turn. Then, in the new body coordinate system, it re-plans a curve that smoothly transitions from the current position "001" to the center of the platform. This curve is shifted to the right by an angle compared to the curve of the previous cycle, and an additional roll correction is added to the right, guiding the drone to cut in at a lower slope.

[0306] Command issuance: The latest path point sequence is encapsulated into control commands and sent to "001" and "002" respectively via channels 01 and 02. The UAV flight control system adjusts the control surfaces according to the commands, accurately flying towards the constantly moving landing point.

[0307] Through the technical solution of this embodiment, the vehicle-mounted multi-channel guidance system utilizes advanced phased array technology and complex path planning algorithms to overcome various interferences during mobile platform recovery, achieving high-precision guidance in parallel multi-channel mode, and providing the last and most critical guarantee for the safe landing of UAVs.

[0308] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions 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 the present invention.

Claims

1. A method for dense recovery of a fixed-wing unmanned aerial vehicle (UAV) swarm, characterized in that, Includes the following steps: The vehicle control center receives recovery request information from the drone swarm and assigns a unique guidance channel and initial recovery access point to each drone requesting recovery based on the current status information of the drone swarm and the real-time movement information of the vehicle. The vehicle-mounted multi-channel guidance system sends high-precision relative positioning signals to the UAV through the allocated guidance channels. After receiving the signals, the UAV calculates its real-time spatial position relative to the vehicle. The vehicle-mounted control center generates a dynamically adjusted recovery path for each UAV based on the real-time spatial position and the vehicle's movement trajectory. The vehicle-mounted control center executes a dense recovery scheduling algorithm. Based on the real-time spatial location of each drone, the remaining battery power, and the conflict detection results of the recovery path, it determines the recovery order and time window of the drone swarm and sends the recovery order and time window to the corresponding drones. The drone flies to the vehicle-mounted recovery platform along a dynamically adjusted recovery path according to the recovery sequence and time window. The vehicle-mounted recovery platform controls the mechanical structure to physically lock and interface with the drone based on the landing signal triggered by the drone. The vehicle-mounted energy management system dynamically allocates charging power based on the battery status of the docked drone and the overall energy load of the vehicle. After charging is completed, it updates the drone status according to mission requirements and either releases the drone or keeps it parked.

2. The method for dense recovery of a fixed-wing UAV swarm according to claim 1, characterized in that, The vehicle-mounted control center receives recovery request information from the drone swarm and, based on the current status information of the drone swarm and the real-time movement information of the vehicle, assigns a unique guidance channel and initial recovery access point to each drone requesting recovery, including: When each drone in the drone swarm detects that its battery level is below a preset threshold or receives a ground command, it sends a recovery request to the vehicle control center, which includes its identification, current location coordinates, remaining battery level, and current flight speed. After receiving the recovery request information, the vehicle control center, based on the vehicle's current speed, direction of travel, and body posture and vibration status, predicts the vehicle's trajectory envelope within a preset time period in the future. Based on the motion trajectory envelope and the current position coordinates of the drone swarm, the estimated arrival time of each drone to the vehicle recovery area is calculated. Then, according to the order of the estimated arrival time or the preset priority rules, each drone is assigned a non-interfering guidance channel number and the corresponding initial access point spatial coordinates.

3. The method for dense recovery of a fixed-wing UAV swarm according to claim 1, characterized in that, The vehicle-mounted multi-channel guidance system sends high-precision relative positioning signals to the UAV through allocated guidance channels. After receiving the signals, the UAV calculates its real-time spatial position relative to the vehicle. Based on the real-time spatial position and the vehicle's movement trajectory, the vehicle control center generates a dynamically adjusted recovery path for each UAV, including: The vehicle-mounted multi-channel guidance system uses beamforming technology to send ultra-wideband pulse signals or laser guidance signals with timestamps and angle information to the corresponding UAV on each guidance channel. After receiving the guidance signal, the UAV calculates its three-dimensional relative coordinates with respect to the vehicle coordinate system by using the signal propagation delay and phase difference, and combines it with its own inertial navigation data for filtering and fusion to obtain a high-precision relative position estimate. The vehicle control center receives the high-precision relative position estimate transmitted back by the UAV, combines it with the position compensation amount generated by the real-time movement of the vehicle, and uses a spline interpolation algorithm to generate a dynamic recovery path with the initial access point as the starting point and the vehicle recovery platform locking point as the ending point.

4. The method for dense recovery of a fixed-wing UAV swarm according to claim 1, characterized in that, The vehicle-mounted control center executes a dense recovery scheduling algorithm, which determines the recovery order and time window of the drone swarm based on the real-time spatial location of each drone, its remaining battery power, and the conflict detection results of the recovery path. The vehicle-mounted control center constructs a scheduling optimization model with the objective functions of minimizing recovery time and balancing battery consumption. The input parameters include the real-time spatial location of all drones to be recovered, the remaining battery power, and the generated dynamic recovery path. The scheduling optimization model was used to detect spatiotemporal conflicts in the dynamic recovery paths of all UAVs to be recovered, and to identify path segments that overlapped in three-dimensional space and time. Based on the collision detection results and the priority of remaining battery power, the path replanning or time window offset is performed on drones with overlapping paths, generating a recovery sequence table containing the precise recovery time and flight speed adjustment instructions for each drone.

5. The method for dense recovery of a fixed-wing UAV swarm according to claim 1, characterized in that, The UAV flies to the vehicle-mounted recovery platform along a dynamically adjusted recovery path according to the recovery sequence and time window. The vehicle-mounted recovery platform controls the mechanical structure to physically lock and interface with the UAV based on the landing signal triggered by the UAV, including: The drone flies to a hovering point above the recovery platform, uses visual sensors to identify the marked graphics on the recovery platform, adjusts its attitude to make the belly of the drone parallel to the plane of the recovery platform, and slowly descends until the landing gear touches the ground and triggers the landing signal. After receiving the landing signal, the vehicle-mounted recovery platform immediately activates the electromagnetic locking device or mechanical gripper on the platform surface to quickly and physically fix the drone in the bumpy environment of the vehicle, preventing the drone from slipping or overturning. After the physical locking is completed, the vehicle-mounted recovery platform controls the telescopic interface probe to rise upwards and blindly connect with the charging interface at the bottom of the drone, and confirms the reliability of the electrical connection through the interface detection circuit.

6. The method for dense recovery of a fixed-wing UAV swarm according to claim 1, characterized in that, The onboard energy management system implements dynamic charging power allocation based on the docked drone's battery status and the vehicle's overall energy load, including: The vehicle-mounted energy management system reads data from the battery management system of each locked drone to obtain the current battery temperature, remaining power, and maximum allowable charging current. The system monitors the output power margin of the vehicle's engine or generator set and the remaining power of the on-board auxiliary energy storage unit in real time, and calculates the maximum total power that can be used to charge the drone. Based on the battery status of each drone and the calculated total power limit, the charging current of each drone is dynamically adjusted using a power allocation algorithm. Under the premise of ensuring that the total power limit is not exceeded, the drone with the lowest battery level is given priority in being allocated the maximum charging power.

7. The method for dense recovery of a fixed-wing UAV swarm according to claim 4, characterized in that, The method of using this scheduling optimization model to perform spatiotemporal conflict detection on the dynamic recovery paths of all UAVs to be recovered, and identifying path segments that overlap in three-dimensional space and time, includes: The dynamic recovery path of each drone is discretized on the time axis to generate a four-dimensional path point sequence containing position coordinates and timestamps; Calculate the Euclidean distance between any two drone path point sequences. If the Euclidean distance is less than a preset safety threshold and the time difference is less than a preset time threshold, it is determined that the two drones have a spatial conflict. Based on the disturbance effect of vehicle movement on the spatial position of the recovery platform, the path point sequence is dynamically corrected, and conflict detection is re-performed on the corrected path until it is ensured that all paths do not overlap in time and space.

8. The method for dense recovery of a fixed-wing UAV swarm according to claim 6, characterized in that, The step of dynamically adjusting the charging current of each drone based on its battery status and the calculated total power limit using a power allocation algorithm includes: Set a battery temperature protection threshold. When the battery temperature of a drone is detected to exceed the protection threshold, set the charging power of the drone to zero or a very low trickle charging mode. A pulse charging strategy is adopted, in which a short discharge depolarization pulse is inserted during the charging cycle. The internal impedance of the battery is evaluated in real time by monitoring the voltage drop rate after the depolarization pulse, and the charging current amplitude of the next cycle is adjusted accordingly. Establish a correlation model between charging efficiency and battery temperature. During the process of dynamically adjusting the charging current, maintain the charging efficiency within the preset high-efficiency range, while ensuring that the battery temperature rise rate is controlled within a safe range.

9. A dense recovery system for a fixed-wing unmanned aerial vehicle swarm, applied to mobile vehicles, characterized in that, include: The vehicle-mounted control center is used to receive recovery request information sent by the drone swarm, and to allocate a unique guidance channel and initial recovery access point to each drone requesting recovery based on the current status information of the drone swarm and the real-time movement information of the vehicle; it is also used to execute a dense recovery scheduling algorithm to determine the recovery order and time window of the drone swarm based on the real-time spatial location of each drone, the remaining battery power, and the conflict detection results of the recovery path. The vehicle-mounted multi-channel guidance system is used to send high-precision relative positioning signals to the UAV through the allocated guidance channels and receive the relative position information transmitted back by the UAV. Combined with the vehicle's movement trajectory, it generates a dynamically adjusted recovery path for each UAV. The vehicle-mounted recovery platform is used to control the mechanical structure to physically lock and interface with the drone based on the landing signal triggered by the drone. The vehicle-mounted energy management system is used to dynamically allocate charging power based on the battery status of the docked drone and the overall energy load of the vehicle, in conjunction with the vehicle-mounted control center.

10. The dense recovery system for a fixed-wing UAV swarm according to claim 9, characterized in that, The vehicle-mounted multi-channel guidance system includes a phased array antenna array and a path generation module; The phased array antenna array is used to send ultra-wideband pulse signals or laser guidance signals with timestamps and angle information to the corresponding UAV on each guidance channel through beamforming technology. The path generation module is used to receive the high-precision relative position estimate calculated by the UAV based on the guidance signal, combine it with the position compensation amount generated by the real-time movement of the vehicle, and use spline interpolation algorithm to generate a dynamic recovery path with the initial access point as the starting point and the vehicle recovery platform locking point as the ending point.

Images