Unmanned aerial vehicle nest adaptive ad hoc network communication method and device

Abstract

The invention discloses an unmanned aerial vehicle nest adaptive ad hoc network communication method and device, a communication interface module sends received external equipment data to a scheduling module through a main control module, and the scheduling module adds the external equipment data into queues with different priorities according to the type of the external equipment data; the scheduling module carries out network bandwidth allocation on the data in the queue by using a bandwidth allocation algorithm combined with multi-task deep reinforcement learning and a generative adversarial network model according to the available bandwidth resource of the current frequency band to obtain a bandwidth allocation strategy, and if the data is in a transmitting time slot of the communication device at present, the scheduling module carries out network bandwidth allocation on the data in the queue by using the bandwidth allocation algorithm combined with the multi-task deep reinforcement learning and the generative adversarial network model; according to the technical scheme of the invention, the data in the queue is sent to the baseband module according to the bandwidth allocation strategy and the priority of the queue, and the baseband module transmits the data in the queue through the radio frequency module, so that the reliability and the real-time performance of key service transmission are preferentially ensured, the bandwidth utilization rate is effectively improved, and the efficiency and the reliability of the nest communication of the unmanned aerial vehicle are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle (UAV) data transmission, and in particular to a method and device for UAV nest adaptive self-organizing network communication. Background Art

[0002] Drones are flexible and convenient, finding widespread use in areas such as power inspections and emergency rescue. Drone nests (also known as airports, hangars, and aircraft carriers) can store, dispatch, and power drones, enabling unmanned, autonomous operations. However, existing microwave communication remote control methods for drones cannot traverse mountains or buildings. Drones can lose communication when obstructed by mountains, and in severe cases, can crash. Relay self-organizing networking technology extends the communication range and increases signal coverage by setting up multiple relay nodes, thereby improving the stability and reliability of data transmission. However, most existing drones only support single-node networking. There are no unified standards or methods for optimizing communication links or scheduling communication node data during multi-node networking. Mountainous areas are prone to complex and changing climate and terrain, and relay node connections and transmission paths can change frequently. Ensuring that drone communication links are adaptive and reliable, ensuring safe and reliable flight operations, is the core issue addressed in this application.

[0003] Traditional self-organizing networking methods mostly consider aerial networking, failing to fully leverage the advantages of ground-based networking in drone nests. Patent document CN 202410680497.X proposes a multi-UAV communication resource allocation method based on multi-agent reinforcement learning. It introduces global and local rewards and task decomposition to optimize the multi-agent deep deterministic policy gradient algorithm. This method comprehensively considers the overall communication performance of multiple drones and the different communication mission requirements of drones, improving the communication efficiency of the multi-UAV communication system. Patent document CN202411865580.0 proposes a drone-to-drone communication method and device based on self-organizing networking that supports hierarchical audio and video transmission, providing communication reliability through hierarchical audio and video transmission. It does not consider reliable communication methods between task drones in the case of multi-UAV + drone nest networking. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method and device for adaptive self-organizing network communication of unmanned aerial vehicle (UAV) nests, which can improve the efficiency and reliability of UAV nest communication.

[0005] In order to solve the above technical problems, a technical solution adopted by the present invention is: A method for communicating in an adaptive self-organizing network of a drone nest, comprising the steps of: The communication interface module sends the received external device data to the scheduling module through the main control module; The scheduling module adds the external device data to queues of different priorities according to the type of the external device data, and determines available bandwidth resources in the current frequency band; The scheduling module allocates network bandwidth to the data in the queue according to the available bandwidth resources of the current frequency band using a bandwidth allocation algorithm that combines multi-task deep reinforcement learning and a generative adversarial network model to obtain a bandwidth allocation strategy; The scheduling module determines whether the communication device is currently in a transmission time slot, and if so, sends the data in the queue to the baseband module according to the bandwidth allocation strategy and the priority of the queue; The baseband module transmits the data in the queue through the radio frequency module.

[0006] In order to solve the above technical problems, another technical solution adopted by the present invention is: A drone nest adaptive self-organizing network communication device includes a communication interface module, a scheduling module, a baseband module, and a main control module. The communication interface module includes a first memory, a first processor, and a first computer program stored in the first memory and executable on the first processor. The scheduling module includes a second memory, a second processor, and a second computer program stored in the second memory and executable on the second processor. The baseband module includes a third memory, a third processor, and a third computer program stored in the third memory and executable on the third processor. The main control module includes a fourth memory, a fourth processor, and a fourth computer program stored in the fourth memory and executable on the fourth processor. When the first processor executes the first computer program, the following steps are implemented: Sending the received external device data to the scheduling module through the main control module; When the second processor executes the second computer program, the following steps are implemented: Adding the external device data to queues of different priorities according to the type of the external device data, and determining available bandwidth resources in the current frequency band; Allocate network bandwidth for the data in the queue according to the available bandwidth resources of the current frequency band using a bandwidth allocation algorithm that combines multi-task deep reinforcement learning and a generative adversarial network model to obtain a bandwidth allocation strategy; Determining whether the communication device is currently in a transmission time slot, and if so, sending the data in the queue to the baseband module according to the bandwidth allocation strategy and the priority of the queue; When the third processor executes the third computer program, the following steps are implemented: The data in the queue is transmitted through a radio frequency module.

[0007] The beneficial effects of the present invention are as follows: the communication interface module sends the received external device data to the scheduling module through the main control module. The scheduling module adds the external device data to queues of different priorities according to the type of the external device data and determines the available bandwidth resources of the current frequency band. The scheduling module uses a bandwidth allocation algorithm that combines multi-task deep reinforcement learning and a generative adversarial network model to allocate network bandwidth to the data in the queue according to the available bandwidth resources of the current frequency band to obtain a bandwidth allocation strategy. If the communication device is currently in a transmission time slot, the data in the queue is sent to the baseband module according to the bandwidth allocation strategy and the priority of the queue. The baseband module transmits the data in the queue through the radio frequency module, thereby adding the data to queues of different priorities, giving priority to ensuring the reliability and real-time performance of key business transmission. The bandwidth allocation algorithm that combines multi-task deep reinforcement learning and a generative adversarial network model is used to allocate network bandwidth, effectively improving bandwidth utilization. The data in the queue is sent according to the bandwidth allocation strategy and priority, which can flexibly adapt to changes in wireless channel quality and ensure reliability in harsh communication environments. In particular, when channel performance degrades, key data is transmitted first to avoid system communication interruption, thereby improving the efficiency and reliability of drone nest communication. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 This is a flowchart of the steps of a drone nest adaptive self-organizing network communication method according to an embodiment of the present invention; Figure 2 This is a schematic structural diagram of a drone nest adaptive ad hoc network communication device according to an embodiment of the present invention; Figure 3 Schematic diagram of a typical application scenario of multiple drone nests and drone combination adaptive networking in a drone nest adaptive self-organizing network communication method according to an embodiment of the present invention; Figure 4 Schematic diagram showing throughput comparison of schemes in a method for adaptive self-organizing network communication of drone nests according to an embodiment of the present invention; Figure 5 The figure is a schematic diagram of the bandwidth allocation effect in a drone nest adaptive self-organizing network communication method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0009] To illustrate the technical content, achieved objectives and effects of the present invention in detail, the following description is given in conjunction with the embodiments and accompanying drawings.

[0010] Please refer to Figure 1 , a drone nest adaptive self-organizing network communication method, comprising the steps of: The communication interface module sends the received external device data to the scheduling module through the main control module; The scheduling module adds the external device data to queues of different priorities according to the type of the external device data, and determines available bandwidth resources in the current frequency band; The scheduling module allocates network bandwidth to the data in the queue according to the available bandwidth resources of the current frequency band using a bandwidth allocation algorithm that combines multi-task deep reinforcement learning and a generative adversarial network model to obtain a bandwidth allocation strategy; The scheduling module determines whether the communication device is currently in a transmission time slot, and if so, sends the data in the queue to the baseband module according to the bandwidth allocation strategy and the priority of the queue; The baseband module transmits the data in the queue through the radio frequency module.

[0011] From the above description, it can be seen that the beneficial effect of the present invention is that: the communication interface module sends the received external device data to the scheduling module through the main control module, and the scheduling module adds the external device data to queues of different priorities according to the type of the external device data, and determines the available bandwidth resources of the current frequency band. The scheduling module uses a bandwidth allocation algorithm that combines multi-task deep reinforcement learning and a generative adversarial network model to allocate network bandwidth to the data in the queue according to the available bandwidth resources of the current frequency band, and obtains a bandwidth allocation strategy. If the current transmission time slot of the communication device is in the transmission time slot, the data in the queue is sent to the baseband module according to the bandwidth allocation strategy and the priority of the queue. The baseband module transmits the data in the queue through the radio frequency module, thereby adding the data to queues of different priorities, giving priority to ensuring the reliability and real-time performance of key business transmission, and using the bandwidth allocation algorithm that combines multi-task deep reinforcement learning and a generative adversarial network model to allocate network bandwidth, effectively improving bandwidth utilization, sending data in the queue according to the bandwidth allocation strategy and priority, can flexibly adapt to changes in wireless channel quality, ensure reliability in harsh communication environments, especially when channel performance degrades, by prioritizing the transmission of key data, avoiding system communication interruption, thereby improving the efficiency and reliability of drone nest communication.

[0012] Furthermore, the scheduling module allocates network bandwidth to the data in the queue according to the available bandwidth resources of the current frequency band using a bandwidth allocation algorithm that combines multi-task deep reinforcement learning and a generative adversarial network model, and obtains a bandwidth allocation strategy including: The scheduling module initializes a network model of a deep deterministic policy gradient algorithm, wherein the network model includes a policy network and a value network; Initialize the bandwidth allocation strategy and define the reward function of the network model; Setting hyperparameters of the network model; Acquire the current status based on the real-time network status, wherein the status includes the available bandwidth resources of the current frequency band, the used bandwidth, the current network delay, the current packet loss rate, the priority requirements of each traffic type, and the predicted bandwidth requirements of each traffic type; selecting a bandwidth allocation action according to the state and an output of the policy network, the bandwidth allocation action including bandwidth allocation for each traffic type; Dynamically adjust the priorities of different traffic types according to the bandwidth allocation action and allocate bandwidth in real time according to network conditions; After executing the bandwidth allocation action, obtaining a change in the network state, and using the reward function to calculate a new state and its corresponding reward based on the change; Use the generative adversarial network model to predict future bandwidth demand based on historical traffic data to obtain the predicted bandwidth demand; The network model is updated using a deep deterministic policy gradient algorithm based on the state, the bandwidth allocation action, the reward, and the new state, and the step of obtaining the current state based on the real-time network state is returned to be executed based on the predicted bandwidth demand and the updated network model to perform multiple iterative training until the model converges and an optimal bandwidth allocation strategy is obtained.

[0013] As can be seen from the above description, the combination of the deep deterministic policy gradient algorithm and the generative adversarial network model can accurately predict future bandwidth demand and dynamically adjust bandwidth allocation, effectively improving the utilization efficiency of network resources and avoiding the waste and over-allocation of bandwidth resources. It also obtains the network status in real time and selects bandwidth allocation actions based on the status and policy network output. It can flexibly respond to the changing priority requirements of different traffic in the network, ensure the transmission quality of key services, and enhance the adaptability and stability of the network. Multiple iterative training until the model converges can continuously optimize the bandwidth allocation strategy to make it more suitable for the actual network environment and further improve network performance.

[0014] Furthermore, the reward function is specifically: R(st,at)=α1×P Score +α2×B Utilization -α3×D Penalty -α4×L Penalty ; Where R(st,at) represents the reward function, α1 represents the priority matching score weight, α2 represents the bandwidth utilization weight, α3 represents the network delay penalty weight, α4 represents the packet loss rate penalty weight, and P Score Indicates the scoring index of bandwidth allocation for different priority traffic in the current state, B Utilization Indicates the utilization efficiency of the total bandwidth resources of the system, D PenaltyIndicates the transmission delay of various types of traffic in the network, L Penalty Indicates the network packet loss rate.

[0015] From the above description, we can see that the reward includes P Score ,The bandwidth allocation strategy can be adjusted according to the traffic priority. The higher the bandwidth allocation for high-priority traffic (such as video and voice), the greater the reward. Utilization Improves the utilization of network bandwidth and avoids bandwidth resource waste. Penalty and L Penalty By penalizing high latency and packet loss rates, the quality of real-time traffic is avoided, and the reward function can effectively ensure the quality of bandwidth allocation.

[0016] Furthermore, the baseband module transmitting the data in the queue through the radio frequency module includes: The baseband module selects a modulation method according to the current channel environment and requirements to encode and modulate the data in the queue to obtain a baseband signal, and transmits the baseband signal to the radio frequency module; The radio frequency module converts the baseband signal into a radio frequency signal and transmits the radio frequency signal through an antenna.

[0017] As can be seen from the above description, the coordinated work of the baseband module and the RF module enables efficient, reliable and flexible data transmission.

[0018] Furthermore, the baseband module selects a modulation method according to the current channel environment and requirements to encode and modulate the data in the queue to obtain a baseband signal, including: After adding protection time slot data to the data in the queue, the baseband module selects a modulation method according to the current channel environment and requirements to perform coding and modulation on the data in the queue to obtain a baseband signal.

[0019] As can be seen from the above description, adding guard time slot data to the data in the queue first and then coding and modulating the data in the queue can resist multipath effects and reduce inter-symbol interference, thereby improving the anti-interference capability of the system.

[0020] Furthermore, determining the available bandwidth resources of the current frequency band includes: The scheduling module detects the signal strength and noise strength of the external device data, and determines the signal power and the noise power according to the signal strength and the noise strength; The scheduling module calculates a signal-to-noise ratio according to the signal power and the noise power, and calculates available bandwidth resources of the current frequency band according to the signal-to-noise ratio and the total bandwidth of the current frequency band.

[0021] As can be seen from the above description, calculating the signal-to-noise ratio can more accurately evaluate the communication quality of the current frequency band. Calculating available bandwidth resources based on the signal-to-noise ratio and the total bandwidth of the current frequency band makes bandwidth allocation more dynamic and flexible. By accurately evaluating available bandwidth resources, bandwidth can be allocated more reasonably, improving overall network performance. This not only improves data transmission efficiency, but also reduces delays and packet loss caused by insufficient or excessive bandwidth allocation.

[0022] Furthermore, the scheduling module calculates a signal-to-noise ratio according to the signal power and the noise power, specifically: SNR = 10 × log 10 (P signal / P noise ); Where SNR represents signal-to-noise ratio, P signal Indicates signal power, P noise Represents the noise power.

[0023] As can be seen from the above description, a high SNR means better signal quality, low communication bit error rate, and high communication quality. By calculating the signal-to-noise ratio, the relative strength of the signal and noise can be quantified, thereby effectively evaluating communication performance.

[0024] Furthermore, the calculation of the available bandwidth resources of the current frequency band according to the signal-to-noise ratio and the total bandwidth of the current frequency band is specifically as follows: B available =B total ×log2(1+SNR); Where B available Indicates the available bandwidth resources of the current frequency band, B total Indicates the total bandwidth of the current frequency band.

[0025] As can be seen from the above description, calculating the available bandwidth resources in the current frequency band is beneficial for subsequent more accurate utilization of spectrum resources and improving data transmission efficiency.

[0026] Furthermore, before the communication interface module sends the received external device data to the scheduling module through the main control module, it also includes: The radio frequency module receives radio frequency signals from the outside world and filters the radio frequency signals to obtain filtered radio frequency signals; The RF module converts the RF signal into a baseband signal and transmits the baseband signal to the baseband module; The baseband module demodulates the baseband signal to obtain a demodulated signal, and decodes the demodulated signal to obtain payload data; The baseband module transmits the payload data to the scheduling module; The scheduling module determines whether the payload data is sent to the communication device. If not, the scheduling module adds the payload data to queues of different priorities according to the type of the payload data. If so, the scheduling module extracts payload data from the payload data and transmits the payload data to the main control module. The main control module sends the payload data to the communication interface module; The communication interface module sends the payload data to an external device according to a protocol corresponding to the payload data; The communication interface module receives external device data sent by the external device.

[0027] From the above description, it can be seen that the scheduling module adds payload data to queues of different priorities according to its type, realizing flexible scheduling and management of data. This priority mechanism can ensure the priority processing and transmission of critical data, improve the response speed and reliability. The scheduling module determines whether the data is sent to the communication device and performs corresponding processing. The use of intelligent distribution mechanism ensures that the data can accurately reach the target module, improving the overall communication efficiency and stability.

[0028] Please refer to Figure 2 , a drone nest adaptive self-organizing network communication device, including a communication interface module, a scheduling module, a baseband module and a main control module, the communication interface module including a first memory, a first processor and a first computer program stored on the first memory and capable of running on the first processor, the scheduling module including a second memory, a second processor and a second computer program stored on the second memory and capable of running on the second processor, the baseband module including a third memory, a third processor and a third computer program stored on the third memory and capable of running on the third processor, the main control module including a fourth memory, a fourth processor and a fourth computer program stored on the fourth memory and capable of running on the fourth processor, when the processor executes the computer program, it implements the steps of the corresponding modules in the above-mentioned drone nest adaptive self-organizing network communication method.

[0029] The above-mentioned drone nest adaptive self-organizing network communication method and device of the present invention can be applied to drone nest communication scenarios, and the following is an explanation through specific implementation methods: Please refer to Figure 1 , embodiment 1 of the present invention is: A method for communicating in an adaptive self-organizing network of a drone nest, comprising the steps of: S1. The radio frequency module receives a radio frequency signal from the outside, and filters the radio frequency signal to obtain a filtered radio frequency signal.

[0030] The filtering processing removes noise and interference signals in the signal and retains signals in an effective frequency band.

[0031] S2, the radio frequency module converts the radio frequency signal into a baseband signal and transmits the baseband signal to the baseband module.

[0032] S3, the baseband module demodulates the baseband signal to obtain a demodulated signal, and decodes the demodulated signal to obtain payload data (such as video, audio, remote control instruction, etc.).

[0033] The demodulation can extract digital information in the signal.

[0034] In an optional embodiment, further comprising: The baseband module detects the received signal strength and noise strength of the baseband signal. The baseband module calculates signal power and noise power according to the received signal strength and the noise strength. The baseband module calculates the signal-to-noise ratio according to the signal power and the noise power, specifically: SNR=10×log 10 (P signal / P noise ); In the formula, SNR represents the signal-to-noise ratio, P signal represents the signal power, and P noise represents the noise power.

[0035] S4, the baseband module transmits the payload data to the scheduling module.

[0036] In an optional embodiment, the baseband module transmits the payload data and the signal-to-noise ratio to the scheduling module.

[0037] S5, the scheduling module determines whether the payload data is sent to a communication device, if not, the payload data is added to a queue with different priorities according to the type of the payload data, if yes, the payload data is extracted from the payload data, and the payload data is transmitted to the master module.

[0038] Specifically, the scheduling module determines whether the payload data is sent to a communication device according to the network identifier in the payload data, if not, the payload data is added to a queue with different priorities according to the type of the payload data, if yes, the payload data is extracted from the payload data, and the payload data is transmitted to the master module.

[0039] The queue with different priorities includes a narrowband control data queue, a voice data queue, and a wideband video data queue.

[0040] In an alternative embodiment, after the step of adding the payload data into queues of different priorities according to the type of the payload data, the method further comprises: The scheduling module calculates available bandwidth resources of the current frequency band according to the signal-to-noise ratio and total bandwidth of the current frequency band. The scheduling module uses a bandwidth allocation algorithm based on multi-task deep reinforcement learning to allocate network bandwidth to the data in the queues according to the available bandwidth resources of the current frequency band, to obtain a bandwidth allocation strategy. The scheduling module determines whether the current time is a transmission time slot of the communication device, and if so, transmits the data in the queues to the baseband module according to the bandwidth allocation strategy and the priority of the queues. The baseband module transmits the data in the queues through the radio frequency module.

[0041] The scheduling module calculates available bandwidth resources of the current frequency band according to the signal-to-noise ratio and total bandwidth of the current frequency band, specifically: B available =B total ×log2(1+SNR); In the formula, B available represents the available bandwidth resources of the current frequency band, B total represents the total bandwidth of the current frequency band.

[0042] S6, the master module transmits the payload data to the communication interface module.

[0043] S7, the communication interface module transmits the payload data to the external device according to the protocol corresponding to the payload data.

[0044] S8, the communication interface module receives external device data sent by the external device.

[0045] S9, the communication interface module transmits the received external device data to the scheduling module through the master module.

[0046] Specifically, the communication interface module transmits the external device data to the master module, and the master module encapsulates the external device data according to the type of the external device data, fills in the corresponding allocated network identifier, and transmits the data to the scheduling module.

[0047] S10, the scheduling module adds the external device data into queues of different priorities according to the type of the external device data, and determines the available bandwidth resources of the current frequency band, specifically comprising: S10.1, the scheduling module adds the external device data into queues of different priorities according to the type of the external device data.

[0048] S10.2. The scheduling module detects the signal strength and noise strength of the external device data, and determines the signal power and the noise power according to the signal strength and the noise strength.

[0049] S10.3. The scheduling module calculates a signal-to-noise ratio according to the signal power and the noise power, and calculates available bandwidth resources of the current frequency band according to the signal-to-noise ratio and the total bandwidth of the current frequency band.

[0050] The specific calculation method of the signal-to-noise ratio and the available bandwidth resources of the current frequency band is as described above and will not be repeated here.

[0051] S11. The scheduling module allocates network bandwidth to the data in the queue based on the available bandwidth resources in the current frequency band using a bandwidth allocation algorithm that combines multi-task deep reinforcement learning and a generative adversarial network model to obtain a bandwidth allocation strategy, specifically including S11.1-S11.9: S11.1. The scheduling module initializes a network model of a deep deterministic policy gradient (DDPG) algorithm, where the network model includes a policy network and a value network.

[0052] S11.2. Initialize the bandwidth allocation strategy and define the reward function of the network model.

[0053] In order to ensure the quality of bandwidth allocation, the reward function is R(st,at): R(st,at)=α1×P Score +α2×B Utilization -α3×D Penalty -α4×L Penalty ; Where α1, α2, α3, and α4 are weighted coefficients (weight parameters) in the reward function, which are used to measure the relative importance of different performance indicators in the total reward. They are key hyperparameters when designing reinforcement learning systems. α1 represents the priority matching score weight, α2 represents the bandwidth utilization weight, α3 represents the network delay penalty weight, α4 represents the packet loss rate penalty weight, and P Score Indicates the scoring indicator of bandwidth allocation for traffic of different priorities in the current state. It is used to measure whether resources are allocated reasonably according to the predetermined traffic priority (control > voice > video, etc.). The bandwidth allocation strategy is adjusted according to the traffic priority. The higher the bandwidth allocation for high-priority traffic (such as video and voice), the greater the reward. Utilization Indicates the utilization efficiency of the total bandwidth resources of the system, which can improve the utilization of network bandwidth and avoid bandwidth resource waste. Penalty Indicates the transmission delay of various types of traffic in the network, L PenaltyIndicates the network packet loss rate. It penalizes high latency and packet loss rate to avoid affecting the quality of real-time traffic.

[0054] S11.3. Set the hyperparameters of the network model, including learning rate, discount factor, and exploration rate.

[0055] S11.4. Obtain the current state according to the real-time network state, wherein the state includes the available bandwidth resources b of the current frequency band. totalb , used bandwidth b usedb , current network delay, current packet loss rate, priority requirements of each traffic type (including priority requirements of control traffic p control , voice traffic priority requirement p voice and the priority requirement of video traffic p video ) and the predicted bandwidth demand for each traffic type (the predicted bandwidth demand for control traffic control , predicted bandwidth demand for voice traffic voice and the predicted bandwidth demand for video traffic video ).

[0056] Among them, the state s t Represents the status information of the UAV communication network, s t =(b totalb , b usedb ,delay,loss,p control ,p voice ,p video ,demand control ,demand voice ,demand video ).

[0057] S11.5. Select a bandwidth allocation action according to the state and the output of the policy network, where the bandwidth allocation action includes bandwidth allocation for each traffic type.

[0058] Wherein, the bandwidth allocation action a t =(b control ,b voice ,b video ), b control Indicates the bandwidth allocation for control traffic, b voice Indicates the bandwidth allocation for voice traffic, b video represents the bandwidth allocation for video traffic, b control +b voice +b video ≤b totalb , that is, to ensure that the total bandwidth allocation does not exceed the available bandwidth.

[0059] S11.6. Dynamically adjust the priorities of different traffic types according to the bandwidth allocation action, and allocate bandwidth in real time according to network conditions.

[0060] S11.7. After executing the bandwidth allocation action, obtain the change in the network state and use the reward function to calculate the new state s according to the change. t+1 and their corresponding rewards.

[0061] S11.8. Use the generative adversarial network model to predict future bandwidth demand based on historical traffic data to obtain predicted bandwidth demand.

[0062] S11.9. Update the network model using a deep deterministic policy gradient algorithm based on the state, the bandwidth allocation action, the reward, and the new state, and return to execute S11.4 based on the predicted bandwidth demand and the updated network model to perform multiple iterative training until the model converges and obtains the optimal bandwidth allocation strategy.

[0063] S12. The scheduling module determines whether the current time slot is a transmission time slot of the communication device. If so, the scheduling module sends the data in the queue to the baseband module according to the bandwidth allocation strategy and the priority of the queue.

[0064] The transmission time slot of the communication device includes a protection sub-time slot, a control data sub-time slot, a voice data sub-time slot and a video data sub-time slot, and the corresponding data is subsequently sent in the corresponding time slot.

[0065] S13. The baseband module transmits the data in the queue through the radio frequency module, specifically including S13.1-S13.2: S13.1. The baseband module selects a modulation method according to the current channel environment and requirements to encode and modulate the data in the queue to obtain a baseband signal, and transmits the baseband signal to the radio frequency module.

[0066] Specifically, after adding protection time slot data to the data in the queue, the baseband module selects a modulation method according to the current channel environment and requirements to encode and modulate the data in the queue to obtain a baseband signal, and transmits the baseband signal to the RF module.

[0067] S13.2. The RF module converts the baseband signal into a RF signal, and transmits the RF signal through an antenna.

[0068] In order to optimize the bandwidth allocation of different traffic types at the same time, a multi-task learning framework is adopted, under which each traffic type (control, voice, video) corresponds to a task, and the reinforcement learning algorithm considers the goals of these tasks to find the globally optimal bandwidth allocation strategy. In the specific implementation, the deep deterministic policy gradient algorithm is used to learn each task, but these tasks share a main network model to accelerate the learning process and share information between traffic demands. In addition to the basic traffic priority consideration, a dynamic priority adjustment mechanism is introduced to dynamically adjust the priority of traffic according to the real-time state of the network, for example, when the network bandwidth is tight, the priority is automatically adjusted according to the real-time performance (such as delay and packet loss rate) of the traffic, to preferentially guarantee delay-sensitive traffic (such as voice and video), and reduce the bandwidth allocation of low-priority traffic (such as control traffic). In addition, in order to more accurately predict the bandwidth demand fluctuations of each traffic type, a generative adversarial network model is combined to generate future bandwidth demand predictions based on historical traffic data. The introduction of this generative model can provide accurate traffic demand predictions for the deep deterministic policy gradient algorithm, avoiding excessive reliance on real-time measurements, thereby improving the forward-looking and accuracy of the bandwidth allocation strategy. The deep deterministic policy gradient algorithm is used to optimize the bandwidth allocation strategy, which is suitable for continuous action space and optimizes the strategy and value function through an actor-critic structure.

[0069] As Figure 3 shown, the unmanned aerial vehicle nest is connected with the ad hoc network communication device through wired / wireless connection, and the typical frequency is 800MHz-5.4GHz. The unmanned aerial vehicle is connected with the ad hoc network communication device through wired connection. The ad hoc network communication devices can communicate bidirectionally. The unmanned aerial vehicle nest and the unmanned aerial vehicle can transmit unmanned aerial vehicle, image transmission and data transmission data. The unmanned aerial vehicle nests can transmit airport conditions and unmanned aerial vehicle route information. The unmanned aerial vehicles can transmit position information, height information, flight conditions and image transmission information. The typical ad hoc network mode is that the unmanned aerial vehicle a is connected with the unmanned aerial vehicle nest 1 through the ad hoc network device 1 and the ad hoc network device a for communication. The unmanned aerial vehicle nest 2 is connected with the unmanned aerial vehicle b through the ad hoc network device 2 and the ad hoc network device b for communication. When the unmanned aerial vehicle c is flying in the air for operation, the surrounding ad hoc network device communication signals are analyzed according to the priority of the scheduling module, and the ad hoc network device and the nest or the unmanned aerial vehicle are preferentially accessed for communication.

[0070] The above method of the application is experimentally evaluated as follows.

[0071] It is assumed that the total bandwidth is 10Mbps, of which the control traffic is expected to be 1Mbps, the voice traffic is expected to be 2Mbps, and the video traffic is expected to be 7Mbps. The control data delay is less than 100ms, and the voice and video traffic delay is less than 200ms.

[0072] As Figure 4 shown, Figure 4The simulation experiment comparison of different network packet loss rates is shown. It can be seen that the method of the application (i.e. the MRL scheme shown in the figure) has better network throughput than the traditional static allocation scheme, indicating better bandwidth utilization.

[0073] As Figure 5 shown, Figure 5 It is shown that in the bandwidth of about 12Mbps, the packet loss rate is 0.1, the HoloWAN device is used to collect and simulate the unmanned aerial vehicle transmission environment of the WiFi band, and the bandwidth allocation of the above method of the application changes with time. After 400ms, the overall bandwidth allocation tends to be stable, because the algorithm collects enough historical data and better predicts the changes of the network.

[0074] In view of the problem that the dynamic change of the transmission distance between the aerial unmanned aerial vehicle node and the ground node in emergency communication may cause the wireless channel quality to deteriorate significantly, which not only affects the unified transmission performance of video, audio, data transmission and remote control data and other services, but also may cause the overall communication efficiency to decline and the delay or interruption of the front and rear command coordination. In order to solve this problem, the application supports independent transmission of different types of services through intelligent service differentiation and priority management, ensures that in the case of simultaneous operation of multiple services, the resource allocation and transmission strategy are dynamically adjusted according to the actual demand, different priorities are set for each service, which can not only preferentially guarantee the reliability and real-time performance of key services (such as remote control data and data transmission), but also take into account the quality requirements of video and audio services, thereby significantly improving the adaptability and overall performance of the emergency communication system in complex dynamic scenarios. In addition, the transmission parameters of different services can be flexibly adjusted according to the change of the channel condition, which provides more stable and efficient protection for emergency communication.

[0075] Please refer to Figure 2 , the second embodiment of the application is: A self-adaptive self-organizing network communication device of an unmanned aerial vehicle nest, comprising a communication interface module (i.e. Figure 2The unmanned aerial vehicle / nest communication interface, the communication interface, the network interface, the USB interface and the media interface, a scheduling module, a baseband module, a master control module, a radio frequency module and a power module, the communication interface module includes a first memory, a first processor and a first computer program stored in the first memory and executable on the first processor, the scheduling module includes a second memory, a second processor and a second computer program stored in the second memory and executable on the second processor, the baseband module includes a third memory, a third processor and a third computer program stored in the third memory and executable on the third processor, the master control module includes a fourth memory, a fourth processor and a fourth computer program stored in the fourth memory and executable on the fourth processor, the radio frequency module includes a fifth memory, a fifth processor and a fifth computer program stored in the fifth memory and executable on the fifth processor, and the processors execute the computer programs to implement the steps of the corresponding modules in the unmanned aerial vehicle nest adaptive ad hoc network communication method in Embodiment I.

[0076] The master control module is responsible for data interaction management between all modules, control of other modules and network routing information update and maintenance. The scheduling module works in a time division mode, is responsible for packaging processing of data to be sent according to a transmission protocol, unpackaging processing of received data, management of matching of data and bandwidth resources and management of priorities of different types of data transmission. The baseband module encodes and modulates transmitted data and decodes and demodulates received data, and the baseband module calculates RSSI and SNR according to received data. The radio frequency module is responsible for conversion processing between baseband data and electromagnetic wave signals. The communication interface module is responsible for data communication with peripheral devices. The power module provides power supply for the hardware single board.

[0077] The scheduling module transmits different types of data in a time-division manner and prioritizes different data types to ensure system communication reliability even when channel quality deteriorates. The scheduling module receives a reference clock signal from the clock module and divides transmission resources into time slots. Its own time slots are divided into protection subslots, control data subslots, voice data subslots, and video data subslots. Based on the system's unique network ID, the scheduling module matches data packets to the subslot corresponding to its own ID. Voice data subslots carry voice signals, with a data rate of 10-100 kbps. Control data subslots carry drone data transmission and remote control data, with a data rate of 10-100 kbps. Video data subslots carry video data, with a data rate of 100 kbps to 5 Mbps. The priority of scheduled data is, from highest to lowest, voice data subslots, control data subslots, and video data subslots. The scheduling module calculates the actual available bandwidth based on the SNR and the current frequency bandwidth, prioritizing voice data, control data (data transmission and remote control data), and video data. When the distance between the drone and ground nodes decreases, causing received power to decrease and signal quality to deteriorate, voice data transmission is prioritized. This allows for rapid distance between the drone and ground network nodes, ensuring that even video communication terminals can maintain normal voice communication between the front and rear.

[0078] The baseband module converts the digital data to be transmitted by the upper layer into an analog signal suitable for transmission in the wireless channel (modulation) and restores the received analog signal to digital data (demodulation). It supports multiple modulation methods to adapt to different channel conditions and rate requirements.

[0079] The RF module amplifies the digital signal generated by the baseband and converts it into an analog signal, or converts the analog signal of the antenna back into a digital signal for the antenna to use for transmission and reception.

[0080] The power module is responsible for powering the main control, RF and other amplifiers. It adopts Buck DC-DC step-down topology, with a typical input value of 12V and an output value of 3.3V-5V.

[0081] In summary, the present invention provides a method and device for adaptive self-organizing network communication of unmanned aerial vehicle nests, wherein the communication interface module sends the received external device data to the scheduling module through the main control module, the scheduling module adds the external device data to queues of different priorities according to the type of the external device data, and determines the available bandwidth resources of the current frequency band, and the scheduling module uses a bandwidth allocation algorithm that combines multi-task deep reinforcement learning and a generative adversarial network model to allocate network bandwidth to the data in the queue according to the available bandwidth resources of the current frequency band, and obtains a bandwidth allocation strategy, if the current time is in the transmission time slot of the communication device, the data in the queue is sent to the baseband module according to the bandwidth allocation strategy and the priority of the queue, and the baseband module transmits the data in the queue through the radio frequency module, thereby adding the data to queues of different priorities, giving priority to ensuring the reliability and real-time performance of key business transmission, and using a bandwidth allocation algorithm that combines multi-task deep reinforcement learning and a generative adversarial network model to allocate network bandwidth, It effectively improves bandwidth utilization. According to the bandwidth allocation strategy and priority of sending data in the queue, it can flexibly adapt to changes in wireless channel quality and ensure reliability in harsh communication environments. Especially when the channel performance degrades, it avoids system communication interruption by prioritizing the transmission of key data, thereby improving the efficiency and reliability of drone nest communication. In addition, through the combination of deep deterministic policy gradient algorithm and generative adversarial network model, it can accurately predict future bandwidth requirements and dynamically adjust bandwidth allocation, effectively improving the utilization efficiency of network resources and avoiding waste and over-allocation of bandwidth resources. It can also obtain network status in real time and select bandwidth allocation actions based on status and policy network output, which can flexibly respond to changes in priority requirements of different traffic in the network, ensure the transmission quality of key services, enhance the adaptability and stability of the network, and iterate training multiple times until the model converges, which can continuously optimize the bandwidth allocation strategy to make it more suitable for the actual network environment and further improve network performance.

[0082] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent transformations made using the contents of the present invention's description and drawings, or directly or indirectly applied in related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A drone nest adaptive self-organizing network communication method, characterized in that: Including steps: The communication interface module sends the received external device data to the scheduling module through the main control module; The scheduling module adds the external device data to queues of different priorities according to the type of the external device data, and determines available bandwidth resources in the current frequency band; The scheduling module allocates network bandwidth to the data in the queue according to the available bandwidth resources of the current frequency band using a bandwidth allocation algorithm that combines multi-task deep reinforcement learning and a generative adversarial network model to obtain a bandwidth allocation strategy; The scheduling module determines whether the communication device is currently in a transmission time slot, and if so, sends the data in the queue to the baseband module according to the bandwidth allocation strategy and the priority of the queue; The baseband module transmits the data in the queue through the radio frequency module.

2. The method for adaptive self-organizing network communication of drone nests according to claim 1, characterized in that: The scheduling module allocates network bandwidth to the data in the queue according to the available bandwidth resources of the current frequency band using a bandwidth allocation algorithm that combines multi-task deep reinforcement learning and a generative adversarial network model. The bandwidth allocation strategy obtained includes: The scheduling module initializes a network model of a deep deterministic policy gradient algorithm, wherein the network model includes a policy network and a value network; Initialize the bandwidth allocation strategy and define the reward function of the network model; Setting hyperparameters of the network model; Acquire the current status based on the real-time network status, wherein the status includes the available bandwidth resources of the current frequency band, the used bandwidth, the current network delay, the current packet loss rate, the priority requirements of each traffic type, and the predicted bandwidth requirements of each traffic type; selecting a bandwidth allocation action according to the state and an output of the policy network, the bandwidth allocation action including bandwidth allocation for each traffic type; Dynamically adjust the priorities of different traffic types according to the bandwidth allocation action and allocate bandwidth in real time according to network conditions; After executing the bandwidth allocation action, obtaining a change in the network state, and using the reward function to calculate a new state and its corresponding reward based on the change; Use the generative adversarial network model to predict future bandwidth demand based on historical traffic data to obtain the predicted bandwidth demand; The network model is updated using a deep deterministic policy gradient algorithm based on the state, the bandwidth allocation action, the reward, and the new state, and the step of obtaining the current state based on the real-time network state is returned to be executed based on the predicted bandwidth demand and the updated network model to perform multiple iterative training until the model converges and an optimal bandwidth allocation strategy is obtained.

3. The method for adaptive self-organizing network communication of drone nests according to claim 2, characterized in that: The reward function is specifically: R(st,at)=α1×P Score +α2×B Utilization -α3×D Penalty -α4×L Penalty ; Where R(st,at) represents the reward function, α1 represents the priority matching score weight, α2 represents the bandwidth utilization weight, α3 represents the network delay penalty weight, α4 represents the packet loss rate penalty weight, and P Score Indicates the scoring index of bandwidth allocation for different priority traffic in the current state, B Utilization Indicates the utilization efficiency of the total bandwidth resources of the system, D Penalty Indicates the transmission delay of various types of traffic in the network, L Penalty Indicates the network packet loss rate.

4. The method for adaptive self-organizing network communication of drone nests according to claim 1, characterized in that: The baseband module transmitting the data in the queue through the radio frequency module includes: The baseband module selects a modulation method according to the current channel environment and requirements to encode and modulate the data in the queue to obtain a baseband signal, and transmits the baseband signal to the radio frequency module; The radio frequency module converts the baseband signal into a radio frequency signal and transmits the radio frequency signal through an antenna.

5. The method for adaptive self-organizing network communication of drone nests according to claim 4, characterized in that: The baseband module selects a modulation method according to the current channel environment and requirements to encode and modulate the data in the queue to obtain a baseband signal, which includes: After adding protection time slot data to the data in the queue, the baseband module selects a modulation method according to the current channel environment and requirements to perform coding and modulation on the data in the queue to obtain a baseband signal.

6. The method for adaptive self-organizing network communication of drone nests according to claim 1, characterized in that: Determining the available bandwidth resources of the current frequency band includes: The scheduling module detects the signal strength and noise strength of the external device data, and determines the signal power and the noise power according to the signal strength and the noise strength; The scheduling module calculates a signal-to-noise ratio according to the signal power and the noise power, and calculates available bandwidth resources of the current frequency band according to the signal-to-noise ratio and the total bandwidth of the current frequency band.

7. The method for adaptive self-organizing network communication of drone nests according to claim 5, characterized in that: The scheduling module calculates the signal-to-noise ratio according to the signal power and the noise power, specifically: SNR=10×log 10 (P signal / P noise ); Where SNR represents signal-to-noise ratio, P signal Indicates signal power, P noise Represents the noise power.

8. The method for adaptive self-organizing network communication of drone nests according to claim 7, characterized in that: The calculating of the available bandwidth resources of the current frequency band according to the signal-to-noise ratio and the total bandwidth of the current frequency band is specifically as follows: B available =B total ×log2(1+SNR); Where B available Indicates the available bandwidth resources of the current frequency band, B total Indicates the total bandwidth of the current frequency band.

9. The method for adaptive self-organizing network communication of drone nests according to claim 1, characterized in that: Before the communication interface module sends the received external device data to the scheduling module through the main control module, it also includes: The radio frequency module receives radio frequency signals from the outside world and filters the radio frequency signals to obtain filtered radio frequency signals; The RF module converts the RF signal into a baseband signal and transmits the baseband signal to the baseband module; The baseband module demodulates the baseband signal to obtain a demodulated signal, and decodes the demodulated signal to obtain payload data; The baseband module transmits the payload data to the scheduling module; The scheduling module determines whether the payload data is sent to the communication device. If not, the scheduling module adds the payload data to queues of different priorities according to the type of the payload data. If so, the scheduling module extracts payload data from the payload data and transmits the payload data to the main control module. The main control module sends the payload data to the communication interface module; The communication interface module sends the payload data to an external device according to a protocol corresponding to the payload data; The communication interface module receives external device data sent by the external device.

10. A drone nest adaptive self-organizing network communication device, comprising a communication interface module, a scheduling module, a baseband module, and a main control module, wherein the communication interface module comprises a first memory, a first processor, and a first computer program stored in the first memory and executable on the first processor, the scheduling module comprises a second memory, a second processor, and a second computer program stored in the second memory and executable on the second processor, the baseband module comprises a third memory, a third processor, and a third computer program stored in the third memory and executable on the third processor, and the main control module comprises a fourth memory, a fourth processor, and a fourth computer program stored in the fourth memory and executable on the fourth processor, characterized in that: When the processor executes the computer program, the steps of the corresponding modules in the drone nest adaptive self-organizing network communication method according to any one of claims 1 to 9 are implemented.

Images