5G communication data transmission management system

Abstract

The invention discloses a 5G communication data transmission management system, which relates to the technical field of communication data transmission and comprises a real-time resource monitoring module, an intelligent evaluation module, a feature extraction module, a transmission process division module, a normal transmission management module and an overload transmission regulation and control module, through a deployed network monitoring tool and performance monitoring software, the resource occupation condition in the video monitoring data transmission process is monitored in real time. According to the invention, the system intelligently evaluates the resource occupation condition through real-time monitoring and a machine learning model, and dynamically regulates and controls the resources, reduces the delay, accelerates the analysis speed and guarantees the emergency response in the case of high load by using the queue expansion index, the link fluctuation index and the queue theory model. Through transmission process division, efficient management in a normal stage and dynamic optimization in an overload stage are realized, timely response and resource release are ensured, waste is avoided, and the stability and efficiency of the system under various loads are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of communication data transmission, and in particular to a 5G communication data transmission management system. Background Art

[0002] 5G communication data transmission management refers to the planning, monitoring, optimization and guarantee of the data transmission process in the 5G network to achieve efficient, stable and secure data communication. With the introduction of 5G technology, the network transmission rate has been significantly improved, the latency has been reduced, and large-scale device connections are supported. Data transmission management mainly involves bandwidth allocation, flow control, channel resource optimization, load balancing and data priority scheduling. In complex scenarios, such as the Internet of Things, large-scale sensor networks and smart city applications, efficient data transmission management ensures the rational use of network resources, avoids congestion and transmission delays, and thus ensures the quality of service (QoS) of diversified applications. In addition, through real-time allocation of resources through intelligent algorithms, the transmission strategy can be dynamically adjusted according to business needs and network conditions to improve user experience.

[0003] In 5G data transmission management, core tasks also include network slicing, edge computing, and security management. Network slicing divides the physical network into multiple virtual networks, and each slice provides differentiated services based on different business needs to ensure the stability of key businesses. Edge computing sinks data processing capabilities to network edge nodes, reducing core network pressure and latency. At the same time, transmission management must strictly monitor data security to prevent illegal data intrusion and leakage. 5G data transmission management also uses intelligent technologies such as machine learning to optimize transmission paths and predict network status, so that the network remains efficient and stable in a high-concurrency environment. This systematic management is crucial in promoting the widespread implementation of applications in future fields such as Internet of Vehicles, telemedicine, and industrial automation.

[0004] The prior art has the following deficiencies:

[0005] Although slices are logically independent, they still share physical resources such as bandwidth, computing power, and storage space at the bottom layer. Existing technologies usually use fixed resource allocation. Although it can meet basic needs in most cases, when a slice requires a large amount of computing resources in a short period of time, the fixed allocation may not be able to cope. In the large-scale monitoring network of a smart city, multiple cameras transmit data at the same time. Especially when a security incident (such as a fire or natural disaster) occurs, centralized viewing of the screen will cause a surge in data volume. In this case, the decoding and real-time analysis of multiple videos requires a lot of computing power, bandwidth, and storage resources, which puts tremendous pressure on the system. If resources are insufficient, delays in decoding and analysis will result in the inability to identify security threats in a timely manner, and emergency response time will be extended, which may cause serious casualties and property losses.

[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0007] The purpose of the present invention is to provide a 5G communication data transmission management system. Through real-time monitoring tools and machine learning models, the system can intelligently evaluate resource occupancy, and combine queue expansion index, link fluctuation index and queuing theory model to timely adjust resources under high load, reduce delays, speed up decoding analysis, ensure emergency response speed, and reduce security risks. At the same time, differentiated resource management is achieved by dividing the transmission process: maintaining preset allocation in the normal stage to improve resource utilization; dynamically optimizing configuration in the overload stage to ensure timely response when demand surges and releasing excess resources when demand decreases. This flexible management mechanism avoids waste of resources and ensures that the system operates efficiently and stably under various loads to solve the problems in the above-mentioned background technology.

[0008] In order to achieve the above-mentioned purpose, the present invention provides the following technical solutions: a 5G communication data transmission management system, including a real-time resource monitoring module, an intelligent evaluation module and a feature extraction module, a transmission process division module, a normal transmission management module and an overload transmission control module:

[0009] The real-time resource monitoring module monitors the resource usage during the transmission of video surveillance data in real time through the deployed network monitoring tools and performance monitoring software;

[0010] The intelligent evaluation module and feature extraction module extract features from the aggregated resource occupancy data and use the pre-learned machine learning model to intelligently evaluate changes in the video surveillance data transmission process based on the extracted resource occupancy data features.

[0011] The transmission process division module divides the video surveillance data transmission process into different stages, including the normal transmission process and the overload transmission process, according to the results of the intelligent evaluation;

[0012] The normal transmission management module manages the normal transmission process according to the pre-set resource allocation plan and continues to monitor the system in real time to ensure that resource usage remains within a reasonable range;

[0013] The overload transmission control module uses the dynamic change information collected in real time to intelligently control resource allocation based on the pre-set resource allocation plan in the overload transmission process through the queuing theory model.

[0014] Preferably, feature extraction is performed on the aggregated resource occupancy data, and the extracted features include the expansion speed of the length of the queue of data packets waiting to be processed in the transmission node and the frequent changes in the network link status. After acquisition, under the monitoring window, the expansion speed of the length of the queue of data packets waiting to be processed in the transmission node and the frequent changes in the network link status are analyzed, and a queue expansion index and a link fluctuation index are generated respectively. The queue expansion index and the link fluctuation index are input into a pre-learned machine learning model, and a resource occupancy variation coefficient is generated by the machine learning model. The aggregated resource occupancy data is comprehensively evaluated by the resource occupancy coefficient.

[0015] Preferably, the queue expansion index is generated after analyzing the expansion speed of the queue length of the data packets waiting to be processed in the transmission node, and the specific steps are as follows:

[0016] In the detection window, the queue length data of the transmission node is collected, and the collected queue length data is calibrated as the queue length data time series Q(t), where Q(t) = {t i ,q i}={(t1, q1), (t2, q2), (t3, q3), ..., (t n ,q n )}, where Q(t) is the collected queue length time series, t is the timestamp, and t i is the timestamp of the i-th time point, a timestamp t i Marks a time point i, q in the time series i is the queue length at the ith time point, and n is the total number of time points;

[0017] To measure the speed at which the queue length changes at different time points, a discrete differential equation is used to calculate the instantaneous rate of change. The instantaneous rate of change is used to capture the rapid change of the queue length in a short period of time and to perform sensitive detection of processing capacity bottlenecks or packet backlogs. The calculation expression is as follows:

[0018]

[0019] , where ΔQ(t i ) is between two adjacent timestamps t i and t i+1 The queue length change rate between i+1 and q i is the queue length between time points i and i+1, t i and t i+1 are two adjacent timestamps;

[0020] Since the state of the system at similar time points can better reflect the current load pressure, the weighted expansion sum is constructed using exponentially increasing weights to amplify the changes at similar moments. The calculation expression is as follows:

[0021]

[0022] WQ(t) is the weighted expansion sum within the detection window, which represents the cumulative sum of the queue change rate at all times within the window. θ(i-n) It is an exponentially increasing weight, which is used to assign different weights to the change rate at different time points in the window. e is the natural base, and θ is the increasing rate adjustment coefficient, which controls the speed of weight decay.

[0023] Autoregressive residual analysis is used to detect the abnormality of queue length changes. The residual is calculated through the queue length data time series. The residual can capture the short-term abnormal fluctuations in the system. It is combined with the weighted inflation sum to generate the final queue inflation index. The calculation expression is as follows:

[0024]

[0025] , where QExpI is the queue expansion index, μ is the residual sensitivity adjustment coefficient, which controls the influence of the residual on the queue expansion index QExpI, ∈(t) is the residual, that is, the error between the actual value of the queue length and the trend fitting value. The residual calculation expression is as follows: ∈(t) = Q(t)-f(t), where f(t) is the trend fitting value.

[0026] Preferably, the link fluctuation index is generated after analyzing the number of frequent changes in the network link status, and the specific steps are as follows:

[0027] In the detection window, the network monitoring tool collects the changes of multiple performance indicators in the link in real time. Each state change triggers a record with a timestamp and change range.

[0028] The rate of change of the link status depends not only on the magnitude of the change, but also on the time interval between two changes. In order to capture frequent fluctuations in a short period of time, a time-sensitive factor is introduced to more accurately calculate the rate of each change. The calculation expression is as follows:

[0029]

[0030] , where R j represents the rate of change of the link at the jth state change, ΔS j is the magnitude of the state change, indicating the magnitude of the link state change in the jth state change event, T j is the timestamp of the jth change event, T j-1is the timestamp of the j-1th change event, T j -T j-1 represents the time interval between the jth and j-1th change events, and P is a sensitive adjustment parameter of the time interval;

[0031] In order to reflect the cumulative and nonlinear characteristics of state changes, an exponential model is used to accumulate the rates of multiple state changes. At the same time, each change is given a weight to reflect the importance of different types of changes. The calculation expression is as follows:

[0032]

[0033] , where E is the cumulative effect of state change, m is the total number of state changes, is the weight of the jth state change, ψ is the exponential amplification factor, which adjusts the sensitivity of the cumulative effect, and e is the natural base;

[0034] The cumulative effect E and the time characteristics are combined to calculate the final link fluctuation index. The calculation expression is as follows:

[0035]

[0036] , where LFI is the link fluctuation index, δ is the time adjustment factor, and ∈ is a constant.

[0037] Preferably, the resource occupancy variation coefficient generated after analyzing the aggregated resource occupancy data under the monitoring window is compared and analyzed with a preset resource occupancy variation coefficient reference threshold, and the video monitoring data transmission process is divided, and the division steps are as follows:

[0038] If the resource occupancy change coefficient is greater than or equal to a preset resource occupancy change coefficient reference threshold, the video surveillance data transmission process is classified as an overload transmission process;

[0039] If the resource occupancy variation coefficient is less than a preset resource occupancy variation coefficient reference threshold, the video surveillance data transmission process is classified as a normal transmission process.

[0040] Preferably, for the overload transmission process, the process of intelligent resource regulation and resource allocation based on a pre-set resource allocation scheme by using the dynamic change information collected in real time through a queuing theory model is as follows:

[0041] In the overload state, it is necessary to calculate the amount of resources that need to be increased based on the real-time queue expansion index QExpI and link fluctuation index LFI to ensure that the processing capacity meets the demand. This process adopts a multi-objective optimization model to balance the node processing capacity and link quality. The calculation expression is as follows:

[0042]

[0043] , where ΔR is the amount of resources that need to be increased, QExpI is the queue expansion index, and QExpI threshold is the preset reference threshold of the queue expansion index, LFI is the link fluctuation index, threshold is the preset reference threshold of the link fluctuation index, α and β are weight coefficients, which respectively control the impact of the queue expansion index QExpI and the link fluctuation index LFI on resource regulation;

[0044] According to the calculated resource control amount ΔR, update the resource allocation of nodes and links. At the same time, the new resource quota should not only cover the current load, but also reserve redundancy to cope with further burst traffic. The calculation expression is as follows: R new =R allocated +ΔR, where R new is the adjusted resource quota, R allocated It is a pre-set resource allocation plan;

[0045] By increasing resource quotas, we ensure that the system can quickly process backlogged data packets and improve the latency and jitter of the transmission link.

[0046] Resource allocation is subject to the physical limitations of the system. Therefore, it is necessary to optimize the calculated new resource quota to prevent it from exceeding the system capacity and ensure that resource allocation does not exceed the physical limit of the system to avoid insufficient resources for other services due to over-allocation. The calculation expression is as follows: R final =min(R new ,R max ), where R final is the actual amount of resources allocated, R max is the maximum amount of resources that can be provided;

[0047] After the overload state is released, the system needs to gradually release excess resources to prevent long-term resource occupation. The calculation expression is as follows:

[0048]

[0049] , where R release is the amount of resources after release, w1 and w2 are weight coefficients for controlling resource release, which are used to regulate the impact of queue expansion index QExpI and link fluctuation index LFI respectively.

[0050] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0051] The present invention uses real-time monitoring tools and performance monitoring software, combined with machine learning models, to intelligently evaluate resource occupancy, thereby effectively improving the dynamic regulation capability of the system. In the face of sudden security incidents, the status of transmission nodes can be monitored in real time, especially through the queue expansion index and link fluctuation index, to identify the trend of overload transmission in advance. Using the queuing theory model, the length of the queue of data packets waiting to be processed in the node can be regulated to ensure that the system can effectively manage resource allocation under high load, reduce transmission delays, and increase the speed of decoding and analysis, so as to identify security threats in a timely manner. This significantly reduces the delay problem caused by unreasonable resource allocation, shortens the emergency response time, avoids the potential safety hazards caused by lagging video analysis at critical moments, and reduces the risk of casualties and property losses.

[0052] The present invention can perform intelligent resource scheduling for different transmission states by dividing the transmission process. In the normal transmission stage, management is performed according to the preset resource allocation plan to avoid over-allocation of resources when they are not needed, thereby improving resource utilization efficiency. In the overload transmission stage, resources are temporarily optimized and configured through real-time dynamic information combined with intelligent control means. In this process, the resource occupancy change coefficient is continuously analyzed to ensure that resources can be allocated in time when sudden demand increases, and that excess resources are quickly released after demand slows down. This flexible resource management mechanism not only ensures efficient operation, but also avoids idleness and waste of resources due to fixed allocation, thereby improving the overall resource utilization efficiency and ensuring that the system can remain stable and reliable under high load conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0054] Figure 1 This is a system flow chart of a 5G communication data transmission management system of the present invention. DETAILED DESCRIPTION

[0055] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0056] The present invention provides Figure 1A 5G communication data transmission management system shown includes the following steps: real-time resource monitoring module, intelligent evaluation module and feature extraction module, transmission process division module, normal transmission management module and overload transmission control module:

[0057] The real-time resource monitoring module monitors the resource usage during the transmission of video surveillance data in real time through the deployed network monitoring tools and performance monitoring software;

[0058] By deploying appropriate network monitoring tools, you can achieve real-time monitoring of multiple resources in the video surveillance system, including bandwidth usage, CPU / GPU load, storage usage, and network traffic status. These monitoring tools can identify bottlenecks in complex data transmission environments, ensure stable transmission of video data streams, and avoid transmission delays and data loss in emergencies.

[0059] PRTG and SolarWinds NPM can monitor bandwidth usage and traffic conditions in video surveillance systems in real time. PRTG collects bandwidth usage of each camera and its connected devices through sensors such as SNMP and NetFlow to identify whether there is insufficient bandwidth or traffic congestion. NetFlow Analyzer further analyzes the specific source and destination of traffic to help locate high-traffic cameras or abnormal traffic paths. SolarWinds NPM not only monitors bandwidth, but also dynamically adjusts network paths to ensure smooth transmission.

[0060] Zabbix and Prometheus focus on monitoring CPU, GPU load, and storage space. Zabbix can monitor the computing power usage of each server and node, and issue an alert in time when the CPU or GPU utilization is close to saturation. Prometheus helps identify trend changes in computing load through high-frequency sampling and time series analysis. When combined with these tools, Grafana can display monitoring data in the form of visual charts, helping to understand storage consumption in real time and ensure that video data is not lost or interrupted due to insufficient storage.

[0061] The intelligent evaluation module and feature extraction module extract features from the aggregated resource occupancy data and use the pre-learned machine learning model to intelligently evaluate changes in the video surveillance data transmission process based on the extracted resource occupancy data features.

[0062] Extract features from the aggregated resource occupancy data. The extracted features include the expansion speed of the length of the queue of data packets waiting to be processed in the transmission node and the number of frequent changes in the network link status. After acquisition, analyze the expansion speed of the length of the queue of data packets waiting to be processed in the transmission node and the number of frequent changes in the network link status in the monitoring window, and generate a queue expansion index and a link fluctuation index respectively. Input the queue expansion index and the link fluctuation index into a pre-learned machine learning model, generate a resource occupancy variation coefficient through the machine learning model, and comprehensively evaluate the aggregated resource occupancy data through the resource occupancy coefficient.

[0063] In the aggregated resource usage data, the rapid expansion rate of the queue length of data packets waiting to be processed in the transmission node usually indicates that the system is in a high-load transmission state. This is because when data packets are transmitted between nodes in the network path, they are ideally processed and forwarded quickly. However, when the processing capacity or bandwidth resources of the node reach the upper limit, the newly incoming data packets cannot be processed in time, resulting in a rapid increase in the queue length. This queue expansion phenomenon reflects that the traffic demand in the network exceeds the resource supply, which is a direct manifestation of network congestion and high-load transmission. Especially in video surveillance systems, the transmitted video data has high throughput and low latency requirements. Therefore, when an emergency (such as a fire) causes a surge in traffic, the node will accumulate a large number of data packets, further increasing the processing delay. The rapid expansion of the queue length will not only slow down the transmission speed of the video stream, but also may cause data loss, transmission delay or packet loss retransmission, thereby significantly reducing the real-time and quality of the monitoring screen and affecting the emergency response capability of the system. Therefore, the queue expansion rate is an important indicator for identifying high-load transmission, indicating that the system is in a state of resource shortage and needs to be regulated and optimized in a timely manner.

[0064] In the monitoring window, the expansion speed of the queue length of the data packets waiting to be processed in the transmission node is analyzed to generate the queue expansion index. The specific steps are as follows:

[0065] In the detection window, the queue length data of the transmission node is collected, and the collected queue length data is calibrated as the queue length data time series Q(t), where Q(t) = {t i ,q i}={(t1, q1), (t2, q2), (t3, q3), ..., (t n ,q n )}, where Q(t) is the collected queue length time series, t is the timestamp, and t i is the timestamp of the i-th time point. A timestamp ti marks a time point i in the time series. i is the queue length at the ith time point, in units of packets, and n is the total number of time points;

[0066] To measure the speed at which the queue length changes at different time points, a discrete differential equation is used to calculate the instantaneous rate of change. The instantaneous rate of change is used to capture the rapid change of the queue length in a short period of time and to perform sensitive detection of processing capacity bottlenecks or packet backlogs. The calculation expression is as follows:

[0067]

[0068] , where ΔQ(t i ) is between two adjacent timestamps t i and t i+1 The queue length change rate between i+1 and q i is the queue length between time points i and i+1, t i and t i+1 are two adjacent timestamps;

[0069] A discrete differential equation is a mathematical expression that describes the state changes of a discrete system at adjacent time points. It reveals the dynamic evolution of a variable over time by calculating the change in a variable (such as queue length) between two adjacent time points. In , we calculated the queue length change rate at two consecutive time points to detect the instantaneous expansion trend. This method is suitable for real-time systems because the data is usually discretely sampled and can reflect subtle changes, which helps to respond quickly in dynamic networks.

[0070] Discrete differential equations are used to calculate instantaneous changes because in actual network environments, changes in queue length and resource status are sampled data at discrete time points rather than functions of continuous changes. Traditional continuous differentials cannot be used directly on such discrete data, so discrete differential equations can be used to calculate the rate of change between adjacent time points and quantify the expansion speed and trend of the packet queue. In addition, discrete differential equations can capture sudden changes in a short period of time, such as traffic surges or congestion, helping the system to more sensitively identify potential abnormal loads and provide accurate data support for subsequent dynamic scheduling and resource optimization. This method is suitable for processing high-frequency sampled data in transmission nodes, ensuring that the system responds promptly in a rapidly changing network environment.

[0071] Since the state of the system at similar time points can better reflect the current load pressure, the weighted expansion sum is constructed using exponentially increasing weights to amplify the changes at similar moments. The calculation expression is as follows:

[0072]

[0073] WQ(t) is the weighted expansion sum within the detection window, which represents the cumulative sum of the queue change rate at all times within the window. θ(i-n)It is an exponentially increasing weight, which is used to assign different weights to the change rate at different time points in the window. e is the natural base, and θ is the increasing rate adjustment coefficient, which controls the speed of weight decay.

[0074] Close time points refer to time points that are close to the current time point, and are close to the most recent data sampling moment in the time series. The states of these close time points are usually more valuable for reference because they can more accurately reflect the actual situation of the current system load. Since the state of the system changes dynamically over time, the data closer to the current moment can better reflect the current operating status and load trend of the system.

[0075] Exponential Increasing Weight is a strategy that assigns gradually increasing weights to different data points in a time series. In this scheme, the closer the data point is to the current time point, the greater the weight, reflecting its greater influence on the current system state. Compared with linear weight, exponential increasing weight is more sensitive to changes in a short period of time, amplifying the recent expansion rate, helping the system to quickly identify congestion and anomalies, so that the system can more effectively schedule resources and manage overloads.

[0076] The purpose of constructing the weighted queue expansion sum is to accurately quantify the overall trend of queue length changes and highlight the impact of changes in the near time points on the system load. In the transmission node, the queue expansion in the recent time period often better reflects the pressure and potential risks of the current system, while data far away from the current time point is less valuable for judging the load status. By using the weighted expansion sum, the impact of changes in the near time can be enhanced, and the upcoming overload state can be effectively captured, avoiding the system relying only on long-term average data, thereby improving the timeliness of response.

[0077] Autoregressive residual analysis is used to detect the abnormality of queue length changes. The residual (i.e., the error between the actual value and the fitted value) is calculated through the queue length data time series. The residual can capture short-term abnormal fluctuations in the system and is combined with the weighted inflation sum to generate the final queue inflation index. The calculation expression is as follows:

[0078]

[0079] , where QExpI is the queue expansion index, μ is the residual sensitivity adjustment coefficient, which controls the influence of the residual on the queue expansion index QExpI, ∈(t) is the residual, that is, the error between the actual value of the queue length and the trend fitting value. The residual calculation expression is as follows: ∈(t) = Q(t)-f(t), where f(t) is the trend fitting value;

[0080] Obtaining the trend fitting value f(t) fits historical data based on the time series analysis model to capture the long-term trend of queue length and exclude short-term fluctuations. The following are commonly used methods and their processes:

[0081] First, select a suitable time series model. Common models include autoregressive moving average model (ARMA), autoregressive integrated moving average model (ARIMA), weighted moving average model (WMA), or more complex deep learning models such as LSTM (Long Short-Term Memory Neural Network). These models can generate expected queue lengths at future time points by analyzing the trends and periodicity of time series data.

[0082] Secondly, the historical queue length data Q(t) is preprocessed, such as processing missing values, outliers, and smoothing the data to eliminate noise. Then, the data is input into the selected time series model for training or fitting. For simple models (such as ARIMA), the optimal model structure can be found by adjusting the parameters (p, d, q). For complex models (such as LSTM), larger-scale training data and computing resources are required to learn nonlinear patterns in time series by training the model.

[0083] After fitting, the model generates a trend value f(t) of the time series, which represents the expected queue length of the node at different time points. This fitted value is used to compare with the actual observed value Q(t) to detect short-term anomalies in the system and support overload detection and resource scheduling.

[0084] In the monitoring window, the calculation expression of the queue expansion index shows that the larger the value of the queue expansion index generated by analyzing the expansion speed of the length of the queue of data packets waiting to be processed in the transmission node, the greater the probability that the system is in high-load transmission during the period. High-load transmission usually means that the transmission node cannot process all incoming data packets in a short period of time, resulting in a rapid increase in the queue length. The higher the index value, the more serious the congestion in the network path, which may be caused by emergencies or traffic surges. For example, during an emergency, when multiple cameras transmit high-definition video at the same time and the node processing capacity and bandwidth reach their limits, data packets will accumulate, causing the queue expansion index to rise significantly. On the contrary, if the queue expansion index is low, it means that the data packets are processed in a timely manner, the network resources are sufficient, the transmission process is smoother, and the system is less likely to be in a high-load state.

[0085] An abnormal increase in the number of frequent changes in the state of a network link usually indicates that the video surveillance data transmission is in a high-load transmission state. This change reflects the dynamic instability of the link, which may be the result of link rerouting triggered by a surge in traffic leading to bandwidth resource exhaustion, node overload, network congestion, or frequent packet loss. Under high load conditions, the system frequently adjusts the transmission path or switches links to cope with sudden traffic demands, which causes the state of the network link to fluctuate frequently. At the same time, link instability may also mean insufficient processing power between nodes, resulting in packet queuing, increased latency, and even the need for retransmission. Frequent link switching and state changes will further increase the load on the network, because each switch will increase the time cost of transmission of control information and reconnection. This will not only reduce transmission efficiency, but may also cause video data jamming, delay, or loss, affecting the real-time and reliability of the video surveillance system.

[0086] In the monitoring window, the link fluctuation index is generated by analyzing the number of frequent changes in the network link status. The specific steps are as follows:

[0087] In the detection window, network monitoring tools are used to collect changes in multiple performance indicators in the link in real time. These indicators may include latency, packet loss rate, bandwidth usage, node response time, etc. Each state change will trigger a record with a timestamp and change magnitude.

[0088] The rate of change of the link status depends not only on the magnitude of the change, but also on the time interval between two changes. In order to capture frequent fluctuations in a short period of time, a time-sensitive factor is introduced to more accurately calculate the rate of each change. The calculation expression is as follows:

[0089]

[0090] , where R j represents the rate of change of the link at the jth state change, ΔS j is the magnitude of the state change, indicating the magnitude of the link state change in the jth state change event, T j is the timestamp of the jth change event, T j-1 is the timestamp of the j-1th change event, T j -T j-1 It represents the time interval between the jth and j-1th change events. P is a sensitive adjustment parameter of the time interval, which is used to control the influence of the time interval on the change rate. It is usually a positive value.

[0091] In order to reflect the cumulative and nonlinear characteristics of state changes, an exponential model is used to accumulate the rate of multiple state changes. This model can sensitively capture frequent changes and ensure that the cumulative effect is significant when the fluctuation amplitude is large. At the same time, each change is given a weight to reflect the importance of different types of changes. The calculation expression is as follows:

[0092]

[0093] , where E is the cumulative effect of state change, m is the total number of state changes, is the weight of the jth state change, ψ is the exponential amplification factor, which adjusts the sensitivity of the cumulative effect, and e is the natural base;

[0094] The cumulative effect E and the time characteristics are combined to calculate the final link fluctuation index. The calculation expression is as follows:

[0095]

[0096] , where LFI is the link fluctuation index, δ is the time adjustment factor used to control the impact of the time interval on the index, and ∈ is a constant, a very small constant to avoid the denominator being zero;

[0097] Under the monitoring window, the link fluctuation index calculation expression shows that the larger the value of the link fluctuation index generated by analyzing the number of frequent changes in the network link status, the higher the probability that the video surveillance data transmission is in high-load transmission, and vice versa. The link fluctuation index reflects the stability of the transmission link by analyzing the number of frequent changes in the network link status within the detection window. Under high-load transmission conditions, network resources (such as bandwidth and computing power) may be over-occupied, resulting in frequent link switching or rerouting to cope with the surge in instantaneous traffic and node overload. This frequent fluctuation in link status will increase the number of queuing, loss, and retransmission of data packets, further aggravating the system burden and affecting the real-time and reliability of transmission. On the contrary, when the link fluctuation index is small, it means that the network link is stable, the data transmission path is relatively smooth, and the resource usage is within a reasonable range. Therefore, the link fluctuation index is an important indicator for judging whether the video surveillance system is in a high-load transmission state. The higher the index value, the more dramatic the dynamic changes of the link and the greater the system pressure; the lower the index value, the more stable the transmission state, indicating that the current transmission load is light.

[0098] The machine learning model is not limited here, and any machine learning model that can analyze the queue expansion index QExpI and the link fluctuation index LFI to generate the resource occupancy variation coefficient RUVCoef can be used. In order to implement the technical solution of the present invention, the present invention provides a specific implementation method:

[0099] The calculation expression for the resource occupancy variation coefficient RUVCoef is as follows:

[0100]

[0101] , where k1 and k2 are the preset proportional coefficients of the queue expansion index QFxpI and the link fluctuation index LFI, respectively, and both k1 and k2 are greater than 0.

[0102] It can be seen from the calculation expression of the resource occupancy variation coefficient that under the detection window, the larger the value of the queue expansion index generated after analyzing the expansion speed of the length of the queue of data packets waiting to be processed in the transmission node, and the larger the value of the link fluctuation index generated after analyzing the frequent changes in the network link status, the larger the value of the resource occupancy variation coefficient generated under the detection window, indicating that the probability of video surveillance data transmission being in high-load transmission is greater, and vice versa, the smaller the probability of video surveillance data transmission being in high-load transmission is.

[0103] The transmission process division module divides the video surveillance data transmission process into different stages, including the normal transmission process and the overload transmission process, according to the results of the intelligent evaluation;

[0104] The resource occupancy change coefficient generated by analyzing the aggregated resource occupancy data under the monitoring window is compared and analyzed with the preset resource occupancy change coefficient reference threshold, and the video monitoring data transmission process is divided. The division steps are as follows:

[0105] If the resource occupancy change coefficient is greater than or equal to a preset resource occupancy change coefficient reference threshold, the video surveillance data transmission process is classified as an overload transmission process;

[0106] If the resource occupancy change coefficient is less than a preset resource occupancy change coefficient reference threshold, the video surveillance data transmission process is classified as a normal transmission process;

[0107] Normal transmission process: refers to the process of video surveillance data transmission, in which the use of system resources (such as bandwidth, computing power, storage) remains within the preset range, the network and nodes are stable, and data can be processed and transmitted normally. Overload transmission process: refers to the situation in which the system resource demand suddenly exceeds the preset threshold, resulting in insufficient bandwidth, increased latency, or node overload, resulting in reduced data transmission efficiency, video freeze or loss, and affecting the normal operation of the system;

[0108] The normal transmission management module manages the normal transmission process according to the pre-set resource allocation plan and continues to monitor the system in real time to ensure that resource usage remains within a reasonable range;

[0109] During normal transmission, the system is managed according to the pre-set resource allocation plan to ensure that the allocation of key resources such as bandwidth, computing power and storage resources meets the daily operation requirements. Although the system load is at a normal level at this time, in order to prevent the accumulation of potential problems or the occurrence of emergencies, continuous real-time monitoring is required. The monitoring tool will track the resource usage of each node and promptly record any deviation from the expected trend. If the resource usage is close to the set upper limit, the system will issue an alarm in advance to prompt the administrator to adjust the resource configuration to prevent overload;

[0110] The overload transmission control module uses the dynamic change information collected in real time to intelligently control resource allocation based on the pre-set resource allocation plan in the overload transmission process through the queuing theory model;

[0111] For the overload transmission process, the process of intelligent resource regulation and resource allocation based on the pre-set resource allocation plan is as follows, using the dynamic change information collected in real time and the queuing theory model:

[0112] In the overload state, it is necessary to calculate the amount of resources that need to be increased based on the real-time queue expansion index QExpI and link fluctuation index LFI to ensure that the processing capacity meets the demand. This process adopts a multi-objective optimization model to balance the node processing capacity and link quality. The calculation expression is as follows:

[0113]

[0114] , where ΔR is the amount of resources that need to be increased, such as bandwidth, CPU, storage space, etc., QExpI is the queue expansion index, QExpI threshold is the preset reference threshold of the queue expansion index, LFI is the link fluctuation index, threshold is the preset reference threshold of the link fluctuation index, α and β are weight coefficients, which respectively control the impact of the queue expansion index QExpI and the link fluctuation index LFI on resource regulation;

[0115] According to the calculated resource control amount ΔR, update the resource allocation of nodes and links. At the same time, the new resource quota should not only cover the current load, but also reserve redundancy to cope with further burst traffic. The calculation expression is as follows: R new =R allocated +ΔR, where R new is the adjusted resource quota, that is, the total amount of resources required by the system under overload conditions, R allocated It is a pre-set resource allocation plan;

[0116] By increasing resource quotas, we ensure that the system can quickly process backlogged data packets and improve the latency and jitter of the transmission link.

[0117] Resource allocation is subject to the physical limitations of the system (such as maximum bandwidth, number of CPU cores, etc.). Therefore, it is necessary to optimize the calculated new resource quota to prevent it from exceeding the system capacity and ensure that resource allocation does not exceed the physical limit of the system to avoid insufficient resources for other services due to over-allocation. The calculation expression is as follows: R final =min(R new ,R max ), where R final is the actual amount of resources allocated, ensuring that the physical resource limit is not exceeded, R max is the maximum amount of resources that can be provided (such as maximum bandwidth, number of CPU cores, etc.);

[0118] After the overload state is released, the system needs to gradually release excess resources to prevent long-term resource occupation. The calculation expression is as follows:

[0119]

[0120] , where R release is the amount of resources after release, w1 and w2 are weight coefficients for controlling resource release, which are used to regulate the impact of queue expansion index QExpI and link fluctuation index LFI respectively.

[0121] The present invention uses real-time monitoring tools and performance monitoring software, combined with machine learning models, to intelligently evaluate resource occupancy, thereby effectively improving the dynamic regulation capability of the system. In the face of sudden security incidents, the status of transmission nodes can be monitored in real time, especially through the queue expansion index and link fluctuation index, to identify the trend of overload transmission in advance. Using the queuing theory model, the length of the queue of data packets waiting to be processed in the node can be regulated to ensure that the system can effectively manage resource allocation under high load, reduce transmission delays, and increase the speed of decoding and analysis, so as to identify security threats in a timely manner. This significantly reduces the delay problem caused by unreasonable resource allocation, shortens the emergency response time, avoids the potential safety hazards caused by lagging video analysis at critical moments, and reduces the risk of casualties and property losses.

[0122] The present invention can perform intelligent resource scheduling for different transmission states by dividing the transmission process. In the normal transmission stage, management is performed according to the preset resource allocation plan to avoid over-allocation of resources when they are not needed, thereby improving resource utilization efficiency. In the overload transmission stage, resources are temporarily optimized and configured through real-time dynamic information combined with intelligent control means. In this process, the resource occupancy change coefficient is continuously analyzed to ensure that resources can be allocated in time when sudden demand increases, and that excess resources are quickly released after demand slows down. This flexible resource management mechanism not only ensures efficient operation, but also avoids idleness and waste of resources due to fixed allocation, thereby improving the overall resource utilization efficiency and ensuring that the system can remain stable and reliable under high load conditions.

[0123] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A 5G communication data transmission management system, characterized in that: It includes real-time resource monitoring module, intelligent evaluation module and feature extraction module, transmission process division module, normal transmission management module and overload transmission control module: The real-time resource monitoring module monitors the resource usage during the transmission of video surveillance data in real time through the deployed network monitoring tools and performance monitoring software; The intelligent evaluation module and feature extraction module extract features from the aggregated resource occupancy data and use the pre-learned machine learning model to intelligently evaluate changes in the video surveillance data transmission process based on the extracted resource occupancy data features. The transmission process division module divides the video surveillance data transmission process into different stages, including the normal transmission process and the overload transmission process, according to the results of the intelligent evaluation; The normal transmission management module manages the normal transmission process according to the pre-set resource allocation plan and continues to monitor the system in real time to ensure that resource usage remains within a reasonable range; The overload transmission control module uses the dynamic change information collected in real time to intelligently control resource allocation based on the pre-set resource allocation plan in the overload transmission process through the queuing theory model.

2. A 5G communication data transmission management system according to claim 1, characterized in that: Feature extraction is performed on the summarized resource occupancy data. The extracted features include the expansion speed of the length of the queue of data packets waiting to be processed in the transmission node and the frequent changes in the network link status. After acquisition, the expansion speed of the length of the queue of data packets waiting to be processed in the transmission node and the frequent changes in the network link status are analyzed under the monitoring window, and a queue expansion index and a link fluctuation index are generated respectively. The queue expansion index and the link fluctuation index are input into a pre-learned machine learning model, and a resource occupancy variation coefficient is generated through the machine learning model. The summarized resource occupancy data is comprehensively evaluated through the resource occupancy coefficient.

3. A 5G communication data transmission management system according to claim 2, characterized in that: In the monitoring window, the expansion speed of the queue length of the data packets waiting to be processed in the transmission node is analyzed to generate the queue expansion index. The specific steps are as follows: In the detection window, the queue length data of the transmission node is collected, and the collected queue length data is calibrated as the queue length data time series Q(t), where Q(t) = {t i ,q i }={(t1, q1), (t2, q2), (t3, q3), ..., (t n ,q n )}, where Q(t) is the collected queue length time series, t is the timestamp, and t i is the timestamp of the i-th time point, a timestamp t i Marks a time point i, q in the time series i is the queue length at the ith time point, and n is the total number of time points; To measure the speed at which the queue length changes at different time points, a discrete differential equation is used to calculate the instantaneous rate of change. The instantaneous rate of change is used to capture the rapid change of the queue length in a short period of time and to perform sensitive detection of processing capacity bottlenecks or packet backlogs. The calculation expression is as follows: ， In the formula, ΔQ(t i ) is between two adjacent timestamps t i and t i+1 The queue length change rate between i+1 and q i is the queue length between time points i and i+1, t i and t i+1 are two adjacent timestamps; Since the state of the system at similar time points can better reflect the current load pressure, the weighted expansion sum is constructed using exponentially increasing weights to amplify the changes at similar moments. The calculation expression is as follows: WQ(t) is the weighted expansion sum within the detection window, which represents the cumulative sum of the queue change rate at all times within the window. θ(i-n) It is an exponentially increasing weight, which is used to assign different weights to the change rate at different time points in the window. e is the natural base, and θ is the increasing rate adjustment coefficient, which controls the speed of weight decay. Autoregressive residual analysis is used to detect the abnormality of queue length changes. The residual is calculated through the queue length data time series. The residual can capture the short-term abnormal fluctuations in the system. It is combined with the weighted inflation sum to generate the final queue inflation index. The calculation expression is as follows: ， Where QExpI is the queue expansion index, μ is the residual sensitivity adjustment coefficient, which controls the influence of the residual on the queue expansion index QExpI, ∈(t) is the residual, that is, the error between the actual value of the queue length and the trend fitting value. The residual calculation expression is as follows: ∈(t) = Q(t) - f(t), where f(t) is the trend fitting value.

4. A 5G communication data transmission management system according to claim 2, characterized in that: In the monitoring window, the link fluctuation index is generated by analyzing the number of frequent changes in the network link status. The specific steps are as follows: In the detection window, the network monitoring tool collects the changes of multiple performance indicators in the link in real time. Each state change triggers a record with a timestamp and change range. The rate of change of the link status depends not only on the magnitude of the change, but also on the time interval between two changes. In order to capture frequent fluctuations in a short period of time, a time-sensitive factor is introduced to more accurately calculate the rate of each change. The calculation expression is as follows: ， In the formula, R j represents the rate of change of the link at the jth state change, ΔS j is the magnitude of the state change, indicating the magnitude of the link state change in the jth state change event, T j is the timestamp of the jth change event, T j-1 is the timestamp of the j-1th change event, T j -T j-1 represents the time interval between the jth and j-1th change events, and P is a sensitive adjustment parameter of the time interval; In order to reflect the cumulative and nonlinear characteristics of state changes, an exponential model is used to accumulate the rates of multiple state changes. At the same time, each change is given a weight to reflect the importance of different types of changes. The calculation expression is as follows: ， Where E is the cumulative effect of state change, m is the total number of state changes, is the weight of the jth state change, ψ is the exponential amplification factor, which adjusts the sensitivity of the cumulative effect, and e is the natural base; The cumulative effect E and the time characteristics are combined to calculate the final link fluctuation index. The calculation expression is as follows: 。 Where LFI is the link fluctuation index, δ is the time adjustment factor, and ∈ is a constant.

5. A 5G communication data transmission management system according to claim 2, characterized in that: The resource occupancy change coefficient generated by analyzing the aggregated resource occupancy data under the monitoring window is compared and analyzed with the preset resource occupancy change coefficient reference threshold, and the video monitoring data transmission process is divided. The division steps are as follows: If the resource occupancy change coefficient is greater than or equal to a preset resource occupancy change coefficient reference threshold, the video surveillance data transmission process is classified as an overload transmission process; If the resource occupancy variation coefficient is less than a preset resource occupancy variation coefficient reference threshold, the video surveillance data transmission process is classified as a normal transmission process.

6. A 5G communication data transmission management system according to claim 5, characterized in that: For the overload transmission process, the process of intelligent resource regulation and resource allocation based on the pre-set resource allocation plan is as follows, using the dynamic change information collected in real time and the queuing theory model: In the overload state, it is necessary to calculate the amount of resources that need to be increased based on the real-time queue expansion index QExpI and link fluctuation index LFI to ensure that the processing capacity meets the demand. This process adopts a multi-objective optimization model to balance the node processing capacity and link quality. The calculation expression is as follows: 。 Where ΔR is the amount of resources that need to be increased, QExpI is the queue expansion index, and QExpI threshold is the preset reference threshold of the queue expansion index, LFI is the link fluctuation index, LFI threshold is the preset reference threshold of the link fluctuation index, α and β are weight coefficients, which respectively control the impact of the queue expansion index QExpI and the link fluctuation index LFI on resource regulation; According to the calculated resource control amount ΔR, update the resource allocation of nodes and links. At the same time, the new resource quota should not only cover the current load, but also reserve redundancy to cope with further burst traffic. The calculation expression is as follows: R new =R allocated +ΔR, where R new is the adjusted resource quota, R allocated It is a pre-set resource allocation plan; By increasing resource quotas, we ensure that the system can quickly process backlogged data packets and improve the latency and jitter of the transmission link. Resource allocation is subject to the physical limitations of the system. Therefore, it is necessary to optimize the calculated new resource quota to prevent it from exceeding the system capacity and ensure that resource allocation does not exceed the physical limit of the system to avoid insufficient resources for other services due to over-allocation. The calculation expression is as follows: R final =min(R new ,R max ), where R final is the actual amount of resources allocated, R max is the maximum amount of resources that can be provided; After the overload state is released, the system needs to gradually release excess resources to prevent long-term resource occupation. The calculation expression is as follows: 。 In the formula, R release is the amount of resources after release, w1 and w2 are weight coefficients for controlling resource release, which are used to regulate the impact of queue expansion index QExpI and link fluctuation index LFI respectively.

Images