An adaptive scheduling method for a communication drone swarm

Abstract

The application discloses a kind of adaptive scheduling methods of communication unmanned aerial vehicle group, it is related to unmanned aerial vehicle control field, by multimodal perception module real-time acquisition water surface dynamic data, constructs water surface dynamic digital twin model;Based on the output data of water surface dynamic digital twin model, using time-space graph convolution network generates electromagnetic environment simulation atlas;Identify the edge of strong interference area in electromagnetic environment simulation atlas, deploy deployable intelligent reflector array, and dynamically optimize the phase configuration of intelligent reflector unit by deep reinforcement learning algorithm;According to the spatial distribution of strong interference area, the unmanned aerial vehicle group is divided into communication subgroup and monitoring subgroup;Monitoring subgroup scans strong interference area, obtains local fluctuation parameter, and real-time feedback data to digital twin model;Using distributed model predictive control framework to optimize formation of communication subgroup, solve multi-objective optimization problem, generate unmanned aerial vehicle position adjustment instruction.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) control, specifically to an adaptive scheduling method for a swarm of communication UAVs. Background Technology

[0002] Current drone swarm scheduling technologies primarily rely on radio communication networks (such as ad hoc networks and software-defined networks) to achieve collaborative operations among drones, combined with swarm intelligence algorithms (such as ant colony optimization and particle swarm optimization) for path planning and task allocation. In terms of environmental perception, traditional methods often use a single sensor (such as radar or lidar) to collect data and construct static 3D models using digital twin technology to support decision-making. For example, the Gaoyou irrigation district uses drone oblique photography to construct a digital twin model of its water conservancy project. Anti-interference technologies mainly employ spectrum switching or low-power communication strategies, such as the "firefly"-inspired optical communication drone proposed by Northwestern Polytechnical University, which utilizes the electromagnetic interference resistance of optical signals to achieve short-range information transmission. Furthermore, existing drone formation optimization is mostly based on genetic algorithms or centralized control, and energy management adopts a fixed energy consumption allocation model.

[0003] In this context, existing communication drone swarm scheduling technologies have at least the following problems: (1) Insufficient communication robustness. Radio communication is susceptible to electromagnetic interference, resulting in a high risk of link interruption. Although optical communication is resistant to interference, it is limited by transmission distance and ambient light, making it difficult to cover a large area. Digital twin models have poor dynamics. Traditional modeling methods rely on offline data updates and cannot synchronize physical environment changes (such as water surface fluctuations and turbulent disturbances) in real time, resulting in lagging simulation results and failing to support dynamic scheduling decisions. (2) The drone swarm functions are fixed (such as dedicated communication or monitoring drones), making it impossible to dynamically adjust the sensor network density according to task requirements, resulting in insufficient data acquisition accuracy or resource waste in local areas. (3) Global optimization methods such as genetic algorithms are computationally time-consuming and cannot meet the real-time adjustment requirements of drone formations in dynamic environments. Summary of the Invention

[0004] The purpose of this invention is to provide an adaptive scheduling method for communication drone swarms, which aims to overcome the aforementioned problems existing in the prior art.

[0005] To achieve the objective, the present invention provides the following technical solution:

[0006] An adaptive scheduling method for a swarm of communication drones includes the following steps:

[0007] Step (1): Collect dynamic water surface data in real time through the multimodal perception module of the UAV swarm, and construct a dynamic digital twin model of the water surface.

[0008] Step (2): Based on the output data of the dynamic digital twin model of the water surface, use the spatiotemporal graph convolutional network to extract the spatial correlation features and temporal evolution law of wave propagation, generate an electromagnetic environment simulation map containing signal reflection path, interference intensity distribution and multipath fading features, and mark the signal interference level at each spatial location.

[0009] Step (3): Identify the regions in the electromagnetic environment simulation map where the signal interference intensity exceeds the first threshold and mark them as strong interference regions; deploy a deployable smart reflective surface array at the edge of the strong interference region, and dynamically optimize the phase configuration of the smart reflective surface unit through a deep reinforcement learning algorithm to form a beam null for suppressing reflection interference.

[0010] Step (4): Based on the spatial distribution of the strong interference area, the UAV swarm is dynamically divided into a communication sub-group and a monitoring sub-group; among them, the monitoring sub-group uses a topology reconstruction algorithm to form a high-density sensor array, performs adaptive focusing scanning on the strong interference area, obtains local fluctuation parameters with sub-meter accuracy, and feeds back the data to the digital twin model in real time for incremental learning.

[0011] Step (5): Use a distributed model predictive control framework to optimize the formation of the communication subgroup, solve the multi-objective optimization problem, and generate UAV position adjustment commands; among which, the optimization objectives include communication quality indicators, energy consumption balance and topology stability.

[0012] Step (6): Dynamically configure the communication parameters of the UAV based on the spectrum situation knowledge graph. When residual interference is detected to exceed the second threshold, trigger the federated learning mechanism to collaboratively decide on the optimal frequency band switching scheme and update the phase distribution of the intelligent reflector synchronously.

[0013] Furthermore, in step (1), the multimodal sensing module includes a millimeter-wave radar array, a distributed laser ranging unit, a surface acoustic wave sensor network, and a polarized light sensor; wherein, the millimeter-wave radar array measures the slope change rate and three-dimensional motion vector of the water surface wave crest, the laser ranging unit detects the water surface spectral characteristics through coherent interference, the surface acoustic wave sensor network captures the air turbulence velocity field at a height of 10 cm above the water surface, and the polarized light sensor inverts the dynamic change of the complex refractive index of the water surface medium; and these multi-source heterogeneous data are fused to construct a dynamic digital twin model of the water surface.

[0014] Furthermore, the dynamic digital twin model of the water surface includes a dynamic mapping layer, an evolution prediction layer, and a physical-signal coupling layer. The dynamic mapping layer uses a spatial topology matching algorithm to align the aforementioned multi-source heterogeneous data with the three-dimensional water surface mesh model in real time. The evolution prediction layer simulates the dynamic evolution of waves based on the finite element method and computational fluid dynamics, and combines Kalman filtering for data fusion and anomaly correction to predict the trend of water surface fluctuations. The physical-signal coupling layer embeds an electromagnetic wave propagation model to quantify the interference of water surface tilt angle changes on the signal reflection path and generate a dynamic electromagnetic environment map.

[0015] Furthermore, step (3) specifically includes:

[0016] Step (301): A ring array is formed around the strong interference area by using the deployable smart reflector mounted on the UAV. The spacing between adjacent smart reflector units satisfies d < λ / 2, where λ is the signal wavelength and the phase resolution is not less than 4 bits.

[0017] Step (302): Establish a mapping model between the phase distribution of the intelligent reflector unit and the orientation of the interference source:

[0018]

[0019] in, For the phase of the nth intelligent reflective surface unit, For direct channel response, The reflection channel response of the intelligent reflector;

[0020] Step (303): Based on the dual-delay deep deterministic strategy gradient algorithm, update the phase distribution according to the real-time electromagnetic environment feedback, and the spacing between adjacent smart reflective surface units satisfies d<λ / 2, and the phase resolution is not less than 4 bits; where λ is the signal wavelength.

[0021] Furthermore, the calculation method for the direct-channel response:

[0022] ;in, This is the direct path attenuation coefficient (including path loss and shadow fading). For the direct path delay, This represents the specific frequency value of the kth frequency sampling point;

[0023] How to calculate the reflection channel response:

[0024] ;in, Let m be the reflection coefficient of the m-th intelligent reflective surface unit. Let m be the initial phase offset of the m-th intelligent reflective surface unit. Let m be the propagation delay from the transmitter to the m-th intelligent reflector unit. This represents the equivalent time delay adjustment introduced by the phase configuration θ.

[0025] Furthermore, the objective function of the multi-objective optimization problem in step (5) is:

[0026]

[0027] Where X represents the position matrix coefficients of the UAV swarm; α, β, γ are the multi-target weights, and α+β+γ=1; QoS represents the quality of communication service, Energy represents the energy consumption balance of the swarm, and Topo_Stab represents the formation deformation index.

[0028] The constraints include:

[0029] drone spacing ;in, , Represents the three-dimensional coordinates of the i / j-th UAV; Indicates the minimum safe distance;

[0030] Signal strength received by ground users ;in, This represents the received signal strength of the k-th ground user. Indicates the communication guarantee threshold;

[0031] Maximum battery capacity difference ;in, This indicates the maximum difference in battery capacity among drones within the group.

[0032] Furthermore, the method for calculating the quality of communication service is as follows:

[0033] Where SINR is the signal-to-interference-plus-noise ratio, Throughput is the effective throughput, ω1 and ω2 are weighting coefficients, and ω1+ω2=1;

[0034] How to calculate group energy balance:

[0035] ;in, The remaining battery percentage of the i-th drone. The average remaining battery level for the group;

[0036] Calculation method of formation deformation index:

[0037] ;in, Let Frobenius norm be the matrix. The time interval between adjacent optimization cycles.

[0038] Furthermore, in step (6), the spectral situation knowledge graph is constructed through a graph attention network.

[0039] Compared with the prior art, the present invention has the following advantages:

[0040] This solution addresses the communication vulnerability, model lag, and resource rigidity issues in traditional UAV swarm scheduling by employing core technologies such as multi-physics sensing, intelligent reflective surface collaboration, and digital twin dynamic evolution. Field tests show that communication availability reaches 98% under level 5 wind and wave conditions, and formation optimization efficiency is significantly improved, providing a robust and adaptive scheduling solution for scenarios such as environmental reconnaissance and disaster monitoring.

[0041] Firstly, this invention integrates multi-source data from millimeter-wave radar, laser ranging, and surface acoustic wave sensors to construct a high-precision dynamic digital twin model of the water surface. This model updates three-dimensional deformation features and electromagnetic interference predictions in real time, fusing multimodal sensing with dynamic digital twin technology. Compared to traditional static modeling, data update latency is reduced to the second level, model accuracy is improved to sub-meter level, and dynamic generation of electromagnetic environment simulation maps is supported.

[0042] Secondly, this invention combines intelligent reflective surface (RIS) active beamforming with optical communication technology to construct a dual-mode anti-interference communication architecture, forming an "active elimination + passive avoidance" mechanism, which can effectively suppress water surface reflection interference.

[0043] Thirdly, in this invention, the UAV swarm is divided into a communication sub-swarm and a monitoring sub-swarm as needed. The monitoring sub-swarm forms a high-density sensor array through topology reconstruction, improving the accuracy of local data acquisition. Employing a distributed model predictive control (DMPC) framework and a parallel particle swarm optimization algorithm, the formation optimization response time is reduced to the millisecond level, while simultaneously achieving multi-objective optimization of communication quality, energy consumption balance, and topology stability. Therefore, this invention enables dynamic resource reconstruction and distributed optimization. Attached Figure Description

[0044] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0045] Specific embodiments of the present invention will now be described with reference to the accompanying drawings. Many details are described below to provide a comprehensive understanding of the invention; however, those skilled in the art will be able to implement the invention without these details.

[0046] like Figure 1 As shown, an adaptive scheduling method for a swarm of communication drones includes the following steps:

[0047] Step (1): Collect dynamic water surface data in real time through the multimodal perception module of the UAV swarm, and construct a dynamic digital twin model of the water surface.

[0048] Specifically, the multimodal sensing module includes a millimeter-wave radar array, a distributed laser ranging unit, a surface acoustic wave sensor network, and a polarization light sensor.

[0049] The process involves several key components: a millimeter-wave radar array measures the rate of change of the slope of water surface wave crests and their three-dimensional motion vectors to quantify the three-dimensional deformation characteristics of the waves; a laser ranging unit detects the spectral characteristics of the water surface through coherent interferometry to identify the frequency distribution and energy density of the waves; a surface acoustic wave sensor network captures the air turbulence velocity field at a height of 10 cm above the water surface to analyze the impact of turbulence disturbances on signal propagation; and a polarized light sensor inverts the dynamic changes in the complex refractive index of the water surface medium to reflect water quality and medium properties. Finally, these multi-source heterogeneous data are fused to construct a dynamic digital twin model of the water surface.

[0050] The dynamic digital twin model of the water surface includes a dynamic mapping layer, an evolution prediction layer, and a physical-signal coupling layer.

[0051] Dynamic mapping layer: The spatial topology matching algorithm is used to align the above multi-source heterogeneous data with the three-dimensional water surface mesh model in real time, so as to achieve millimeter-level precision mapping from physical water surface to digital space.

[0052] Evolution prediction layer: Based on the finite element method (FEM) and computational fluid dynamics (CFD), the dynamic evolution of waves is simulated, and Kalman filtering is used for data fusion and anomaly correction to predict the water surface fluctuation trend in the next few minutes (e.g., five minutes).

[0053] Physics-signal coupling layer: embeds electromagnetic wave propagation models (such as ray tracing), quantifies the interference of water surface tilt angle changes on signal reflection paths, and generates dynamic electromagnetic environment maps.

[0054] Step (2): Based on the output data of the dynamic digital twin model of the water surface, use the temporal-space graph convolutional network (ST-GCN) to extract the spatial correlation features and temporal evolution law of wave propagation, generate an electromagnetic environment simulation map containing signal reflection path, interference intensity distribution and multipath fading features, and mark the signal interference level at each spatial location.

[0055] The Spatiotemporal Graph Convolutional Network (ST-GCN) generates electromagnetic environment simulation maps through the following steps:

[0056] Step 1: Based on the wave data of the water surface grid nodes, extract spatial correlation features through graph convolution (GCN).

[0057] Specifically, the three-dimensional grid node data (such as wave crest phase and turbulent velocity) output by the dynamic digital twin model of the water surface is used as input. The graph convolutional layer (GCN) models the water surface grid as a graph structure, where nodes represent spatial locations and edge weights reflect the correlation of fluctuations in adjacent regions, and outputs a spatial correlation matrix (such as the inter-regional fluctuation synchronicity index).

[0058] Step 2: Based on historical time series data, extract the wave evolution time series patterns through a Long Short-Term Memory (LSTM) network.

[0059] Specifically, using historical time-series data (such as the wave spectrum evolution of the past 10 seconds) as input, the Long Short-Term Memory Network (LSTM) extracts the periodicity and abrupt changes in the historical time-series data and outputs a time evolution feature vector (such as the fluctuation prediction of the next 5 seconds).

[0060] Step 3: Integrate spatial correlation features, temporal patterns, and electromagnetic parameters, and calculate signal reflection paths, interference intensity distribution, and multipath fading characteristics through a fully connected layer to generate an electromagnetic environment simulation map.

[0061] Step (3): Identify regions in the electromagnetic environment simulation map where the signal interference intensity exceeds the first threshold (e.g., -90dBm) and mark them as strong interference regions; deploy deployable smart reflector (RIS) arrays at the edge of strong interference regions, and dynamically optimize the phase configuration of the smart reflector units through deep reinforcement learning algorithms to form beam nulls for suppressing reflection interference.

[0062] Specifically, it includes:

[0063] Step (301): A ring array is formed around the strong interference area using deployable intelligent reflective surface units mounted on a UAV, thus forming an intelligent reflective surface array. The spacing between adjacent intelligent reflective surface units satisfies d < λ / 2, where λ is the signal wavelength, and the phase resolution is not less than 4 bits.

[0064] Step (302): Establish a mapping model between the phase distribution of the intelligent reflector unit and the orientation of the interference source:

[0065] .

[0066] in, The phase of the nth intelligent reflective surface unit;

[0067] For the direct channel response, the calculation method is as follows:

[0068] ;in, This is the direct path attenuation coefficient (including path loss and shadow fading). For the direct path delay, This represents the specific frequency value of the kth frequency sampling point;

[0069] The calculation method for the reflection channel response of the intelligent reflector is as follows:

[0070] ;in, Let m be the reflection coefficient of the m-th intelligent reflective surface unit. Let m be the initial phase offset of the m-th intelligent reflective surface unit. Let m be the propagation delay from the transmitter to the m-th intelligent reflector unit. This represents the equivalent time delay adjustment introduced by the phase configuration θ.

[0071] Step (303): Based on the dual-delay deep deterministic policy gradient algorithm (TD3), the phase distribution is updated according to the real-time electromagnetic environment feedback, and the spacing between adjacent smart reflector units satisfies d < λ / 2, with a phase resolution of not less than 4 bits; where λ is the signal wavelength. The dual-delay deep deterministic policy gradient (TD3) algorithm is used to update the phase distribution of the smart reflector units in real time.

[0072] Step (4): Based on the spatial distribution of the strong interference area, the UAV swarm is dynamically divided into a communication sub-swarm and a monitoring sub-swarm. The monitoring sub-swarm uses a topology reconstruction algorithm to form a high-density sensor array, performs adaptive focusing scanning on the strong interference area, obtains local fluctuation parameters with sub-meter accuracy, and feeds back the data to the digital twin model in real time for incremental learning.

[0073] Step (5): The distributed model predictive control framework is used to optimize the formation of the communication subgroup, solve the multi-objective optimization problem, and generate UAV position adjustment commands. The optimization objectives include communication quality indicators, energy consumption balance, and topology stability.

[0074] Specifically, the objective function of the multi-objective optimization problem in step (5) is:

[0075]

[0076] Where X represents the coefficients of the UAV swarm position matrix, X∈R N×3 These are the three-dimensional coordinates of N drones.

[0077] α, β, and γ are the weights for multiple objectives, and α + β + γ = 1. The default values ​​are α = 0.5, β = 0.3, and γ = 0.2. Of course, α, β, and γ can be dynamically adjusted according to task requirements.

[0078] QoS stands for Quality of Service, and is calculated as follows:

[0079] Where SINR is the signal-to-interference-plus-noise ratio, Throughput is the effective throughput, ω1 and ω2 are weighting coefficients, and ω1+ω2=1, with the default ω1=0.6 and ω2=0.4.

[0080] Energy represents the energy balance of a group, calculated as follows:

[0081] ;in, The remaining battery percentage of the i-th drone. This represents the average remaining battery level for the group.

[0082] Topo_Stab represents the formation deformation index, calculated as follows:

[0083] ;in, Let Frobenius norm be the matrix. The time interval between adjacent optimization cycles.

[0084] The constraints of the multi-objective optimization problem include the following three.

[0085] Constraint 1 is used to prevent collisions between drones:

[0086] ;in, , , representing the three-dimensional coordinates of the i / j-th UAV; : Indicates the minimum safe distance.

[0087] Constraint 2 is used to ensure user service quality: ;in, This represents the received signal strength of the k-th ground user. This indicates the communication guarantee threshold, such as -90dBm.

[0088] Constraint 3 is designed to prevent individual drones from running out of power prematurely:

[0089] ;in, This indicates the maximum difference in battery capacity among drones within the group.

[0090] Step (6): Dynamically configure the communication parameters of the UAV based on the spectrum situation knowledge graph. When residual interference is detected to exceed the second threshold (e.g., -80dBm), trigger the federated learning mechanism to collaboratively decide on the optimal frequency band switching scheme and synchronously update the phase distribution of the intelligent reflector.

[0091] Specifically, the spectral situational knowledge graph is constructed using a graph attention network (GAT) and includes:

[0092] Historical interference feature database, used to store the <frequency, polarization, interference mode> triple;

[0093] A dynamic association rule base is used to establish the mapping relationship between interference features and <available frequency bands, modulation and coding schemes>.

[0094] The federated learning framework is a local spectrum decision model that aggregates multiple drone swarms to generate globally optimal frequency band switching strategies.

[0095] Step (7): When a UAV fails in the monitoring subgroup, perform the following: (a) re-divide the sensor array responsibility area based on the Voronoi diagram; (b) activate the neighboring UAV sensor redundancy mode and increase the sampling frequency to 1.5 times the baseline value; (c) limit the formation optimization search space radius to 60% of the initial value to ensure real-time response speed.

[0096] Experimental testing and verification

[0097] I. Comparison of Communication Availability Data

[0098]

[0099] II. Comparison of Formation Optimization Efficiency Data

[0100]

[0101] III. Experimental Design and Statistical Validation

[0102] 1. Variable control

[0103] (1) Fixed environmental parameters: wind speed 14m / s (level 5 wind and waves), water temperature 20±2℃, humidity 60±5%.

[0104] (2) The comparison group used the same hardware platform (DJI Matrice 300 RTK UAV, H20T sensor payload).

[0105] (3) Eliminate traffic fluctuation interference: Only analyze the communication data of the same user group (reconnaissance task group).

[0106] 2. Significance test

[0107] (1) A two-sample t-test was used (α=0.05):

[0108] The difference in communication availability is p = 1.2 × 10⁻⁶. -9 (Much less than 0.05)

[0109] The optimized response time difference is p = 3.7 × 10⁻⁶. -12

[0110] (2) Power analysis (Power=0.8): The minimum required sample size is 512 groups, and the actual measured sample size is 864 groups, which meets the requirements.

[0111] 3. Robustness verification

[0112] (1) Extreme conditions test: Communication availability remains at 96.2% under the scenario of sudden gust (instantaneous wind speed of 18m / s).

[0113] (2) Fault injection test: After randomly removing 20% ​​of the drone nodes, the success rate of formation optimization only decreased by 4.7% (compared to a 31% decrease in the traditional solution).

[0114] IV. Comparison Benchmark with Industry Standards

[0115]

[0116] The above data shows that, through innovative designs such as multimodal perception fusion and intelligent reflective surface collaboration, this invention has achieved a breakthrough improvement in communication reliability and formation efficiency in complex water environments.

[0117] The above are merely specific embodiments of the present invention, but the design concept of the present invention is not limited thereto. Any non-substantial modifications made to the present invention using this concept shall be considered as infringing upon the protection scope of the present invention.

Claims

1. An adaptive scheduling method for a communication drone swarm, characterized in that: The method comprises the following steps: Step (1), real-time collection of water surface dynamic data by a multi-modal sensing module of the UAV group to construct a water surface dynamic digital twin model; Step (2), based on the output data of the water surface dynamic digital twin model, a space-time graph convolution network is used to extract the spatial correlation characteristics and time evolution law of wave propagation, to generate an electromagnetic environment simulation atlas containing signal reflection path, interference intensity distribution and multipath fading characteristics, and to label the signal interference level of each spatial position; Step (3), identifying the area where the signal interference intensity exceeds the first threshold value in the electromagnetic environment simulation atlas, and marking it as a strong interference area; deploying an expandable intelligent reflectarray on the edge of the strong interference area, and dynamically optimizing the phase configuration of the intelligent reflectarray unit through a deep reinforcement learning algorithm to form a beam null for suppressing reflected interference; Step (4), according to the spatial distribution of the strong interference area, the UAV group is dynamically divided into a communication subgroup and a monitoring subgroup; the monitoring subgroup forms a high-density sensing array by using a topology reconstruction algorithm to perform adaptive focused scanning on the strong interference area, to obtain sub-meter precision local fluctuation parameters, and to feed back data to the digital twin model for incremental learning in real time; Step (5), using a distributed model predictive control framework to optimize the formation of the communication subgroup, to solve a multi-objective optimization problem, and to generate UAV position adjustment instructions; the optimization objectives include communication quality indicators, energy consumption balance, and topology stability; Step (6), dynamically configuring the communication parameters of the UAV based on a spectrum situation knowledge graph, when residual interference exceeding a second threshold value is detected, triggering a federated learning mechanism to cooperatively decide an optimal frequency band switching scheme, and synchronously updating the phase distribution of the intelligent reflectarray.

2. The method of claim 1, wherein: In step (1), the multi-modal sensing module includes a millimeter wave radar array, a distributed laser ranging unit, a surface acoustic wave sensor network, and a polarized light sensor; the millimeter wave radar array measures the slope change rate and three-dimensional motion vector of the water surface wave peak, the laser ranging unit detects the water surface spectral characteristics through coherent interference, the surface acoustic wave sensor network captures the air turbulence velocity field of the 10 cm height layer near the water surface, and the polarized light sensor inverts the dynamic change of the complex refractive index of the water surface medium; the multi-source heterogeneous data is fused to construct the water surface dynamic digital twin model. 3.The method of claim 2, wherein: The water surface dynamic digital twin model comprises a dynamic mapping layer, an evolution prediction layer, and a physical-signal coupling layer; the dynamic mapping layer uses a spatial topology matching algorithm to real-time align the multi-source heterogeneous data with a three-dimensional water surface grid model; the evolution prediction layer simulates wave dynamic evolution based on the finite element method and computational fluid dynamics, combines Kalman filtering for data fusion and anomaly correction, and predicts the water surface fluctuation trend; the physical-signal coupling layer embeds an electromagnetic wave propagation model to quantify the interference of water surface inclination change on signal reflection path, and generates a dynamic electromagnetic environment atlas.

4. The method of claim 1, wherein: Step (3) specifically comprises: Step (301), forming a ring array outside the strong interference area by the expandable intelligent reflectarray mounted on the UAV; Step (302), establishing a mapping relationship model between the phase distribution of the intelligent reflectarray unit and the direction of the interference source: wherein, is a phase of the n-th intelligent reflecting surface element, is a direct channel response, is a reflection channel response of the intelligent reflecting surface; Step (303), updating the phase distribution according to the real-time electromagnetic environment feedback based on the double-delay deep deterministic policy gradient algorithm, and the spacing of adjacent intelligent reflecting surface units satisfies d < λ / 2, and the phase resolution is not less than 4 bits; wherein, λ is the signal wavelength.

5. The adaptive scheduling method of the communication UAV group according to claim 4, characterized in that: The calculation method of the direct channel response is: ; where, is the direct path attenuation coefficient, including path loss and shadow fading, is the direct path delay, is the specific frequency value of the kth frequency sample point; The calculation method of the reflected channel response is: ; wherein is a reflection coefficient of the mthIRS unit, is an initial phase offset of the mthIRS unit, is a propagation delay from the transmitting end to the mthIRS unit, denotes an equivalent delay adjustment introduced by the phase configuration θ.

6. The method of adaptive scheduling of a swarm of communication drones according to claim 1, wherein: The objective function of the multi-objective optimization problem in the step (5) is Wherein, X represents the position matrix coefficient of the UAV group; α, β, γ are multi-objective weights, and α+β+γ=1; QoS represents the communication service quality, Energy represents the group energy consumption balance degree, and Topo_Stab represents the formation change degree index; The constraint conditions include: Inter- drone distance ; wherein, , represents the three-dimensional coordinates of the i / jth drone; represents the minimum safety distance; ground user received signal strength ; wherein, represents the received signal strength of the kth ground user, represents a communication assurance threshold; maximum battery power difference ; wherein, represents the maximum battery power difference among the drones within the group.

7. The method of adaptive scheduling of a swarm of communication drones according to claim 6, wherein: The calculation method of the communication service quality is: ; wherein SINR is signal to interference and noise ratio, Throughput is effective throughput, and ω1, ω2 are weight coefficients, and ω1+ω2=1. The calculation method of the group energy consumption balance degree is: ; wherein, is the percentage of the remaining battery of the ith drone, is the group average remaining power; The calculation method of the formation change degree index is: ; wherein, is the matrix Frobenius norm, is the time interval of adjacent optimization periods.

8. The method of adaptive scheduling of a swarm of communication drones according to claim 1, wherein: In the step (6), the spectrum situation knowledge graph is constructed through a graph attention network.

Images