A job data transmission processing method suitable for a heterogeneous unmanned aerial vehicle cluster

Abstract

This invention discloses a method for processing operational data transmission in heterogeneous UAV swarms, relating to the field of UAV application technology. The method includes constructing a transmission prediction model based on task trajectory and communication capabilities. Before the task is executed, a structured estimate of link quality change trends and potential channel contention is formed. During operation, the operational indicators of the protocol's physical layer and MAC layer are monitored, and trends are extrapolated through a sliding window to transform them into continuous environmental evolution trends. The extrapolated trend results are compared with the output of the initial prediction model in the same dimension, and attribution calculations are further performed using the consistency of changes in physical layer and MAC layer indicators. Each UAV makes differentiated adjustments based on its own attribution failure type, thereby improving overall resource utilization efficiency. This invention achieves targeted and adaptive heterogeneous swarm transmission strategies through attribution-based parameter adjustment and online model correction.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) application technology, specifically a method for processing operational data transmission in heterogeneous UAV swarms. Background Technology

[0002] In low-altitude collaborative operation environments, the communication of heterogeneous UAV swarms is mainly affected by the combined effects of three types of objective constraints. First, communication links are affected by factors such as building obstruction, terrain undulations, and multipath propagation. The link gain of some UAVs in the swarm may exhibit significant short-term fluctuations due to environmental obstruction. Second, there is the issue of channel resource competition. During multi-UAV collaborative operations, the number of UAVs, their spatial distribution, and the pace of task execution are constantly changing, leading to continuous reconstruction of concurrent access relationships on the same channel.

[0003] Third, the communication performance of drones in a heterogeneous swarm varies. Different drones exhibit significant differences in perceived data scale, encoding and processing capabilities, link conditions, and remaining energy. These differences directly translate into varying degrees of data compressibility and channel utilization efficiency during communication. Therefore, in scenarios of concurrent transmission, these differences can easily lead to some drones over-consuming resources and operating at low efficiency, while other capable drones are underutilized.

[0004] Existing methods that employ periodic communication scheduling or allocate communication resources for each UAV at the start of a mission often implicitly assume that the transmission conditions of multiple UAVs are relatively stable within the scheduling period. However, in actual execution, the number of concurrent heterogeneous UAVs transmitting, communication link conditions, and competition relationships are constantly changing, leading to problems such as invalid transmission, increased channel conflicts, or increased data latency. These methods are insufficient to meet the requirements of heterogeneous UAV swarms for transmission efficiency and stability in highly dynamic environments. Summary of the Invention

[0005] (a) Technical problems to be solved This invention provides a method for processing operational data transmission in heterogeneous drone swarms, which can make differentiated adjustments to data transmission based on the actual constraints of each drone in the heterogeneous swarm.

[0006] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: a method for processing operational data transmission in heterogeneous UAV swarms, comprising the following steps: Obtain the task transmission constraints for collaborative operations of heterogeneous UAV clusters, including upper and lower limits of throughput and maximum allowable latency, and construct a transmission prediction model based on the inherent characteristics of each UAV's communication hardware and the preset task trajectory. The transmission prediction model transforms the preset mission trajectories of each UAV into spatiotemporally related spatial relative positions to predict the environmental interference at future time points. Then, it maps and generates a reference transmission parameter set corresponding to each time point. The environmental interference includes link gain and channel collision probability, and the reference transmission parameter set includes data compression ratio and channel access probability. During the operation, each UAV encapsulates and transmits operation data based on the corresponding reference transmission parameter set within the set sliding observation time window, and simultaneously monitors the operation indicators of the physical layer and media access control layer of the communication protocol stack; wherein, the signal-to-noise ratio and signal strength changes are extracted at the physical layer, and the number of collision backoffs and ACK timeouts are recorded at the media access control layer. Within the observation time window, calculate the rate of change of each of the aforementioned operational indicators, use the trend extrapolation method to predict the link gain and channel collision probability at several future time points, and construct the environmental interference prediction quantity. The predicted environmental interference quantity is compared with the environmental interference quantity at the corresponding future time point in the same dimension. When the deviation exceeds the preset confidence interval, the failure probability caused by channel quality degradation and channel contention intensification is estimated and the failure type is determined based on the correlation between the operating indicators of the physical layer and the medium access control layer. Based on the failure type, the current data compression ratio and channel access probability are corrected and covered respectively; and the deviation is used as a feedback quantity to correct the prediction weights of link gain and channel collision probability in the transmission prediction model.

[0007] In some feasible embodiments, the following steps are performed when constructing the transmission prediction model for the heterogeneous drone swarm: The preset mission trajectories of each UAV are imported onto a unified time axis, and time discretization processing is performed to calculate the time-varying relative distance between any two UAVs. The inherent characteristics of the communication hardware of each UAV are obtained, including transmit power, antenna gain and receive sensitivity. The time-varying relative distance between any two UAVs is converted into the link transmission loss characteristics at the corresponding time point, and a predictive relationship of link gain changing with time is established. Identify the potential interference sets of each UAV at different time points, and combine them with the preset channel access initial strategy when performing the task to construct a channel conflict probability prediction mechanism for the communication link competition relationship of each UAV in a heterogeneous UAV cluster. The prediction relationship of link gain and the channel conflict probability prediction mechanism are integrated to form a transmission prediction model that predicts changes in the communication environment at future points in time.

[0008] In some feasible embodiments, when the transmission prediction model performs prediction, the preset mission trajectory of each UAV is mapped to a spatial position sequence at continuous time points, and the relative distance change relationship between each UAV is calculated based on the spatial position sequence. Based on the relative distance change relationship, and combined with the transmission capability and reception conditions in the communication hardware characteristics, the link propagation status at each time point is predicted, and the corresponding link gain change trend is obtained. Simultaneously, based on the relative distance changes of each UAV at each time point, the potential UAV set that generates channel competition within the link coverage area is identified, and the channel conflict probability at each time slice is predicted. The link gain generated at each time point and the channel collision probability constitute the environmental interference quantity.

[0009] In some feasible embodiments, when generating the reference transmission parameter set using the predicted interference amount mapping, at each predicted time point, the data compression ratio that satisfies the minimum transmission requirement is calculated through a utility function based on the lower limit of throughput in the task transmission constraints and the predicted link gain; and the channel access probability is calculated based on the maximum allowable delay and the predicted channel collision probability.

[0010] In some feasible embodiments, during collaborative operations within a heterogeneous UAV swarm, each UAV, within a set sliding observation time window, encapsulates and transmits the operational data to the receiving end based on the corresponding reference transmission parameter set, and simultaneously collects operational metrics of each layer of the communication protocol stack during data transmission; wherein, At the physical layer, the signal-to-noise ratio and received signal strength of each data packet are obtained by measuring the received job data in real time. In the media access control layer, the number of collision backoffs triggered in each job data transmission and the ACK timeout frequency caused by not receiving acknowledgment data from the receiver are recorded. Within each observation time window, first-order difference operations are performed on the signal-to-noise ratio, signal strength, number of collision backoffs, and ACK timeout frequency to obtain the rate of change of each operating index within the observation time window.

[0011] In some feasible embodiments, when using the trend extrapolation method to predict and construct the environmental interference prediction quantity, for the rate of change of the signal-to-noise ratio and signal strength of the physical layer, a time-weighted trend extension method is used to predict the direction and magnitude of change of these values ​​at several future time points, and the prediction results are mapped to the link gain change trend at the corresponding time points. Based on the rate of change of the number of collision backoffs and the ACK timeout frequency at the medium access control layer, the channel contention level at several future time points is extrapolated based on their growth rate, and the channel collision probability change trend at the corresponding time points is obtained.

[0012] In some feasible embodiments, after obtaining the predicted amount of environmental interference based on trend extrapolation, it is time-aligned with the amount of environmental interference output by the transmission prediction model at the corresponding future time point, and the same-dimensional deviation calculation is performed. The deviation measures of each dimension are compared with the preset confidence intervals. When any deviation continues to exceed the risk range, or shows a monotonically increasing trend over multiple consecutive time points, the current environmental prediction results are determined to be mismatched, triggering failure attribution calculation.

[0013] In some feasible embodiments, after triggering the attribution calculation, the changing trends of signal-to-noise ratio and signal strength within the observation time window are aligned with the link gain prediction deviation, and the degree of matching between the two in terms of consistency of change direction and change magnitude is calculated to obtain a first attribution metric characterizing the impact of channel quality degradation; the changing trends of conflict backoff times and ACK timeout frequency are aligned with the channel conflict probability prediction deviation, and the degree of matching between the two in terms of consistency of change direction and change magnitude is calculated to obtain a second attribution metric characterizing the impact of increased channel competition. In some feasible embodiments, the first attribution metric and the second attribution metric are normalized to obtain the failure attribution probabilities corresponding to channel quality degradation and channel contention aggravation, respectively. Based on the relative magnitude of each attribution probability, the dominant failure type of the current environment prediction mismatch is determined.

[0014] (III) Beneficial Effects: Compared with the prior art, this invention has the following beneficial effects: This invention constructs a transmission prediction model based on task trajectory and communication capabilities in the initial stage, so that subsequent transmission parameters (compression ratio, access probability) are not statically set, but are established with spatiotemporal correlation with changes in spatial location and potential environmental interference. This allows for a preliminary structured prediction of the link quality change trend of cluster communication and the potential channel competition degree among UAVs in the cluster.

[0015] During operation, cross-layer operational metrics (signal-to-noise ratio and signal strength at the physical layer, and backoff count and ACK timeout rate at the MAC layer) are introduced. By using a sliding time window and trend extrapolation, discrete transmission results are transformed into continuous environmental evolution trends. By comparing the extrapolated trend results with the output of the initial prediction model in the same dimension, and further utilizing the consistency of changes in physical layer and MAC layer metrics for attribution calculation, the reasons for the degradation of cluster communication performance are decomposed into communication link quality issues or multi-machine channel contention issues. Compared with adjusting strategies based on throughput or packet loss rate, this approach avoids confusion between different failure mechanisms, thereby reducing system oscillations caused by incorrect adjustments. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating a method for processing operational data in a heterogeneous drone swarm, as provided in an embodiment of the present invention. Figure 2 This is a flowchart illustrating the process of obtaining the task constraints of the heterogeneous drone cluster and the initial transmission strategy of each drone in the cluster in a data transmission processing method for heterogeneous drone clusters provided in an embodiment of the present invention. Figure 3 This is a flowchart illustrating the mapping sequence at each level in a job data transmission processing method for heterogeneous drone clusters provided in an embodiment of the present invention, when the heterogeneous drone cluster performs job transmission tasks. Detailed Implementation

[0017] 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. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0018] It should be noted that, where there is no conflict, the features in the embodiments of the present invention can be combined with each other.

[0019] In low-altitude mission environments, UAV communication faces frequent obstructions, such as unavoidable factors like buildings, terrain, and vegetation. Additionally, UAVs typically move at high speeds during missions, and when multiple UAVs are performing the same task concurrently, bandwidth contention occurs, limiting the data transmission from each UAV and ultimately restricting the actual effective data throughput of the UAV.

[0020] Environmental interference (buildings, vegetation obstruction, etc.) is indeed an objective physical constraint, but in communication protocols, existing scheduling algorithms often make decisions based on "outdated" state information. By the time the central scheduling center or inter-machine protocol completes a resource allocation, the actual channel state may have deteriorated, leading to a large number of retransmissions and packet losses. The delay of this feedback loop far exceeds the impact of the environmental interference itself.

[0021] When link states change rapidly in low-altitude environments due to building obstruction, maneuvering, or fluctuations in node density, combined with Figures 1 to 3 The method shown is a data transmission processing method for heterogeneous UAV swarms. This method continuously corrects the judgment of the environment within each sliding window based on actual operating indicators (signal-to-noise ratio, backoff behavior, ACK feedback, etc.), so that the transmission strategy of the heterogeneous UAV swarm is always adjusted around the evolution trend of the current real state, thereby avoiding continuous misjudgment caused by prediction failure.

[0022] In the initial phase of heterogeneous drone swarm mission execution, it is necessary to acquire mission operation information and conduct a preliminary assessment of the communication capabilities of each drone during mission execution. (Refer to the reference...) Figure 2 Specifically, step S10 involves obtaining the task transmission constraints for the collaborative operation of the heterogeneous UAV cluster, including the upper and lower limits of throughput and the maximum allowable latency, and constructing a transmission prediction model based on the inherent characteristics of each UAV's communication hardware and the preset task trajectory.

[0023] First, the task's operational requirements need to be received from the task awareness platform, including the upper and lower limits of throughput and data transmission latency constraints. The upper throughput limit, i.e., the maximum amount of data the task can receive (e.g., 1GB per hour), is used to control the total output of the drone and avoid channel congestion.

[0024] The minimum throughput limit can be understood as the amount of critical data that the task requires to be prioritized, such as at least 500MB of key target images or sensor measurement data, to ensure that the core objectives of the task are met. The data transmission latency constraint, i.e., the average data transmission latency acceptable to the task, such as ≤100ms, ensures the real-time performance or time-series dependence of the task.

[0025] After obtaining the above task constraints, which are often macro-level objectives—such as the maximum amount of data allowed to be transmitted, which data portions cannot be lost or slowed down, and the overall tolerable time delay—it is necessary to further transform these task constraints into executable data transmission constraints in wireless transmission systems.

[0026] First, based on the upper limit of data volume, in the embodiments of the present invention, the total upper limit of data needs to be divided first, including but not limited to the total amount of data allowed within the task cycle, the upper limit of data that can be sent within a unit time window, and the theoretical maximum allocation value of a single drone within that window.

[0027] After converting the acquired task constraints into relevant constraints for wireless system transmission, it is necessary to also convert the acquired capability indicators of each UAV in the cluster into an evaluation of the transmission capability of each UAV relative to this transmission operation.

[0028] For each UAV in the heterogeneous cluster, the link transmission loss characteristics at each time point are calculated based on the inherent characteristics of each UAV's communication hardware, including reflection power, antenna gain, and receiver sensitivity, thereby identifying the correlation between the changes in communication link gain over time during the flight of each UAV.

[0029] The first step in execution is not to directly calculate communication, but to transform the mission trajectory into a spatial relationship on a discrete time series. Specifically, this involves discretizing the preset flight trajectory of each UAV in time, for example, according to data transmission time slots or fixed time steps. Taking time steps as an example, the trajectories of all UAVs are expanded along a unified time axis, such as taking a state point every 1 second.

[0030] Then, at each discrete time point, the relative distances between drones and between drones and communication nodes are calculated, such as the Euclidean distance between them, and environmental information is used to mark whether the link is in an obstructed state (marked as LOS) or an unobstructed state (marked as NLOS).

[0031] After determining the distance and path type, the link calculation begins. This step is executed continuously. Generally, for each drone, the path loss of the communication link is calculated first, followed by the power of the received signal and the link gain.

[0032] When calculating path loss, the above calculations are combined at each time step. The calculation of any two drones and The relative distance between them (using Euclidean distance as an example) Calculate the path loss for each link at each time point: ; In the formula The path loss index is typically set based on the complexity of the task environment. The decay index is generally set based on the complexity distribution within the scope of the task environment.

[0033] Then, even considering the inherent characteristic parameters of each UAV's communication hardware, the link gain is calculated using the Friis transmission formula, which governs the link gain: ; Among them It refers to the transmission power of each drone. and These are the gain of the drone's transmitting antenna and receiving antenna, respectively.

[0034] Then, combined with the drone's receiver sensitivity... By obtaining the following parameters at each time step: whether communication is possible (link availability) and signal-to-noise ratio, the link gain sequence can be obtained. .

[0035] For heterogeneous drone swarm transmission operations, each drone has an initial access probability, and each drone attempts to send data in each time slot. Therefore, it can be understood that at each transmission time point, based on link relationships, it's possible to determine which drones are within the same channel coverage area and which drones might interfere with each other during transmission. (UAVs...) For example, based on the link competition relationship, a competition set will be obtained. That is, with drones The set of all drone nodes that have experienced transmission contention conflicts.

[0036] When calculating the probability of conflict, based on the drone The probability of itself attempting to send. And the probability of sending when no one else in the competing set uses the time. The probability of it having no contention is now set as: ; We can then deduce the probability of a conflict occurring as follows: ; At this stage of the calculation, it can be understood that the problem of competition for channel resources is equivalent to transforming it into a quantifiable and computable probabilistic problem. Furthermore, the conflict between drones in the cluster is no longer a matter of whether it occurred or not, but rather a predictable probability distribution at each point in time.

[0037] The next step is to perform S20: The transmission prediction model transforms the preset mission trajectories of each UAV into spatiotemporally related spatial relative positions to predict the environmental interference at each future time point. Then, it maps and generates the reference transmission parameter set corresponding to each time point. The environmental interference includes link gain and channel collision probability, and the reference transmission parameter set includes data compression ratio and channel access probability.

[0038] In the prediction model constructed by integrating the channel conflict probability prediction mechanism based on the communication link competition relationship, and based on the flight mission trajectory of each UAV and the communication hardware of each UAV, two types of interference are generated simultaneously for each future time point: one is the physical interference represented by path loss, and the other is the channel competition interference represented by the conflict probability.

[0039] The interference calculation for path loss is based on the relative distance obtained earlier. Path status (whether occluded), calculate path loss at that moment. Then, by combining the transmit power and antenna gain, the receive power is obtained, and further, the link gain or signal-to-noise ratio is obtained. It should be noted that the same link will have a continuously changing link capability curve at different points in time. This curve reflects how the link will improve or deteriorate if the drone flies along the trajectory.

[0040] Regarding the calculation of the contention interference amount for the collision probability, at the same time point (simultaneously with the calculation of path loss), based on the spatial distribution of the UAVs, determine which nodes are in the same channel coverage area (competition set) and the access probability of each UAV node (initial setting), and then calculate the collision probability of the UAV at that moment. The calculation formula is as described above.

[0041] After obtaining the interference amount, a first-level mapping is performed in the pre-simulation phase. Here, we first refer to... Figure 3 Specifically, environmental interference is transformed into constraints on transmission behavior, and the corresponding initial predicted transmission parameter set is output based on the amount of interference, including the data compression ratio and the initial channel access probability.

[0042] Regarding the calculation of the data transmission compression ratio, it is first based on the lower limit of throughput in the task constraints. and the link capacity for the current time period. , here It is obtained from the link gain: ; Among them This represents the signal-to-noise ratio at that time point.

[0043] Considering that the compression ratio is mainly used to adjust the amount of data that each drone needs to send to match the current link capacity, it generally satisfies the following: ; Here For the original data transmission rate, it needs to meet the requirement of being greater than the lower limit of throughput and less than the capacity of the link itself. Therefore, a feasible range can be obtained: ; It is important to note that this range is not chosen arbitrarily, but rather the optimal compression ratio is determined through a utility function. For example, the compression ratio can be selected within this range based on the information fidelity required by the task.

[0044] The probability of access is generated based on the maximum allowable latency in the task constraints. Based on the above calculations, the probability of conflict at that point in time is obtained. .

[0045] Considering that the communication mechanism of heterogeneous drone swarms is generally a random access mechanism, the expected time for a drone to successfully transmit data can be approximated as: ; This expected time needs to meet the task's latency requirements, that is: ; Substituting into the above equation, we get: ; Based on the above calculations, finally at each time point Each of these can yield a set of parameters: ; Among them This refers to the data compression ratio. This represents the access time slot probability; as time changes, the transmission parameter sequence for each UAV can be predicted.

[0046] After calculating the predicted transmission parameters for each drone in the aforementioned preliminary prediction of the heterogeneous drone cluster, the next step is S30: During operation, each drone encapsulates and transmits operation data based on the corresponding reference transmission parameter set within the set sliding observation time window, and simultaneously monitors the operating indicators of the physical layer and media access control layer of the communication protocol stack; among them, the signal-to-noise ratio and signal strength changes are extracted at the physical layer, and the number of collision backoffs and ACK timeout rates are recorded at the media access control layer.

[0047] During the operation, each drone encodes the data according to the generated baseline parameters, uses a compression ratio, attempts to send data according to the access probability, and then receives feedback and updates its own recorded status.

[0048] First, it is necessary to explain why the physical layer and MAC layer metrics in the protocol are different. The physical layer metrics are used to directly reflect the quality of the communication link, while the MAC layer metrics reflect the channel resource contention caused by multi-machine transmission.

[0049] Regarding the acquisition of physical layer operational metrics, firstly, within a certain time slot, the UAV generates data units according to a predetermined compression ratio and decides whether to initiate transmission based on the access probability. Once the transmission phase begins, the signal will be affected by physical environmental factors such as path loss, obstruction, and multipath propagation during its spatial propagation. These physical processes directly determine the signal form received by the receiver.

[0050] At the signal receiving end, while demodulating the signal, two quantities are estimated simultaneously: one is the Received Signal Strength (RSSI), which can be directly measured as power intensity when receiving the signal.

[0051] Another is the signal-to-noise ratio (SNR), which is estimated during demodulation for each successfully received data packet.

[0052] It is important to note that these two quantities are generated simultaneously within the same receiving action, and are not separately calculated metrics.

[0053] Next, we move on to the MAC layer behavior metrics. First, it's important to understand that when a drone initiates a transmission, success is not guaranteed (refer to the access probability calculation above), as other drones may also be attempting to access the channel at the same time. If the channel is detected to be occupied or a collision occurs during transmission, a backoff mechanism is triggered. This process manifests as: transmission attempt fails, random backoff occurs, and a second attempt is made after a delay.

[0054] Each time such a process is triggered, it is recorded as a collision backoff at the transmitting end. Therefore, by counting the number of collision backoffs within a time window (consisting of one or more time slots), we can reflect how many other nodes (drones) are competing for the channel at the same time, that is, the intensity of the channel competition.

[0055] If the transmission does not fail directly at the physical layer or get blocked during the access phase, the data packet will continue to be transmitted to the receiving end and wait for an ACK confirmation. If no ACK is received within the specified time, it will be recorded as an ACK timeout. The ACK timeout rate can be obtained by calculating the ratio of ACK timeouts to transmissions within the time window.

[0056] It's important to note that ACK timeouts don't differentiate between specific causes. They can stem from two main scenarios: one is insufficient link quality, where data packets are corrupted or lost during transmission; the other is that although transmission was successful, collisions or interference prevented the receiver from decoding correctly. Therefore, the ACK timeout occurs precisely at the intersection of the physical layer and contention behavior within the execution link, essentially reflecting all ultimately unsuccessful data packet transmissions.

[0057] Up to this point, a set of raw observations has been generated for each time window: Physical layer: SNR and RSSI; MAC layer: number of collision backoff attempts, ACK timeout rate (whether to attempt to send / whether backoff was triggered / whether to receive successfully).

[0058] As the operation continues, a sliding time window is introduced to accumulate the results from multiple consecutive time slots. Within this window, the obtained metrics are arranged in chronological order, and their changes are calculated. Specifically, the differences between metrics at adjacent time points within the window are used to obtain the first-order difference, such as the change in the average SNR in the current window compared to the previous window, or the increase in the total number of collision backoffs relative to the previous window. The impact of the changing trends of each metric on transmission parameters (data compression ratio, access probability) is shown in Table 1 below.

[0059] Table 1. Impact of Measured Communication Performance Indicators at Each Layer on Transmission Parameters As can be seen from the table above, if the SNR continues to decline, even if it is currently still within the usable range, it indicates that the link is deteriorating. If the number of conflict avoidances continues to increase, it indicates that competition is intensifying; If the ACK timeout rate increases simultaneously, it indicates that the transmission problem has evolved from localized congestion to overall transmission failure.

[0060] In this way, through sliding window and differential calculation, the instantaneous behaviors that were originally scattered in each layer are transformed into a set of dynamic signals with time directionality. Finally, a set of information that can simultaneously reflect link changes, contention intensity and overall transmission results and has an evolution direction is output.

[0061] The next step is S40: within the observation time window, calculate the rate of change of each operating indicator, use trend extrapolation to predict the link gain and channel collision probability at several future time points, and construct the environmental interference prediction quantity.

[0062] In the previous stage (S30), each UAV has been operating for a period of time according to the initially predicted transmission parameters during mission execution, and has obtained a set of continuous operational metrics through the physical layer and MAC layer: SNR, RSSI, backoff count, ACK timeout rate, and their respective rates of change within the time window (first-order difference). Furthermore, in some embodiments, their acceleration (second-order difference) is also calculated. During UAV communication transmission, link attenuation is reflected in a decrease in the signal-to-noise ratio, and node congestion is reflected in an increase in the number of backoff counts.

[0063] However, it is important to note that these are still just discrete results (statistics within each time period). If we directly compare the current statistics with the prediction model, we will only get a conclusive result of whether there is a deviation, but we will not know whether the deviation is expanding / stabilizing / mitigating.

[0064] Therefore, when entering the third-level mapping (mapping the trend to environmental intervention), the observations in multiple consecutive time slots are first concatenated by a sliding time window, and then first-order and second-order difference processing is performed on these sequences.

[0065] However, using these rates of change directly still results in instability because sporadic disturbances in a single time slot (such as momentary blockages or a sudden collision) can cause abrupt changes, and if these are extrapolated directly, such momentary disturbances will be mistaken for trends.

[0066] Therefore, after obtaining the rate of change, in some embodiments of the present invention, a time-weighted mechanism is introduced, such as the exponential weighting (EWMA) of the first-order rate of change of a certain operating index as described below: in, The weights they represent are determined by historical trends.

[0067] Only after obtaining a smooth and stable trend does the extrapolation process begin. It's important to note that this extrapolation doesn't extend the original indicator, but rather the trend itself. For example, suppose we assume that the current direction and rate of change of a certain indicator will remain constant for a short period. Based on this assumption, we can advance the current indicator value along its direction of change by several time slots to obtain the indicator's trajectory in the future short term.

[0068] It is important to note that the extrapolation prediction window set here must be short-term, such as a few to a dozen time slots. This is because the MAC competition structure will not be drastically restructured in the short term, and long-term extrapolation errors will accumulate rapidly.

[0069] After completing the above trend extrapolation, at this point on the same time axis, for each time slot of the future prediction period ( There are two sets of sequences: One set consists of environmental disturbance quantities given in prior knowledge by the prediction model: and ; One set consists of actual measured data from the work, and the interference amount obtained by trend extrapolation: and .

[0070] The next step is S50: comparing the predicted environmental interference with the corresponding future environmental interference at the same time point in the same dimension. When the deviation exceeds the preset confidence interval, based on the correlation between the operating indicators of the physical layer and the medium access control layer, the failure probability caused by channel quality degradation and channel contention intensification are estimated and the failure type is determined.

[0071] After standardizing the interference quantities of each dimension and unifying their dimensions, the deviation between the predicted and extrapolated quantities of each dimension's indicators is calculated on a time-slot basis. and In some embodiments, to avoid [the situation in the observation window]... If a single point of fluctuation error occurs within the window, the deviation will continue to be aggregated and processed within the window:

[0072] ; ; Obtain their respective corresponding deviation measures and When the respective deviation measures exceed the corresponding confidence interval, attribution calculation is triggered. The confidence interval here is not a fixed threshold, but rather a tolerance limit for the normal error range of the model. When the deviation falls within this range, the prediction can be considered consistent with reality. However, if the following conditions occur, the current prediction mechanism is determined to have failed, and the failure type needs to be identified based on subsequent attribution calculations, thereby executing the corresponding prediction mechanism correction based on the type.

[0073] Specifically, we first look at the link direction. For the link gain, we align its predicted deviation sequence with the SNR and RSSI trends of the physical layer. The alignment dimensions include the direction and magnitude of the deviation trend and the trend of the operating indicators.

[0074] Regarding the assessment of the directional consistency of trends, we mainly look at the directional components of the link gain deviation and the rate of change of the signal-to-noise ratio. For example, if the link gain is predicted to increase, but the SNR trend is decreasing, it indicates that the prediction direction is incorrect.

[0075] Regarding amplitude matching, the correlation coefficient between the link gain deviation and the rate of change of the signal-to-noise ratio can be calculated, for example, by calculating the Pearson correlation coefficient: ; in, The set scale matching coefficient can be understood as a normalized coefficient that maps the rate of change of SNR to the dimension of link gain. This is a set constant to prevent overfitting. If the SNR decrease is close to the link gain deviation, it indicates a strong correlation between the two.

[0076] In summary, the above-mentioned directional consistency components and amplitude matching components The product of the two is now used to form the first attribution metric. If the prediction error of the link gain can be explained by the rate of change of SNR / RSSI in both direction and intensity, then the value of the first attribution metric is high.

[0077] Similarly, the calculation principle of the first attribution metric for access probability is the same as above.

[0078] Next, we examine the calculation of the attribution metric for increased channel contention (MAC layer). The collision probability prediction bias is aligned with the trends in MAC layer collision backoff counts and ACK timeout rates using the same method described above.

[0079] If the trend of conflict backoff growth is highly consistent with the deviation in conflict probability in both direction and magnitude, it indicates that the prediction mismatch mainly stems from changes in competitive relationships (e.g., access strategies are no longer suitable). This leads to the second attribution metric, which is obtained by multiplying the two components. Taking the number of conflict backoffs as an example, if the increase in the deviation in conflict probability is synchronously supported by a surge in the number of backoffs, then the corresponding second attribution metric value will be large.

[0080] Finally, the two metrics are normalized to obtain the failure probability caused by signal quality degradation and the failure probability caused by increased channel contention. Based on the magnitude of the attribution probability and whether it exceeds the set confidence interval, the type of dominant failure (physical link-dominated failure / contention-congestion-dominated failure) is determined.

[0081] Then, based on the failure type (physical link-driven failure / contention-driven failure), the current data compression ratio and channel access probability are corrected and covered respectively; and the deviation is used as feedback to correct the prediction weights of link gain and channel collision probability in the transmission prediction model (S60).

[0082] For each identified failure type, immediately adjust the operating parameters of the current protocol stack.

[0083] If the problem is attributed to physical failure, it indicates a deterioration in link quality. While meeting the minimum throughput requirement of the task, the link deviation needs to be mapped to an adjustment amount requiring compression. This is achieved by reducing the payload size of individual data packets to offset the increase in bit error rate caused by the drop in link gain. Regarding the compression adjustment amount... Calculation: ; in, This represents a calculation function that maps the link gain deviation to the proportion of data that needs to be reduced; for example, linear and logarithmic functions can be selected. To adjust the step size, The attribution probability is obtained above.

[0084] If the failure is attributed to contention, it indicates that the problem primarily stems from channel contention during multi-machine transmission. This is mainly due to concurrent transmission by multiple machines, leading to an increased probability of collisions. Therefore, based on the estimated collision probability, the initial channel access probability is adjusted. The principle of this probability adjustment algorithm is similar to the compression calculation described above. Channel collapse is mitigated by reducing access frequency requests.

[0085] It is important to note that the revised parameter set is immediately sent to the communication control unit of each UAV, overriding the original baseline transmission parameters, to ensure that the protocol stack has been switched in the next transmission cycle.

[0086] While adjusting the communication parameters as described above, the residual vector (difference) between the measured environmental interference prediction and the model's expected interference is used as a feedback signal. The transmission prediction model uses this feedback to update its internal link gain and collision probability prediction weight coefficients for that trajectory point (spatial coordinates). Regarding the correction of the prediction weights, taking link deviation as an example, a gradient correction method is used:

[0087] ; in, The formula represents the corresponding prediction weights in the original prediction model. It can be understood as follows: the greater the contribution of a factor to the bias, the greater its weight; conversely, the weaker the prediction, the weaker its weight. Furthermore, in some implementations, if gradient correction is not used, the corresponding correction can be made based on the proportion of each factor's contribution to the bias (calculated from residual decomposition).

[0088] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. The scope of protection of the present invention is defined by the claims. Similarly, any equivalent structural changes made based on the description and drawings of the present invention should also be included within the scope of protection of the present invention.

Claims

1. A method for processing operational data transmission in heterogeneous unmanned aerial vehicle (UAV) swarms, characterized in that, This includes performing the following steps: Obtain the task transmission constraints for collaborative operations of heterogeneous UAV clusters, including upper and lower limits of throughput and maximum allowable latency, and construct a transmission prediction model based on the inherent characteristics of each UAV's communication hardware and the preset task trajectory. The transmission prediction model transforms the preset mission trajectories of each UAV into spatiotemporally related spatial relative positions to predict the environmental interference at future time points. Then, it maps and generates a reference transmission parameter set corresponding to each time point. The environmental interference includes link gain and channel collision probability, and the reference transmission parameter set includes data compression ratio and channel access probability. During the operation, each UAV encapsulates and transmits operation data based on the corresponding reference transmission parameter set within the set sliding observation time window, and simultaneously monitors the operation indicators of the physical layer and media access control layer of the communication protocol stack; wherein, the signal-to-noise ratio and signal strength changes are extracted at the physical layer, and the number of collision backoffs and ACK timeouts are recorded at the media access control layer. Within the observation time window, calculate the rate of change of each of the aforementioned operational indicators, use the trend extrapolation method to predict the link gain and channel collision probability at several future time points, and construct the environmental interference prediction quantity. The predicted environmental interference quantity is compared with the environmental interference quantity at the corresponding future time point in the same dimension. When the deviation exceeds the preset confidence interval, the failure probability caused by channel quality degradation and channel contention intensification is estimated and the failure type is determined based on the correlation between the operating indicators of the physical layer and the medium access control layer. Based on the failure type, the current data compression ratio and channel access probability are corrected and covered respectively; and the deviation is used as a feedback quantity to correct the prediction weights of link gain and channel collision probability in the transmission prediction model.

2. The method for processing operational data transmission in heterogeneous UAV swarms according to claim 1, characterized in that, When constructing the transmission prediction model for the heterogeneous drone cluster, the following steps are performed: The preset mission trajectories of each UAV are imported onto a unified time axis, and time discretization processing is performed to calculate the time-varying relative distance between any two UAVs. The inherent characteristics of the communication hardware of each UAV are obtained, including transmit power, antenna gain and receive sensitivity. The time-varying relative distance between any two UAVs is converted into the link transmission loss characteristics at the corresponding time point, and a predictive relationship of link gain changing with time is established. Identify the potential interference sets of each UAV at different time points, and combine them with the preset channel access initial strategy when performing the task to construct a channel conflict probability prediction mechanism for the communication link competition relationship of each UAV in a heterogeneous UAV cluster. The prediction relationship of link gain and the channel conflict probability prediction mechanism are integrated to form a transmission prediction model that predicts changes in the communication environment at future points in time.

3. The method for processing operational data transmission in heterogeneous UAV swarms according to claim 2, characterized in that, When the transmission prediction model performs prediction, the preset mission trajectory of each UAV is mapped to a spatial position sequence at continuous time points, and the relative distance change relationship between each UAV is calculated based on the spatial position sequence. Based on the relative distance change relationship, and combined with the transmission capability and reception conditions in the communication hardware characteristics, the link propagation status at each time point is predicted, and the corresponding link gain change trend is obtained. Simultaneously, based on the relative distance changes of each UAV at each time point, the potential UAV set that generates channel competition within the link coverage area is identified, and the channel conflict probability at each time slice is predicted. The link gain generated at each time point and the channel collision probability constitute the environmental interference quantity.

4. The method for processing operational data transmission in heterogeneous UAV swarms according to claim 1, characterized in that, When generating the reference transmission parameter set using the predicted interference mapping, at each predicted time point, the data compression ratio that satisfies the minimum transmission requirement is calculated using a utility function based on the lower limit of throughput in the task transmission constraints and the predicted link gain; and the channel access probability is calculated based on the maximum allowable delay and the predicted channel collision probability.

5. The method for processing operational data transmission in heterogeneous UAV swarms according to claim 1, characterized in that, In a heterogeneous drone swarm, during collaborative operations, each drone, within a set sliding observation time window, encapsulates and transmits its operational data to the receiving end based on the corresponding reference transmission parameter set. Simultaneously, during data transmission, it collects operational metrics from each layer of the communication protocol stack. At the physical layer, the signal-to-noise ratio and received signal strength of each data packet are obtained by measuring the received job data in real time. In the media access control layer, the number of collision backoffs triggered in each job data transmission and the ACK timeout frequency caused by not receiving acknowledgment data from the receiver are recorded. Within each observation time window, first-order difference operations are performed on the signal-to-noise ratio, signal strength, number of collision backoffs, and ACK timeout frequency to obtain the rate of change of each operating index within the observation time window.

6. The method for processing operational data transmission in heterogeneous UAV swarms according to claim 5, characterized in that, When using the trend extrapolation method to predict and construct the environmental interference prediction quantity, for the rate of change of the signal-to-noise ratio and signal strength of the physical layer, the time-weighted trend extension method is used to predict the direction and magnitude of change of these values ​​at several future time points, and the prediction results are mapped to the link gain change trend at the corresponding time points. Based on the rate of change of the number of collision backoffs and the ACK timeout frequency at the medium access control layer, the channel contention level at several future time points is extrapolated based on their growth rate, and the channel collision probability change trend at the corresponding time points is obtained.

7. The method for processing operational data transmission in heterogeneous UAV swarms according to claim 1, characterized in that, After obtaining the predicted environmental interference based on trend extrapolation, it is time-aligned with the environmental interference output by the transmission prediction model at the corresponding future time point, and the same-dimensional deviation calculation is performed. The deviation measures of each dimension are compared with the preset confidence intervals. When any deviation continues to exceed the risk range, or shows a monotonically increasing trend over multiple consecutive time points, the current environmental prediction results are determined to be mismatched, triggering failure attribution calculation.

8. A method for processing operational data transmission in heterogeneous UAV swarms according to claim 7, characterized in that, After triggering the attribution calculation, the changing trends of signal-to-noise ratio and signal strength within the observation time window are aligned with the link gain prediction deviation. The degree of matching between the two in terms of consistency in the direction of change and the magnitude of change is calculated to obtain the first attribution metric characterizing the degree of impact of channel quality degradation. Align the changing trends of the number of conflict backoffs and the ACK timeout frequency with the channel conflict probability prediction deviation, calculate the degree of matching between the two in terms of the consistency of the direction of change and the magnitude of change, and obtain a second attribution metric that characterizes the degree of impact of intensified channel competition.

9. A method for processing operational data transmission in heterogeneous UAV swarms according to claim 8, characterized in that, The first and second attribution metrics are normalized to obtain the failure attribution probabilities corresponding to channel quality degradation and channel contention aggravation, respectively. Based on the relative magnitude of each attribution probability, the dominant failure type of the current environment prediction mismatch is determined.

Images