Method for optimizing ai-based aerial drone communication networks
By employing AI-based methods for intelligent perception and data processing, decision-making and optimization, communication and task collaborative management, and resource and energy consumption optimization, the problem of signal interruption in UAV communication networks in complex terrain was solved, thereby improving the continuity and efficiency of UAV communication and extending its flight time.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA TOWER CO LTD
- Filing Date
- 2026-04-03
- Publication Date
- 2026-06-05
AI Technical Summary
During flight, drones are affected by terrain-induced multipath traps, which can cause signal interruptions and a sharp drop in communication quality, especially in scenarios such as urban canyons, valleys, and tunnels, where it is difficult to predict and resolve.
Employing AI-based intelligent sensing and data processing, AI decision-making and optimization, communication and task collaborative management, and resource and energy consumption optimization methods, this system collects data in real time through multiple sensors, optimizes the UAV communication network using reinforcement learning and deep learning algorithms, dynamically adjusts channel and power resources, predicts energy consumption and formulates energy-saving strategies, thereby achieving collaborative operation and resource optimization of multiple UAVs.
It effectively avoids signal interruption caused by terrain-induced multipath traps, improves the continuity and efficiency of UAV communication, extends flight time, and enhances communication quality and mission completion rate in complex environments.
Smart Images
Figure CN122160277A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of communication network technology, and in particular to an optimization method for an AI-based aerial drone communication network. Background Technology
[0002] When a drone is flying in the air, if it is affected by the terrain and encounters a terrain-induced multipath trap, it can cause problems with the drone's communication network, for example: When a drone flies over a canyon, the signal is reflected by the mountains on both sides, creating multipath interference. At a certain flight altitude (such as an integer multiple of 1 / 2 wavelength of the canyon width), deep fading occurs, causing communication interruption. The streets lined with high-rise buildings form signal corridors, and drones flying along these streets may fall into areas of strong interference due to multipath superposition. Terrain-induced multipath traps refer to the phenomenon where, when a drone flies in a specific geographical environment, the terrain structure causes wireless signals to propagate through multiple paths, creating complex interference effects and ultimately leading to a sudden drop in communication quality or even an interruption at a specific location. This problem is characterized by its high degree of concealment and unpredictability, and is particularly prominent in scenarios such as urban canyons, valleys, and tunnels.
[0003] Therefore, how to select the appropriate transverse radio wave signal transmission mode based on the flight terrain in order to prevent communication network failures during UAV flight is a problem that needs to be improved.
[0004] Therefore, an optimization method for aerial drone communication networks based on AI is proposed to address the aforementioned problems. Summary of the Invention
[0005] The purpose of this invention is to propose an optimization method for an AI-based aerial drone communication network in order to solve the above-mentioned problems.
[0006] To achieve the above objectives, the present invention adopts the following technical solution: AI-based optimization methods for aerial drone communication networks include: Intelligent sensing and data processing: Real-time collection of environmental and network status data through multiple sensors, followed by noise reduction, feature extraction and fusion, to provide accurate data support for subsequent decision-making; AI Decision-Making and Optimization Core: Utilize algorithms such as reinforcement learning and deep learning to build intelligent models, achieve lightweight deployment after training, and quickly output drone communication and resource optimization strategies based on network status; Communication and Task Collaboration Management: Dynamically allocate wireless communication resources, optimize network topology, and achieve intelligent allocation and collaborative operation of multiple UAV tasks through multi-agent reinforcement learning, thereby improving network performance and task efficiency; Resource and energy consumption optimization: predict drone energy consumption, formulate energy-saving strategies, and flexibly schedule computing and storage resources.
[0007] Preferably, the intelligent sensing and data processing specifically includes: The drone is equipped with lidar, high-definition cameras, high-precision GPS, and inertial measurement unit sensors; The system collects communication parameters such as channel gain, signal strength, and signal-to-noise ratio through software-defined wireless devices, and monitors network performance indicators in real time; it also has built-in monitoring tools to obtain information on the drone's energy consumption and computing resource usage. Kalman filtering algorithm is used to fuse GPS and IMU data to eliminate noise interference; wavelet transform technology is used to denoise the time series data collected by the sensors. Convolutional neural networks are used to extract features from image data captured by cameras to identify obstacles and target objects.
[0008] Preferably, the AI decision-making and optimization core specifically includes: In UAV path planning, a deep Q-network algorithm is used to learn the optimal flight path; for resource allocation problems, a near-end strategy optimization algorithm is used to dynamically adjust channel and power resources. Graph neural networks are used to model the topology of UAV networks and predict link state changes. When the flight area of the UAV meets the communication network environment corresponding to the terrain-induced multipath trap, the communication network interference of the UAV is analyzed to determine the most suitable transverse radio wave signal transmission mode.
[0009] Preferably, when the flight area of the UAV satisfies the communication network environment corresponding to the terrain-induced multipath trap, the communication network interference of the UAV is analyzed to determine the most suitable transverse radio wave signal transmission mode. The process includes: The canyon terrain is approximated as an open rectangular waveguide, and the width and depth of the canyon terrain are obtained; The formula for calculating the cutoff frequency of the mode is: ; in At the speed of light, The width of the canyon. The depth of the canyon; Calculate the cutoff frequency for each mode in sequence; to obtain the cutoff frequency corresponding to each mode, divide the UAV communication operating frequency by the cutoff frequency corresponding to each mode to obtain the basic satisfaction.
[0010] Preferably, the method further includes: The width and height of the canyon at the current location of the drone are obtained at preset time intervals, and the corresponding cutoff frequency and its corresponding basic satisfaction are output. A preset basic satisfaction threshold is set. If the basic satisfaction level is less than the basic satisfaction threshold, then... Module switching; sequentially obtain each The cutoff frequency of the modulus and the corresponding basic satisfaction level, with the basic satisfaction level greater than the basic satisfaction threshold being considered. The mode is used as the transmission mode.
[0011] Preferably, the drone's flight path, and the canyon height and width at the path's endpoint, are calculated based on the drone's flight speed and a determined time interval. Using the current position of the UAV as the center and the calculated flight path length of the UAV as the radius, construct a cylindrical scanning model along the predicted flight path, obtain the cutoff frequency at each point on the surface of the scanning model, and the basic satisfaction degree corresponding to the cutoff frequency; When the drone reaches the end of its path, if the baseline satisfaction level at the end of the path is less than the baseline satisfaction threshold, then an early warning will be issued within the corresponding time interval during the drone's flight. Mode switching allows the drone to complete the process when it reaches the end of the path within the time interval. Module switching.
[0012] Preferably, the communication and task collaborative management specifically includes: Deep reinforcement learning algorithms are used to optimize subcarrier allocation in orthogonal frequency division multiple access. By monitoring the signal-to-noise ratio of each link in real time, the subcarrier allocation scheme is dynamically adjusted to reduce co-channel interference and improve spectrum utilization. Analysis of UAV network topology based on graph neural network; By using a multi-agent reinforcement learning algorithm, data acquisition, transmission, or relay tasks are dynamically allocated based on the computing power, communication load, and remaining battery power of each UAV.
[0013] Preferably, the resource and energy consumption optimization specifically includes: A drone energy consumption prediction model is built based on LSTM neural network. Input parameters are used to predict the energy consumption under different mission scenarios. Develop dynamic energy-saving strategies based on task priorities; When the computational load on the drone is too high, the AI model will offload the task to a nearby node or edge server through transfer learning;
[0014] Based on data importance and access frequency, deep reinforcement learning is used to dynamically adjust the data caching strategy.
[0015] Preferably, the security monitoring and feedback optimization involves: using machine learning algorithms to detect network security threats, ensuring communication security, monitoring system performance indicators in real time, and iteratively optimizing the AI model based on operational data to ensure stable and efficient system operation.
[0016] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1. This invention uses lidar to scan canyon terrain in real time, calculates the cutoff frequency using a rectangular waveguide model, dynamically adjusts the transverse radio wave transmission mode, predicts terrain changes in advance and triggers mode switching, and, with buffer and rollback mechanisms, avoids signal interruption caused by terrain-induced multipath traps; when the canyon width changes abruptly, mode switching is initiated in advance to ensure that parameters are updated before the UAV reaches the critical area, reducing the risk of communication interruption, while the rollback mechanism can correct erroneous switching and ensure link continuity.
[0017] 2. This invention utilizes multi-agent reinforcement learning to dynamically allocate communication resources and task loads, combined with an LSTM energy consumption prediction model, to achieve collaborative operation and energy consumption optimization of UAV swarms. In disaster relief, the AI model can allocate image acquisition tasks to the optimal UAV based on parameters such as power and computing power, thereby improving the task completion rate. At the same time, the energy consumption prediction error rate is small, and the dynamic energy-saving strategy extends the flight time. Furthermore, task offloading reduces computational latency. Attached Figure Description
[0018] Further details, features, and advantages of this application are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which: Figure 1 This is a system structure diagram of the present invention. Detailed Implementation
[0019] Several embodiments of this application will now be described in more detail with reference to the accompanying drawings to enable those skilled in the art to implement this application. This application may be embodied in many different forms and for various purposes and should not be limited to the embodiments set forth herein. These embodiments are provided to make this application thorough and complete, and to fully convey the scope of this application to those skilled in the art. The embodiments described do not limit this application.
[0020] Unless otherwise defined, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It will be further understood that terms such as those defined in commonly used dictionaries shall be interpreted as having a meaning consistent with their meaning in the relevant field and / or the context of this specification, and shall not be interpreted in an idealized or overly formal sense unless expressly defined herein.
[0021] Example 1
[0022] Its specific implementation method is combined with the appendix Figure 1 Please provide a detailed explanation.
[0023] Appendix Figure 1 The flowchart of the optimization method for an AI-based aerial drone communication network provided in this embodiment of the invention illustrates the complete steps from intelligent sensing and data processing to resource and energy consumption optimization.
[0024] In this embodiment, it includes: Intelligent sensing and data processing: Real-time collection of environmental and network status data through multiple sensors, followed by noise reduction, feature extraction and fusion, to provide accurate data support for subsequent decision-making; Specifically, it includes: The drone is equipped with lidar, high-definition camera, high-precision GPS (positioning accuracy down to the centimeter level), and inertial measurement unit (IMU) sensors. Taking forest fire monitoring as an example, lidar can scan the terrain and vegetation distribution in real time, the camera can identify fire sources through computer vision technology, and the IMU helps the drone maintain a stable flight attitude. The system collects communication parameters such as channel gain, signal strength, and signal-to-noise ratio through software-defined wireless devices, and uses network probes to monitor network performance indicators such as packet loss rate and latency in real time. It also has built-in power sensors and CPU / GPU usage monitoring tools to obtain information on the drone's energy consumption and computing resource usage. Kalman filtering algorithm is used to fuse GPS and IMU data to eliminate noise interference and improve the accuracy of location information; wavelet transform technology is used to denoise the time-series data collected by the sensors to ensure data reliability. Convolutional neural networks are used to extract features from image data captured by cameras to identify obstacles and target objects. Long Short-Term Memory (LSTM) networks are used to process temporal data of network states to predict link quality change trends. A federated learning framework is employed to achieve collaborative data processing among multiple drones, enabling data fusion while protecting data privacy.
[0025] AI Decision-Making and Optimization Core: Utilize algorithms such as reinforcement learning and deep learning to build intelligent models, achieve lightweight deployment after training, and quickly output drone communication and resource optimization strategies based on network status;
[0026] Specifically, it includes: In the UAV path planning problem, the Deep Q-Network (DQN) algorithm is adopted to learn the optimal flight path with the goal of maximizing the task completion rate and minimizing energy consumption; for the resource allocation problem, the Proximal Policy Optimization (PPO) algorithm is used to dynamically adjust channel and power resources. We utilize graph neural networks to model the topology of UAV networks and predict changes in link states; we also use the Transformer architecture to process long-sequence communication data, thereby improving the accuracy of predicting network traffic changes. When the flight area of the UAV meets the communication network environment corresponding to the terrain-induced multipath trap, the interference of the UAV's communication network is analyzed to determine the most suitable transverse radio wave signal transmission mode, so as to optimize the UAV's communication network. The process includes: The canyon terrain is approximated as an open rectangular waveguide, and the width and depth of the canyon terrain are obtained; mold( , The order of the pattern. Corresponding width direction mode number, The formula for calculating the cutoff frequency (corresponding to the number of modes in the depth direction) is: ; in For the speed of light ( ), The width of the canyon. The depth of the canyon; Calculate the cutoff frequency for each mode in turn, for example To obtain the cutoff frequency corresponding to each mode, divide the UAV communication operating frequency by the cutoff frequency corresponding to each mode to obtain the basic satisfaction level. Also includes: Since the width and depth of the canyon terrain corresponding to the drone will change during the flight, the width and height of the canyon at the current location of the drone are obtained at preset time intervals, and the corresponding cutoff frequency and its corresponding basic satisfaction are output; the width and height data of the canyon at the location of the drone are obtained by the lidar installed on the drone. A preset basic satisfaction threshold is set. If the basic satisfaction level is less than the basic satisfaction threshold, then... Module switching; sequentially obtain each The cutoff frequency of the modulus and the corresponding basic satisfaction level, with the basic satisfaction level greater than the basic satisfaction threshold being considered. Module as a transmission mode; Based on the drone's flight speed and a defined time interval, the corresponding drone flight path, as well as the height and width of the canyon at the path's endpoint, are calculated. Using the current position of the UAV as the center and the calculated flight path length of the UAV as the radius, construct a cylindrical scanning model along the predicted flight path, obtain the cutoff frequency at each point on the surface of the scanning model, and the basic satisfaction degree corresponding to the cutoff frequency; When the drone reaches the end of its path, if the baseline satisfaction level at the end of the path is less than the baseline satisfaction threshold, then an early warning will be issued within the corresponding time interval during the drone's flight. Mode switching allows the drone to complete the process when it reaches the end of the path within the time interval. Module switching; If mode switching is performed only after reaching the end of the path, signal transmission may be interrupted due to time delays in the switching process (such as protocol negotiation and parameter reconfiguration). Initiating the switching in advance within the time interval ensures that the UAV has completed the mode update by the time it reaches the end, maintaining communication continuity. If the drone's flight speed is v and the time interval is T, then the advance switching time window is T, which is sufficient to complete the mode switching process (such as calculating the new cutoff frequency, filtering candidate modes, and updating parameters) and avoid the risk of disconnection upon arrival. Since frequent mode switching can increase communication protocol overhead (such as renegotiating transmission parameters), a switching buffer is set up so that when the basic satisfaction level approaches a threshold (e.g., if the threshold is 0.8, an alert is triggered when the calculated value is ≤0.9), the mode candidate pool is activated in advance to reduce the computational delay during the actual switching; and when If the signal does not improve after mode switching, roll back to the original mode; If the basic satisfaction of the new mode does not meet expectations after switching (e.g., calculation deviation caused by terrain data errors), the rollback mechanism can immediately restore the original mode, avoiding further signal attenuation due to erroneous switching. This mechanism is similar to a "fault-tolerant switch," ensuring that the system always maintains optimal communication status in dynamic environments.
[0027] Application scenario: When the lidar obtains incorrect canyon width data due to light or dust interference, the cutoff frequency calculated based on the incorrect data may lead to incorrect switching. The rollback mechanism can quickly correct this and ensure communication reliability. When the rollback mechanism is combined with AI algorithms, the prediction model can be optimized by recording switching and rollback events. For example, if the switching success rate is low under a certain type of terrain, the AI can adjust the switching threshold or candidate mode priority under that terrain to form a closed-loop optimization. It supports both offline and online training modes. During offline training, historical data is used to train large-scale models on cloud servers; during online training, the drone updates model parameters in real time through edge computing devices to adapt to dynamic changes in the environment. The trained AI model is quantized and pruned, and then deployed to the embedded computing platform of the drone (such as the NVIDIA Jetson series) to reduce the model inference time. Based on the current network status, the AI model quickly generates optimization strategies, such as adjusting the three-dimensional spatial position of the drone to optimize communication coverage, dynamically allocating channel resources to avoid interference, and controlling transmission power to balance communication quality and energy consumption. Communication and Task Collaboration Management: Dynamically allocate wireless communication resources, optimize network topology, and achieve intelligent allocation and collaborative operation of multiple UAV tasks through multi-agent reinforcement learning, thereby improving network performance and task efficiency; Specifically, it includes: Deep reinforcement learning algorithms are used to optimize subcarrier allocation in orthogonal frequency division multiple access. By monitoring the signal-to-noise ratio of each link in real time, the subcarrier allocation scheme is dynamically adjusted to reduce co-channel interference and improve spectrum utilization. Based on graph neural network analysis of drone network topology, when communication demand surges in a certain area, the AI model dispatches nearby drones to move to the hotspot area, forming temporary relay nodes, expanding network coverage and balancing the load. Using multi-agent reinforcement learning algorithms, data acquisition, transmission, or relay tasks can be dynamically allocated based on each drone's computing power, communication load, and remaining battery power. For example, in disaster relief scenarios, image acquisition tasks can be prioritized for drones with sufficient battery power that are close to the disaster area. Through a distributed decision-making framework, each UAV trains its model locally and shares parameters, enabling collaborative flight and communication among multiple UAV swarms. For example, multiple UAVs can form a virtual MIMO array, increasing link capacity and enhancing data transmission reliability. Resource and energy consumption optimization: Utilize models such as LSTM to predict UAV energy consumption, formulate energy-saving strategies, and flexibly schedule computing and storage resources to improve resource utilization and extend network service time; Specifically, it includes: A drone energy consumption prediction model is built based on LSTM neural network. By inputting parameters such as flight speed, altitude, communication power, and computing load, the model can predict energy consumption under different mission scenarios. For example, it can predict the power consumption of drones when performing high-altitude patrol missions and plan the return path in advance. Dynamic energy-saving strategies are developed based on task priorities. When the urgency of a task is low, the drone's flight speed and communication frequency are reduced. For drones that support wireless charging, reinforcement learning is used to optimize the location of charging nodes and the charging sequence to improve energy replenishment efficiency. When the computational load on a drone is too high, the AI model can offload the task to a nearby node or edge server through transfer learning; for example, when processing high-definition image recognition tasks, some computational tasks can be transferred to drones or ground base stations with stronger computing power to reduce processing latency. Based on data importance and access frequency, deep reinforcement learning is used to dynamically adjust the data caching strategy; for frequently accessed real-time monitoring data, it is preferentially stored in the cache; for historical data, it is migrated to a larger capacity storage device to reduce redundant storage.
[0028] It also includes security monitoring and feedback optimization: Machine learning algorithms are used to detect cybersecurity threats, ensure communication security, monitor system performance indicators in real time, and iteratively optimize AI models based on operational data to ensure stable and efficient system operation. Specifically, it includes: Machine learning algorithms such as isolated forests and autoencoders are used to detect anomalies in network traffic and identify malicious attacks and data tampering. Classification models (such as SVM and random forest) are trained using historical attack data to improve the accuracy of attack identification. A generative adversarial network (GAN) is used to dynamically generate encryption keys, and federated learning is combined to achieve distributed key management and prevent key leakage. In electromagnetic interference environments, reinforcement learning is used to optimize the frequency hopping strategy and adaptively select the optimal communication frequency band to ensure the security of the communication link. The system provides a real-time visualization of key performance indicators such as network throughput, end-to-end latency, drone survivability, and mission completion rate. When these indicators deviate from thresholds (e.g., packet loss rate exceeding 10%), the AI model is triggered to replan its strategy and generate an early warning report. Collect real-world operational data and update AI model parameters using online learning, incremental learning, or transfer learning techniques. Simultaneously, leverage digital twin technology to construct a virtual network environment to simulate and test new strategies, verifying optimization effectiveness before actual deployment to mitigate system risks.
[0029] The above formulas are derived from software simulations using a large amount of data and are selected to be close to the actual values. The influence weight factors and specific coefficient values in the formulas are set by those skilled in the art based on the actual situation and can be adjusted and modified in the future.
[0030] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device.
[0031] The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means (e.g., infrared, wireless, microwave, etc.).
[0032] The computer-readable storage medium can be any available medium that a computer can access, or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium, an optical medium, or a semiconductor medium. A semiconductor medium can be a solid-state drive (SSD).
[0033] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.
[0034] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatus and methods can be implemented in other ways.
[0035] For example, the device embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be indirect couplings or communication connections between devices or units through some interfaces, and may be electrical, mechanical, or other forms.
[0036] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0037] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.
[0038] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application.
[0039] The aforementioned storage media include various media that can store program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.
[0040] The above description of the embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. An optimization method for aerial drone communication networks based on AI, characterized in that, include: Intelligent sensing and data processing: Real-time collection of environmental and network status data through multiple sensors, followed by noise reduction, feature extraction and fusion, to provide accurate data support for subsequent decision-making; AI Decision-Making and Optimization Core: Utilize algorithms such as reinforcement learning and deep learning to build intelligent models, achieve lightweight deployment after training, and quickly output drone communication and resource optimization strategies based on network status; Communication and Task Collaboration Management: Dynamically allocate wireless communication resources, optimize network topology, and achieve intelligent allocation and collaborative operation of multiple UAV tasks through multi-agent reinforcement learning, thereby improving network performance and task efficiency; Resource and energy consumption optimization: predict drone energy consumption, formulate energy-saving strategies, and flexibly schedule computing and storage resources.
2. The optimization method for AI-based aerial drone communication networks according to claim 1, characterized in that, Intelligent sensing and data processing specifically include: The drone is equipped with lidar, high-definition cameras, high-precision GPS, and inertial measurement unit sensors; The system collects communication parameters such as channel gain, signal strength, and signal-to-noise ratio through software-defined wireless devices, and monitors network performance indicators in real time; it also has built-in monitoring tools to obtain information on the drone's energy consumption and computing resource usage. Kalman filtering algorithm is used to fuse GPS and IMU data to eliminate noise interference; wavelet transform technology is used to denoise the time series data collected by the sensors. Convolutional neural networks are used to extract features from image data captured by cameras to identify obstacles and target objects.
3. The optimization method for AI-based aerial drone communication networks according to claim 1, characterized in that, The core of AI decision-making and optimization includes: In UAV path planning, a deep Q-network algorithm is used to learn the optimal flight path; for resource allocation problems, a near-end strategy optimization algorithm is used to dynamically adjust channel and power resources. Graph neural networks are used to model the topology of UAV networks and predict link state changes. When the flight area of the UAV meets the communication network environment corresponding to the terrain-induced multipath trap, the communication network interference of the UAV is analyzed to determine the most suitable transverse radio wave signal transmission mode.
4. The optimization method for AI-based aerial drone communication networks according to claim 3, characterized in that, When the UAV's flight area meets the communication network environment corresponding to terrain-induced multipath traps, the communication network interference of the UAV is analyzed to determine the most suitable transverse radio wave signal transmission mode. The process includes: The canyon terrain is approximated as an open rectangular waveguide, and the width and depth of the canyon terrain are obtained; The formula for calculating the cutoff frequency of the mode is: ; in At the speed of light, The width of the canyon. The depth of the canyon; Calculate the cutoff frequency for each mode in sequence; to obtain the cutoff frequency corresponding to each mode, divide the UAV communication operating frequency by the cutoff frequency corresponding to each mode to obtain the basic satisfaction.
5. The optimization method for AI-based aerial drone communication networks according to claim 4, characterized in that, Also includes: The width and height of the canyon at the current location of the drone are obtained at preset time intervals, and the corresponding cutoff frequency and its corresponding basic satisfaction are output. A preset basic satisfaction threshold is set. If the basic satisfaction level is less than the basic satisfaction threshold, then... Module switching; sequentially obtain each The cutoff frequency of the modulus and the corresponding basic satisfaction level, with the basic satisfaction level greater than the basic satisfaction threshold being considered. The mode is used as the transmission mode.
6. The optimization method for an AI-based aerial drone communication network according to claim 5, characterized in that, Based on the drone's flight speed and a defined time interval, the corresponding drone flight path, as well as the height and width of the canyon at the path's endpoint, are calculated. Using the current position of the UAV as the center and the calculated flight path length of the UAV as the radius, construct a cylindrical scanning model along the predicted flight path, obtain the cutoff frequency at each point on the surface of the scanning model, and the basic satisfaction degree corresponding to the cutoff frequency; When the drone reaches the end of its path, if the baseline satisfaction level at the end of the path is less than the baseline satisfaction threshold, then an early warning will be issued within the corresponding time interval during the drone's flight. Mode switching allows the drone to complete the process when it reaches the end of the path within the time interval. Module switching.
7. The optimization method for an AI-based aerial drone communication network according to claim 1, characterized in that, Communication and task collaboration management, specifically including: Deep reinforcement learning algorithms are used to optimize subcarrier allocation in orthogonal frequency division multiple access. By monitoring the signal-to-noise ratio of each link in real time, the subcarrier allocation scheme is dynamically adjusted to reduce co-channel interference and improve spectrum utilization. Analysis of UAV network topology based on graph neural network; By using a multi-agent reinforcement learning algorithm, data acquisition, transmission, or relay tasks are dynamically allocated based on the computing power, communication load, and remaining battery power of each UAV.
8. The optimization method for an AI-based aerial drone communication network according to claim 1, characterized in that, Resource and energy consumption optimization specifically includes: A drone energy consumption prediction model is built based on LSTM neural network. Input parameters are used to predict the energy consumption under different mission scenarios. Develop dynamic energy-saving strategies based on task priorities; When the computational load on the drone is too high, the AI model will offload the task to a nearby node or edge server through transfer learning; Based on data importance and access frequency, deep reinforcement learning is used to dynamically adjust the data caching strategy.
9. The optimization method for an AI-based aerial drone communication network according to claim 1, characterized in that, It also includes security monitoring and feedback optimization: Machine learning algorithms are used to detect cybersecurity threats, ensure communication security, monitor system performance indicators in real time, and iteratively optimize AI models based on operational data to ensure stable and efficient system operation.