Low-altitude flight communication network resource allocation method based on large model algorithm

Abstract

The invention relates to the technical field of resource allocation, and particularly discloses a low-altitude flight communication network resource allocation method based on a large model algorithm, which comprises the following steps: acquiring an unmanned aerial vehicle task instruction, carrying out LLM model-based intention analysis and demand prediction on the unmanned aerial vehicle task instruction to capture communication network resource demand characteristics, and meanwhile, carrying out LLM model-based demand prediction on the communication network resource demand characteristics; according to the method, network state data and environment data are further combined, network state features, environment state features and resource demand features are fused to construct comprehensive state expression of an unmanned aerial vehicle task execution scene, intelligent resource allocation decision making is carried out on the basis, and a network resource allocation instruction is generated and executed. Through the mode, intelligent allocation of low-altitude flight communication network resources is realized, the resource utilization efficiency and the task execution capability are improved, the accuracy and the real-time performance of resource allocation can be effectively ensured, and the intelligent level of resource allocation is further improved.

Description

Technical Field

[0001] This application relates to the technical field of resource allocation, and more specifically, to a method for allocating low-altitude flight communication network resources based on large model algorithms. Background Art

[0002] In the context of the rapid development of low-altitude flight communication networks, the allocation of UAV mission resources has become an important research direction in the communication field. The efficient allocation of communication resources is crucial for ensuring the real-time, reliable, and flexible execution of UAV missions. Traditional communication resource allocation methods mainly rely on the input of fixed quality of service (QoS) parameters, such as bandwidth, delay, and reliability requirements. These methods have obvious limitations when faced with complex and changing UAV mission scenarios, and their ability to respond to dynamic changes and environmental conditions (such as weather, interference, network congestion) during the mission execution phase is insufficient, and they cannot adjust resource allocation strategies in real time to meet mission requirements.

[0003] Specifically, traditional resource allocation methods mainly rely on the QoS parameter hard-coding mechanism, and achieve network resource scheduling by manually configuring fixed thresholds such as bandwidth, delay, and reliability. Such methods are limited by the preset fixed thresholds and are difficult to adapt to the dynamically changing UAV mission requirements, resulting in low resource allocation efficiency and inability to meet the diverse requirements in complex mission scenarios. That is to say, due to the highly time-varying nature of low-altitude UAV mission requirements (such as sudden high-definition video transmission back, emergency obstacle avoidance instructions), when faced with dynamic environmental changes, the existing technology lacks the adaptive ability for dynamic resource reallocation, which may lead to "oversupply" or "undersupply" of resources. For example, the sudden increase in high-bandwidth requirements during a reconnaissance mission when a suspicious target is discovered may cause the interruption of critical data transmission due to network congestion.

[0004] Therefore, an optimized method for allocating low-altitude flight communication network resources based on large model algorithms is expected. Summary of the Invention

[0005] To solve the above technical problems, the present application is proposed. An embodiment of the present application provides a method for allocating low-altitude flight communication network resources based on a large model algorithm, which obtains a drone task instruction, and performs intention parsing and demand prediction based on the LLM model on it to capture the communication network resource demand characteristics. At the same time, further combining network status data and environmental data, by fusing network status characteristics, environmental status characteristics and resource demand characteristics, a comprehensive state expression of the drone task execution scenario is constructed, and then an intelligent resource allocation decision is made on this basis, generating and executing a network resource allocation instruction. In this way, the intelligent allocation of low-altitude flight communication network resources is realized, the resource utilization efficiency and task execution ability are improved, the accuracy and real-time performance of resource allocation can be effectively ensured, and the intelligent level of resource allocation is further improved.

[0006] According to one aspect of the present application, there is provided a method for allocating low-altitude flight communication network resources based on a large model algorithm, which includes: Obtain a drone task instruction; Perform intention parsing and demand prediction based on the LLM model on the drone task instruction to obtain a structured encoding vector of future resource demands; Obtain network status data and environmental data; Use a network status embedding encoding matrix and an environmental data embedding encoding matrix to perform structured processing on the network status data and the environmental data to obtain a network status structured encoding vector and an environmental status structured encoding vector; Fuse the structured encoding vector of future resource demands, the network status structured encoding vector and the environmental status structured encoding vector to obtain a comprehensive state feature vector; Input the comprehensive state feature vector into an intelligent resource allocation decision module to obtain a network resource allocation instruction, and the network resource allocation instruction is transmitted to a network controller for execution.

[0007] Compared with the prior art, the method for allocating low-altitude flight communication network resources based on a large model algorithm provided by the present application obtains a drone task instruction, and performs intention parsing and demand prediction based on the LLM model on it to capture the communication network resource demand characteristics. At the same time, further combining network status data and environmental data, by fusing network status characteristics, environmental status characteristics and resource demand characteristics, a comprehensive state expression of the drone task execution scenario is constructed, and then an intelligent resource allocation decision is made on this basis, generating and executing a network resource allocation instruction. In this way, the intelligent allocation of low-altitude flight communication network resources is realized, the resource utilization efficiency and task execution ability are improved, the accuracy and real-time performance of resource allocation can be effectively ensured, and the intelligent level of resource allocation is further improved. Description of the Drawings

[0008] The above and other objects, features, and advantages of the present application will become more apparent by describing the embodiments of the present application in more detail with reference to the accompanying drawings. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation to the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0009] Figure 1 It is a flowchart of a method for allocating low-altitude flight communication network resources based on a large model algorithm according to an embodiment of the present application.

[0010] Figure 2 It is a schematic diagram of data flow of a method for allocating low-altitude flight communication network resources based on a large model algorithm according to an embodiment of the present application.

[0011] Figure 3 It is a flowchart of sub-step S2 of a method for allocating low-altitude flight communication network resources based on a large model algorithm according to an embodiment of the present application.

[0012] Figure 4 It is a flowchart of sub-step S22 of a method for allocating low-altitude flight communication network resources based on a large model algorithm according to an embodiment of the present application.

[0013] Figure 5 It is a flowchart of sub-step S222 of a method for allocating low-altitude flight communication network resources based on a large model algorithm according to an embodiment of the present application. Detailed implementation manners

[0014] As shown in the present application and the claims, unless the context clearly indicates otherwise, words such as "a", "an", "one", and / or "the" are not specifically singular and may also include plural. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of the clearly identified steps and elements, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0015] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, however, any number of different modules can be used and run on the user terminal and / or the server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0016] In this application, flowcharts are used to illustrate the operations performed by the systems according to the embodiments of this application. It should be understood that the operations before or below do not necessarily need to be executed precisely in sequence. Instead, according to the needs, various steps can be processed in reverse order or simultaneously. At the same time, other operations can also be added to these processes, or one or several operations can be removed from these processes.

[0017] Next, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments of this application. It should be understood that this application is not limited by the exemplary embodiments described herein.

[0018] It is worth noting that in this application, all actions of obtaining data are carried out on the premise of complying with the corresponding data protection regulations and policies of the country where it is located and obtaining authorization from the corresponding device owner.

[0019] In response to the technical problems described in the above background art, this application proposes a method for resource utilization of mixed solid waste. By obtaining drone task instructions and performing intent parsing and demand prediction based on the LLM model to capture the communication network resource demand characteristics. At the same time, further combining network status data and environmental data, by fusing network status characteristics, environmental status characteristics, and resource demand characteristics, to construct a comprehensive state expression of the drone task execution scenario. Then, based on this, an intelligent resource allocation decision is made, generating and executing network resource allocation instructions. In this way, the intelligent allocation of low-altitude flight communication network resources is achieved, improving the resource utilization efficiency and task execution ability, effectively ensuring the accuracy and real-time nature of resource allocation, and further improving the intelligent level of resource allocation.

[0020] Figure 1 FIG. is a flowchart of a method for allocating low-altitude flight communication network resources based on a large model algorithm according to an embodiment of this application. Figure 2 FIG. is a schematic diagram of data flow of a method for allocating low-altitude flight communication network resources based on a large model algorithm according to an embodiment of this application. As Figure 1 and Figure 2As shown, the method for allocating low-altitude flight communication network resources based on large model algorithms includes the steps of: S1, obtaining drone task instructions; S2, performing intent parsing and demand prediction on the drone task instructions based on the LLM model to obtain a structured encoding vector of future resource requirements; S3, obtaining network status data and environmental data; S4, using a network status embedding encoding matrix and an environmental data embedding encoding matrix to perform structured processing on the network status data and the environmental data to obtain a network status structured encoding vector and an environmental status structured encoding vector; S5, fusing the structured encoding vector of future resource requirements, the network status structured encoding vector, and the environmental status structured encoding vector to obtain a comprehensive status feature vector; S6, inputting the comprehensive status feature vector into an intelligent resource allocation decision module to obtain a network resource allocation instruction, and transmitting the network resource allocation instruction to a network controller for execution.

[0021] In the above method for allocating low-altitude flight communication network resources based on large model algorithms, in step S1, drone task instructions are obtained. It should be understood that since the drone task scenario is highly dynamic, for example, sudden high-definition video backhaul or emergency obstacle avoidance instructions will cause drastic fluctuations in bandwidth requirements, and relying solely on preset QoS parameters cannot respond to such changes in a timely manner. Therefore, in this application, the original input of drone task instructions (such as task type, target parameters, operation sequence, etc.) is captured in real time and converted into an initial demand signal that can be understood by the system, providing basic data support for subsequent resource demand prediction and dynamic adjustment, enabling the system to perceive the dynamic changes in the task execution stage in real time (such as sudden high bandwidth requirements or priority transitions of emergency obstacle avoidance instructions), and avoiding resource allocation lags or failures caused by relying on fixed thresholds. For example, when a drone triggers high-definition video backhaul due to detecting a suspicious target, the immediate acquisition of task instructions can trigger the priority reconstruction of the resource allocation process, thereby reducing the risk of critical data transmission interruption, while reducing the redundant resource occupancy of non-emergency tasks and improving the overall network efficiency and task reliability.

[0022] In the process of specific implementation, in order to obtain accurate task instructions, an efficient mechanism needs to be established to capture this information in real time. This mechanism first needs to design a reliable communication link to ensure stable data exchange between the drone and the ground control station. This usually involves using advanced wireless communication technologies such as 5G networks to maintain high-quality data transmission even in complex and changing environments. In this way, regardless of the geographical location of the drone or the weather conditions it faces, the task instructions sent by it can be received in a timely and accurate manner. In addition, considering the highly dynamic nature of the drone task scenario, such as the need for high-definition video transmission in case of emergencies or emergency obstacle avoidance instructions, this communication link also needs to have the ability to respond quickly, be able to adapt to environmental changes in a short time, and quickly adjust its own state to meet new task requirements.

[0023] Furthermore, in the process of specific implementation, the drone may receive multiple different types of task instructions simultaneously. For example, on the one hand, it is required to continuously monitor a certain area and regularly upload monitoring data, and on the other hand, it may suddenly encounter an obstacle and needs to take immediate obstacle avoidance measures. In this case, the instruction acquisition mechanism not only needs to be able to distinguish different types of instructions, but also needs to sort these instructions according to the preset priority rules to ensure that high-priority tasks can be processed first. For example, when the drone detects an obstacle ahead, the relevant obstacle avoidance instructions should be immediately recognized and trigger the corresponding action sequence, such as adjusting the flight path or changing the flight speed. At the same time, for those non-urgent task instructions, they can be continued to be executed or temporarily shelved on the premise of ensuring safety.

[0024] In addition to the task instructions directly obtained from the drone, the system's database may also store a large number of data records on past similar tasks. These historical data are of great significance for understanding the context of the current task instructions. For example, by studying the instructions generated during past executions of similar tasks, certain task patterns or rules can be discovered. Using this information can help the system more accurately predict the possible future task types and their specific requirements. For example, if the analysis results show that emergency obstacle avoidance events frequently occur in a certain area during a specific time period, then this area can be considered to be bypassed when planning future flight routes, thereby reducing the probability of such events. Such an approach not only helps to improve the safety and efficiency of task execution, but also provides an important reference basis for subsequent resource allocation decisions.

[0025] It is worth noting that in the process of implementing drone mission instruction acquisition, special attention should be paid to how to effectively integrate information from different sources. In addition to the historical data mentioned above, it also includes environmental data collected in real time (such as weather forecasts, air traffic conditions) and network status data (such as current available bandwidth, network latency). All of this information together constitutes the complete background information required for drones to perform tasks. Therefore, building a platform that can automatically collect, organize and analyze this diverse information can not only help the system better understand the specific needs of each independent task, but also dig out some common trends or characteristics from it, thereby guiding future mission planning and resource allocation strategies. For example, if the system detects that the current network congestion is serious and may affect the upcoming high-definition video backhaul mission, then corresponding adjustments can be made in advance, such as optimizing route selection or reserving more bandwidth resources to ensure the smooth transmission of critical data.

[0026] In the above-mentioned low-altitude flight communication network resource allocation method based on the large model algorithm, the step S2 performs intent analysis and demand prediction on the drone mission instructions based on the LLM model to obtain a structured coding vector of future resource requirements. Specifically, in the present application, the LLM model (Large Language Model) is used to perform intent analysis and demand prediction on drone mission instructions, which can deeply understand the mission semantics and mine the implicit dynamic resource demand patterns, convert the mission objectives into quantifiable future resource demand structured coding vectors, and combine historical data with real-time context to predict the resource demand inflection points that may occur during the execution of the mission, thereby achieving real-time matching of network resource supply and dynamic demand in complex and changeable low-altitude scenarios, effectively reducing the risk of transmission interruption caused by resource estimation deviations, while improving network resource utilization and mission execution success rate. Among them, Figure 3 FIG. 1 is a flowchart of sub-step S2 of the low-altitude flight communication network resource allocation method based on the large model algorithm according to an embodiment of the present application. Figure 3 As shown, the step S2 includes the steps of: S21, performing word segmentation on the UAV task instruction and then inputting it into a text encoder to obtain a sequence distribution of UAV task word granularity semantic embedding coding vectors; S22, performing UAV task word granularity context semantic association coding on the sequence distribution of the UAV task word granularity semantic embedding coding vectors to obtain a UAV task semantic coding vector; S23, inputting the UAV task semantic coding vector into a demand predictor based on an LLM model to obtain the future resource demand structured coding vector.

[0027] Specifically, in step S21, after performing word segmentation on the UAV task instruction, it is input into a text encoder to obtain a sequence distribution of UAV task word-level semantic embedding coding vectors. Specifically, since UAV task instructions often describe complex requirements in natural language form, their semantic connotations have characteristics of multi-granularity, strong correlation, and dynamic context dependence. Therefore, in this application, the continuous UAV task instruction is first disassembled into word-level units with independent semantics through word segmentation and input into the text encoder to generate a sequence distribution of word-level UAV task word-level semantic embedding coding vectors, which can transform unstructured text instructions into computable mathematical representations while retaining the grammatical relationships and semantic hierarchies between the words in the instructions, thus providing a fine-grained and semantically consistent basic input for the fusion of network status and environmental data. In a specific example of this application, the text encoder is a word embedding coding model based on the BERT model. It should be understood that the BERT model is a pre-trained language representation model that can capture complex semantic relationships and context dependencies between words through unsupervised learning on a large-scale corpus and generate high-quality word-level semantic embedding codes. On this basis, this application uses the BERT model to deeply understand the semantics of each word unit in the UAV task instruction, transforms the key information in the task description (such as task type, target parameters, operation sequence, etc.) from unstructured text into vector representations in a high-dimensional semantic space, and obtains a sequence distribution of UAV task word-level semantic embedding coding vectors for subsequent task intention understanding and requirement prediction.

[0028] Specifically, in step S22, context semantic association coding for the UAV task word-level semantic embedding coding vector sequence distribution is performed to obtain a UAV task semantic coding vector. It should be understood that due to the dynamic temporal sequence and multi-dimensional correlation of the semantics of UAV task instructions, it is difficult for the sequence distribution of single-word-level UAV task word-level semantic embedding coding vectors to capture the spatio-temporal evolution relationship between key semantic units in the task scenario. Therefore, this application further performs context semantic association coding based on spatio-temporal collaborative constraints on the sequence distribution of UAV task word-level semantic embedding coding vectors to enhance the accuracy of task intention parsing, dynamically analyze the semantic dependence between words in the task instruction, and map discrete word-level semantics to continuous UAV task semantic coding vectors, providing fine-grained semantic input with spatio-temporal reasoning ability for the resource allocation module. Among them, Figure 4 is a flowchart of sub-step S22 of the low-altitude flight communication network resource allocation method based on the large model algorithm according to the embodiment of this application. As Figure 4As shown, step S22 includes steps: S221, inputting the sequence distribution of the drone task word granularity semantic embedding coding vectors into a sequence encoder based on a recurrent neural network to obtain the sequence distribution of the drone task semantic transmission coding vectors; S222, calculating the message passing spatio-temporal collaborative constraint factors of each drone task semantic transmission coding vector in the sequence distribution of the drone task semantic transmission coding vectors; S223, dynamically adjusting and aggregating each drone task semantic transmission coding vector based on the message passing spatio-temporal collaborative constraint factors to obtain the drone task semantic coding vector.

[0029] More specifically, step S221 is expressed by the formula: Where represents the sequence distribution of the drone task word granularity semantic embedding coding vectors, , , and respectively represent the 1st, 2nd, th, and th drone task word granularity semantic embedding coding vectors in the sequence distribution of the drone task word granularity semantic embedding coding vectors, , , and respectively represent the 1st, 2nd, th, and th drone task semantic transmission coding vectors in the sequence distribution of the drone task semantic transmission coding vectors, represents the recurrent neural network.

[0030] That is, through the RNN model, the sequence distribution of the drone task word granularity semantic embedding coding vectors is passed layer by layer and feature accumulation is performed to extract the deep implicit patterns of the task requirements in the temporal context, generating a comprehensive feature expression with temporal context awareness ability, that is, the sequence distribution of the drone task semantic transmission coding vectors. Through this dynamic feature extraction method, it provides an interpretable and traceable context information basis for subsequent intelligent resource allocation decisions.

[0031] Figure 5 is the flowchart of sub-step S222 of the low-altitude flight communication network resource allocation method based on the large model algorithm according to the embodiment of the present application. As Figure 5As shown, step S222 includes steps: S2221, calculating the temporal confidence constraint factor of each UAV mission semantic transfer coding vector in the sequence distribution of the UAV mission semantic transfer coding vectors; S2222, calculating the spatial confidence constraint factor of each UAV mission semantic transfer coding vector in the sequence distribution of the UAV mission semantic transfer coding vectors; S2223, constructing the message passing spatio-temporal collaborative constraint factor of each UAV mission semantic transfer coding vector based on the spatial confidence constraint factor and the temporal confidence constraint factor of each UAV mission semantic transfer coding vector.

[0032] In a specific example of the present application, step S2221 is expressed by the formula: Where, and respectively represent different weight parameters, and respectively represent different weight matrices, represents the hyperbolic tangent function, represents matrix multiplication, represents the UAV mission semantic temporal context transfer feature importance scoring conversion vector, represents the UAV mission semantic transfer temporal context confidence factor of, represents the softmax normalization function, represents the temporal confidence constraint factor of.

[0033] That is, by performing weighted fusion on the original UAV mission word granularity semantic embedding coding vector and the UAV mission semantic transfer coding vector after context transfer coding, and utilizing the non-linear mapping ability of the neural network to capture the correlation features between the two, the confidence weight of each UAV mission word granularity semantic embedding coding vector in the context is dynamically evaluated. Specifically, based on the generated temporal confidence constraint factor, noise suppression and information enhancement can be achieved, that is, by reducing the word granularity weight with low confidence, reducing its interference with global resource allocation, and at the same time amplifying the key features of the word granularity with high confidence, providing high-quality feature input for subsequent resource demand prediction, thereby significantly optimizing the information screening efficiency and decision reliability.

[0034] In a specific example of the present application, step S2222 is expressed by the formula: Where, represents calculating the spatial confidence score of, Denotes the exponential function operation with base e, Denotes The spatial confidence constraint factor of

[0035] That is, through structured information embedding, the topological awareness ability of resource allocation is enhanced. Specifically, the spatial confidence constraint factor guides the model to prioritize the resource requirements of high-structural-value nodes in subsequent decisions by quantifying the importance of node spatial roles, while suppressing redundant requests from low-value nodes. In this structured way, the dynamic adaptation ability of the model is optimized, enabling the resource allocation strategy to synergize with the node spatial functions and avoiding local congestion or global inefficiency caused by ignoring structural heterogeneity. The spatial confidence constraint factor generated based on this significantly improves the topological rationality and propagation efficiency of network resource allocation. In addition, through the priority allocation of high-confidence nodes, the efficient transmission of information on the critical path is promoted, while the noise propagation of non-critical nodes is suppressed, enabling subsequent resource allocation decisions to generate more accurate network scheduling instructions based on reliable structural features, ultimately achieving efficient and robust resource management of the low-altitude flight communication network under complex dynamic topologies.

[0036] In a specific example of this application, the step S2223 includes: weighting and fusing the spatial confidence constraint factor and the temporal confidence constraint factor of the UAV task semantic transfer coding vector, and then inputting them into the sigmoid function for normalization processing to obtain the message passing spatio-temporal collaborative constraint factor of the UAV task semantic transfer coding vector, which is expressed by the formula: Wherein, And Respectively represent different constraint weight parameters, Represents the sigmoid function, Denotes The message passing spatio-temporal collaborative constraint factor of

[0037] That is, through spatio-temporal collaborative weight fusion, the confidence weights of the UAV task semantic transfer coding vector in spatial position and temporal context can be effectively balanced, thus more comprehensively evaluating the comprehensive importance of each UAV task semantic transfer coding vector in a complex dynamic environment. In addition, through the normalization processing of the sigmoid function, the message passing spatio-temporal collaborative constraint factor of the fused UAV task semantic transfer coding vector is further mapped to the [0,1] interval to achieve numerical stability and provide a standardized weight input for subsequent message passing coding.

[0038] More specifically, in a preferred example of the present application, the step S223 includes: First, perform space-time coupling compensation optimization based on a multi-dimensional constraint manifold on the message-passing space-time collaborative constraint factor of each UAV mission semantic transmission encoding vector to obtain the optimized message-passing space-time collaborative constraint factor of each UAV mission semantic transmission encoding vector, which is expressed by the formula: Among them, represents the space-time confidence balance factor between the temporal confidence constraint factor and the spatial confidence constraint factor, represents the space-time constraint curvature compensation factor between the temporal confidence constraint factor and the spatial confidence constraint factor represents the sine function, represents the message-passing space-time constraint coupling compensation factor of represents the corresponding optimized message-passing space-time collaborative constraint factor.

[0039] That is, through geometric constraint optimization, eliminate the distortion effect of space-time dimension coupling, and enhance the collaborative efficiency of the temporal confidence constraint factor and the spatial confidence constraint factor in the fusion space. Specifically, the multi-dimensional constraint manifold compensation mechanism forces the space-time coupling attraction to converge towards the flat space form by dynamically adjusting the geometric properties of the temporal confidence constraint factor and the spatial confidence constraint factor. Through this manifold optimization method, the model can adaptively balance the space-time curvature difference, so that the optimized message-passing space-time collaborative constraint factor after fusion maintains a stable attractor structure in the flat space, avoiding misjudgment of resource allocation caused by geometric distortion. In this way, through multi-dimensional constraint manifold compensation optimization, the model realizes geometric distortion suppression, and its compensation mechanism can correct the curvature deviation of the message-passing space-time collaborative constraint factor in real time, avoiding the invalidation of the fusion weight caused by geometric alienation. Secondly, by forcing the space-time coupling attraction to converge towards the flat space, strengthen the complementarity of the temporal confidence constraint factor and the spatial confidence constraint factor, and enhance the logical consistency of the resource allocation strategy. Based on the generated optimized message-passing space-time collaborative constraint factor, it has stronger fault tolerance for sudden changes in the network state, providing a weight basis with geometric stability for subsequent message-passing encoding.

[0040] Then, based on the optimized message-passing spatio-temporal collaborative constraint factors of the respective UAV mission semantic transmission encoding vectors, perform message-passing structure modulation on the respective UAV mission semantic transmission encoding vectors to obtain the sequence distribution of the UAV mission semantic transmission structured modulation encoding vectors; calculate the position-wise sum of the sequence distribution of the UAV mission semantic transmission structured modulation encoding vectors to obtain the UAV mission semantic encoding vector, which is expressed by the formula: where represents the UAV mission semantic encoding vector.

[0041] That is, through the optimization of structured information propagation and aggregation, a comprehensive semantic representation with spatio-temporal consistency is generated. Specifically, message-passing structure modulation realizes information screening driven by spatio-temporal constraints through a weighting strategy, ensuring the reliability of resource allocation decisions. In this way of position-wise summation, the sequence distribution of the UAV mission semantic transmission structured modulation encoding vectors is further compressed into the UAV mission semantic encoding vector, capturing the global context semantic features of the UAV mission, and through adaptive weight constraints, focusing on key task semantics, which helps to improve the accuracy of task understanding and network demand prediction.

[0042] Specifically, in step S23, input the UAV mission semantic encoding vector into a demand predictor based on the LLM model to obtain the future resource demand structured encoding vector. Specifically, the LLM model is a deep learning-based natural language processing (NLP) model. It uses self-supervised learning for pre-training on large-scale text data, learns the statistical laws and semantic information of the language, and performs supervised fine-tuning through a small amount of labeled data to adapt to specific tasks. In this application, through the context reasoning of the UAV mission semantic encoding vector by the demand predictor based on the LLM model, the implicit intent semantics in the task instructions can be deeply analyzed, and relying on the powerful context reasoning and prediction ability of the LLM model, the future resource demand structured encoding vector is dynamically generated, converting the discrete task requirements into continuous UAV mission semantic encoding vectors, thereby providing a forward-looking data basis for subsequent resource allocation decisions, enhancing the matching accuracy of resource allocation and task objectives, and ultimately improving the real-time response ability and resource utilization efficiency of the low-altitude network in complex scenarios.

[0043] In the above method for allocating low-altitude flight communication network resources based on large model algorithms, in step S3, network status data and environmental data are obtained. It should be understood that since the low-altitude flight communication network faces dynamically changing task requirements and complex environmental interference, and traditional static resource allocation methods cannot perceive network load, channel quality, and external environmental changes in real time, resulting in a disconnection between resource supply and task requirements. Therefore, in this application, network status data (such as bandwidth occupancy rate, node load, channel quality) and environmental data (such as meteorological information, geographical obstacles, interference source distribution) are obtained in real time to dynamically evaluate the current network carrying capacity and potential risk factors. In this way, real-time environmental disturbances and network performance indicators can be incorporated into the resource allocation decision model, enabling the model to predict the changing trend of resource requirements based on global scenario characteristics and achieve precise adaptation of resource supply to dynamic task requirements.

[0044] Specifically, in terms of network status data, real-time monitoring of the bandwidth occupancy rate has become an important means of understanding the current network load situation. By deploying sensors on key nodes of the network, the bandwidth usage of each node can be continuously collected and this information can be uploaded to the central processing unit for analysis. This approach not only helps to identify bottleneck positions in the network but also provides a basis for subsequent adjustment of the resource allocation strategy. For example, when it is found that the bandwidth occupancy rate in a specific area has increased abnormally, it indicates that there may be a large amount of data transmission demand in this area. At this time, it may be necessary to reallocate existing resources to avoid task interruption or delay caused by network congestion.

[0045] At the same time, node load is also an important parameter for measuring the health status of the network. By monitoring key indicators such as the CPU usage rate and memory occupancy of each network node, potential overload risks can be detected in a timely manner. Once it is detected that the load of a certain node is approaching its maximum bearing capacity, the system can take corresponding measures, such as dispersing some tasks to other nodes with lower load, or optimizing the routing selection to relieve the pressure on the high-load node. In addition, machine learning algorithms can be used to predict the changing trend of node load in the next period of time, so as to make preparations for resource scheduling in advance and ensure that the network always maintains an efficient operation state.

[0046] As one of the key factors affecting data transmission efficiency, channel quality also needs to be fully concerned. Using advanced signal processing technologies, parameters such as signal strength and bit error rate of each communication link can be monitored in real time. Based on these data, the system can quickly determine which channels are in good condition and which may have quality problems. For the latter, the communication effect can be improved by switching to a backup channel or increasing the signal transmission power. At the same time, by analyzing historical data, the possible fluctuations in channel quality at different time periods can also be predicted, so as to formulate a more scientific and reasonable resource allocation plan.

[0047] In addition to network status data, the acquisition of environmental data is also an indispensable part. Considering the complex natural conditions and human interference factors in the low-altitude flight environment, a complete environmental monitoring system must be established. First of all, the real-time update of meteorological information is of great significance to ensure the safe flight of drones. Using the latest weather forecast data provided by the meteorological station, combined with satellite remote sensing images and feedback from other meteorological observation equipment, a detailed meteorological map can be constructed to show the wind speed, rainfall, temperature changes and other conditions in the current area. Based on this information, it can not only guide drones to choose the best flight route and avoid bad weather areas, but also help predict the network performance degradation that may be caused by climate change, and take countermeasures in advance.

[0048] The presence of geographical obstacles will also have a significant impact on the flight path planning and communication links of drones. To this end, it is necessary to rely on high-precision map services and advanced sensing technologies such as LiDAR to accurately map the terrain and geomorphic features. In this way, when planning the flight route of drones, obstacles such as mountains and buildings can be effectively avoided to reduce the risk of flight accidents. At the same time, geographical obstacles may also block radio signals and cause local communication blind spots. Therefore, understanding the distribution of obstacles can also help optimize the base station layout and ensure maximum network coverage.

[0049] The distribution of interference sources is another environmental factor that needs to be focused on. With the increasing number of wireless communication devices, electromagnetic interference is becoming increasingly serious, which poses a huge challenge to low-altitude flight communication networks. In order to effectively identify and locate interference sources, spectrum monitoring equipment can be used to scan the entire working frequency band and record the location and strength of all abnormal signals. Based on this data, not only can the location of the specific interference source be tracked, but also its specific impact on network performance can be further analyzed. On this basis, targeted measures are taken, such as adjusting frequency settings or adding anti-interference coding, to reduce the negative impact of interference.

[0050] In the above method for allocating low-altitude flight communication network resources based on large model algorithms, in step S4, the network state embedding coding matrix and the environmental data embedding coding matrix are used to structurally process the network state data and the environmental data to obtain a network state structured coding vector and an environmental state structured coding vector. It should be understood that through the network state embedding coding matrix and the environmental data embedding coding matrix, the original network state data and environmental data are mapped to a high-dimensional semantic space to generate a network state structured coding vector and an environmental state structured coding vector with dynamic perception ability. In this way, discrete and heterogeneous network performance indicators and environmental perturbation parameters can be transformed into continuous vector representations, so that the model can capture the implicit spatio-temporal correlation and task coupling relationship between the data, thereby dynamically identifying the spatio-temporal evolution law of network bottlenecks and the potential conflicts of task requirements, and ultimately supporting the intelligent decision-making module to achieve joint optimization and adaptive allocation of multi-dimensional resource constraints in complex scenarios. In the embodiments of the present application, the network state embedding coding matrix and the environmental data embedding coding matrix are trained by a pre-trained word vector model such as Word2vec.

[0051] In the above method for allocating low-altitude flight communication network resources based on large model algorithms, in step S5, the future resource demand structured coding vector, the network state structured coding vector, and the environmental state structured coding vector are fused to obtain a comprehensive state feature vector. In a specific example of the present application, step S5 includes: concatenating and fusing the future resource demand structured coding vector, the network state structured coding vector, and the environmental state structured coding vector to obtain the comprehensive state feature vector. It should be understood that by concatenating and fusing the future resource demand structured coding vector, the network state structured coding vector, and the environmental state structured coding vector, a collaborative representation framework for multi-dimensional information can be constructed. On the one hand, the concatenation mechanism can retain the time evolution sequence of task requirements, the topological features of network states, and the spatial heterogeneity of environmental conditions, avoiding feature dilution of single-dimensional information during the fusion process. On the other hand, high-order semantic associations are formed through cross-modal concatenation, enabling the comprehensive state feature vector to have context reasoning ability. Through this concatenation method, the model can dynamically perceive the chain reaction among the three, thereby generating a resource allocation strategy with strong robustness, achieving network load balancing, zero interruption of critical tasks, and Pareto optimization of resource utilization in complex scenarios.

[0052] In the above method for allocating low-altitude flight communication network resources based on large model algorithms, in step S6, the comprehensive state feature vector is input into the intelligent resource allocation decision module to obtain a network resource allocation instruction, and the network resource allocation instruction is transmitted to the network controller for execution. In a specific example of the present application, step S6 includes: inputting the comprehensive state feature vector into the intelligent resource allocation decision module based on a decoder to obtain a network resource allocation instruction. Specifically, through the autoregressive generation mechanism of the decoder architecture in the present application, the spatio-temporal coupling relationship between task requirements, network state, and environmental changes can be decoupled layer by layer. By using the semantic understanding ability of the pre-trained large model, multi-modal features are mapped into hierarchical resource allocation strategies (such as channel allocation, power control, bandwidth adjustment, etc.), thereby realizing refined management and efficient utilization of resources and constructing a continuous decision-making closed-loop in a dynamic environment. In this way, while capturing sudden task requirements in real time, the spatial distribution characteristics of network congestion nodes and meteorological interference areas are synchronously sensed, enabling the intelligent resource allocation decision module to generate a resilient allocation scheme with anti-interference ability based on the global resource topology, significantly improving the context awareness accuracy of resource scheduling, helping to eliminate decision-making blind spots caused by artificial feature engineering in traditional methods, and realizing dynamic optimization of network resource allocation while ensuring low-latency constraints.

[0053] After the generation of network resource allocation instructions, these instructions need to be transmitted to the network controller to perform corresponding resource adjustment and management operations. It should be noted that various unexpected situations or challenges may be encountered during the execution of the instructions. For example, due to the dynamic changes in the network environment, a previously planned path may suddenly become unavailable, or some key devices malfunction and cannot respond to the instructions normally. To address these issues, the system should have a certain fault tolerance mechanism. On the one hand, the overall system robustness can be enhanced by setting up alternative paths or redundant devices; on the other hand, when an abnormal situation is detected, the alarm mechanism is triggered in a timely manner and detailed information is reported to the administrator so that remedial measures can be taken as soon as possible. In addition, a self-healing function can be introduced to enable the system to automatically solve some common problems within a certain range, reducing the need for manual intervention.

[0054] In summary, the method for allocating low-altitude flight communication network resources based on large model algorithms according to the embodiments of the present application is elucidated. It obtains drone mission instructions, and performs intent parsing and demand prediction based on the LLM model on them to capture the communication network resource demand characteristics. At the same time, further combining network status data and environmental data, through fusing network status characteristics, environmental status characteristics, and resource demand characteristics, to construct a comprehensive state expression of the drone mission execution scenario, and then making an intelligent resource allocation decision based on this, generating and executing network resource allocation instructions. In this way, the intelligent allocation of low-altitude flight communication network resources is achieved, improving the resource utilization efficiency and mission execution ability, effectively ensuring the accuracy and real-time nature of resource allocation, and further enhancing the intelligent level of resource allocation.

[0055] The basic principles of the present invention have been described above in combination with specific embodiments. However, it should be noted that the advantages, benefits, effects, etc. mentioned in the present invention are only examples and not limitations. It cannot be considered that these advantages, benefits, effects, etc. are essential for each embodiment of the present invention. Additionally, the specific details of the above embodiments are only for illustrative and easy-to-understand purposes, not limitations. The above details do not limit the present invention to necessarily adopt the above specific details for implementation.

[0056] In the above embodiments, the descriptions of each embodiment have their own emphases. For parts not detailed or recorded in a certain embodiment, reference can be made to the relevant descriptions of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are only illustrative. For example, the unit division is only a logical function division, and there can be other division methods in actual implementation. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0057] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any perspective, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, it is intended to encompass all changes falling within the meaning and scope of the equivalent elements of the claims in the present invention. Any associated drawing marks in the claims should not be regarded as limiting the claimed rights.

[0058] In addition, it is obvious that the term "including" does not exclude other units or steps, and the singular does not exclude the plural. A plurality of units stated in the system claims can also be implemented by one unit through software or hardware.

[0059] Finally, it should be noted that the above description has been given for purposes of illustration and description. In addition, the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for allocating resources of a low-altitude flight communication network based on a large model algorithm, characterized in that, Including: Obtain the UAV mission instruction; Perform intent parsing and demand prediction on the UAV mission instruction based on the LLM model to obtain a structured encoding vector of future resource requirements; Obtain network status data and environmental data; Use the network status embedding encoding matrix and the environmental data embedding encoding matrix to perform structured processing on the network status data and the environmental data to obtain a network status structured encoding vector and an environmental status structured encoding vector; Fuse the structured encoding vector of future resource requirements, the network status structured encoding vector, and the environmental status structured encoding vector to obtain a comprehensive status feature vector; Input the comprehensive status feature vector into the intelligent resource allocation decision module to obtain a network resource allocation instruction, and the network resource allocation instruction is passed to the network controller for execution.

2. The method for allocating resources of a low-altitude flight communication network based on a large model algorithm according to claim 1, characterized in that, Performing intent parsing and demand prediction on the UAV mission instruction based on the LLM model to obtain a structured encoding vector of future resource requirements, including: Perform word segmentation on the UAV mission instruction and then input it into the text encoder to obtain a sequence distribution of UAV mission word-level semantic embedding encoding vectors; Perform UAV mission word-level context semantic association encoding on the sequence distribution of the UAV mission word-level semantic embedding encoding vectors to obtain a UAV mission semantic encoding vector; Input the UAV mission semantic encoding vector into the demand predictor based on the LLM model to obtain the structured encoding vector of future resource requirements.

3. The method for allocating resources of a low-altitude flight communication network based on a large model algorithm according to claim 2, characterized in that, The text encoder is a word embedding encoding model based on the BERT model.

4. The method for allocating resources of a low-altitude flight communication network based on a large model algorithm according to claim 3, characterized in that, Performing UAV mission word-level context semantic association encoding on the sequence distribution of the UAV mission word-level semantic embedding encoding vectors to obtain a UAV mission semantic encoding vector, including: Input the sequence distribution of the UAV mission word-level semantic embedding encoding vectors into a sequence encoder based on a recurrent neural network to obtain a sequence distribution of UAV mission semantic transfer encoding vectors; Calculate the message passing spatio-temporal coordination constraint factor of each UAV mission semantic transfer encoding vector in the sequence distribution of the UAV mission semantic transfer encoding vectors; Based on the message passing spatio-temporal coordination constraint factor, dynamically adjust and aggregate each UAV mission semantic transfer encoding vector to obtain the UAV mission semantic encoding vector.

5. The method for allocating resources of a low-altitude flight communication network based on a large model algorithm according to claim 4, characterized in that, Calculating the message passing spatio-temporal coordination constraint factor of each UAV mission semantic transfer encoding vector in the sequence distribution of the UAV mission semantic transfer encoding vectors, including: Calculate the temporal confidence constraint factor of each UAV mission semantic transfer encoding vector in the sequence distribution of the UAV mission semantic transfer encoding vectors; Calculate the spatial confidence constraint factor of each UAV mission semantic transfer encoding vector in the sequence distribution of the UAV mission semantic transfer encoding vectors; Based on the spatial confidence constraint factor and the temporal confidence constraint factor of each UAV mission semantic transfer encoding vector, construct the message passing spatio-temporal coordination constraint factor of each UAV mission semantic transfer encoding vector.

6. The method for allocating resources of a low-altitude flight communication network based on a large model algorithm according to claim 5, characterized in that, Construct the message-passing spatio-temporal collaborative constraint factor of each UAV mission semantic transmission coding vector based on the spatial confidence constraint factor and the temporal confidence constraint factor of each UAV mission semantic transmission coding vector, including: Input the weighted fusion of the spatial confidence constraint factor and the temporal confidence constraint factor of the UAV mission semantic transmission coding vector into the sigmoid function for normalization processing to obtain the message-passing spatio-temporal collaborative constraint factor of the UAV mission semantic transmission coding vector.

7. The method for allocating resources of a low-altitude flight communication network based on a large model algorithm according to claim 6, characterized in that, Based on the message-passing spatio-temporal collaborative constraint factor, dynamically adjust and aggregate each UAV mission semantic transmission coding vector to obtain the UAV mission semantic coding vector, including: Perform spatio-temporal coupling compensation optimization based on a multi-dimensional constraint manifold on the message-passing spatio-temporal collaborative constraint factor of each UAV mission semantic transmission coding vector to obtain the optimized message-passing spatio-temporal collaborative constraint factor of each UAV mission semantic transmission coding vector; Based on the optimized message-passing spatio-temporal collaborative constraint factor of each UAV mission semantic transmission coding vector, perform message-passing structure modulation on each UAV mission semantic transmission coding vector to obtain the sequence distribution of the UAV mission semantic transmission structure modulation coding vector; Calculate the position-wise sum of the sequence distribution of the UAV mission semantic transmission structure modulation coding vector to obtain the UAV mission semantic coding vector.

8. The method for allocating resources of a low-altitude flight communication network based on a large model algorithm according to claim 7, characterized in that,Fuse the future resource demand structured coding vector, the network state structured coding vector, and the environmental state structured coding vector to obtain the comprehensive state feature vector, including: Perform concatenation fusion on the future resource demand structured coding vector, the network state structured coding vector, and the environmental state structured coding vector to obtain the comprehensive state feature vector.

9. The method for allocating low-altitude flight communication network resources based on the large model algorithm according to claim 8, characterized in that, Input the comprehensive state feature vector into the intelligent resource allocation decision module to obtain a network resource allocation instruction, and the network resource allocation instruction is transmitted to the network controller for execution, including: Input the comprehensive state feature vector into the intelligent resource allocation decision module based on the decoder to obtain a network resource allocation instruction.

Images