An intelligent AI combined unmanned aerial vehicle air-ground integrated communication method

Abstract

The application discloses an unmanned aerial vehicle air-ground integrated communication method combined with intelligence AI, relates to the unmanned aerial vehicle communication technical field, and is used for solving the problem of cooperation confusion in unmanned aerial vehicle cluster communication; the application acquires real-time positions, speeds, postures and signal quality parameters of unmanned aerial vehicles through a distributed sensing unit, and constructs a dynamic network topological atlas; intelligent decision-making combining centralization and distribution is adopted to perform role allocation, space-time resource allocation and ground-air link switching; high-reliability data transmission is realized based on multi-unmanned aerial vehicle cooperative beam forming, adaptive modulation and coding, priority queue scheduling and hybrid automatic repeat request; performance indexes are collected in real time through an AI learning unit, and parameters are dynamically adjusted, so that a self-optimization closed loop is formed; and the cooperation and resource utilization efficiency of the unmanned aerial vehicle cluster communication are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) communication technology, specifically to an integrated air-to-ground communication method for UAVs that combines intelligent AI. Background Technology

[0002] With the accelerated development of integrated air-space-ground information networks, the application of drone swarms in logistics and emergency communications is shifting from single-platform control to intelligent swarms with cloud-edge-device collaboration. Simultaneously, the demand for high reliability, low latency, and autonomous collaborative capabilities is surging in areas such as smart cities, integrated security systems, and instant delivery. Dynamic collaborative communication technology, integrating intelligent decision-making and anti-interference technologies, has become a core direction for the transformation of drone communication and control systems. However, the existing technology system still suffers from many structural deficiencies:

[0003] In traditional technologies, the quantitative fusion logic of communication mechanisms based on preset rules and intelligent anti-interference algorithms is insufficient. The communication control strategy does not integrate the collaborative constraint parameters of physical layer channel state and network layer topology. It lacks a dynamic adaptation and scheduling mechanism for airborne trunk communication resources by the server in environments with strong electromagnetic countermeasures and complex terrain. In addition, the communication security verification does not fully consider the compliance correlation and real-time distribution of national cryptographic algorithms and transmission protocols in high-speed mobile scenarios between machines and between machines and ground. As a result, the communication coordination and resource utilization efficiency are low when facing a collaborative perception and command and control integration process with wide coverage, high dynamism and multiple tasks.

[0004] To address the aforementioned shortcomings, a technical solution is provided. Summary of the Invention

[0005] The purpose of this invention is to solve the problems of low reliability and low mission response efficiency in UAV swarm communication, and to propose an intelligent AI-integrated UAV air-ground communication method.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] A method for integrated air-to-ground communication of unmanned aerial vehicles (UAVs) combining intelligent AI, comprising:

[0008] S1. Holographic perception of network status: Real-time position, velocity vector, attitude parameters and signal quality parameters of ground-to-air communication links of UAV clusters are obtained through distributed sensing units to construct a dynamic network topology map;

[0009] S2. Resource Management and Collaborative Control: Based on the dynamic network topology map from step S1, intelligent decision-making combining centralized and distributed decision-making is employed to achieve integrated collaborative control of communication roles, spatiotemporal resources, and communication links. This generates communication role allocation strategies, spatiotemporal resource allocation strategies, and ground-to-air link switching strategies. The centralized decision-making involves global integrated collaborative control of communication roles, spatiotemporal resources, and communication links based on the dynamic network topology map. The distributed decision-making is manifested in each UAV completing local link quality perception, node status acquisition, local parameter calculation, and communication resource adjustment.

[0010] S3. Cooperative transmission execution: Based on the role allocation, spatiotemporal resource allocation and ground-to-air link switching strategies determined in step S2, high-reliability data transmission is executed through multi-UAV cooperative beamforming, adaptive modulation and coding, priority queue scheduling and hybrid automatic repeat mechanism.

[0011] S4. Communication Quality Closed-Loop Optimization: By collecting communication performance indicators in real time, the collaborative communication effect is evaluated. Based on network status, adjustable communication parameters, and comprehensive performance indicators, the link quality index is updated and communication parameters are adjusted through deep reinforcement learning algorithms using AI learning units, forming a self-optimizing closed loop.

[0012] As a further improvement of the present invention, the specific implementation process of step S1 is as follows:

[0013] Raw data is collected based on distributed sensing units, which include a BeiDou and GPS dual-mode positioning unit, an IMU inertial measurement unit, and a signal receiving unit. The raw data includes the UAV's three-dimensional coordinates, velocity vector, and signal strength indication value.

[0014] Doppler frequency shift is obtained based on velocity vector and Doppler effect; packet error rate is obtained based on data packet reception results;

[0015] The link quality index is obtained by combining the signal strength indication value, Doppler frequency shift and packet error rate with a weighted formula.

[0016] Each UAV periodically broadcasts beacon frames including its own status information and updates its dynamic neighbor list based on received neighbor beacon frames; it also predicts the distance between UAVs at future moments based on the UAVs' position and velocity vectors.

[0017] A dynamic adjacency matrix is ​​constructed based on the link quality index. If the link quality index is greater than the minimum available link quality index, the matrix element represents the same value as the link quality index. If the link quality index is less than or equal to the minimum available link quality index, the value is zero.

[0018] Based on the dynamic adjacency matrix, the Dijkstra algorithm is used to obtain the shortest path hop count from each UAV to the ground base station, and the node centrality index is obtained by combining the remaining battery power of the UAV and the number of neighboring nodes.

[0019] As a further improvement of the present invention, the specific operation process of step S2 includes:

[0020] Role Assignment:

[0021] Based on a dynamic network topology map, the UAV's working airspace is divided into a three-dimensional virtual grid. Based on the UAV's grid index and unique ID, candidate roles are obtained. A role suitability function is constructed by combining node centrality index and link quality index. The result of this function assigns a communication role to each UAV, including cluster head node, relay node, and member node. The node centrality index is calculated based on the shortest path hop count obtained from the Dijkstra algorithm, the UAV's remaining battery power, and the number of neighboring nodes. The link quality index is calculated based on signal strength indicator, Doppler shift, and packet error rate. The equilibrium state is determined based on the variance of the UAV cluster's remaining battery power. When the variance is less than a preset threshold, the basic rotation frequency is maintained; when the variance is greater than the preset threshold, the rotation cycle is automatically shortened. An adaptive role rotation mechanism is initiated based on the equilibrium state determination result to adjust the rotation cycle.

[0022] Resource allocation:

[0023] In the time domain, the communication cycle is divided into multiple time slices, and corresponding time slices are allocated to nodes with different communication roles;

[0024] In the frequency domain, multiple sub-channels are divided based on a preset spectrum range, and a center frequency is assigned to each UAV based on the spatial distribution of the UAVs and combined with a hash algorithm, thus constructing a space-frequency mapping matrix that satisfies the constraints of minimum frequency reuse distance and protection bandwidth.

[0025] In the airspace, the three-dimensional space is divided into multiple communication regions and assigned independent spatial codes. Based on virtual MIMO technology, multiple spatially distributed UAVs are combined into a distributed antenna array, and the cluster head node calculates and distributes beamforming weights based on the maximum ratio combining criterion.

[0026] As a further improvement of the present invention, the specific operation process of step S2 also includes:

[0027] Based on the inherent priority of the task, the remaining power ratio, the urgency of the delay, and the data packet length factor, a multi-dimensional priority evaluation system is constructed to obtain the comprehensive priority index of the data packet.

[0028] Based on the current position and velocity vector of the drone, predict the signal strength indication value at future moments, and set the switching trigger threshold and execution threshold;

[0029] When the predicted signal strength indicator value is lower than the handover trigger threshold, the handover preparation process is initiated. Available candidate communication targets are scanned in parallel. Based on the signal strength indicator value, the real-time distance between the UAV and the candidate target, and the load rate of the candidate target, a comprehensive score is calculated using a weighted formula after normalization. The target with the best comprehensive score is selected to initiate a handover request. The handover target node is based on the sum of the new load and its own load. If it exceeds its own maximum load threshold, it will refuse access and trigger a reselection of the communication target with the second highest comprehensive score as the handover target.

[0030] Once the target network agrees to accept the handover, dual connections are maintained during the handover transition until the new link quality is stable and meets the handover execution threshold, at which point the original connection is officially disconnected and the occupied resources are released.

[0031] As a further improvement of the present invention, the specific operation process of step S3 is as follows:

[0032] Cooperative beamforming transmission: For scenarios requiring data transmission to ground base stations, multiple UAVs within a preset cooperative range of spatial distance from the cluster head node are selected to form a virtual antenna array; the cluster head node broadcasts synchronization instructions to the cooperative UAVs, including the target direction vector, carrier center frequency, and transmission time reference; each cooperative UAV synchronizes its clock based on a two-way time synchronization protocol, and calculates the amplitude and phase of the transmitted signal based on the beamforming weights distributed by the cluster head node, transmitting the signal simultaneously at the synchronization time reference, so that the signal can be superimposed in phase and combined with power at the receiving end;

[0033] Adaptive modulation and coding steps: Each UAV dynamically selects the modulation scheme and coding rate based on the real-time link quality index and signal strength indicator; the signal-to-noise ratio (SNR) is obtained based on the signal strength indicator and noise power, and a mapping table between the SNR and the modulation and coding scheme is constructed. At the same time, the SNR threshold is dynamically adjusted based on the modulation scheme; the transmitting end embeds the selected scheme identifier in the data frame header, and the receiving end adjusts the demodulation and decoding parameters based on the frame header.

[0034] As a further improvement of the present invention, the specific operation steps of the priority queue scheduling and hybrid automatic retransmission are as follows:

[0035] Priority queue scheduling and transmission: Each UAV maintains a local transmission queue and sorts data packets in descending order based on the overall priority index of the data packets; within the allocated communication time slots, high-priority data packets are sent first, and fragmented transmission of excessively long data packets is allowed; when a high-priority urgent data packet arrives, a preemptive scheduling mechanism is activated to interrupt the transmission of low-priority data packets or request insertion into an urgent time slot from a neighboring node.

[0036] Hybrid Automatic Repeat: The sender assigns a unique sequence number to each data packet it sends and starts a retransmission timer;

[0037] The receiving end performs cyclic redundancy check on the received data packets; if the check passes, it replies with an acknowledgment frame and delivers the data packet; if the check fails, it buffers soft information and replies with a negative acknowledgment frame.

[0038] The sending end selects either an additional redundant retransmission or a full retransmission strategy based on the number of retransmissions and the link quality index, until success or the maximum number of retransmissions is reached; if an acknowledgment frame is received, the timer is stopped and the sending buffer is cleared.

[0039] As a further improvement of the present invention, the specific operation process of step S4 is as follows:

[0040] The communication performance monitoring unit collects key performance indicators in real time, including end-to-end latency, data packet delivery rate, throughput, handover success rate, and energy efficiency; the collected key performance indicators are normalized and combined with a weighted formula to obtain comprehensive performance indicators.

[0041] The current network state is used as the state space, adjustable communication parameters as the action space, and comprehensive performance indicators as the reward function. The link quality index is updated based on a deep reinforcement learning algorithm and then distributed to each UAV communication unit for adaptive adjustment. The AI ​​learning unit is incrementally trained at fixed intervals based on the accumulated new data.

[0042] When the overall performance index is detected to be less than the preset threshold and the duration exceeds the maximum allowable time, the alarm mechanism is triggered; the emergency plan is executed, including suspending non-critical data transmission, switching to the most robust BPSK modulation mode, increasing the transmission power to the upper limit, and shortening the switching decision time; at the same time, an anomaly report including fault characteristics, scope of impact, and measures already taken is sent to the ground control center.

[0043] Compared with the prior art, the beneficial effects of the present invention are:

[0044] This invention constructs a mapping relationship between a dynamic network topology map of UAVs and multidimensional communication parameters, and combines link quality variation range analysis to quantify the impact of network dynamism on communication quality, generating a link quality index and a node centrality index. Based on the link quality index, a resource allocation strategy is generated, and the communication role allocation and data scheduling mechanism are optimized by combining a role suitability function and a priority evaluation system. This effectively solves the problems of link instability and low resource allocation efficiency caused by high-speed node movement and complex environments in traditional UAV communication, and significantly improves the reliability and resource utilization efficiency of UAV swarm communication. Attached Figure Description

[0045] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] Example:

[0048] like Figure 1 As shown, an intelligent AI-integrated UAV air-ground communication method includes network status holographic perception, resource management and collaborative control, collaborative transmission execution, and closed-loop optimization of communication quality.

[0049] S1. Holographic Network Status Perception: Real-time position, velocity vector, attitude parameters, and signal quality parameters of the air-to-ground communication link of the UAV swarm are acquired through distributed sensing units to construct a dynamic network topology map. The specific implementation process is as follows:

[0050] Raw data is collected through a distributed sensing unit carried by each drone; the distributed sensing unit includes a Beidou and GPS dual-mode positioning unit, an IMU inertial measurement unit, and a signal receiving unit;

[0051] The BeiDou and GPS dual-mode positioning unit is used to acquire the three-dimensional coordinates of the UAV. The IMU data is fused using Kalman filtering to obtain smooth position and velocity vectors. The signal receiving unit continuously monitors signal quality parameters from the ground base station and neighboring UAVs. Simultaneously, the signal receiving unit continuously acquires signal strength data from the ground base station and neighboring UAVs to obtain signal strength indication values. ;

[0052] Through formula The Doppler frequency shift is calculated, where, Indicates the center frequency of the carrier. Represents the speed of light. Represents the velocity vector. Indicates the angle between the direction of motion and the direction of signal propagation;

[0053] Continuous data acquisition based on sliding window mechanism The packet error rate is calculated based on the ratio of the number of received erroneous packets to the total number of packets. ;

[0054] Based on the signal strength indication value, Doppler frequency shift, and packet error rate, after normalization, the formula is entered. The link quality index is calculated. ,in, Indicates drone To drones The signal strength indication value, This indicates that the drone is in an ideal state. To drones Maximum signal strength indication value, Indicates the Doppler frequency shift threshold. Indicates signal power quality. Indicates the correct packet transmission rate. Represents the Doppler effect term. These represent the weighting factors for signal power quality, packet correct transmission rate, and the Doppler effect term, respectively.

[0055] Each drone periodically broadcasts a beacon frame, including its own ID, location, speed, and remaining battery power; the beacon frame uses BPSK modulation to ensure maximum coverage.

[0056] When the drone receives a neighbor beacon frame, it calculates the beacon connection duration based on the current time and the last time the same neighbor beacon was received. If the beacon connection duration is greater than a preset threshold, it determines that the current neighbor beacon node is disconnected and removes it from the neighbor beacon list.

[0057] Based on the position and velocity vectors of the UAV, using the formula Predicting the future The distance between drones at any given moment; They represent the current time. drones Position vector in three-dimensional space They represent drones velocity vector Indicates the prediction time interval;

[0058] Based on real-time collected link quality indices, a dynamic adjacency matrix is ​​constructed, where the matrix elements... , Indicates the current time drones To drones The value of the link quality index in the dynamic adjacency matrix. Indicates the current time drones To drones The link quality index, Indicates the minimum available link quality index;

[0059] Based on the dynamic adjacency matrix and combined with Dijkstra's algorithm, the shortest path hop count from each UAV to the ground base station is obtained. Through formula The node centrality index is calculated, where, These represent the drone's current remaining battery power and maximum battery power, respectively. This indicates the number of neighboring beacon nodes. This represents the total number of nodes in the network.

[0060] S2. Resource Management and Collaborative Control: Based on the dynamic network topology map from step S1, intelligent decision-making combining centralized and distributed decision-making is employed to achieve integrated collaborative control of communication roles, spatiotemporal resources, and communication links. This generates communication role allocation strategies, spatiotemporal resource allocation strategies, and ground-to-air link switching strategies. The centralized decision-making involves global integrated collaborative control of communication roles, spatiotemporal resources, and communication links based on the dynamic network topology map. The distributed decision-making is manifested in each UAV completing local link quality perception, node status acquisition, local parameter calculation, and communication resource adjustment. The specific implementation process is as follows:

[0061] S201, Role Assignment:

[0062] The working airspace of the UAV is divided into several three-dimensional virtual grids. The grid index of the UAV is determined based on the UAV's three-dimensional coordinates, the horizontal grid side length and the vertical grid height of the three-dimensional virtual grid. Then, a hash operation is performed on the grid index and unique ID of each UAV to obtain the role candidate value.

[0063] Based on the node center index, link quality index, and role candidate value, a role adaptability function is constructed, and three types of communication roles are defined, including cluster head node, relay node, and member node.

[0064] Cluster head nodes are responsible for global resource scheduling and data aggregation and forwarding, relay nodes are responsible for data forwarding between non-cluster head nodes and cluster head nodes, and member nodes are only responsible for local data collection and direct transmission.

[0065] Within each virtual grid, the drone with the highest role suitability is selected as the cluster head node:

[0066] Traverse all drones within each virtual grid, calculate the role fit of each drone based on the role fit function, and select the drone with the maximum fit as the cluster head node of the current grid. If there are drones with the same fit, select the drone with the higher node centrality index as the cluster head node. If the node centrality index is still the same, select the drone with the smaller unique ID value as the cluster head node.

[0067] For non-cluster head nodes, the determination is based on the distance to the nearest cluster head and the link quality index: if the distance between the node and the cluster head is greater than the preset communication threshold, or the direct link quality index between the node and the cluster head node is lower than the minimum available link quality index, and the remaining power of the current node is higher than the minimum power threshold of the relay node, then the current node is assigned as a relay node; otherwise, it is defined as a member node.

[0068] After role assignment is completed, cluster head nodes form the backbone layer and communicate directly with ground base stations, relay nodes form the forwarding layer, and member nodes access the network through the nearest relay node or cluster head.

[0069] When an energy imbalance is detected due to the long-term fixation of communication roles, an adaptive role rotation mechanism is activated:

[0070] When the variance of the remaining power is less than the preset threshold, it is determined to be a power balance state, and the basic rotation cycle is maintained.

[0071] When the variance of remaining battery power exceeds a preset threshold, it is determined to be a state of battery imbalance, and the rotation cycle is automatically shortened using the formula. The adjusted rotation cycle is calculated dynamically, whereby... Indicates the current rotation cycle. Indicates the basic rotation cycle. Indicates the adjustment factor. This represents the variance of the remaining battery power of the drone swarm. Indicates the preset power balancing threshold;

[0072] S202, Resource Allocation:

[0073] In the time domain, the communication cycle is divided into The time slices are allocated in order, with priority time slices reserved for cluster head nodes and numbered sequentially; for relay nodes and member nodes, time slices are allocated based on their respective clusters.

[0074] In the frequency domain, several sub-channels are uniformly divided within a preset available spectrum range. Based on the spatial distribution of the UAVs, a center frequency is assigned to each UAV using a hash algorithm. A space-frequency mapping matrix is ​​constructed based on the UAV's location and its assigned center frequency, satisfying the following constraints:

[0075] For any two drones in the network, if the three-dimensional Euclidean distance between any two drones is less than the preset minimum frequency reuse distance, then the difference in the center frequencies of the two drones is greater than or equal to the preset protection bandwidth.

[0076] In the airspace, the three-dimensional space is divided into multiple communication regions, and each communication region is assigned an independent spatial code. Based on virtual MIMO technology, multiple spatially distributed UAVs are combined into a distributed antenna array. The cluster head node collects the channel state information of each member node UAV, including the channel state vector, the conjugate transpose of the channel vector, and the magnitude of the channel vector. Based on the maximum ratio combining criterion, the beamforming weight of each node is obtained, and the weight coefficients are distributed to each UAV.

[0077] Construct a multi-dimensional priority evaluation system, using formulas The overall priority index of the data packets is calculated. This indicates the inherent priority of a task, which is the basic priority determined by the task type to which the data packet belongs. This indicates the percentage of remaining battery power, which is the ratio of the drone's remaining battery power to its maximum battery power. Indicating the urgency of the delay, it refers to the proportion of the already delayed time to the maximum tolerable delay. This represents the packet length factor, which is the ratio of the current packet length to the maximum packet length supported by the network. These represent the weighting factors of the task's inherent priority, remaining battery percentage, delay urgency, and data packet length factor, respectively.

[0078] S203, Ground-to-Air Link Switching:

[0079] Based on the UAV's current position and velocity vector, using the formula Predicting the future The drone signal strength indicator value at any given time; where, This indicates the current drone signal strength value. Representing the present moment and the future moment respectively. Communication distance at any given moment; Represents the decay function; Indicates the future The altitude of the drone at all times ;

[0080] Set a handover trigger threshold and a handover execution threshold respectively, with the handover trigger threshold being greater than the handover execution threshold; when the predicted drone signal strength indicator value is less than the handover trigger threshold, immediately start the handover preparation process;

[0081] During the handover preparation phase, the UAV scans all available alternative communication targets in parallel, including other ground base stations and airborne cluster head nodes, to build a complete list of candidate targets. Based on the current signal strength of the candidate targets, the real-time distance between the UAV and the candidate targets, and the current load rate of the candidate targets, a comprehensive score for each candidate target is obtained using a weighted formula. The comprehensive scores of the candidate targets are sorted, and the candidate target with the highest comprehensive score is selected as the handover target. A handover request frame is sent to the handover target, which includes the current connection status, data cache information, and handover reason code.

[0082] After receiving the request, the target to switch assesses its own resource availability:

[0083] If the sum of the current load of the target node to be switched and the new load of the new request is greater than the maximum load threshold, the switch is rejected and a switch rejection frame is sent to the requesting drone. After receiving the rejection response, the requesting drone selects the candidate target with the second highest comprehensive score as the switch target. The target node to be switched determines whether to reject or accept based on the load result.

[0084] If the load is less than the maximum load threshold, acceptance is granted, a handover confirmation is sent, and necessary communication resources are reserved; during the handover transition, connections with both the original target and the newly created target are maintained simultaneously.

[0085] When the link quality index between the current UAV and the switching target is stable and the signal strength indicator value is greater than the switching execution threshold for a continuous period of time, the connection with the original target is officially disconnected, and a link release frame is sent to the original target to release the occupied resources.

[0086] S3. Cooperative Transmission Execution: Based on the role allocation, spatiotemporal resource allocation, and ground-to-air link handover strategies determined in step S2, high-reliability data transmission is executed through multi-UAV cooperative beamforming, adaptive modulation and coding, priority queue scheduling, and hybrid automatic repeater. The specific implementation process is as follows:

[0087] S301, Cooperative beamforming transmission:

[0088] For scenarios where cluster head nodes need to transmit data to ground base stations, multiple UAVs within a preset coordination range are selected to form a virtual antenna array; based on the known location coordinates of the ground base station, the unit vector of direction from the cluster head node to the ground base station is calculated;

[0089] The cluster head node broadcasts a synchronization command frame to each participating UAV. The frame includes the target direction vector, carrier center frequency, transmission time reference, and data to be transmitted.

[0090] After receiving the synchronization command frame, each collaborative drone synchronizes its clock based on a two-way time synchronization protocol: the initiating node synchronizes its clock at time... Send a time synchronization request, and the cooperating node will synchronize at time... Receive and immediately at the moment Reply, the initiating node at time Receive reply; via formula The one-way propagation delay is calculated using the formula. The clock deviation is calculated, and each coordinating node adjusts its local clock based on the clock deviation to achieve the preset clock synchronization accuracy.

[0091] Based on the beamforming weights already distributed to the cluster head nodes in step S2, each cooperative UAV calculates the amplitude and phase of the transmitted signal;

[0092] Amplitude: Calculated by taking the square root of the current drone's transmit power and the total transmit power of all participating drones.

[0093] Phase: via formula Calculations are performed to obtain the first The channel phase of the collaborative UAV, among which, Indicates the center frequency of the carrier. Indicates the first The position vector of the collaborative UAV in three-dimensional space. This represents the target direction vector pointing towards the target ground base station. Represents the speed of light;

[0094] Each coordinating drone simultaneously transmits amplitude- and phase-adjusted signals based on a synchronization time reference. At the ground base station receiver, the signals from multiple drones are superimposed in space and analyzed using the formula... The received signal is obtained through calculation, where, Indicates the number of collaborative drones. Indicates the first The channel phase from the drone to the ground base station, Represents Euler's formula; Indicates the first The range of the drone Indicates the first The baseband signal transmitted by the UAV; based on phase pre-compensation, the signals from each path are superimposed in phase at the receiving end to achieve power combining gain;

[0095] S302, Adaptive Modulation and Coding Selection:

[0096] Each UAV dynamically selects its modulation scheme and coding rate based on real-time link quality index and signal strength indicator values; using formulas... The signal-to-noise ratio (SNR) is calculated, where NP represents the noise power; a mapping table between the SNR and the modulation and coding scheme is constructed.

[0097] For example: when the signal-to-noise ratio SNR > 20 dB, 64QAM modulation and coding rate are adopted ; when 15 dB < SNR ≤ 20 dB, 16QAM modulation and coding rate are adopted ; when 10 dB < SNR ≤ 15 dB, QPSK modulation and coding rate are adopted ; when SNR ≤ 10 dB, BPSK modulation and coding rate are adopted ;

[0098] When the current modulation mode is high-order modulation, the signal-to-noise ratio threshold for switching to low-order modulation is reduced by a preset value; when the current modulation mode is low-order modulation, the signal-to-noise ratio threshold for switching to high-order modulation is increased by a preset value;

[0099] The sender embeds the identifier of the currently selected modulation and coding scheme in the modulation and coding indication field of the MAC frame header; after the receiver analyzes the frame header, it adjusts the parameters of the demodulator and decoder;

[0100] S303. Priority queue scheduling and transmission:

[0101] Each UAV maintains a local transmission queue, sorts the data packets in the queue in descending order based on the comprehensive priority index of the data packets, and the data packets with a higher priority index are arranged at the front of the queue;

[0102] When the time slot allocated to the current UAV arrives, the data packet is taken from the head of the queue for transmission; if the remaining duration of the current time slot is not sufficient to transmit the complete data packet at the head of the queue, it is judged whether fragmentation transmission is allowed:

[0103] If the length of the data packet exceeds 1.5 times the maximum transmission unit, fragmentation processing is performed, the data packet is split into multiple fragments, the length of each fragment does not exceed the maximum transmission unit, the fragment sequence number and the total number of fragments are marked in the fragment header, the first fragment is sent in the current time slot, and the subsequent fragments are sent in the next time slot;

[0104] If the length of the data packet is less than or equal to 1.5 times the maximum transmission unit, the complete data packet is retained and waiting to be sent in the next allocated time slot;

[0105] When it is detected that a high-priority emergency data packet arrives, a preemptive scheduling mechanism is started:

[0106] If the current UAV is in the transmission state and the transmitted data packet is a low-priority data packet, the transmitted ratio is judged. If the transmitted ratio is less than 30%, the current transmission is interrupted, the untransmitted part is returned to the queue, and the high-priority data packet is immediately sent; if the transmitted ratio is greater than or equal to 30%, the high-priority data packet is sent after the current data packet is transmitted;

[0107] If the drone is currently idle, it broadcasts an emergency time slot request frame to neighboring nodes, requesting that an emergency time slot be inserted in the current communication cycle. After receiving the request, the neighboring node checks whether there is any data to be sent with the same or higher priority in its own queue. If not, it replies with a time slot yield confirmation frame and suspends data transmission in this time slot. When a sufficient number of yield confirmation frames are collected, the requesting node sends high-priority data packets in the inserted emergency time slot.

[0108] S304, Hybrid Automatic Repeat:

[0109] The sending end assigns a unique sequence number to each transmitted data packet and starts a retransmission timer, with the timer duration set to [specify duration]. ,in, This represents the average end-to-end latency. Indicates the maximum allowable delay;

[0110] The receiving end performs cyclic redundancy check on the received data packets; if the check passes, it replies with an acknowledgment frame to the sending end, which includes the sequence number of the successfully received data packet, and simultaneously submits the data packet to the upper layer.

[0111] If the verification fails, the erroneous data packet is not immediately discarded. Instead, the soft information is stored in the buffer and a negative acknowledgment frame is sent back to the sender. The frame includes the sequence number of the erroneous data packet and the current link quality index.

[0112] After receiving a negative acknowledgment frame, the sending end selects a retransmission strategy based on the current retransmission count:

[0113] If this is the first retransmission and the link quality index is greater than the preset threshold, an additional redundant retransmission strategy is adopted, which only sends extra redundant check bits and does not retransmit the original data; the receiving end combines the received redundant check bits with the buffered soft information, performs a log-likelihood ratio calculation, and re-attempts decoding.

[0114] If the current retransmission is a second retransmission or the link quality index is less than the preset threshold, the full retransmission strategy is adopted to retransmit the complete data packet.

[0115] If the retransmission attempt fails after reaching the maximum retransmission limit, a transmission failure report is sent to the upper layer, and the sequence number, retransmission count, and final link quality index of the failed data packet are recorded.

[0116] After receiving the acknowledgment frame, the sending end stops the retransmission timer, deletes the acknowledged data packet from the sending buffer, and releases the buffer space.

[0117] S4. Communication Quality Closed-Loop Optimization: The effectiveness of collaborative communication is evaluated through real-time collected communication performance indicators. Based on network status, adjustable communication parameters, and comprehensive performance indicators, the link quality index is updated and communication parameters are adjusted using a deep reinforcement learning algorithm combined with an AI learning unit, forming a self-optimizing closed loop. The specific implementation process is as follows:

[0118] Based on the established communication performance monitoring unit, communication performance indicators are collected in real time, including end-to-end latency, data packet delivery rate, throughput, handover success rate, and energy efficiency.

[0119] After normalizing the collected key performance indicators, they were input into the formula. The comprehensive performance index is obtained through calculation, among which, Indicates the key performance indicator (KPI) The normalized value of each component. This indicates the total number of key performance indicators (KPIs). Indicates the key performance indicator (KPI) The influence weighting factor of each sub-item;

[0120] Using the current network state as the state space, adjustable parameters as the action space, and comprehensive performance index as the reward function, the link quality index is updated based on a deep reinforcement learning algorithm. The adjustable communication parameters include transmit power, modulation and coding scheme, time slice allocation ratio, beamforming weight, and handover decision time. The AI ​​learning unit uses a deep reinforcement learning algorithm to combine the network state in the state space and the adjustable communication parameters in the action space, and uses the comprehensive performance index as the reward signal to iteratively optimize the parameter adjustment strategy, determine the optimal communication parameter configuration for each UAV, and bind the optimal communication parameters to the updated link quality index.

[0121] The updated link quality index and corresponding optimal communication parameter configuration are sent to each UAV communication unit. After receiving the configuration, each UAV communication unit automatically adjusts its communication parameters accordingly. At fixed intervals, the AI ​​learning unit is incrementally trained based on the accumulated new data to continuously optimize the parameter adjustment strategy and maintain the timeliness of the model.

[0122] When the overall performance index is detected to be less than the preset threshold and the duration exceeds the preset maximum allowable time, the alarm mechanism is triggered and the emergency plan is automatically executed: suspend non-critical data transmission, switch to the most robust BPSK modulation, increase the transmission power to the upper limit, and shorten the switching decision time; at the same time, an anomaly report is sent to the ground control center, including the fault characteristics, the scope of impact, and the measures already taken.

[0123] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for integrated air-to-ground communication of unmanned aerial vehicles (UAVs) combining intelligent AI, characterized in that: include: S1: Obtain the real-time position, velocity vector, attitude parameters, and signal quality parameters of the ground-to-air communication link of the UAV cluster through distributed sensing units, and construct a dynamic network topology map; The specific operation process of step S1 is as follows: Raw data is collected based on distributed sensing units, which include a BeiDou and GPS dual-mode positioning unit, an IMU inertial measurement unit, and a signal receiving unit. The raw data includes the UAV's three-dimensional coordinates, velocity vector, and signal strength indication value. Doppler frequency shift is obtained based on velocity vector and Doppler effect; The packet error rate is obtained based on the data packet reception results; The link quality index is obtained by combining the signal strength indication value, Doppler frequency shift and packet error rate with a weighted formula. Each UAV periodically broadcasts beacon frames including its own status information and updates its dynamic neighbor list based on received neighbor beacon frames; it also predicts the distance between UAVs at future moments based on the UAVs' position and velocity vectors. A dynamic adjacency matrix is ​​constructed based on the link quality index. If the link quality index is greater than the minimum available link quality index, the matrix element represents the same value as the link quality index. If the link quality index is less than or equal to the minimum available link quality index, the value is zero. Based on the dynamic adjacency matrix, the Dijkstra algorithm is used to obtain the shortest path hop count from each UAV to the ground base station, and the node centrality index is obtained by combining the remaining battery power of the UAV and the number of neighboring nodes. S2: Based on the dynamic network topology map of step S1, intelligent decision-making combining centralized and distributed decision-making is adopted to perform integrated and coordinated control of communication roles, spatiotemporal resources and communication links, and to generate communication role allocation strategy, spatiotemporal resource allocation strategy and ground-to-air link switching strategy. The centralized decision-making is based on the dynamic network topology map to perform global integrated and coordinated control of communication roles, spatiotemporal resources and communication links. The distributed decision-making is reflected in each UAV completing local link quality perception, node status acquisition, local parameter calculation and communication resource adjustment; S3: Based on the role allocation, spatiotemporal resource allocation and ground-to-air link switching strategies determined in step S2, high-reliability data transmission is performed through multi-UAV collaborative beamforming, adaptive modulation and coding, priority queue scheduling and hybrid automatic repeat mechanism; S4: By collecting communication performance indicators in real time, the collaborative communication effect is evaluated. Based on network status, adjustable communication parameters, and comprehensive performance indicators, the link quality index is updated and communication parameters are adjusted through deep reinforcement learning algorithms using the AI ​​learning unit, forming a self-optimizing closed loop.

2. The intelligent AI-integrated UAV air-to-ground communication method according to claim 1, characterized in that, The specific operation process of step S2 includes: The specific steps for generating the communication role allocation strategy are as follows: Based on a dynamic network topology map, the UAV's working airspace is divided into a three-dimensional virtual grid. Based on the UAV's grid index and unique ID, candidate roles are obtained. A role suitability function is constructed by combining node centrality index and link quality index. The result of this function assigns a communication role to each UAV, including cluster head node, relay node, and member node. The node centrality index is calculated based on the shortest path hop count obtained from the Dijkstra algorithm, the UAV's remaining battery power, and the number of neighboring nodes. The link quality index is calculated based on signal strength indicator, Doppler shift, and packet error rate. The equilibrium state is determined based on the variance of the UAV cluster's remaining battery power. When the variance is less than a preset threshold, the basic rotation frequency is maintained; when the variance is greater than the preset threshold, the rotation cycle is automatically shortened. An adaptive role rotation mechanism is initiated based on the equilibrium state determination result to adjust the rotation cycle. The specific operational steps of the spatiotemporal resource allocation strategy are as follows: In the time domain, the communication cycle is divided into multiple time slices, and corresponding time slices are allocated to nodes with different communication roles; In the frequency domain, multiple sub-channels are divided based on a preset spectrum range, and a center frequency is assigned to each UAV based on the spatial distribution of the UAVs and combined with a hash algorithm, thus constructing a space-frequency mapping matrix that satisfies the constraints of minimum frequency reuse distance and protection bandwidth. In the airspace, the three-dimensional space is divided into multiple communication regions and assigned independent spatial codes. Based on virtual MIMO technology, multiple spatially distributed UAVs are combined into a distributed antenna array, and the cluster head node calculates and distributes beamforming weights based on the maximum ratio combining criterion.

3. The intelligent AI-integrated UAV air-to-ground communication method according to claim 1, characterized in that, The specific operation process of step S2 also includes: Based on the inherent priority of the task, the remaining power ratio, the urgency of the delay, and the data packet length factor, a multi-dimensional priority evaluation system is constructed to obtain the comprehensive priority index of the data packet. Based on the current position and velocity vector of the drone, predict the signal strength indication value at future moments, and set the switching trigger threshold and execution threshold; When the predicted signal strength indicator value is lower than the handover trigger threshold, the handover preparation process is initiated. Available candidate communication targets are scanned in parallel. Based on the signal strength indicator value, the real-time distance between the UAV and the candidate target, and the load rate of the candidate target, a comprehensive score is calculated using a weighted formula after normalization. The target with the best comprehensive score is selected to initiate a handover request. The handover target node is based on the sum of the new load and its own load. If it exceeds its own maximum load threshold, it will refuse access and trigger a reselection of the communication target with the second highest comprehensive score as the handover target. Once the target network agrees to accept the handover, dual connections are maintained during the handover transition until the new link quality is stable and meets the handover execution threshold, at which point the original connection is officially disconnected and the occupied resources are released.

4. The intelligent AI-integrated UAV air-to-ground communication method according to claim 1, characterized in that, The specific operation process of step S3 is as follows: For scenarios requiring data transmission to ground base stations, multiple drones within a preset coordination range of the cluster head node are selected to form a virtual antenna array. The cluster head node broadcasts synchronization instructions, including the target direction vector, carrier center frequency, and transmission time reference, to the coordinating drones. Each coordinating drone synchronizes its clock based on a two-way time synchronization protocol and calculates the amplitude and phase of the transmitted signal based on the beamforming weights distributed by the cluster head node. The signals are transmitted simultaneously at the synchronization time reference, enabling the signals to be superimposed in phase and combined with power at the receiving end. Each UAV dynamically selects the modulation scheme and coding rate based on the real-time link quality index and signal strength indicator; the signal-to-noise ratio (SNR) is obtained based on the signal strength indicator and noise power, and a mapping table between the SNR and the modulation and coding scheme is constructed. At the same time, the SNR threshold is dynamically adjusted based on the modulation scheme; the transmitting end embeds the selected scheme identifier in the data frame header, and the receiving end adjusts the demodulation and decoding parameters based on the frame header.

5. The intelligent AI-integrated UAV air-to-ground integrated communication method according to claim 1, characterized in that, The specific operation steps of the priority queue scheduling and hybrid automatic retransmission mechanism are as follows: The specific steps for priority queue scheduling are as follows: Each UAV maintains its local transmission queue and sorts data packets in descending order based on the overall priority index. Within the allocated communication time slots, high-priority data packets are sent first, and fragmented transmission of excessively long data packets is allowed. When a high-priority urgent data packet arrives, a preemptive scheduling mechanism is activated to interrupt the transmission of low-priority data packets or request insertion into an urgent time slot from a neighboring node. The specific operation steps of the hybrid automatic repeater mechanism are as follows: The sending end assigns a unique sequence number to each data packet it sends and starts a retransmission timer; The receiving end performs cyclic redundancy check on the received data packets; if the check passes, it replies with an acknowledgment frame and delivers the data packets. If the verification fails, cache the soft information and reply with a negative confirmation frame; The sending end selects either an additional redundant retransmission or a full retransmission strategy based on the number of retransmissions and the link quality index, until success or the maximum number of retransmissions is reached; if an acknowledgment frame is received, the timer is stopped and the sending buffer is cleared.

6. The intelligent AI-integrated UAV air-to-ground integrated communication method according to claim 1, characterized in that, The specific operation process of step S4 is as follows: The communication performance monitoring unit collects key performance indicators in real time, including end-to-end latency, data packet delivery rate, throughput, handover success rate, and energy efficiency; the collected key performance indicators are normalized and combined with a weighted formula to obtain comprehensive performance indicators. The link quality index is updated based on a deep reinforcement learning algorithm, using the current network state as the state space, adjustable communication parameters as the action space, and comprehensive performance indicators as the reward function. The data is then distributed to each UAV communication unit for adaptive adjustments. The AI ​​learning unit is incrementally trained based on accumulated new data at fixed intervals. When the overall performance index is detected to be less than the preset threshold and the duration exceeds the maximum allowable time, the alarm mechanism is triggered; the emergency plan is executed, including suspending non-critical data transmission, switching to the most robust BPSK modulation mode, increasing the transmission power to the upper limit, and shortening the switching decision time; at the same time, an anomaly report including fault characteristics, scope of impact, and measures already taken is sent to the ground control center.

Images