An edge-computing-based unmanned aerial vehicle cluster scheduling and anti-collision method

Abstract

The application discloses an unmanned aerial vehicle cluster cooperative task scheduling and anti-collision method based on edge computing driving. The method comprises the following steps: collecting sensing data by an unmanned aerial vehicle and unloading the sensing data to an edge computing MEC node; making centralized anti-collision decision by the MEC node, and generating avoidance instructions; making global cooperative scheduling by the MEC node, and allocating conflict-free air routes for cluster unmanned aerial vehicles; and issuing instructions by the MEC node, and dynamically adjusting air routes of other associated unmanned aerial vehicles. The application can effectively improve the operation safety and efficiency of the unmanned aerial vehicle cluster and enhance the communication robustness by unloading the decision task to the network edge and utilizing the low-latency characteristics of the 5G-A network to realize centralized cooperative control.

Description

Technical Field

[0001] This application belongs to the field of unmanned aerial vehicle (UAV) control, communication and edge computing technology, specifically involving a method, system and related equipment for collaborative task scheduling and real-time collision avoidance control of large-scale UAV swarms using 5G-A sensing integration and edge computing MEC technology. Background Technology

[0002] With the rapid development of the low-altitude economy, drones are increasingly being used in logistics, power line inspection, and agricultural plant protection, and large-scale drone swarm operations have become an important trend. However, existing technologies still face many challenges in supporting the efficient and safe operation of large-scale drone swarms.

[0003] The bottleneck of single-machine computing power is prominent. Due to limitations in size, power consumption, and cost, the computing power of onboard processors for drones is limited. When faced with complex flight environments, especially in sudden situations requiring real-time obstacle avoidance path planning, the onboard computing power is insufficient to handle high-intensity computing tasks, leading to decision delays and significantly increasing the risk of collisions.

[0004] Excessive network communication latency. Traditional 4G or ordinary 5G networks typically have end-to-end communication latency between 10 milliseconds and 300 milliseconds. For high-speed flying drones, collision avoidance control is a typical latency-sensitive service. Excessive network latency prevents remote control commands from being delivered in a timely manner, which may cause the drone to miss the optimal avoidance window when it receives the avoidance command, resulting in an accident.

[0005] Lack of swarm collaboration capabilities. Most existing drones adopt a single-drone autonomous planning or simple ground station control mode, lacking a global airspace situational awareness and collaborative scheduling mechanism. In scenarios where multiple drones operate intensively, each drone plans its own route independently, which easily leads to route conflicts and mid-air collisions, severely restricting the application scale and operational density of drone swarms.

[0006] Therefore, how to solve the technical pain points of insufficient onboard computing power, high network latency, and lack of cluster collaborative scheduling in UAVs is an urgent technical problem to be solved in the current UAV technology field. Summary of the Invention

[0007] To address the problems mentioned in the background art, this application provides a method, system, and device for collaborative task scheduling and collision avoidance of UAV swarms based on edge computing. The core idea is to offload the high-computing-power-consuming decision-making and planning tasks from the UAV to the network edge, and utilize the ultra-low latency characteristics of the 5G-A network to achieve centralized and collaborative intelligent control, thereby effectively improving the operational safety and efficiency of UAV swarms.

[0008] To achieve the above objectives, this application provides a method for collaborative task scheduling and collision avoidance of UAV swarms based on edge computing, including: The data acquisition and offloading steps involve the 5G-A module and sensors on the drone acquiring real-time perception data of surrounding obstacles; the drone then offloads the perception data to an edge computing MEC node deployed on the side of the 5G-A base station via the 5G-A network. The centralized collision avoidance decision-making process involves the MEC node receiving and analyzing the perception data to classify the risk level of the obstacle; when the obstacle is determined to be at a high risk level, the MEC node generates an emergency avoidance command. The global collaborative scheduling step involves the MEC node acquiring the location, battery level, and target point information of all UAVs in the cluster, and using a preset path planning algorithm to uniformly allocate a dedicated route without route conflicts for each UAV. The instruction issuance and dynamic adjustment steps involve the MEC node issuing the avoidance instruction and / or flight path instruction to the corresponding UAV via the 5G-A network; when any UAV performs an avoidance action, the MEC node updates the flight paths of other associated UAVs in real time to maintain cluster coordination.

[0009] Further, in a preferred embodiment, the centralized collision avoidance decision-making step specifically includes: the MEC node classifies the risk level into four levels—Level I (Safe), Level II (Attention), Level III (Warning), and Level IV (Emergency)—based on the relative distance between the UAV and the obstacle in the perception data; when the risk level is Level I, the UAV maintains its current flight path; when the risk level is Level II, the MEC node generates a deceleration command; when the risk level is Level III, the MEC node generates a detour command; when the risk level is Level IV, the MEC node generates an emergency avoidance command, which includes a maneuvering flight action at maximum acceleration.

[0010] Furthermore, in a preferred embodiment, when the risk level is IV, the centralized collision avoidance decision-making step further includes flight control-beam joint scheduling, specifically: while generating the emergency avoidance command, the MEC node calculates the attitude angle change sequence of the UAV in the future time series when executing the emergency avoidance command based on the flight dynamics model of the UAV; based on the attitude angle change sequence and the antenna model of the 5G-A base station, the MEC node pre-generates a 5G-A base station beamforming weight sequence that corresponds to the attitude angle change sequence in time; after the MEC node binds the emergency avoidance command with the beamforming weight sequence in time, it sends them to the UAV and the 5G-A base station respectively to achieve beam compensation based on attitude feedforward.

[0011] Furthermore, in a preferred embodiment, in the global cooperative scheduling step, the path planning algorithm is an improved A... Algorithm, the improved A The algorithm incorporates communication link quality as a penalty term into the cost function.

[0012] Furthermore, in a preferred embodiment, the calculation of the communication link quality includes: the MEC node pre-stores a 3D urban model of the target airspace; during the search process of the path planning algorithm, for each candidate path node, the MEC node calls a ray tracing algorithm to calculate the expected received signal power between the candidate path node and the serving base station based on the 3D urban model. and viewing distance Probability; the MEC node will receive the expected signal power. and the aforementioned line of sight The weighted combination of probabilities is used as the quality of the communication link, and is superimposed on A. The algorithm's cost function penalizes paths with poor signal quality.

[0013] To achieve the above objectives, this application also provides a drone swarm collaborative scheduling and collision avoidance system, comprising: At least one drone is equipped with a 5G-A module and sensors. The sensors are used to collect real-time perception data of surrounding obstacles, and the 5G-A module is used to offload the perception data through a 5G-A network. An edge computing MEC node is deployed on the side of a 5G-A base station and communicates with the drone. The MEC node is configured to: receive and analyze the perception data, classify the obstacles into risk levels, and generate an emergency avoidance command when the risk level is high; obtain the location, battery level, and target point information of all drones in the cluster, and use a preset path planning algorithm to uniformly allocate a dedicated route without flight path conflicts to each drone; send the avoidance command and / or flight path command to the corresponding drone through the 5G-A network, and update the flight paths of other associated drones in real time when any drone performs an avoidance action.

[0014] This application also provides a drone swarm collaborative scheduling and collision avoidance device for implementing the method described in any of the foregoing.

[0015] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method of any one of claims 1 to 7.

[0016] Compared with the prior art, the beneficial effects of this application are as follows: 1. By completely offloading the complex collision avoidance decision and path planning calculation tasks to the MEC node, the onboard computing resources of the drone are freed up and the power consumption of the onboard processor is reduced, which can increase the drone's flight time by 15%-20%.

[0017] 2. By combining the near-field computing power of MEC with the ultra-low latency of less than 1 millisecond of 5G-A network, a high-speed closed-loop control link was constructed, which shortened the response time of path update and collision avoidance commands to less than 100 milliseconds. The success rate of high-risk collision avoidance can reach more than 99%, meeting the latency-sensitive requirements of high-speed flight scenarios.

[0018] 3. As a centralized control center, the MEC node can acquire global airspace situational awareness and plan conflict-free flight paths for all UAVs within the cluster. During dynamic adjustments, the response time remains within 100 milliseconds, reducing the flight path conflict rate of a cluster of thousands of UAVs to below 1%, enabling large-scale, high-density safe collaborative flight.

[0019] 4. By using a flight control-beam joint scheduling mechanism and a channel-constrained path planning algorithm, the stability of the control link and the command arrival rate are ensured when the UAV performs large-scale emergency avoidance maneuvers or traverses complex building environments. The command arrival rate in dense building environments is increased to 99.99%, avoiding the risk of loss of control due to communication interruption. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the hardware architecture of the drone swarm collaborative scheduling and collision avoidance system provided in the embodiments of this application.

[0022] Figure 2 This is a flowchart of the drone swarm collaborative scheduling and collision avoidance method provided in the embodiments of this application.

[0023] Figure 3 This is a detailed flowchart of the flight control-beam joint scheduling method provided in the embodiments of this application.

[0024] Figure 4 This is a schematic diagram of the channel-constrained path planning principle provided in the embodiments of this application.

[0025] In the diagram: 100 - Drone, 101 - Sensor, 102 - 5G-A Module, 200 - 5G-A Network, 201 - 5G-A Base Station, 300 - Edge Computing MEC Node, 301 - Data Reception and Analysis Module, 302 - Collision Avoidance Decision Module, 303 - Path Planning Module, 304 - Command Generation and Issuance Module, 305 - Cluster Status Management Module, 401 - City 3D Electronic Map, 402a - Candidate Node A, 402b - Candidate Node B, 403a - Final Selected Route (High-Quality Link), 403b - Penalized Candidate Route (Extremely Poor Signal), 404 - Communication Blind Spot, 405 - Airspace with Better Signal Coverage. Detailed Implementation

[0026] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0027] Example 1

[0028] This embodiment provides a method for collaborative task scheduling and collision avoidance of unmanned aerial vehicle (UAV) swarms based on edge computing. For example... Figure 1 As shown, the method is executed in a system consisting of a drone 100, a 5G-A (5G-Advanced, fifth-generation mobile communication technology evolution) network 200, and an edge computing MEC node 300.

[0029] The drone 100, acting as the sensing and execution unit, only needs to carry lightweight sensors and communication modules. The 5G-A network 200 provides a high-speed, low-latency communication link between the drone 100 and the MEC node 300, with the 5G-A base station 201 as its core equipment. The edge computing MEC node 300, serving as the system's computing center, centrally processes all complex decision-making and planning tasks. It is physically deployed on the side of the 5G-A base station 201 to minimize communication distance and reduce network latency.

[0030] like Figure 2 As shown, the method in this embodiment mainly includes the following steps: Step S201: Data acquisition and unloading.

[0031] In this step, each drone in the cluster 100 uses its onboard sensors and 5G-A sensing module to collect real-time perception data of the surrounding environment during flight. This perception data includes not only basic information such as the distance, size, and relative speed of obstacles, but also the drone's own attitude and speed information, forming airspace point cloud and Doppler shift data with precise timestamps.

[0032] Subsequently, the drone 100 performs computation offloading, meaning it avoids complex local calculations and instead reports the massive amounts of collected sensing data via the 5G-A network 200. Specifically, the data is transmitted to the 5G-A base station 201 via an ultra-low latency uplink configured with a 120kHz subcarrier spacing. After receiving the data, the User Plane Function (UPF) network element within base station 201 directly routes the demodulated raw data to the MEC node 300 at high speed without bypassing the core network, thus establishing an ultra-fast data path from sensing to computation.

[0033] Step S202: Centralized collision avoidance decision.

[0034] After receiving perception data from one or more UAVs 100, the MEC node 300 performs real-time analysis using its powerful computing capabilities. The obstacle classification module within the MEC node 300 first processes the data, identifies potential obstacles, and calculates the collision time TTC.

[0035] Based on the calculation results, MEC node 300 makes graded, time-sensitive collision avoidance decisions for obstacles. Specifically, based on the relative distance between the UAV 100 and the obstacle, the risk is divided into four levels: Level I Safety and Level II Concern: When the distance to the obstacle is greater than 50 meters, the threat level is low. MEC node 300 determines that no intervention is needed and instructs UAV 100 to maintain the current flight path and continue flying.

[0036] Level III Warning: When an obstacle enters the range of 20-50 meters, there is a potential collision risk. MEC node 300 generates deceleration or detour commands, for example, instructing UAV 100 to reduce its current speed by 30%-50% and make minor course adjustments to move away from the obstacle.

[0037] Level IV Emergency: When the obstacle distance is less than 20 meters, or the calculated TTC is less than a preset threshold, it is determined to be a high-risk emergency. MEC node 300 immediately generates an emergency avoidance command, which includes a maneuvering flight action that causes the UAV 100 to perform at maximum acceleration, such as rapid vertical climb, descent, or sharp turn, to escape the collision path as quickly as possible.

[0038] In a specially optimized scenario, when a Level IV emergency response is triggered, to ensure that the communication link between the UAV 100 and the MEC node 300 remains uninterrupted during violent maneuvers, the system will perform the following... Figure 3 The flight control-beam joint scheduling process is shown below: Step S301: While generating the emergency avoidance command, MEC node 300 calls the preset UAV flight dynamics model and, based on the UAV's current velocity vector... And avoidance commands, quickly calculate the drone's attitude angle (roll angle) when executing the command within a short time window in the future. Pitch angle and yaw angle A sequence that varies over time. This sequence can be represented as the expected fuselage attitude angle sequence. At the same time, the corresponding expected trajectory matrix can also be obtained. .

[0039] Step S302: MEC node 300 acquires the attitude angle change sequence. Subsequently, based on this sequence and the antenna model of the active antenna unit (AAU) of the 5G-A base station 201, the beamforming weights required by the base station AAU at corresponding time points to continuously and stably cover the rapidly moving UAV 100 were pre-calculated. This process aims to maximize the signal strength received by the UAV, and its weight calculation objective function can be expressed as:

[0040] in, At any moment Beamforming weight vector, It is the predicted channel matrix. It is a correction function used to describe the radiation pattern of an UAV's onboard antenna array. The receiver gain change caused by attitude angle variation. By solving this optimization problem, MEC node 300 generates an optimal transmit beamforming weight vector sequence that precisely corresponds in time to the attitude angle change sequence. .

[0041] Step S303: MEC node 300 hard-binds the generated emergency avoidance command with the beamforming weight sequence in terms of timing to form a joint scheduling frame.

[0042] Step S304: MEC node 300 sends the bound joint scheduling frame to UAV 100 and the AAU of 5G-A base station 201 via the 5G-A downlink control channel PDCCH. The flight control system of UAV 100 performs avoidance maneuvers, while the base station AAU synchronously adjusts its transmit beam according to the received weight sequence. This attitude-feedforward-based beam compensation mechanism upgrades traditional passive feedback tracking to active predictive compensation, ensuring that the downlink control link remains stable even when the UAV performs large maneuvers with a roll angle greater than 45 degrees, reducing the usual 15-30 millisecond interruption time to 0 milliseconds.

[0043] Step S203: Global Coordinated Scheduling.

[0044] In addition to dealing with unexpected obstacles, MEC Node 300 is also responsible for the global flight path planning and resource scheduling of the entire drone swarm. MEC Node 300 can obtain real-time status information such as the position, remaining battery power, and mission target point of all drones 100 in the swarm, forming a dynamic global airspace situation map.

[0045] Building upon this, MEC node 300 employs an improved path planning algorithm to uniformly assign a dedicated, conflict-free route to each drone within the cluster. For example... Figure 4 As shown, in order to ensure that the planned route not only avoids physical obstacles but also guarantees high-quality communication, this A The algorithm incorporates communication link quality as a penalty term into the cost function.

[0046] Traditional A The algorithm cost function is ,in It is the actual cost from the starting point to the current node. This is the estimated cost from the current node to the destination. In this embodiment, the cost function is expanded to:

[0047] in, These are the weighting coefficients; It is the value of the repulsive field potential energy constructed based on the dynamic obstacles sensed by MEC node 300, which is used to avoid obstacles; This is the key penalty item for communication links.

[0048] The calculation method is as follows: MEC node 300 pre-loads a 3D electronic map 401 of the city in the target airspace, which contains the precise location, height, and even material information of buildings. In A... During the path search process, for each of the multiple candidate path nodes to be expanded (such as candidate node A, 402a, candidate node B, 402b), MEC node 300 invokes the ray tracing algorithm to simulate the signal propagation path between the node and the 5G-A base station 201 serving it, based on the 3D map 401. This algorithm allows for the real-time calculation of the node's expected received signal power. And the probability of LOS (Loss of Sight) at view distance. For example... Figure 4 As shown, when evaluating different candidate routes, if a node on the penalized candidate route 403b (such as a node located near the communication blind zone 404) has its prediction... When the signal quality is below a preset threshold or the LOS probability is below 60%, it indicates poor signal quality at that location, possibly indicating a communication dead zone (404). The value will increase exponentially, thereby increasing the total surcharge value of the path and forcing A to... The algorithm abandons this search branch and instead explores airspace 405, which has better signal coverage, ultimately determining a better-quality route 403a. In this way, the algorithm ensures that every route planned for the UAV is within a high-bandwidth, low-latency, high-quality signal coverage area.

[0049] Step S204: Issuance of instructions and dynamic adjustment.

[0050] The MEC node 300 will generate emergency avoidance commands and / or global collaborative flight path commands, and then accurately send them to the corresponding UAV 100 for execution via the downlink of the 5G-A network.

[0051] The system is highly dynamic. When any UAV 100 in the cluster deviates from its original flight path due to performing an emergency avoidance maneuver, the MEC node 300 will detect this change in a very short time. To maintain the coordination of the entire cluster, the MEC node 300 will immediately recalculate and update the flight paths of the other affected associated UAVs 100. In one specific implementation, the airspace temporarily occupied by the UAV performing the avoidance maneuver will be marked as a dynamic temporary no-fly zone by the MEC node 300, ensuring that other UAVs will actively avoid this area when replanning their flight paths, thereby ensuring the operational order and safety of the entire cluster.

[0052] Furthermore, the MEC node 300 also possesses closed-loop feedback optimization capabilities, continuously monitoring the drone's command execution results and flight status. If it detects that the planned flight path is inefficient or the avoidance maneuvers are not precise enough, the system will automatically adjust the relevant parameters of its internal algorithm, achieving self-learning and evolution, making the flight logic of the entire cluster increasingly intelligent.

[0053] Example 2

[0054] This embodiment provides a drone swarm collaborative scheduling and collision avoidance system, the hardware architecture of which is as follows: Figure 1 As shown. This system is the physical carrier for implementing the method described in Embodiment 1.

[0055] The system mainly includes: at least one drone 100, a 5G-A network 200, and an edge computing MEC node 300.

[0056] The UAV 100 undergoes specialized hardware and software configuration. In terms of hardware, besides the basic structure including the frame, motors, propellers, and battery, its core flight control motherboard can utilize a processor with edge processing capabilities, such as an embedded processing chip with neural network acceleration, but it primarily executes basic flight control and commands issued by the MEC node 300. Key configurations lie in its onboard sensor 101 and 5G-A module 102. Sensor 101 can be a lightweight millimeter-wave radar or infrared sensor for collecting environmental data. The 5G-A module 102 must support 3GPP Release 18 and subsequent standards, possessing integrated sensing and communication capabilities and ultra-low latency communication capabilities; it is a crucial hardware component for offloading computation.

[0057] The 5G-A network 200 serves as a communication infrastructure, with the core being the 5G-A base station 201, which includes an active antenna unit (AAU) responsible for transmitting and receiving wireless signals and a user plane function (UPF) network element responsible for forwarding data streams.

[0058] The MEC node 300 is the core of this system. In terms of hardware, it is typically a high-performance server deployed in the base station equipment room. This could be a computing server with a general-purpose instruction set architecture, equipped with hardware acceleration modules (such as graphics processing units (GPUs) or dedicated AI accelerator cards) capable of massive parallel computing, or even higher-performance GPUs, to meet the computing power demands of massive concurrent sensory data processing and complex algorithms. In terms of software, the MEC node 300 deploys a series of functional modules, including: The data receiving and analysis module 301 is responsible for receiving the perception data unloaded from the UAV 100, analyzing and fusing it, and extracting key information such as the position and speed of obstacles.

[0059] The collision avoidance decision module 302 is the core of the centralized collision avoidance decision-making process. Its internal configuration is as follows: First, based on the relative distance between the UAV and the obstacle and the Time To Chance (TTC), it classifies the risk level from I to IV. Second, based on different risk levels, it generates corresponding control commands, such as maintaining the flight path, decelerating, detouring, or emergency avoidance. Specifically, when the risk level is determined to be IV, this module is also configured to execute the aforementioned flight control-beamforming joint scheduling logic, i.e., calculate the attitude angle change sequence, pre-generate the beamforming weight sequence, and then distribute it after timing binding.

[0060] Path planning module 303 is responsible for global coordinated scheduling. Its core is the execution of an improved A... Algorithm. This module is configured to incorporate communication link quality as a penalty when calculating path cost. To achieve this, the module is also configured to: pre-load and manage the city's 3D model, and invoke the ray tracing algorithm in real-time during path search to calculate the expected cost of candidate nodes. And LOS probability, ultimately quantifying these communication quality metrics into cost value.

[0061] The instruction generation and distribution module 304 is responsible for packaging the instructions generated by the collision avoidance decision module 302 and the path planning module 303, and sending them to the target UAV 100 via the downlink of the 5G-A network 200. Simultaneously, this module is also responsible for distributing beamforming weights to the AAU of the 5G-A base station 201 when necessary.

[0062] The cluster status management module 305 is responsible for maintaining a real-time database that records information such as the location, speed, battery level, and task status of all drones 100 in the cluster, providing data support for global scheduling.

[0063] The system may also include a ground control station or user terminal for task deployment, monitoring cluster status, and receiving alarm information.

[0064] In this embodiment, the system's workflow is as follows: When a cluster task is received from the ground station, the path planning module 303 of the MEC node 300 first plans an initial conflict-free flight path for all UAVs 100. The UAVs 100 take off and fly along the flight path, while their sensors 101 and 5G-A modules 102 continuously collect and unload environmental data to the MEC node 300. The data receiving and analysis module 301 and the collision avoidance decision module 302 of the MEC node 300 continuously monitor potential collision risks. Once a risk is detected, the collision avoidance decision module 302 intervenes and generates an avoidance command. If an emergency avoidance occurs, the cluster status management module 305 updates the status of the affected UAVs and triggers the path planning module 303 to dynamically adjust the flight paths of the associated UAVs. The entire process forms an efficient, intelligent, and collaborative closed-loop control system.

[0065] It should be noted that the various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for the system and device embodiments, since they basically correspond to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiments.

[0066] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A method for collaborative task scheduling and collision avoidance of unmanned aerial vehicle (UAV) swarms based on edge computing, characterized in that, include: The data acquisition and offloading steps involve the 5G-A module and sensors on the drone acquiring real-time perception data of surrounding obstacles; the drone then offloads the perception data to an edge computing MEC node deployed on the side of the 5G-A base station via the 5G-A network. The centralized collision avoidance decision-making process involves the MEC node receiving and analyzing the perception data to classify the risk level of the obstacle; when the obstacle is determined to be at a high risk level, the MEC node generates an emergency avoidance command. The global collaborative scheduling step involves the MEC node acquiring the location, battery level, and target point information of all UAVs in the cluster, and using a preset path planning algorithm to uniformly allocate a dedicated route without route conflicts for each UAV. The instruction issuance and dynamic adjustment steps involve the MEC node issuing the avoidance instruction and / or flight path instruction to the corresponding UAV via the 5G-A network; when any UAV performs an avoidance action, the MEC node updates the flight paths of other associated UAVs in real time to maintain cluster coordination.

2. The method according to claim 1, characterized in that, The centralized collision avoidance decision-making steps specifically include: The MEC node classifies the risk level into four levels: Level I (Safe), Level II (Pay Attention), Level III (Early Warning), and Level IV (Emergency) based on the relative distance between the UAV and the obstacle in the perception data. When the risk level is Level I, the drone maintains its current flight path; When the risk level is Level II, the MEC node generates a deceleration command; When the risk level is Level III, the MEC node generates a detour instruction; When the risk level is IV, the MEC node generates the emergency avoidance command, which includes maneuvering flight actions at maximum acceleration.

3. The method according to claim 2, characterized in that, When the risk level is IV, the centralized collision avoidance decision-making steps further include flight control-beam joint scheduling, specifically: While generating the emergency avoidance command, the MEC node calculates the sequence of attitude angle changes of the UAV in the future time series when the emergency avoidance command is executed, based on the flight dynamics model of the UAV. Based on the attitude angle change sequence and the antenna model of the 5G-A base station, the MEC node pre-generates a 5G-A base station beamforming weight sequence that corresponds to the attitude angle change sequence in time. The MEC node binds the emergency avoidance command with the beamforming weight sequence in a time sequence and then sends it to the UAV and the 5G-A base station respectively to achieve beam compensation based on attitude feedforward.

4. The method according to claim 1, characterized in that, In the global collaborative scheduling step, the path planning algorithm is an improved A... Algorithm, the improved A The algorithm incorporates communication link quality as a penalty term into the cost function.

5. The method according to claim 4, characterized in that, The methods for calculating the quality of the communication link include: The MEC node pre-stores a 3D model of the city in the target airspace; During the search process of the path planning algorithm, for each candidate path node, the MEC node invokes the ray tracing algorithm to calculate the expected received signal power between the candidate path node and the serving base station based on the city's 3D model. and viewing distance probability; The MEC node will receive the expected signal power. and the aforementioned line of sight The weighted combination of probabilities is used as the quality of the communication link, and is superimposed on A. The algorithm's cost function penalizes paths with poor signal quality.

6. The method according to claim 1, characterized in that, The perception data also includes the relative speed of the obstacle, the attitude and speed information of the drone itself.

7. The method according to claim 1, characterized in that, In the dynamic adjustment step, when updating the flight paths of other associated UAVs, the MEC node marks the airspace occupied by the UAV that performed the avoidance maneuver as a temporary no-fly zone.

8. A drone swarm collaborative scheduling and collision avoidance system, characterized in that, include: At least one drone is equipped with a 5G-A module and sensors. The sensors are used to collect real-time perception data of surrounding obstacles, and the 5G-A module is used to offload the perception data through a 5G-A network. An edge computing MEC node is deployed on the side of a 5G-A base station and communicates with the drone. The MEC node is configured as follows: Receive and analyze the perception data, classify the risk level of the obstacle, and generate an emergency avoidance command when the obstacle is determined to be of a high risk level. The system acquires the location, battery level, and target point information of all drones in the cluster, and uses a preset path planning algorithm to uniformly assign each drone a dedicated route without route conflicts. The avoidance command and / or flight path command are sent to the corresponding UAVs through the 5G-A network, and the flight paths of other associated UAVs are updated in real time when any UAV performs an avoidance action.

9. The system according to claim 8, characterized in that, When classifying the risk level of the obstacle, the MEC node is specifically configured as follows: Based on the relative distance between the drone and the obstacle, the risk level is divided into four levels: Level I (Safe), Level II (Attention), Level III (Warning), and Level IV (Emergency). Based on the different levels, instructions are generated to maintain the current flight path, reduce speed, detour, or make emergency avoidance.

10. The system according to claim 9, characterized in that, When the risk level is determined to be Level IV, the MEC node is also configured to perform flight control-beam joint scheduling, specifically, The sequence of attitude angle changes of the UAV when executing the emergency avoidance command is calculated; The beamforming weight sequence of the 5G-A base station is pre-generated based on the attitude angle change sequence; After the emergency avoidance command is time-bound with the beamforming weight sequence, it is sent to the drone and the 5G-A base station respectively.

11. The system according to claim 8, characterized in that, When the MEC node employs a preset path planning algorithm, it is specifically configured to use an improved A algorithm. Algorithm, the improved A The algorithm incorporates communication link quality as a penalty term into the cost function.

12. The system according to claim 11, characterized in that, The MEC node is also configured to... Based on a pre-stored 3D city model, and using a ray tracing algorithm, the expected received signal power of candidate path nodes is calculated during the path planning process. and viewing distance Probability, and the expected received signal power and the aforementioned line of sight The weighted combination of probabilities is used as the quality of the communication link.

13. A drone swarm collaborative scheduling and collision avoidance device, used to implement the method of claim 1.

14. A computer-readable storage medium having a computer program stored thereon, the computer program, when executed by a processor, implementing the method of claim 1.

Images