5g-based ultra-low latency data map transmission method and system

Abstract

The application discloses a 5G-based ultra-low latency data map transmission method and system, and relates to the technical field of unmanned aerial vehicle data transmission. The method effectively manages the switching of unmanned aerial vehicles between different signal towers through a hybrid neural network and a dynamic programming algorithm to ensure the continuity of data transmission and the stability of signals. The method comprises the following steps: firstly, obtaining flight data of the unmanned aerial vehicle and performing pretreatment; constructing a hybrid neural network comprising spatial feature extraction, time feature extraction and signal attenuation feature extraction; finally, based on real-time flight information, the neural network is used to predict and determine the optimal signal tower switching point. In addition, the application also provides a technology for dynamically adjusting the size of the data window. According to the flight speed of the unmanned aerial vehicle and the signal change, the frequency of data acquisition is dynamically adjusted to adapt to different environmental conditions, thereby improving the data transmission efficiency and reducing the delay.

Description

Technical Field

[0001] This application relates to the field of unmanned aerial vehicle (UAV) data transmission technology, and in particular to a 5G-based ultra-low latency data transmission method and system. Background Technology

[0002] With the rapid expansion of drone applications, particularly in urban surveillance, traffic management, and emergency rescue, the demand for real-time data transmission from drones is increasing. Especially in urban environments, dense buildings and unevenly distributed cell towers pose challenges to drones during missions, including unstable signal coverage and data transmission latency. Traditional data transmission methods often rely on static cell tower connection strategies, lacking the ability to dynamically adjust according to the real-time environment, which limits the efficiency of drone applications in complex environments. To address these issues, 5G communication is a viable solution, as 5G technology offers higher data transmission rates and lower latency. However, even 5G technology can encounter signal instability issues in highly dynamic environments, especially when drones frequently traverse high-rise buildings and multiple 5G base stations.

[0003] For example, Chinese patent CN114885304A discloses a 5G-connected drone data transmission system and method based on FlexE technology. The system includes a drone terminal, a 5G base station, and a drone cloud platform. The drone terminal tags data packets according to their categories and queues the tagged packets. The 5G base station uses FlexE as its bearer network, providing a dedicated data path with sufficient bandwidth and deterministic latency for the tagged data packets, and sends the packets to the core network. The core network then sends the packets to the drone cloud platform. The drone cloud platform receives data transmitted by the drone terminal, processes the received data, and receives operation commands for the drone terminal. Based on this system, the invention also proposes a 5G-connected drone data transmission method based on FlexE technology. This method can reduce latency, save energy, lower costs, increase system capacity, and enable large-scale device connectivity.

[0004] The above methods suffer from the problems described in this background section. Therefore, developing a method that can respond to environmental changes in real time and intelligently manage cell tower switching is crucial for improving the operational efficiency of UAVs and the reliability of data transmission. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the purpose of this application is to provide an ultra-low latency data transmission method and system based on 5G. This method dynamically adjusts the signal switching points of a drone as it moves between various signal towers, taking into account the characteristics of short latency but limited tower coverage and frequent switching under 5G signals, thereby improving the transmission quality of data transmission tasks. The specific solution is as follows:

[0006] Firstly, this application provides a 5G-based ultra-low latency data and image transmission method, including:

[0007] Historical flight data of the UAV within the coverage area of ​​multiple target signal towers is acquired and preprocessed; wherein, the historical flight data includes: quadrature amplitude modulation data, access point data, and signal attenuation data;

[0008] Based on the preprocessed historical flight data, a hybrid neural network is constructed; wherein the hybrid neural network includes: a spatial feature extraction branch, a temporal feature extraction branch, and a signal attenuation feature extraction branch;

[0009] The real-time flight information of the UAV is acquired, and the hybrid neural network is used to determine the signal switching point of the UAV within the coverage area of ​​the multiple target signal towers.

[0010] As an optional implementation, the preprocessing includes:

[0011] Random phase shift is applied to the orthogonal amplitude modulation data to generate target training samples for training the hybrid neural network.

[0012] The actual delay distribution is calculated based on the historical flight data, generating a distribution that includes random delays. The first time series.

[0013] As an optional implementation, constructing the hybrid neural network includes:

[0014] The spatial feature extraction branch is constructed based on a multi-layer convolutional neural network to extract the spatial position and environmental layout features of the UAV relative to the target signal tower and target environmental obstacles;

[0015] Each layer of the convolutional neural network is followed by a batch normalization layer and a ReLU activation function;

[0016] A time feature extraction branch is constructed based on a multilayer long short-term memory network to extract the time dependence and dynamic changes of the historical flight data; wherein, the multilayer long short-term memory network includes: a first time step sliding window method;

[0017] The signal attenuation feature extraction branch is composed of at least one fully connected layer. It analyzes the distance and signal strength data from the UAV to each signal tower to predict signal attenuation and potential handover points. Each fully connected layer is followed by a dropout layer.

[0018] As an optional implementation, the historical flight data further includes: flight speed and network change information; the first time step sliding window method includes: dynamically adjusting the size of the time window based on the UAV's flight speed and network signal change information;

[0019] The dynamic adjustment of the time window size includes:

[0020] Based on the flight mission of the UAV, a scene mode is determined; the scene mode includes a first scene mode, a second scene mode, and a third scene mode.

[0021] When the drone is in a first scene mode, the drone's flight speed is detected, and a first initial time window size is set; in response to the drone's flight speed being greater than or equal to a first speed threshold, the time window size is decreased; in response to the drone's flight speed being less than a second speed threshold, the time window size is increased.

[0022] When the drone is in the second scene mode, the signal-to-noise ratio of the environment in which the drone is located is detected, and a second initial time window size is set. In response to the signal-to-noise ratio being greater than or equal to a first signal-to-noise ratio threshold, the time window size is reduced, and in response to the signal-to-noise ratio being less than the second signal-to-noise ratio threshold, the time window size is increased.

[0023] When the drone is in the third scene mode, the signal strength of the drone is detected, and a third initial time window size is set. In response to the signal strength being greater than or equal to a first signal threshold, the time window size is reduced, and in response to the signal strength being less than a second signal threshold, the time window size is increased.

[0024] As an optional implementation, the output of the hybrid neural network includes: signal quality score, signal attenuation prediction, and interference level assessment; determining the signal switching point of the UAV within the coverage area of ​​the multiple target signal towers includes:

[0025] Using the first dynamic programming scheme, determine the signal switching points of the UAV within the coverage area of ​​multiple target signal towers;

[0026] The first dynamic programming scheme includes: state definition, decision variables, transition equation, objective function, initialization and boundary conditions, and dynamic programming solution.

[0027] Secondly, this application also provides a 5G-based ultra-low latency data transmission system, including:

[0028] The acquisition unit is used to acquire and preprocess historical flight data of the UAV within the coverage area of ​​multiple target signal towers; wherein, the historical flight data includes: quadrature amplitude modulation data, access point data, and signal attenuation data;

[0029] The construction unit constructs a hybrid neural network based on the preprocessed historical flight data; wherein the hybrid neural network includes: a spatial feature extraction branch, a temporal feature extraction branch, and a signal attenuation feature extraction branch;

[0030] The processing unit is used to acquire the real-time flight information of the UAV and, using the hybrid neural network, determine the signal switching point of the UAV within the coverage area of ​​the multiple target signal towers.

[0031] Thirdly, an optional implementation of this application also provides an electronic device, including a processor and a memory, wherein the memory stores machine-readable instructions executable by the processor, and the processor is used to execute the machine-readable instructions stored in the memory. When the machine-readable instructions are executed by the processor, the steps of the above-mentioned 5G-based ultra-low latency data transmission method are performed.

[0032] Fourthly, an optional implementation of this application also provides a computer-readable storage medium storing a computer program, which, when run, executes the steps of the above-described 5G-based ultra-low latency data transmission method.

[0033] Compared with existing technologies, the advantages of this invention are: by introducing hybrid neural networks and dynamic programming algorithms, intelligent signal switching management of UAVs under 5G networks is achieved, significantly improving the stability and continuity of data transmission and reducing signal interruption problems in traditional methods. Simultaneously, by dynamically adjusting the time window size, the data acquisition frequency can be flexibly adjusted according to changes in UAV flight speed and signal quality, reducing transmission latency and improving data transmission efficiency. Attached Figure Description

[0034] Figure 1 A flowchart illustrating the 5G-based ultra-low latency data transmission method provided in this application embodiment;

[0035] Figure 2 A flowchart illustrating a method for dynamically adjusting the size of a time window, as provided in this application embodiment;

[0036] Figure 3 A schematic diagram of the structure of a 5G-based ultra-low latency data transmission system provided in an embodiment of this application;

[0037] Figure 4 This is a schematic diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0038] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0039] 5G networks offer high bandwidth and low latency, enabling drones to transmit high-quality data in real time, such as video streams and sensor information. However, 5G network tower coverage is relatively limited, and drones may need to frequently switch between multiple towers during flight to maintain signal stability and continuity. This necessitates a dynamic adjustment strategy that utilizes hybrid neural networks and dynamic programming to optimize the tower switching process, reduce latency and data loss, and ensure drones can reliably perform missions in complex environments.

[0040] For this reason, see Figure 1 The flowchart shown is for a 5G-based ultra-low latency data transmission method provided in this application embodiment. The method includes steps S101 to S103, wherein:

[0041] S101: Acquire and preprocess flight data of the UAV within the coverage area of ​​multiple target signal towers; wherein, the historical flight data includes: quadrature amplitude modulation data, access point data, and signal attenuation data;

[0042] S102: Based on the preprocessed historical flight data, construct a hybrid neural network; wherein the hybrid neural network includes: a spatial feature extraction branch, a temporal feature extraction branch, and a signal attenuation feature extraction branch;

[0043] S103: Obtain the real-time flight information of the UAV, and use the hybrid neural network to determine the signal switching point of the UAV within the coverage area of ​​the multiple target signal towers.

[0044] In its implementation, this application first collects historical flight data involved in the flight of the UAV, including but not limited to: in-phase and quadrature data (IQ), access point data (AP), and signal attenuation data. The collected data is then cleaned and formatted, such as by removing noise, standardizing, and synchronizing the data to ensure the accuracy and consistency of the data when it is input into the neural network.

[0045] IQ data consists of fundamental signals in wireless communication, including in-phase and quadrature components. This data reflects the amplitude and phase of the signal and is crucial information for signal processing and analysis. AP data involves wireless connectivity information between the UAV and ground or other network infrastructure, including but not limited to the access point's identification, location, signal strength, and network quality. Signal attenuation data describes how the signal strength received by the UAV from signal sources (such as GPS satellites or Wi-Fi access points) decreases with distance or other obstacles.

[0046] There are several ways to acquire the aforementioned IQ data, AP data, and signal attenuation data. Drones can directly measure and record this data using integrated wireless communication hardware such as Wi-Fi modules and cellular network modules. These sensors can capture signal strength and quality, as well as information about cell towers, in real time. Software tools can also be used to analyze the received signal data, such as using algorithms to process raw signal data to estimate signal attenuation and network access point performance. Alternatively, ground control stations or cloud platforms communicating with the drone can be used to obtain detailed data about cell towers and network coverage from external network services. The choice of these methods can be based on the specific application requirements and operating environment of the drone to ensure data accuracy and real-time performance.

[0047] After acquiring and preprocessing the historical flight data of the aforementioned UAVs, this application constructs a hybrid neural network to extract features from the data from multiple different dimensions, including: using a spatial feature extraction branch to process and analyze the spatial position information of the UAV relative to the signal tower, and extracting position-related features; using a time feature extraction branch to analyze the flight dynamics of the UAV, such as speed and trajectory, and extracting time series data features; and using a signal attenuation feature extraction branch to analyze signal strength data and predict potential signal attenuation under different environments.

[0048] In this application, after constructing the aforementioned hybrid neural network, real-time updated flight data is input into the trained hybrid neural network. The hybrid neural network comprehensively considers spatial location, flight time characteristics, and signal attenuation, and outputs a signal tower switching suggestion. This suggestion optimizes the communication efficiency of the UAV based on the predicted signal quality and future flight trajectory.

[0049] As an optional implementation, the preprocessing includes: randomly phase-shifting the IQ data to generate target training samples, and calculating the actual delay distribution based on the flight data to generate samples containing random delays. The first time series.

[0050] In practice, random phase shifting of the IQ data is used to generate target training samples. IQ data contains both amplitude and phase information of the signal; by randomly shifting the phase of the IQ data, different signal interferences and environmental changes can be simulated. This process enhances the diversity of the training data, enabling the network model to maintain high robustness in various environments.

[0051] For example, randomly shifting the phase of the original IQ data by a range of values ​​(such as ±30 degrees) to generate various different signal samples can help improve the model's generalization ability and effectively avoid overfitting.

[0052] In this application, calculating the actual delay distribution based on flight data is used to generate a first time series containing random delays. During UAV flight, signal transmission is affected by various factors, leading to uncertainty in the actual delay distribution. By analyzing flight data, the actual delay distribution is calculated, and a first time series containing random delays is generated based on this distribution. The first time series can accurately simulate real-world signal transmission.

[0053] This provides more representative and diverse training data for hybrid neural networks, thereby improving the model's performance in practical applications.

[0054] Regarding S102 above: Based on the preprocessed historical flight data, a hybrid neural network is constructed; wherein, the hybrid neural network includes: a spatial feature extraction branch, a temporal feature extraction branch, and a signal attenuation feature extraction branch.

[0055] As an optional implementation, constructing a hybrid neural network includes: constructing the spatial feature extraction branch based on a multi-layer convolutional neural network; wherein each convolutional network layer is followed by batch normalization and ReLU activation functions to extract the spatial position and environmental layout features of the UAV relative to the signal tower and target environment obstacles;

[0056] A time feature extraction branch is constructed based on a multilayer long short-term memory network to extract the time dependence and dynamic changes in UAV flight data; the multilayer long short-term memory network includes: the first time step sliding window method;

[0057] Signal attenuation feature extraction branch: consists of at least one fully connected layer, each followed by dropout, used to analyze distance and signal strength data from the drone to each signal tower, predict signal attenuation and potential handover points.

[0058] This application provides a specific hybrid neural network, including a spatial feature extraction branch, a temporal feature extraction branch, and a signal attenuation feature extraction branch.

[0059] S1021: Spatial feature extraction branch, which uses a multi-layer convolutional neural network (CNN) to construct the spatial feature extraction branch. Each convolutional layer is followed by batch normalization and ReLU activation functions to ensure the effectiveness and stability of feature extraction.

[0060] For example, the specific parameters are as follows:

[0061] Convolutional layer: Each layer contains 32 3x3 filters.

[0062] Batch normalization: used to stabilize and accelerate the training process.

[0063] ReLU activation function: used to introduce nonlinearity, thereby enhancing the expressive power of the model.

[0064] This allows for the effective extraction of the spatial position and layout features of the drone relative to signal towers and environmental obstacles, improving the accuracy and robustness of the model.

[0065] S1022: Temporal feature extraction branch, which uses a multi-layer long short-term memory network to process and extract the temporal dependence and dynamic changes in UAV flight data.

[0066] For example, the specific parameters are as follows:

[0067] LSTM layer: Each layer contains 50 hidden units. Two LSTM layers are stacked to capture complex temporal dependencies.

[0068] S1023: Signal attenuation feature extraction branch, consisting of at least one fully connected (Dense Layer), with each fully connected layer followed by a dropout layer to prevent overfitting and enhance the model's generalization ability.

[0069] For example, the specific parameters are as follows:

[0070] Fully connected layer: contains 128 neurons.

[0071] Dropout layer: Set the drop rate to 0.5.

[0072] In this way, through this hybrid neural network, multi-layer convolutional and LSTM networks can efficiently process spatial and temporal features, ensuring rapid data transmission and processing; the signal attenuation feature extraction branch enables UAVs to accurately predict signal attenuation and switching points, thereby optimizing signal switching decisions.

[0073] As one possible implementation, the historical flight data further includes: flight speed and network change information; the first time step sliding window method includes:

[0074] Based on the UAV's flight speed and network signal change information, the window size is dynamically adjusted, including:

[0075] Based on the flight mission of the UAV, a scene mode is determined; the scene mode includes a first scene mode, a second scene mode, and a third scene mode.

[0076] When the drone is in a first scene mode, the drone's flight speed is detected, and a first initial time window size is set; in response to the drone's flight speed being greater than or equal to a first speed threshold, the time window size is decreased; in response to the drone's flight speed being less than a second speed threshold, the time window size is increased.

[0077] When the drone is in the second scene mode, the signal-to-noise ratio of the environment in which the drone is located is detected, and a second initial time window size is set. In response to the signal-to-noise ratio being greater than or equal to a first noise threshold, the time window size is reduced, and in response to the signal-to-noise ratio being less than the second noise threshold, the time window size is increased.

[0078] When the drone is in the third scene mode, the signal strength of the drone is detected, and a third initial time window size is set. In response to the signal strength being greater than or equal to a first signal threshold, the time window size is reduced, and in response to the signal strength being less than a second signal threshold, the time window size is increased.

[0079] In its implementation, this application adopts the first time step sliding window method, i.e., the dynamic adjustment strategy of the time window, to meet the real-time data processing needs of UAVs in different flight scenarios.

[0080] The time window refers to the amount of data collected and processed within a certain period of time. The purpose of dynamically adjusting the time window is to optimize the efficiency of data transmission and processing, flexibly adjusting the time interval between data collection and processing based on real-time flight data and network conditions.

[0081] In practice, flight data includes not only quadrature amplitude modulation (IQ) data, access point (AP) data, and signal attenuation data, but also extended information on flight speed and network changes. Based on this data, the time window size can be dynamically adjusted using the first-time-step sliding window method.

[0082] First, determine the scene mode based on the drone's flight mission. The scene mode determines how to adjust the time window size.

[0083] The scene mode can be determined manually or automatically to determine the drone's flight mission and scene mode.

[0084] For example, users can manually set the scene mode according to predefined flight missions and environmental conditions. For instance, before takeoff, the operator can select the appropriate scene mode based on mission requirements, such as "high-speed cruise mode" or "low-speed reconnaissance mode." This manual setting method ensures the operator's control over the drone and allows them to make the most suitable choice based on specific mission needs.

[0085] For example, drones can also automatically determine scene modes by analyzing flight missions and environmental changes in real time through built-in sensors and intelligent algorithms.

[0086] For example, a drone identifies its current flight mission through a mission management system. By reading mission parameters such as flight path, target area, and flight speed, it determines the nature of the current mission and acquires real-time environmental data through sensors, including flight speed, signal strength, and signal-to-noise ratio. Based on this data, it automatically assesses the complexity and changes in the current flight environment.

[0087] Based on task recognition and environmental perception results, the system automatically selects and switches to the most suitable scene mode.

[0088] In practical implementation, this application provides three specific scenario modes, and the applicable scenarios for each of these three modes are as follows:

[0089] 1. Speed-Sensing Mode: Suitable for scenarios where the drone's flight speed changes significantly, such as tracking fast-moving targets or high-speed cruising. In this mode, the time window adjusts according to the flight speed, enabling a faster response to data changes caused by speed.

[0090] 2. Signal Strength Mode: Suitable for environments with large signal strength fluctuations, such as densely built-up areas in cities or remote areas. This mode helps optimize data capture and processing in areas with weak signals by dynamically adjusting the time window based on signal strength.

[0091] 3. Signal Complexity Mode: Suitable for environments with severe signal interference or multiple signals, such as industrial areas or areas using various communication devices. In complex signal environments, frequent updates to the data window help reduce errors and signal loss.

[0092] By selecting appropriate modes, unmanned aerial vehicle (UAV) systems can more effectively adapt to different operating environments and challenges, thereby improving mission execution efficiency and success rate.

[0093] In practical implementation, when the drone is in speed-sensing mode, this application adjusts the time window by detecting the flight speed. For example, an initial time window size can be set, namely the first initial time window size in this application. When the flight speed is greater than or equal to a first speed threshold, the time window size is reduced based on the initial time window to respond to environmental changes more quickly; when the flight speed is less than a second speed threshold, the time window size is increased to reduce the frequency of data processing.

[0094] In signal strength mode, this application adjusts the time window by detecting the signal-to-noise ratio of the environment in which the drone is located. For example, an initial time window size can be set, namely the second initial time window size in this application. When the signal-to-noise ratio is greater than or equal to a first noise threshold, the time window size is reduced based on the initial time window to ensure that data can be processed quickly in a high-noise environment; when the signal-to-noise ratio is less than the second noise threshold, the time window size is increased to reduce unnecessary data processing burden.

[0095] In the signal complexity mode, this application adjusts the time window by detecting the signal strength of the UAV. For example, an initial time window size can be set, namely the second initial time window size in this application. When the signal strength is greater than or equal to a first signal threshold, the time window size is reduced to cope with changes in signal strength; when the signal strength is less than a second signal threshold, the time window size is increased to ensure sufficient data can be collected for processing when the signal is weak. The purpose of dynamically adjusting the time window is to optimize the efficiency of data acquisition and processing under different flight and network conditions, and to improve the stability and real-time performance of data transmission. This method helps the UAV maintain efficient communication and data processing capabilities in complex and ever-changing environments, ensuring the successful completion of the mission.

[0096] Please see Figure 2 , Figure 2 The flowchart of a method for dynamically adjusting the size of a time window is provided in this application embodiment. In this application, there are three scenario modes: a first scenario mode for drone speed, a second scenario mode for drone signal strength, and a third scenario mode for drone signal complexity. In each scenario mode, this application provides two methods for dynamically adjusting the size of the time window, which are respectively applicable to decreasing and increasing the size of the time window.

[0097] As one possible implementation, when the drone is in a first scene mode, the time window size is set. The formula is:

[0098] ;

[0099] in Represents the flight speed of the drone. It is the size of the first initial time window. It is the speed adjustment coefficient. It is the first speed threshold.

[0100] As an optional implementation, when the drone is in the first scene mode, the time window size is set. The formula is:

[0101] ;

[0102] in It is the speed adjustment coefficient. It is the second speed threshold.

[0103] In the first scenario mode of this application, given the immediate impact of speed changes on data acquisition, the UAV's need for real-time data is less urgent during low-speed flight than during high-speed flight. Therefore, the speed can be lower than the second speed threshold. Increasing the time window can reduce data processing frequency and save energy. Considering that data changes are not significant at lower speeds, a square function can be used to smoothly increase the time window, reducing fluctuations caused by frequent adjustments.

[0104] Conversely, at high speeds, drones may need to update data more frequently to cope with rapidly changing environments. This occurs when the speed exceeds a first speed threshold. In this application, an exponential function is used to rapidly reduce the time window in order to improve the frequency and accuracy of data updates.

[0105] These two formulas combine the specific characteristics of velocity, through and Adjustments to these parameters can flexibly address data transmission needs under different flight conditions. When adjusting these parameters, the actual flight environment and the UAV's mission requirements should be considered to ensure the effectiveness of the time window adjustments.

[0106] For example, in urban and rural areas, aerial monitoring requires different data update frequencies depending on the flight speed.

[0107] For example, high-speed speed regulation coefficient 0.1, used for fast-moving dynamic environments (such as urban areas). Low-speed adjustment coefficient. 0.02, used for stable, low-change environments (such as rural areas or controlled airspace). First velocity threshold. 50 km / h; reaching or exceeding this speed requires rapid data updates. Second speed threshold. 20 km / h; speeds below this can reduce the data update frequency. If the drone speed is 70 km / h (exceeding the first speed threshold);

[0108] This indicates that the data update frequency increases during high-speed flight. At this time, the current time window value can be set to the minimum time window value, for example, the minimum time window value can be set to 0.1 seconds.

[0109] If the drone speed drops to 10 km / h (below the second speed threshold);

[0110] When the drone is flying at low speed, the data update frequency is reduced and the time window is increased to 1.2 seconds.

[0111] By adjusting and This coefficient allows drones to more effectively manage their data collection frequency in response to changes in speed and environmental conditions. Under rapid and highly variable conditions, a higher coefficient is crucial. A larger value allows for a more rapid reduction in the time window, ensuring that data can be updated frequently to capture dynamic changes. Conversely, a smaller value... The value mitigates the increase in window size when less frequent updates are needed, and this flexibility enhances the operational efficiency and responsiveness of drones when performing missions.

[0112] As an optional implementation, when the drone is in the second scene mode, the time window size is set. The formula is:

[0113] ;

[0114] in It's the signal-to-noise ratio. It is the size of the second initial time window. It is the signal-to-noise ratio adjustment coefficient. It is the first signal-to-noise ratio threshold.

[0115] As an optional implementation, when the drone is in the second scene mode, the time window size is set. The formula is:

[0116] ;

[0117] in It is the second signal-to-noise ratio threshold.

[0118] In practical implementation, when the signal-to-noise ratio exceeds At that time, the size of the time window It will decrease as the signal-to-noise ratio increases, and when the signal-to-noise ratio is less than... At that time, time window That's when it will increase.

[0119] This approach more accurately reflects the impact of the signal-to-noise ratio (SNR) on the time window size, ensuring that the UAV's data transmission frequency can be appropriately adjusted when the SNR changes significantly. This adjustment is based on the actual SNR position relative to a set threshold and employs exponential and logarithmic functions to enhance the adjustment range of the window size, making it more sensitive and reasonably responsive to environmental changes.

[0120] In this application, the exponential function To enhance the signal-to-noise ratio When the time window is higher than the threshold The reduction effect. When Exceeding the threshold At that time, that is It is a positive number. Rapid growth, which means that as With further increases in signal strength, the reduction in the time window becomes even more significant. The processing method employed in this application is suitable for scenarios requiring greater sensitivity to high signal-to-noise ratios, ensuring more frequent data transmission when the signal is excellent, thus leveraging the advantages of a favorable signal environment.

[0121] In this application, the logarithmic function Used when Below the second threshold Add time window In this scenario, the logarithmic function grows slowly, which means that even... As the gap widens, the time window also increases gradually, without drastically increasing the window due to a slight decrease in signal strength. This maintains a certain level of stability and data continuity.

[0122] In this way, extreme adjustments caused by sudden signal changes can be avoided, maintaining a more stable data transmission efficiency. Utilizing... and Its mathematical properties allow for a balance between performance and the stability of response speed and time window adjustment, especially in highly dynamic drone operating environments.

[0123] For example, a second initial time window can be set. 1.5 seconds.

[0124] Signal-to-noise ratio adjustment factor : 0.05.

[0125] First signal-to-noise ratio threshold 20 dB.

[0126] If the current signal-to-noise ratio is 23 dB;

[0127] ;

[0128] Thus, when the signal-to-noise ratio (SNR) slightly exceeds the first SNR threshold, the time window is reduced, resulting in a value of 0.49575 seconds. This is a significant decrease compared to the initial second window of 1.5 seconds, demonstrating that as the SNR increases, the data update frequency is increased to better utilize signal conditions and adapt to dynamically changing environments. Similarly, the method used in the first scenario mode can be referenced to set a minimum time window value, which will not be elaborated upon here.

[0129] As an optional implementation, the time window size is set when the drone is in the third scene mode. The formula is:

[0130] ;

[0131] in It's the signal strength. It is the size of the third initial time window. It is the signal strength adjustment coefficient. It is the first signal threshold. It is the rate of change of signal strength.

[0132] As an optional implementation, when the drone is in the third scene mode, the time window size is set. The formula is:

[0133] ;

[0134] in, It is the signal strength adjustment coefficient. It is the second signal threshold.

[0135] In practice, besides considering the signal strength exceeding the first signal threshold, the rate of increase in signal strength can also be taken into account. If the signal strength increases rapidly, it indicates that the drone may be approaching the signal source. In this case, the time window can be reduced more aggressively to increase the data update frequency.

[0136] Similarly, when the signal strength is below the second signal threshold and the signal strength is decreasing, it means that the drone may be leaving the area with a strong signal. In this case, the time window should be increased to prevent data loss.

[0137] In this way, introducing the rate of change of signal strength makes the adjustment of the time window more sensitive and timely, better adapting to the actual flight conditions of the UAV in complex environments. By monitoring rapid changes in signal strength, the data update frequency can be controlled more precisely, ensuring faster data updates when the signal is strong and slower updates when the signal is weak to reduce data loss.

[0138] Regarding S103 above: Obtain the real-time flight information of the UAV, and use the hybrid neural network to determine the signal switching point of the UAV within the coverage area of ​​the multiple target signal towers.

[0139] As one possible implementation, the output of the hybrid neural network includes: signal quality score, signal attenuation prediction, and interference level assessment; determining the signal switching point of the UAV within the coverage area of ​​multiple target signal towers includes:

[0140] Using the first dynamic programming scheme, determine the signal switching points of the UAV within the coverage area of ​​multiple target signal towers;

[0141] The first dynamic programming scheme includes: state definition, decision variables, transition equation, objective function, initialization and boundary conditions, and dynamic programming solution.

[0142] In practical implementation, when the neural network receives real-time flight data, such as flight speed, signal strength, and environmental noise, it immediately processes and predicts the upcoming signal conditions or other flight-related parameters. Based on these predictions, dynamic decisions are made, such as adjusting the time window size. If the predicted signal will worsen, the system can increase the time window to reduce the data update frequency and avoid transmitting too much data when the signal is unstable; conversely, if the predicted signal is stable, the system can decrease the time window and increase the data transmission frequency to improve the real-time performance of data processing. Furthermore, this dynamic adjustment mechanism can also be combined with the three scenarios described above. For example, if the hybrid network includes speed prediction, then the predicted speed from the hybrid network can be combined with the time window adjustment method mentioned above for the speed scenario. The signal-to-noise ratio and signal strength are similarly addressed, and will not be elaborated upon here.

[0143] After the decision-making process is adjusted, the new operational parameters (such as the adjusted time window size) will affect subsequent data acquisition and processing. This newly acquired data is then fed back into the neural network for analysis. In this way, the network continuously receives new inputs and adjusts its output, forming a closed-loop feedback system. This feedback mechanism ensures that the UAV system can continuously learn and adapt to environmental changes, optimizing its flight and data transmission strategies. For example, if the actual signal quality does not match the prediction, the system can further adjust the time window based on the actual situation, ensuring the stability and accuracy of data transmission.

[0144] For example, when a drone is performing urban surveillance tasks, it can be set to the third scene mode, with the following parameters: base time window. Signal strength adjustment coefficient and Signal strength threshold and .

[0145] UAVs collect real-time signal strength According to the formula:

[0146] ;

[0147] ;

[0148] The new time window size is calculated. The set parameters are then substituted into the calculation to obtain the new time window size, which is used to handle changes in signal strength.

[0149] The adjusted time window size is used in subsequent data acquisition cycles. If the signal strength improves in the next cycle, such as with a signal strength increase, the time window will be recalculated and appropriately reduced to increase the data update frequency and improve the real-time performance of the monitoring.

[0150] In this implementation, the reason for adopting a dynamic programming approach is that while hybrid neural networks are powerful and can predict and evaluate signal quality, signal attenuation, and interference levels, they do not directly decide on specific operational steps but rather provide decision support data. Dynamic programming can combine this predictive information and use a series of optimization algorithms to calculate the optimal signal switching strategy. Dynamic programming provides a systematic approach to decision-making problems, especially in complex environments involving multi-step decisions and multi-state considerations. It ensures that throughout the entire flight mission, the UAV's behavior always moves towards the optimal objective (such as minimizing signal switching losses and maximizing signal quality). The output of the hybrid neural network serves as a key parameter in the dynamic programming input, helping to define state transitions and evaluate possible decision effects. This allows the entire system to leverage the powerful predictive capabilities of neural networks while achieving precise optimization decisions through dynamic programming.

[0151] Among them, the state definition is used to define the connection state between the UAV and each signal tower; the decision variable is used to determine whether to switch from one signal tower to another; the transition equation is used to describe the transition logic and conditions between states; the objective function is used to set the optimization objective, such as maximizing signal quality or minimizing switching cost; initialization and boundary adjustment are used to set the initial state and calculate boundary adjustment; dynamic programming solution is used to solve the signal tower switching problem.

[0152] As an optional implementation, the state definition includes: defining the state. ,in, Indicates a point in time The drone connected to the signal tower The situation;

[0153] The decision variables include: definitions ,in, For at a certain point in time At that time, the drone flew from the signal tower. Switch to signal tower The decision variables, among which, It is a binary variable, when the switch occurs. When no handover occurs ;

[0154] The recurrence equation S(t+1,j) The function expression for ) is:

[0155] ;

[0156] in, and The index of the cell tower indicates the cell tower the drone is currently connected to. To from the signal tower Switch to signal tower The cost includes: signal quality difference information, predicted signal attenuation information, and signal interference information; At a certain point in time At that time, the drone was at the signal tower Signal quality score; The set of all signal towers;

[0157] The optimization objective is to maximize signal quality and minimize handover costs throughout the entire flight.

[0158] The objective function is used to maximize signal stability at the end of flight and minimize the total switching cost. The objective function is:

[0159] ;

[0160] Where T is the final time point at the end of the flight, and j is the index of the signal tower;

[0161] The initialization and boundary conditions include: setting an initial state based on the signal tower at the time of UAV startup. During the flight, a connection to a signal tower must be established at each point in time, that is:

[0162] ;

[0163] Where T is the final time point at the end of the flight. For each point in time during the flight, For the index of the signal tower;

[0164] The dynamic programming solution includes:

[0165] from Start by calculating the optimal value for all possible current states based on the previous state and the current signal quality and transfer cost;

[0166] Recursive calculation until The signal switching point from the beginning to the end is determined by backtracking.

[0167] In practice, hybrid neural networks provide signal quality scoring. The outputs, including the predicted future signal conditions, serve as input parameters for dynamic programming. Based on changes in real-time flight and signal status, the input data of the neural network is updated, signal quality and transfer costs are reassessed, and subsequent decision paths are optimized.

[0168] Among them, switching costs The calculation can be determined based on several factors, including: when a drone switches from one cell tower to another, it may experience a brief connection interruption, and this time delay can be translated into cost. During the handover process, if data transmission is not completed, data loss may occur, and this risk can also be quantified as cost. Signal handover may require additional processing and energy expenditure, especially in areas with weak signal strength where searching for new cell towers may be necessary.

[0169] The specific cost calculation method can be a weighted sum of these factors, with the weight allocation depending on the specific requirements of the task and the operating environment. For example, if the task is extremely time-sensitive, the weight of time delay can be higher; if the task has extremely high requirements for data integrity, the cost of data loss risk can be given more weight. In this way, the switching cost can be reasonably assessed according to the actual situation, providing quantitative decision support for dynamic planning.

[0170] In dynamic programming, the calculation is recursively performed until... Determining the signal switching points from start to finish through backtracking is a core step. Here, T represents the planned end time, and the entire process includes two main phases:

[0171] 1. Recursive Calculation: Starting from the initial time point, the optimal state at each time point is calculated step by step using the recurrence relation. For each time point... and every possible signal tower According to the previous time point Given the current state, signal conditions, and cost at the current time, the system calculates the optimal value for each state. This process continues until the task's end time T is reached.

[0172] 2. Backtracking process: Once time T is reached, it will start from... The process begins by tracing the signal switching points backward. This step is performed by examining the optimal decisions made at each point in time, i.e., checking which cellar switch provided the best outcome. By backtracking, it becomes clear how the drone should have switched from one cellar to another throughout the mission, ensuring maximum signal strength and minimum cost across the entire flight mission.

[0173] For example, the specific implementation process of an emergency rescue monitoring mission using a drone in an urban environment.

[0174] First, the drone's mission is to monitor urban areas filled with high-rise buildings, requiring real-time transmission of video data. Because of the uneven distribution of cell towers and the impact of building obstructions on signal strength, dynamic adjustments to signal connections are necessary to ensure data transmission quality.

[0175] At the start of the mission, the estimated monitoring time is set to 120 minutes, defined as T. Upon takeoff, the drone selects the strongest signal tower near the starting point as the initial connection tower. Every minute, the system collects the signal quality data of the current signal tower. and from the current tower The cost of switching to another tower j .

[0176] Based on the output of the hybrid neural network, the dynamic programming scheme proceeds through the following steps: First, define the state. Indicates time Drone connects to signal tower In this case, and set decision variables. To decide whether to leave the tower Switch to tower Then, the state is updated according to the above transition equation, and the optimal handover strategy is selected to maximize signal quality and minimize handover cost.

[0177] Throughout the flight, the objective function optimization aims to maximize signal quality and minimize handover cost, i.e. During this process, adjustments are made in real time based on feedback from signal conditions and neural network predictions to ensure that the drone always connects to the best signal tower.

[0178] In this way, drones can efficiently complete monitoring tasks in complex urban environments, maintain optimal signal connections, and ensure the stability and real-time performance of data transmission. After the mission is completed, the collected data and records of dynamic adjustments during execution are used to analyze and optimize dynamic programming models for future missions, thereby improving the overall performance and reliability of the system.

[0179] For example, consider a drone used for urban traffic monitoring. Assume there are three signal towers A, B, and C.

[0180] Signal quality score A=70, B=50, C=80.

[0181] Switching costs : From A to B = 10, From A to C = 20, From B to A = 5, From B to C = 15, From C to A = 10, From C to B = 10.

[0182] Time point The consideration is the transition from time 1 to time 2.

[0183] Assuming the drone initially connects to cell tower A, its state is defined as follows: (One base value), the rest .

[0184] Calculate the optimal strategy for signal switching from A to B and C:

[0185] Calculate using specific values:

[0186] ;

[0187] Similarly, calculate :

[0188] ;

[0189] At time point 2, connection to signal tower C provided the highest total score. Therefore, the optimal strategy is to switch from cell tower A to cell tower C.

[0190] In this way, the dynamic programming approach not only considers signal quality and switching costs but also adjusts the drone's connectivity strategy in real time, ensuring optimal communication efficiency and stability in a dynamic urban environment. This method allows drones to make the most appropriate cell tower switching decisions based on real-time conditions when performing monitoring tasks.

[0191] Based on the same inventive concept, this application also provides a 5G-based ultra-low latency data transmission system corresponding to the 5G-based ultra-low latency data transmission method. Since the principle of the device in this application is similar to the above-mentioned 5G-based ultra-low latency data transmission method, the implementation of the system can refer to the implementation of the method, and the repeated parts will not be described again.

[0192] Reference Figure 3 The diagram shown is a schematic representation of a 5G-based ultra-low latency data transmission system provided in an embodiment of this application. The system includes:

[0193] The acquisition unit 10 is used to acquire and preprocess historical flight data of the UAV within the coverage area of ​​multiple target signal towers; wherein, the historical flight data includes: quadrature amplitude modulation data, access point data, and signal attenuation data;

[0194] Construction unit 20 constructs a hybrid neural network based on the preprocessed historical flight data; wherein the hybrid neural network includes: a spatial feature extraction branch, a temporal feature extraction branch, and a signal attenuation feature extraction branch;

[0195] The processing unit 30 is used to acquire the real-time flight information of the UAV and use the hybrid neural network to determine the signal switching point of the UAV within the coverage area of ​​the multiple target signal towers.

[0196] This application also provides an electronic device, such as... Figure 4 The diagram shown is a schematic representation of the electronic device structure provided in an embodiment of this application, including:

[0197] Processor 41 and memory 42; the memory 42 stores machine-readable instructions executable by the processor 41, and the processor 41 executes the machine-readable instructions stored in the memory 42. When the machine-readable instructions are executed by the processor 41, the processor 41 performs the following steps:

[0198] Historical flight data of the UAV within the coverage area of ​​multiple target signal towers is acquired and preprocessed; wherein, the historical flight data includes: quadrature amplitude modulation data, access point data, and signal attenuation data;

[0199] Based on the preprocessed historical flight data, a hybrid neural network is constructed; wherein the hybrid neural network includes: a spatial feature extraction branch, a temporal feature extraction branch, and a signal attenuation feature extraction branch;

[0200] The real-time flight information of the UAV is acquired, and the hybrid neural network is used to determine the signal switching point of the UAV within the coverage area of ​​the multiple target signal towers.

[0201] The aforementioned memory 42 includes a main memory 421 and an external memory 422; the main memory 421, also known as internal memory, is used to temporarily store the computational data in the processor 41, as well as the data exchanged with external memory 422 such as a hard disk. The processor 41 exchanges data with the external memory 422 through the main memory 421.

[0202] The specific execution process of the above instructions can be referred to the steps of the 5G-based ultra-low latency data transmission method described in the embodiments of this application, and will not be repeated here.

[0203] This application also provides a computer-readable storage medium storing a computer program. When a processor executes the computer program, it performs the steps of the 5G-based ultra-low latency data transmission method described in the above-described method embodiments. The storage medium can be either volatile or non-volatile computer-readable storage.

[0204] This application also provides a computer program product carrying program code. The program code includes instructions that can be used to execute the steps of the 5G-based ultra-low latency data transmission method described in the above method embodiments. For details, please refer to the above method embodiments, which will not be repeated here.

[0205] The aforementioned computer program product can be implemented through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied in a computer storage medium; in another optional embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0206] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0207] It should be understood that determining B based on A does not mean determining B solely based on A; it also means determining B based on A and / or other information.

[0208] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0209] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0210] The preferred embodiments of the present invention disclosed above are merely illustrative of this application. These preferred embodiments do not exhaustively describe all details, nor do they limit the application to specific implementations. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of this application, thereby enabling those skilled in the art to better understand and utilize this application. This application is limited only by the claims and their full scope and equivalents.

Claims

1. A 5G-based ultra-low latency data and image transmission method, characterized in that, include: Historical flight data of the UAV within the coverage area of ​​multiple target signal towers is acquired and preprocessed; wherein, the historical flight data includes: quadrature amplitude modulation data, access point data, signal attenuation data, and flight speed and network change information; wherein, the quadrature amplitude modulation data is the in-phase component and quadrature component of the wireless link baseband signal; Based on the preprocessed historical flight data, a hybrid neural network is constructed; wherein the hybrid neural network includes: a spatial feature extraction branch, a temporal feature extraction branch, and a signal attenuation feature extraction branch; The spatial feature extraction branch is a multi-layer convolutional neural network with each layer followed by batch normalization and ReLU activation, used to extract the spatial position and environmental layout features of the UAV relative to the target signal tower and target environmental obstacles. The time feature extraction branch is used to extract the time dependence and dynamic changes of the historical flight data. The time feature extraction branch is a multi-layer long short-term memory network and includes a time step sliding window method. The time window size is dynamically adjusted according to a preset threshold rule based on three scene modes corresponding to the UAV flight speed, environmental signal-to-noise ratio and signal strength. The signal attenuation feature extraction branch consists of at least one fully connected layer and each layer is followed by a dropout layer. It is used to extract and analyze distance and signal strength data from the UAV to each signal tower, and to predict signal attenuation and potential switching points. The real-time flight information of the UAV is acquired, and the hybrid neural network is used to determine the signal switching point of the UAV within the coverage area of ​​the multiple target signal towers; The construction of the hybrid neural network includes: after preprocessing the historical flight data, inputting the spatial feature extraction branch, the temporal feature extraction branch, and the signal attenuation feature extraction branch; The preprocessing includes: Random phase shift is applied to the orthogonal amplitude modulation data to generate target training samples for training the hybrid neural network. The actual delay distribution is calculated based on the historical flight data, generating a distribution that includes random delays. The first time series; It also includes: acquiring the real-time flight information of the UAV, using the hybrid neural network to obtain prediction results including at least signal quality score, signal attenuation prediction and interference level assessment; based on the prediction results and the distance from the UAV to each signal tower, signal strength and environmental layout information, according to the first dynamic programming scheme, solving based on state definition, decision variables, transition equation, objective function, initialization and boundary conditions, with the goal of maximizing signal quality and minimizing switching cost, determining the optimal signal switching point sequence of the UAV within the coverage area of ​​multiple target signal towers.

2. The 5G-based ultra-low latency data and image transmission method according to claim 1, characterized in that, The dynamic adjustment of the time window size includes: Based on the flight mission of the UAV, a scene mode is determined; the scene mode includes a first scene mode, a second scene mode, and a third scene mode. When the drone is in a first scene mode, the drone's flight speed is detected, and a first initial time window size is set; in response to the drone's flight speed being greater than or equal to a first speed threshold, the time window size is decreased; in response to the drone's flight speed being less than a second speed threshold, the time window size is increased. When the drone is in the second scene mode, the signal-to-noise ratio of the environment in which the drone is located is detected, and a second initial time window size is set. In response to the signal-to-noise ratio being greater than or equal to a first signal-to-noise ratio threshold, the time window size is reduced, and in response to the signal-to-noise ratio being less than the second signal-to-noise ratio threshold, the time window size is increased. When the drone is in the third scene mode, the signal strength of the drone is detected, and a third initial time window size is set. In response to the signal strength being greater than or equal to a first signal threshold, the time window size is reduced, and in response to the signal strength being less than a second signal threshold, the time window size is increased.

3. A 5G-based ultra-low latency data transmission system, used to implement the 5G-based ultra-low latency data transmission method as described in claim 1 or 2, characterized in that, include: The acquisition unit is used to acquire and preprocess historical flight data of the UAV within the coverage area of ​​multiple target signal towers; wherein, the historical flight data includes: quadrature amplitude modulation data, access point data, signal attenuation data, and flight speed and network change information; wherein, the quadrature amplitude modulation data is the in-phase component and quadrature component of the wireless link baseband signal; A construction unit is used to construct a hybrid neural network based on the preprocessed historical flight data. The hybrid neural network includes a spatial feature extraction branch, a temporal feature extraction branch, and a signal attenuation feature extraction branch. The spatial feature extraction branch is a multi-layer convolutional neural network with each layer followed by batch normalization and ReLU activation, used to extract the spatial position and environmental layout features of the UAV relative to the target signal tower and target environmental obstacles. Constructing the hybrid neural network includes preprocessing the historical flight data and then inputting it into the spatial feature extraction branch, the temporal feature extraction branch, and the signal attenuation feature extraction branch. The time feature extraction branch is used to extract the time dependence and dynamic changes of the historical flight data. The time feature extraction branch is a multi-layer long short-term memory network and includes a time step sliding window method. The time window size is dynamically adjusted according to a preset threshold rule based on three scene modes corresponding to the UAV flight speed, environmental signal-to-noise ratio and signal strength. The signal attenuation feature extraction branch consists of at least one fully connected layer and each layer is followed by a dropout layer. It is used to extract and analyze distance and signal strength data from the UAV to each signal tower, and to predict signal attenuation and potential switching points. The preprocessing includes: Random phase shift is applied to the orthogonal amplitude modulation data to generate target training samples for training the hybrid neural network. The actual delay distribution is calculated based on the historical flight data, generating a distribution that includes random delays. The first time series; The processing unit is used to acquire the real-time flight information of the UAV and, using the hybrid neural network, determine the signal switching point of the UAV within the coverage area of ​​the multiple target signal towers; The processing unit is further configured to acquire real-time flight information of the UAV, and use the hybrid neural network to obtain prediction results including at least signal quality score, signal attenuation prediction, and interference level assessment; based on the prediction results and the distance from the UAV to each signal tower, signal strength, and environmental layout information, according to the first dynamic programming scheme, the system solves the problem based on state definition, decision variables, transition equations, objective functions, initialization, and boundary conditions, with the goal of maximizing signal quality and minimizing switching costs, to determine the optimal signal switching point sequence of the UAV within the coverage area of ​​multiple target signal towers.

4. An electronic device, characterized in that, include: The processor and the memory, wherein the memory stores machine-readable instructions executable by the processor, the processor is used to execute the machine-readable instructions stored in the memory, and when the machine-readable instructions are executed by the processor, the processor performs the steps of the 5G-based ultra-low latency data transmission method as described in claim 1 or 2.

5. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by an electronic device, performs the steps of the 5G-based ultra-low latency data transmission method as described in claim 1 or 2.

Images