Satellite communication resource allocation system and method based on artificial intelligence

Abstract

The application discloses a satellite communication resource allocation system and method based on artificial intelligence, and relates to the technical field of data analysis.The application identifies abnormal behaviors and establishes an abnormal scheduling label set by counting scheduling intervals based on a satellite communication resource scheduling process, collects and quantifies satellite channel quality, service demand and load state, and constructs satellite real-time state features.The application dynamically adjusts a scheduling period through an attention mechanism by fusing the state features, realizes period compression and amplification strategies, constructs a resource optimization model, minimizes resource weights under constraint conditions, supports a scheduling triggering mechanism under a minimum period, solves key technical problems such as rigidity, poor adaptability and response lag of existing period setting, improves flexibility and timeliness of resource scheduling, and realizes energy efficiency optimization of communication resource configuration.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, specifically to a satellite communication resource allocation system and method based on artificial intelligence. Background Technology

[0002] In modern satellite communication systems, resource scheduling efficiency directly determines the overall service capacity and communication quality of the system. With the rapid deployment of low-Earth orbit satellite constellations and the continuous increase in user access density, the traditional fixed-cycle resource scheduling mechanism has gradually exposed problems such as slow response and low resource utilization.

[0003] Existing scheduling mechanisms typically make resource allocation decisions based on preset time intervals, lacking the ability to dynamically perceive and adjust to business status and link conditions. Some systems have introduced scheduling mechanisms based on event-driven or load-aware approaches, but a mature periodic adaptive optimization system has not yet been formed. At the same time, the optimal scheduling period varies for different types of resources.

[0004] Therefore, there is a need for an artificial intelligence-based satellite communication resource allocation system and method that supports adaptive adjustment of scheduling cycles, cross-resource type collaborative modeling, and intelligent algorithm-driven optimization. This will solve the key technical problems of rigid cycle settings, poor adaptability, and delayed response in the current system, improve the flexibility and timeliness of resource scheduling, and achieve energy efficiency optimization of communication resource allocation. Summary of the Invention

[0005] The purpose of this invention is to provide a satellite communication resource allocation system and method based on artificial intelligence to solve the problems raised in the prior art.

[0006] To address the aforementioned technical problems, this invention provides the following technical solution: a satellite communication resource allocation method based on artificial intelligence, the method comprising:

[0007] Step S100: Based on the satellite communication resource scheduling process, statistically analyze the scheduling intervals, identify abnormal behaviors, and establish an abnormal scheduling tag set;

[0008] Step S200: Collect and quantify satellite channel quality, service requirements, and payload status to construct real-time satellite status characteristics;

[0009] Step S300: The fused state features are dynamically adjusted through an attention mechanism to achieve a cycle compression and amplification strategy;

[0010] Step S400: Construct a resource optimization model to minimize resource weights under constraints and support a scheduling triggering mechanism under minimum cycle.

[0011] Furthermore, step S100 includes:

[0012] Step S101: In the satellite communication system, monitor the satellite's payload status, channel quality, and service requirements. When satellite communication resources are reallocated, generate a scheduling record.

[0013] Step S102: Obtain historical scheduling records, collect the time point corresponding to each historical scheduling record, calculate the interval between every two adjacent historical scheduling records, summarize all interval durations, and calculate the average interval duration.

[0014] Step S103: Extract abnormal periodic change points and collect the scheduling behavior corresponding to the abnormal periodic change points to establish a historical abnormal scheduling behavior tag set;

[0015] By monitoring satellite payload status, channel quality, and service requirements in real time, dynamic perception and analysis of communication resources can be achieved, improving scheduling accuracy. Resource allocation strategies can be automatically adjusted according to changing environmental conditions and service priorities, reducing manual intervention and enhancing system adaptability.

[0016] Based on the results of scheduling records and historical data analysis, resource reallocation can be predicted and prepared in advance to avoid resource idleness or conflict. By calculating the average interval duration, the scheduling frequency and trend can be grasped, thereby rationally arranging the scheduling cycle and improving the rhythm and coordination of scheduling behavior.

[0017] By utilizing periodic analysis and the mechanism for extracting abnormal periodic change points, potential unstable factors or sudden events in system operation can be identified in a timely manner. Abnormal scheduling behaviors can be collected and marked, and a set of historical abnormal scheduling behavior tags can be constructed. This helps to build a knowledge graph of scheduling behavior and provides a data foundation for subsequent identification and intervention.

[0018] Based on the analysis of historical abnormal scheduling behavior, intelligent models can be trained to achieve abnormal scheduling early warning, detect potential problems in advance, and build fault-tolerant scheduling strategies for the system to respond quickly to sudden scheduling needs or resource bottlenecks, ensuring that communication links are not interrupted.

[0019] The accumulation of historical scheduling behavior and anomaly labels constitutes the knowledge base for the system to continuously learn and optimize. As data accumulates, the system scheduling algorithm can be continuously updated and evolved, thereby gradually reaching the optimal or suboptimal scheduling strategy, demonstrating a certain degree of "self-optimization" capability.

[0020] With more scientific and reasonable scheduling, the timeliness and reliability of high-priority services can be guaranteed, and the user experience can be improved. At the same time, through the abnormal behavior identification mechanism, the impact of scheduling interruption or resource conflict on communication quality can be avoided.

[0021] Furthermore, step S200 includes:

[0022] Step S201: Collect the real-time channel quality of the satellite and calculate the effective channel quality according to the following formula:

[0023]

[0024] Where Q represents the effective channel quality, and A a Let B be the signal-to-noise-to-interference ratio of the a-th communication link. a Let be the weight of the a-th communication link, and b be the total number of communication links.

[0025] Step S202: Collect real-time service requirements from the satellite and calculate the service score according to the following formula:

[0026]

[0027] Among them, C t Let D be the business score at time t. t Let F be the delay requirement at time t. t Let d represent the bandwidth requirement at time t, and let f represent the weights of the delay requirement and bandwidth requirement, respectively.

[0028] Step S203: Collect the real-time payload status of the satellite and construct the real-time status matrix of the satellite;

[0029] The calculation of effective channel quality Q fully considers the signal-to-noise ratio and link weight of each communication link, which can reflect the communication capability between the satellite and the ground station in a true and objective way. By integrating multi-link information through weighting, it avoids the distortion caused by relying on a single channel indicator, enabling the scheduling system to maintain good judgment in complex and ever-changing channel environments. The service score combines two key performance indicators, delay and bandwidth, and is dynamically adjusted according to the weights to accurately depict the real-time demand of services for communication resources, thereby enhancing the timeliness and priority identification capability of scheduling.

[0030] The construction of the real-time payload state matrix enables a systematic expression of various physical states of the satellite, such as current resource occupancy, processing capacity, and energy consumption level. This provides structured data support for scheduling strategies. The state matrix, as input, can be integrated with machine learning models and scheduling rule engines to achieve state-aware decision-making, which helps to formulate more scientific and reasonable resource scheduling schemes.

[0031] In complex spatial environments, link quality fluctuates frequently. This method dynamically calculates the effective channel quality Q to achieve real-time evaluation of link performance, thereby supporting dynamic resource allocation at the link level. It can prioritize the allocation of communication resources to high-quality links, avoid communication interruptions or data packet loss caused by channel degradation, and improve overall link stability and transmission efficiency.

[0032] The service scoring mechanism can differentiate the sensitivity of different services to bandwidth and latency, providing higher priority scheduling guarantees for critical tasks such as telemetry, remote sensing, and video backhaul, and supporting "demand-based allocation". It can flexibly adjust resource allocation strategies according to real-time changes in service requirements, thereby achieving differentiated guarantee and dynamic optimization of QoS.

[0033] Feature channel quality, service score, and load state matrix serve as high-dimensional inputs to the scheduling model, which can be used to train scheduling optimization algorithms, such as reinforcement learning and graph neural networks. This improves the efficiency and generalization ability of the algorithm in learning scheduling rules, and helps the system to automatically find the best solution and continuously evolve in different task scenarios.

[0034] The multi-source information fusion mechanism avoids the risks associated with judging by a single indicator. It can still make reliable decisions when faced with network disturbances or sudden business demands. Even if some links temporarily fail, the system can still reconfigure resources based on the quality of the remaining links and business scores, demonstrating strong fault tolerance and redundant scheduling capabilities. It serves as an input module for advanced functions such as anomaly detection, trend prediction, and task planning, and has good potential for platformization and modular development.

[0035] Furthermore, step S300 includes:

[0036] Step S301: Concatenate the satellite's effective channel quality, service score, and real-time state matrix and feed them into the attention network to calculate the corrected period T1. Calculate the real-time scheduling period using the following formula:

[0037] T=r×T2+(1-r)×T1, r∈[0, 1];

[0038] Where T represents the real-time scheduling period, T2 represents the average interval duration, and r represents the weight of the real-time scheduling period;

[0039] Step S302: Preset the collection time interval, calculate the change rate of the service score in each time interval, preset the change rate threshold, and if the change rate threshold is exceeded, the compression period is T3 = p × T, where T3 represents the compressed period and p represents the preset compression coefficient.

[0040] Step S303: Collect the satellite's resource utilization rate and preset the resource utilization rate threshold. If the resource utilization rate threshold is not exceeded for k consecutive collection time intervals, the amplification period is T4 = min(q × T, Tmax), where T4 represents the amplified period, q represents the preset amplification coefficient, and Tmax represents the preset maximum period.

[0041] By concatenating the satellite's effective channel quality, service score, and real-time state matrix and inputting them into the attention network, the system can calculate the corrected scheduling cycle in real time, ensuring that the scheduling strategy can be adjusted in a timely manner according to changes in the satellite's state. This dynamic decision-making mechanism can cope with different communication needs and channel environments, ensuring that resources are used efficiently while meeting the timeliness requirements of services.

[0042] By monitoring the rate of change in business scores, if the change exceeds a preset threshold, the scheduling cycle can be compressed in real time to ensure timely processing of sudden business or high-priority tasks. This mechanism ensures rapid response of resource scheduling in high-demand scenarios.

[0043] By compressing and extending the scheduling cycle, the scheduling cycle can be flexibly adjusted according to real-time business needs, resource utilization, and system load. In particular, when there is insufficient resource utilization in the system, the scheduling frequency can be reduced by extending the cycle, thereby reducing the system load and improving the overall system stability and reliability. This flexible cycle adjustment mechanism effectively responds to changes in communication tasks. For example, when the system detects high-frequency changes in business needs, it will compress the scheduling cycle to shorten the resource allocation time; while when resource utilization is low, the system will extend the scheduling cycle to reduce invalid scheduling requests and reduce energy consumption.

[0044] By dynamically adjusting the scheduling cycle, the system can lengthen the scheduling cycle when resource utilization is low, thus avoiding resource waste and redundant computation caused by frequent scheduling. Conversely, when resource utilization is high, the system will compress the scheduling cycle to ensure that resources are allocated in a timely manner under high load, avoiding task failures or packet loss caused by delayed scheduling. Due to the compression and amplification mechanism of the scheduling cycle, the system can effectively avoid performance degradation caused by over-scheduling or resource idleness. The scheduling system can adaptively adjust its behavior, making resource allocation and use more balanced and efficient.

[0045] By utilizing attention networks, the system can accurately assess and adjust the scheduling cycle based on factors such as current channel quality, service requirements, and load status. This mechanism enables the system to automatically identify which factors are more important in complex, multi-dimensional environments, thereby accurately scheduling resources and maximizing communication efficiency.

[0046] By extending the scheduling cycle during periods of low demand, the system can reduce unnecessary scheduling operations, thereby reducing energy consumption. This mechanism is particularly suitable for high-energy-consuming systems such as satellite communications, effectively reducing the system's energy consumption pressure, avoiding over-scheduling, reducing the overload operation of satellite equipment, extending the satellite's lifespan, and improving its long-term operational stability.

[0047] Furthermore, step S400 includes:

[0048] Step S401: When the adjustment period is reached, establish a resource optimization model, wherein the resource optimization model is to minimize... Among them, J i Let L be the weight of the i-th type of resource. i This represents the allocation amount for the i-th type of resource;

[0049] Step S402: The constraint of the optimization model is that the sum of all resource allocations does not exceed the preset resource limit, and the effective channel quality formed after each resource configuration is not lower than the target effective channel quality.

[0050] Step S403: Preset several triggering conditions. If a triggering condition exists, skip the original scheduled scheduling cycle and use the preset minimum cycle for resource scheduling.

[0051] The resource optimization model aims to minimize the total weighted resource usage, fully considering the importance of different types of resources. It enables "on-demand allocation and trade-off optimization" of resources, upgrading scheduling decisions from "whether to schedule" to "how to schedule optimally," achieving a leap from coarse-grained control to fine-grained optimization, and significantly improving resource management efficiency and scheduling accuracy.

[0052] The optimization model introduces constraints to ensure that the total amount of all resource allocation does not exceed the system's resource limit, such as power, bandwidth, and computing power. The effective channel quality achieved by each type of resource after configuration is not lower than the predetermined target value, thus avoiding communication link degradation due to insufficient resource allocation. This dual constraint mechanism enhances the controllability and feasibility of scheduling and ensures that the communication quality baseline during task execution is not breached.

[0053] The triggering mechanism that skips the original scheduling cycle allows the system to immediately trigger the minimum scheduling cycle for resource allocation when faced with emergencies such as a sudden increase in tasks or a sudden degradation of critical links, without waiting for the original cycle. This improves the system's response speed to emergencies, provides higher real-time performance and adaptability, and ensures uninterrupted communication in emergency situations.

[0054] The goal of minimizing resources helps reduce satellite power consumption, alleviate thermal load, and extend payload life. It is a concrete practice of the green communication concept in satellite systems. During non-critical mission periods, the system can achieve "energy-saving scheduling" through minimum allocation, providing strong energy efficiency guarantee for long-cycle missions and low-power satellites.

[0055] To better implement the above methods, an artificial intelligence-based satellite communication resource allocation system is also proposed. The system includes a scheduling record module, a real-time analysis module, an adjustment cycle module, and a resource optimization module.

[0056] Scheduling record module: Based on the satellite communication resource scheduling process, it counts scheduling intervals, identifies abnormal behavior, and establishes an abnormal scheduling tag set;

[0057] Real-time analysis module: Collects and quantifies satellite channel quality, service requirements and payload status, and constructs real-time satellite status characteristics;

[0058] The scheduling cycle adjustment module dynamically adjusts the scheduling cycle through an attention mechanism by fusing state features, thereby implementing a cycle compression and amplification strategy.

[0059] Resource optimization module: Construct a resource optimization model to minimize resource weights under constraints and support scheduling triggering mechanisms under minimum cycles.

[0060] Furthermore, the scheduling recording module includes a unit for calculating the average interval duration and a unit for setting behavior tags:

[0061] The average interval duration calculation unit: In the satellite communication system, the satellite payload status, channel quality, and service requirements are monitored. When satellite communication resources are reallocated, a scheduling record is generated, historical scheduling records are obtained, the time point corresponding to each historical scheduling record is collected, the interval duration between every two adjacent historical scheduling records is calculated, all interval durations are summarized, and the average interval duration is calculated.

[0062] Behavior tag set unit: Extract abnormal periodic change points, collect the scheduling behaviors corresponding to the abnormal periodic change points, and establish a historical abnormal scheduling behavior tag set.

[0063] Furthermore, the real-time analysis module includes a unit for calculating effective channel quality and a unit for calculating service scores.

[0064] Effective channel quality calculation unit: Collects real-time channel quality data from satellites and calculates effective channel quality;

[0065] The service scoring unit collects real-time service requirements from the satellite, calculates service scores, collects real-time payload status of the satellite, and constructs the satellite's real-time status matrix.

[0066] Furthermore, the adjustment cycle module includes a unit for calculating the real-time scheduling cycle and a unit for determining the adjustment cycle:

[0067] Calculate the real-time scheduling cycle unit: Concatenate the satellite's effective channel quality, service score, and real-time state matrix and feed them into the attention network to calculate the corrected cycle, and then calculate the real-time scheduling cycle.

[0068] Determine the adjustment cycle unit: preset the collection time interval, calculate the change rate of the service score in each time interval, preset the change rate threshold, if the change rate threshold is exceeded, the compression cycle is T3 = p × T, collect the satellite resource utilization rate, preset the resource utilization rate threshold, if k consecutive collection time intervals do not exceed the resource utilization rate threshold, the amplification cycle is T4 = min(q × T, Tmax).

[0069] Furthermore, the resource optimization module includes a unit for establishing an optimization model and a unit for determining triggering conditions:

[0070] Establish an optimization model unit: When the adjustment period is reached, a resource optimization model is established. The constraints of the optimization model are that the sum of all resource allocations does not exceed the preset resource limit, and the effective channel quality formed after each resource configuration is not lower than the target effective channel quality.

[0071] Determine trigger condition unit: Several trigger conditions are preset. If a trigger condition exists, the original scheduled scheduling cycle is skipped, and the preset minimum cycle is used for resource scheduling.

[0072] Compared with the prior art, the beneficial effects of the present invention are as follows: In the prior art, satellite resource scheduling is mostly based on preset rules or fixed strategies, which have slow response and poor adaptability; while the present invention fully introduces attention mechanism and multi-source real-time data fusion, enabling scheduling behavior to have intelligent capabilities such as state perception, anomaly recognition and periodic adaptation. Based on the learning of historical scheduling behavior and the construction of label sets, it supports subsequent intelligent prediction and anomaly intervention mechanisms, demonstrating a high degree of self-learning and self-optimization capabilities.

[0073] This invention introduces an attention network model to achieve real-time correction and dynamic adjustment of the scheduling cycle. It can automatically compress or expand the scheduling cycle based on factors such as link quality, service score and system load, thereby improving system responsiveness and resource utilization flexibility.

[0074] This invention simultaneously collects and integrates link channel quality, service score, and real-time load status to construct a multi-dimensional system state matrix, providing more comprehensive and accurate input features for the scheduling algorithm, making the scheduling strategy more precise and reliable, supporting differentiated service assurance and dynamic link adjustment, and improving the accuracy and generalization ability of overall scheduling decisions.

[0075] This invention proposes a scheduling optimization method based on minimizing the resource-weighted objective function, which can realize strategic trade-offs and fine-grained control of resource scheduling. Especially when resources are limited or task priorities conflict, it can ensure the priority of core business and improve the task completion rate and resource utilization efficiency of the system.

[0076] Current systems experience lag in response to sudden link degradation or the insertion of high-priority tasks. This invention addresses this by setting scheduling trigger conditions. When certain emergency conditions are met, the original planned cycle can be skipped, and tasks can be quickly executed in the minimum scheduling cycle. This significantly improves the system's real-time performance and responsiveness, effectively ensuring that critical tasks are not interrupted and communication links do not fail.

[0077] This method proactively extends the scheduling cycle when resource utilization is low, reduces system power consumption and load pressure while ensuring service quality, improves system energy efficiency ratio, supports energy consumption-aware scheduling and satellite equipment lifespan management, and aligns with the development trend of the next generation of "green satellite communication". Attached Figure Description

[0078] Figure 1 This is a flowchart illustrating an artificial intelligence-based satellite communication resource allocation method according to the present invention.

[0079] Figure 2 This is a schematic diagram of the structure of a satellite communication resource allocation system based on artificial intelligence according to the present invention. Detailed Implementation

[0080] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0081] Please see Figure 1 This invention provides a technical solution: a satellite communication resource allocation method based on artificial intelligence, the method comprising:

[0082] Step S100: Based on the satellite communication resource scheduling process, statistically analyze the scheduling intervals, identify abnormal behaviors, and establish an abnormal scheduling tag set;

[0083] Step S100 includes:

[0084] Step S101: In the satellite communication system, monitor the satellite's payload status, channel quality, and service requirements. When satellite communication resources are reallocated, generate a scheduling record.

[0085] Step S102: Obtain historical scheduling records, collect the time point corresponding to each historical scheduling record, calculate the interval between every two adjacent historical scheduling records, summarize all interval durations, and calculate the average interval duration.

[0086] Step S103: Extract abnormal periodic change points and collect the scheduling behavior corresponding to the abnormal periodic change points to establish a historical abnormal scheduling behavior tag set;

[0087] For example, there are 4 historical scheduling records, and the time points corresponding to each historical scheduling record are 12s, 20s, 35s, and 50s. The calculated adjacent scheduling intervals are 12s, 8s, 15s, and 15 seconds, and the calculated average interval duration is 12.5s.

[0088] Step S200: Collect and quantify satellite channel quality, service requirements, and payload status to construct real-time satellite status characteristics;

[0089] Step S200 includes:

[0090] Step S201: Collect the real-time channel quality of the satellite and calculate the effective channel quality according to the following formula:

[0091]

[0092] Where Q represents the effective channel quality, and A a Let B be the signal-to-noise-to-interference ratio of the a-th communication link. a Let be the weight of the a-th communication link, and b be the total number of communication links.

[0093] Step S202: Collect real-time service requirements from the satellite and calculate the service score according to the following formula:

[0094]

[0095] Among them, C t Let D be the business score at time t. t Let F be the delay requirement at time t. t Let d represent the bandwidth requirement at time t, and let f represent the weights of the delay requirement and bandwidth requirement, respectively.

[0096] Step S203: Collect the real-time payload status of the satellite and construct the real-time status matrix of the satellite;

[0097] For example, if the signal-to-noise-interference ratio (SNR) of the first link is 10 and the weight is 0.5, the SNR of the second link is 5 and the weight is 0.3, and the SNR of the third link is 2 and the weight is 0.2, the effective channel quality is calculated to be 0.94.

[0098] The acquisition latency requirement is 8, with a weight of 0.7, and the bandwidth requirement is 5, with a weight of 0.3. The calculated service score is 0.9992.

[0099] Step S300: The fused state features are dynamically adjusted through an attention mechanism to achieve a cycle compression and amplification strategy;

[0100] Step S300 includes:

[0101] Step S301: Concatenate the satellite's effective channel quality, service score, and real-time state matrix and feed them into the attention network to calculate the corrected period T1. Calculate the real-time scheduling period using the following formula:

[0102] T=r×T2+(1-r)×T1, r∈[0, 1];

[0103] Where T represents the real-time scheduling period, T2 represents the average interval duration, and r represents the weight of the real-time scheduling period;

[0104] Step S302: Preset the collection time interval, calculate the change rate of the service score in each time interval, preset the change rate threshold, and if the change rate threshold is exceeded, the compression period is T3 = p × T, where T3 represents the compressed period and p represents the preset compression coefficient.

[0105] Step S303: Collect the satellite's resource utilization rate and preset the resource utilization rate threshold. If the resource utilization rate threshold is not exceeded for k consecutive collection time intervals, the amplification period is T4 = min(q × T, Tmax), where T4 represents the amplified period, q represents the preset amplification coefficient, and Tmax represents the preset maximum period.

[0106] For example, in step S200, the effective channel quality is 0.94, the service score is 0.9992, the remaining power ratio in the real-time satellite status is 0.7, and the remaining channel percentage is 0.3. These values ​​are input into the attention network, and the correction coefficient is calculated to be 0.82. In step S100, the average interval duration is calculated to be 12.5 seconds, the corrected period is calculated to be 10.25 seconds, the weight of the real-time scheduling period is 0.6, and the real-time scheduling period is calculated to be 11.6 seconds.

[0107] The preset data collection time interval is 5 seconds, the change rate threshold is 0.2, the first business score is 0.6, the second business score is 0.85, and the change rate is 0.05. If the change rate threshold is not exceeded, no compression will be performed.

[0108] The utilization rate at the first time point is 0.4, the utilization rate at the second time point is 0.5, the utilization rate at the third time point is 0.55, the resource utilization threshold is 0.6, and the preset k is 3. Then, the amplified period is calculated to be min(12.3, 20), and the amplified period is 12.3s.

[0109] Step S400: Construct a resource optimization model to minimize resource weights under constraints and support a scheduling triggering mechanism under minimum cycle.

[0110] Step S400 includes:

[0111] Step S401: When the adjustment period is reached, establish a resource optimization model, wherein the resource optimization model is to minimize... Among them, J i Let L be the weight of the i-th type of resource. i This represents the allocation amount for the i-th type of resource;

[0112] Step S402: The constraint of the optimization model is that the sum of all resource allocations does not exceed the preset resource limit, and the effective channel quality formed after each resource configuration is not lower than the target effective channel quality.

[0113] Step S403: Several triggering conditions are preset. If a triggering condition exists, the original scheduled scheduling cycle is skipped, and the preset minimum cycle is used for resource scheduling.

[0114] To better implement the above methods, an artificial intelligence-based satellite communication resource allocation system is also proposed. The system includes a scheduling record module, a real-time analysis module, an adjustment cycle module, and a resource optimization module.

[0115] Scheduling record module: Based on the satellite communication resource scheduling process, it counts scheduling intervals, identifies abnormal behavior, and establishes an abnormal scheduling tag set;

[0116] The scheduling record module includes a unit for calculating the average interval duration and a unit for setting behavior tags.

[0117] The average interval duration calculation unit: In the satellite communication system, the satellite payload status, channel quality, and service requirements are monitored. When satellite communication resources are reallocated, a scheduling record is generated, historical scheduling records are obtained, the time point corresponding to each historical scheduling record is collected, the interval duration between every two adjacent historical scheduling records is calculated, all interval durations are summarized, and the average interval duration is calculated.

[0118] Behavior tag set unit: Extract abnormal periodic change points, collect the scheduling behaviors corresponding to the abnormal periodic change points, and establish a historical abnormal scheduling behavior tag set.

[0119] Real-time analysis module: Collects and quantifies satellite channel quality, service requirements and payload status, and constructs real-time satellite status characteristics;

[0120] The real-time analysis module includes a unit for calculating effective channel quality and a unit for calculating service scores.

[0121] Effective channel quality calculation unit: Collects real-time channel quality data from satellites and calculates effective channel quality;

[0122] The service scoring unit collects real-time service requirements from the satellite, calculates service scores, collects real-time payload status of the satellite, and constructs the satellite's real-time status matrix.

[0123] The scheduling cycle adjustment module dynamically adjusts the scheduling cycle through an attention mechanism by fusing state features, thereby implementing a cycle compression and amplification strategy.

[0124] The adjustment cycle module includes a unit for calculating the real-time scheduling cycle and a unit for determining the adjustment cycle.

[0125] Calculate the real-time scheduling cycle unit: Concatenate the satellite's effective channel quality, service score, and real-time state matrix and feed them into the attention network to calculate the corrected cycle, and then calculate the real-time scheduling cycle.

[0126] Determine the adjustment cycle unit: preset the collection time interval, calculate the change rate of the service score in each time interval, preset the change rate threshold, if the change rate threshold is exceeded, the compression cycle is T3 = p × T, collect the satellite resource utilization rate, preset the resource utilization rate threshold, if k consecutive collection time intervals do not exceed the resource utilization rate threshold, the amplification cycle is T4 = min(q × T, Tmax).

[0127] Optimize resource module: Construct a resource optimization model to minimize resource weights under constraints and support scheduling triggering mechanisms under minimum cycles;

[0128] The resource optimization module includes a unit for establishing an optimization model and a unit for determining triggering conditions.

[0129] Establish an optimization model unit: When the adjustment period is reached, a resource optimization model is established. The constraints of the optimization model are that the sum of all resource allocations does not exceed the preset resource limit, and the effective channel quality formed after each resource configuration is not lower than the target effective channel quality.

[0130] Determine trigger condition unit: Several trigger conditions are preset. If a trigger condition exists, the original scheduled scheduling cycle is skipped, and the preset minimum cycle is used for resource scheduling.

[0131] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. An artificial intelligence-based satellite communication resource allocation method, characterized by, The method comprises: Step S100: based on the satellite communication resource scheduling process, the scheduling interval is counted, the abnormal behavior is identified, and the abnormal scheduling label set is established; Step S200: collect and quantify the satellite channel quality, service demand and load state, and construct the satellite real-time state feature; Step S300: the state feature is fused to dynamically adjust the scheduling period through the attention mechanism, and the period compression and amplification strategy is realized; The step S300 comprises the following steps: Step S301: the effective channel quality, service score and real-time state matrix of the satellite are spliced and sent into the attention network, and the corrected period T1 is calculated, and the real-time scheduling period is calculated according to the following formula: ； Wherein, T represents the real-time scheduling period, T2 represents the average interval time, and r represents the weight value of the real-time scheduling period; Step S302: a preset collection time interval is calculated, the change rate of the service score of each time interval is calculated, a preset change rate threshold is set, and if the change rate threshold is exceeded, the compressed period is T3=p×T, wherein T3 represents the compressed period, and p represents the preset compression coefficient; Step S303: the resource utilization rate of the satellite is collected, a preset resource utilization rate threshold is set, and if the resource utilization rate threshold is not exceeded for continuous k collection time intervals, the amplified period is T4=min(q×T, Tmax), wherein T4 represents the amplified period, q represents the preset amplification coefficient, and Tmax represents the preset maximum period; Step S400: a resource optimization model is constructed, the resource weight is minimized under the constraint condition, and the scheduling triggering mechanism under the minimum period is supported.

2. The satellite communication resource allocation method based on artificial intelligence according to claim 1, characterized in that, The step S100 comprises the following steps: Step S101: in the satellite communication system, the load state, channel quality and service demand of the satellite are monitored, and when the satellite communication resource is redistributed, a scheduling record is generated; Step S102: the historical scheduling record is obtained, the time point corresponding to each historical scheduling record is collected, the interval time between each two adjacent historical scheduling records is calculated, all interval times are summarized, and the average interval time is calculated; Step S103: the abnormal period change point is extracted, the scheduling behavior corresponding to the abnormal period change point is collected, and the historical abnormal scheduling behavior label set is established.

3. The method of claim 2, wherein, The step S200 comprises the following steps: Step S201: the real-time channel quality of the satellite is collected, and the effective channel quality is calculated according to the following formula: ； where Q represents the effective channel quality, A a S / N of the a-th communication link, B a weight of the a-th communication link, and b represents the total number of communication links. Step S202: the real-time service demand of the satellite is collected, and the service score is calculated according to the following formula: ； wherein C t denotes the service score at the tth time point, D t denotes the delay requirement at the tth time point, F t denotes the bandwidth requirement at the tth time point, d and f respectively denote the weight values of the delay requirement and the bandwidth requirement; Step S203: the real-time load state of the satellite is collected, and the real-time state matrix of the satellite is constructed.

4. The method of claim 1, wherein, The step S400 comprises the following steps: Step S401: when reaching the adjustment period, a resource optimization model is established, the resource optimization model is to minimize wherein, J i represents the weight of the i-th resource, L i represents the allocation amount of the i-th resource, and m represents the total number of resource types; Step S402: the constraint of the optimization model is that the sum of all resource allocations does not exceed the preset resource upper limit, and the effective channel quality formed after each resource configuration is not less than the target effective channel quality; Step S403: a plurality of trigger conditions are preset, if there is a trigger condition, the original planned scheduling period is skipped, and the resource scheduling is performed using the preset minimum period.

5. An artificial intelligence-based satellite communication resource allocation system for implementing the artificial intelligence-based satellite communication resource allocation method of any one of claims 1-4, characterized in that, The system comprises a scheduling record module, a real-time analysis module, an adjustment period module and an optimized resource module; The scheduling record module: based on satellite communication resource scheduling process, statistics scheduling interval, identify abnormal behavior and establish abnormal scheduling label set; The real-time analysis module: collect and quantify satellite channel quality, business demand and load state, build satellite real-time state characteristics; The adjustment cycle module: fusion state characteristics through attention mechanism dynamic adjustment scheduling cycle, realize cycle compression and amplification strategy; The adjustment cycle module includes a real-time scheduling cycle calculation unit and a determination adjustment cycle unit: The real-time scheduling cycle calculation unit: the effective channel quality of satellite, business score and real-time state matrix are spliced and sent into attention network, the corrected cycle is calculated, and the real-time scheduling cycle is calculated; The determination adjustment cycle unit: preset acquisition time interval, calculate the change rate of business score of each time interval, preset change rate threshold, if it exceeds the change rate threshold, then the compressed cycle is T3=p×T, wherein T3 represents the compressed cycle, p represents the preset compression coefficient, T represents the real-time scheduling cycle; the resource utilization rate of satellite is collected, the resource utilization rate threshold is preset, if the resource utilization rate threshold is not exceeded in continuous k acquisition time intervals, then the amplified cycle is T4=min(q×T, Tmax), wherein T4 represents the amplified cycle, q represents the preset amplification coefficient, Tmax represents the preset maximum cycle; The optimization resource module: build resource optimization model, minimize resource weight under constraint condition, support scheduling trigger mechanism under minimum cycle.

6. The satellite communication resource allocation system based on artificial intelligence according to claim 5, wherein, The scheduling record module includes a calculation average interval duration unit and a behavior label set unit: The calculation average interval duration unit: in satellite communication system, monitor the load state, channel quality and business demand of satellite, generate a scheduling record when satellite communication resource is redistributed, obtain historical scheduling record, collect the time point corresponding to each historical scheduling record, calculate the interval duration between each two adjacent historical scheduling record, summarize all interval durations, and calculate the average interval duration; The behavior label set unit: extract abnormal cycle change point, and collect the scheduling behavior corresponding to the abnormal cycle change point, and establish historical abnormal scheduling behavior label set.

7. The artificial intelligence based satellite communication resource allocation system of claim 5, wherein, The real-time analysis module includes a calculation effective channel quality unit and a calculation business score unit: The calculation effective channel quality unit: collect real-time channel quality of satellite, calculate effective channel quality; The calculation business score unit: collect real-time business demand of satellite, calculate business score, collect real-time load state of satellite, and build real-time state matrix of satellite.

8. The satellite communication resource allocation system based on artificial intelligence according to claim 5, wherein, The optimization resource module includes an optimization model establishment unit and a trigger condition determination unit: The optimization model establishment unit: when the adjustment cycle is reached, the resource optimization model is established, the constraint of optimization model is that the sum of all resource allocation amounts does not exceed the preset resource upper limit, and the effective channel quality formed after each resource configuration is not less than the target effective channel quality; The determining trigger condition unit: preset several trigger conditions, if there is a trigger condition, skip the original plan of the scheduling cycle, using the preset minimum cycle for resource scheduling.

Images