Communication system and method based on multi-source data collaboration and intelligent decision, computer device and storage medium

Abstract

The invention discloses a communication system and method based on multi-source data collaboration and intelligent decision, a computer device and a storage medium, and the method comprises the steps: generating a feature vector according to underlying performance index information and business volume index information collected from a wireless access network, operating a neural network model to process the feature vector, and obtaining strategy instruction information; and executing resource scheduling in response to the strategy instruction information. Multi-source data such as bottom layer performance index information from a physical layer and business volume index information from a link layer are integrated, strategy instruction information is generated through an intelligent decision-making mechanism based on a neural network model, more comprehensive and efficient network optimization is achieved, and meanwhile data safety and strategy reliability are remarkably improved; according to the invention, the physical layer and transmission layer performance of the wireless access network can be monitored in real time and collaboratively optimized, the efficient utilization and dynamic adjustment of network resources are ensured, and the high-performance requirement in a complex network environment is met. The method is widely applied to the technical field of communication.

Description

Technical Field

[0001] The present invention relates to the field of communication technologies, and in particular, to a communication system, method, computer device, and storage medium based on multi-source data collaboration and intelligent decision-making. Background Art

[0002] In wireless communication technologies, user equipment accesses the communication core network by connecting to a radio access network. With the application and popularization of wireless communication technologies, the number of user equipment has increased significantly, that is, the deployment scale of the communication system has increased, and in addition to user equipment such as mobile phones, more and more types of user equipment such as smart watches and smart bracelets are connected, making the network environment complex. The current radio access network is prone to problems when facing large-scale deployment and complex network environments. For example, traditional radio access networks usually adopt a distributed base station architecture. In this architecture, each base station independently processes the transmission of wireless signals and resource management, and there are problems such as low resource utilization efficiency, high operation and maintenance costs, and difficulty in achieving flexible scheduling in large-scale deployment and complex network environments. Summary of the Invention

[0003] Aiming at the technical problems of low resource utilization efficiency, high operation and maintenance costs, and difficulty in achieving flexible scheduling existing in the current communication system, the purpose of the present invention is to provide a communication system, method, computer device, and storage medium based on multi-source data collaboration and intelligent decision-making.

[0004] On the one hand, an embodiment of the present invention includes a communication method based on multi-source data collaboration and intelligent decision-making. The communication method based on multi-source data collaboration and intelligent decision-making includes the following steps:

[0005] A first functional module for collecting underlying performance index information from a radio access network;

[0006] A second functional module for collecting traffic volume index information from the radio access network;

[0007] A third functional module for generating a feature vector according to the underlying performance index information and the traffic volume index information;

[0008] A fourth functional module for running a neural network model to process the feature vector to obtain policy instruction information;

[0009] The second functional module is used to execute resource scheduling in response to the policy instruction information.

[0010] Further, the generating a feature vector according to the underlying performance index information and the traffic volume index information includes:

[0011] Setting the size of a time window;

[0012] Slide the time window so that the time window reaches multiple time positions;

[0013] For any determined time position of the time window, perform statistical feature calculations on the underlying performance metric information and the traffic volume metric information in the time window to obtain the coefficient of variation of the physical resource block utilization rate, the 95th percentile value of the delay metric, and the traffic flow autocorrelation parameter corresponding to the time position, and form a feature vector corresponding to the time position with the coefficient of variation of the physical resource block utilization rate, the 95th percentile value of the delay metric, and the traffic flow autocorrelation parameter.

[0014] Further, the running neural network model processes the feature vector to obtain policy instruction information, including:

[0015] Input the feature vector into the running neural network model for processing;

[0016] Obtain the policy recommendation information output by the running neural network model; the content of the policy recommendation information belongs to one of multiple levels;

[0017] Determine the service level agreement according to the level corresponding to the policy recommendation information;

[0018] Use the service level agreement to verify the policy recommendation information;

[0019] Generate the policy instruction information according to the verification result.

[0020] Further, the fourth functional module is used to perform success rate simulation verification on the policy instruction information through the digital twin engine. When the verification passes, the policy instruction information is sent to the second functional module. When the verification fails, the policy instruction information is regenerated;

[0021] The fourth functional module is used to encode and compress the policy instruction information by ASN.1 PER and send it to the second functional module through a 5QI = 80 high-priority channel.

[0022] Further, the communication system based on multi-source data collaboration and intelligent decision-making further includes a fifth functional module;

[0023] The fifth functional module is used to encrypt and forward the underlying performance metric information collected by the first functional module to the third functional module, collect status feedback information through radio resource control measurement reports, form a policy effect evaluation data set, and send the policy effect evaluation data set to the fourth functional module;

[0024] The fourth functional module is used to dynamically adjust the parameters of the neural network model according to the policy effect evaluation data set.

[0025] Further, the communication system based on multi-source data collaboration and intelligent decision-making further includes a sixth functional module, a seventh functional module, and an eighth functional module;

[0026] The sixth functional module is used to maintain clock synchronization between the first functional module and the second functional module;

[0027] The seventh functional module is used to perform edge computing and adjust the policy instruction information;

[0028] The eighth functional module is used to monitor the execution process of the policy instruction information. When the execution of the policy instruction information is detected to be abnormal, multi-level rollback is performed.

[0029] Further, the first functional module is a CU, the second functional module is a UPF, the third functional module is a NWDAF, the fourth functional module is a PCF, the fifth functional module is an AMF, the sixth functional module is an SMF, the seventh functional module is a NEF, and the eighth functional module is a SEPP.

[0030] On the other hand, an embodiment of the present invention further includes a communication method based on multi-source data collaboration and intelligent decision-making. The communication method based on multi-source data collaboration and intelligent decision-making includes:

[0031] Collecting underlying performance index information from a radio access network;

[0032] Collecting traffic index information from the radio access network;

[0033] Generating a feature vector according to the underlying performance index information and the traffic index information;

[0034] Running a neural network model to process the feature vector to obtain policy instruction information;

[0035] Responding to the policy instruction information and performing resource scheduling.

[0036] On the other hand, an embodiment of the present invention further includes a computer device, including a memory and a processor. The memory is used to store at least one program, and the processor is used to load at least one program to execute the communication method based on multi-source data collaboration and intelligent decision-making in the embodiment.

[0037] On the other hand, an embodiment of the present invention further includes a computer-readable storage medium, in which a program executable by a processor is stored. The program executable by the processor is used to execute the communication method based on multi-source data collaboration and intelligent decision-making in the embodiment when executed by the processor.

[0038] The beneficial effects of the present invention are as follows: The communication system based on multi-source data collaboration and intelligent decision-making in the embodiments can integrate multi-source data such as the underlying performance index information from the physical layer and the traffic index information from the link layer, generate policy instruction information through an intelligent decision-making mechanism based on a neural network model, achieve more comprehensive and efficient network optimization, and significantly improve data security and policy reliability. Specifically, the communication system based on multi-source data collaboration and intelligent decision-making in the present embodiment can monitor and collaboratively optimize the performance of the physical layer and the transport layer of the radio access network in real time, ensure the efficient utilization and dynamic adjustment of network resources, and meet the high-performance requirements in complex network environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a schematic diagram of the communication system based on multi-source data collaboration and intelligent decision-making in the embodiments and the process it executes;

[0040] Figure 2 It is a schematic diagram of the steps of the communication method based on multi-source data collaboration and intelligent decision-making in the embodiments;

[0041] Figure 3 It is a schematic diagram of the principle of the communication method based on multi-source data collaboration and intelligent decision-making in the embodiments. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0042] Term Explanation:

[0043] 1. Radio Access Network (RAN): The part of a mobile communication system that connects the UE to the core network, responsible for the transmission and processing of wireless signals. Specifically, it includes base stations, a Central Unit (CU), and a Distributed Unit (DU), etc., and is a key link for users to access the network;

[0044] 2. Central Unit (CU): A functional module in the Radio Access Network (RAN), responsible for centralized processing of radio resource management, data collection, and some control functions. It communicates with other network units through interfaces to achieve efficient network management and resource allocation;

[0045] 3. Physical Resource Block Utilization: An important indicator for measuring the efficiency of wireless resource usage, representing the proportion of physical resource blocks (PRBs) resources used in the network. It is used to evaluate the resource allocation situation and performance status of the network;

[0046] 4. Channel Quality Indicator (CQI): A parameter measured by a User Equipment (UE) and reported to the base station, which is used to reflect the transmission quality of the wireless channel. It provides a basis for wireless resource allocation and scheduling to help optimize network performance;

[0047] 5. Scheduling queue depth: The number of data packets waiting to be processed during wireless scheduling. It reflects the data backlog in the network and is an important indicator for measuring network scheduling performance and resource utilization efficiency;

[0048] 6. AMF (Access and Mobility Management Function): A functional module in the 5G core network responsible for access and mobility management. As a transit node, it processes functions such as access requests, authentication, and mobility management of user equipment;

[0049] 7. NWDAF (Network Data Analytics Function): A network data analysis functional module in the 5G core network, which is responsible for decrypting, processing, and analyzing the collected network data to provide data support for network optimization and policy generation;

[0050] 8. UPF (User Plane Function): A user plane functional module in the 5G core network, which is responsible for forwarding and processing user data. It provides data support for network performance optimization by monitoring transport layer metrics at the service flow level (such as latency and packet loss rate);

[0051] 9. SMF (Session Management Function): A session management functional module in the 5G core network, which is responsible for functions such as session management, resource allocation, and clock synchronization. It ensures the time synchronization of multi-source data by deploying a precise clock synchronization module;

[0052] 10. PCF (The Policy Control Function): A policy control functional module in the 5G core network, whose main function is to implement user control policy management, including QOS control, service access control, etc.;

[0053] 11. NEF (Network Exposure Function): A network exposure functional module in the 5G core network, which is mainly used to provide network information and services to third-party applications, realize the friendly docking of network capabilities and business requirements, improve the business experience, and optimize network resource configuration;

[0054] 12. Digital Twin Engine: A virtualization technology for network simulation and policy verification. By constructing a refined network model, it simulates the actual network operating state, evaluates the success probability of policies, and ensures the feasibility and effectiveness of policies.

[0055] 13. BiLSTM (Bidirectional Long Short-Term Memory): An improved long short-term memory neural network structure. Through a bidirectional processing mechanism and an attention mechanism, it can analyze feature vectors in real time and generate targeted policy recommendations for in-depth analysis of network performance data and generation of optimization policies.

[0056] 14. SLA (Service Level Agreement): A service quality agreement reached between network service providers and users. It defines key indicators of network performance (latency, throughput) and service quality requirements, and is used to evaluate whether the network meets user needs.

[0057] 15. ATSSS (Access Traffic Steering, Switching and Splitting): An access traffic guidance, switching and splitting mechanism used to dynamically adjust traffic paths, direct latency-sensitive services to edge nodes, and optimize network resource allocation and performance.

[0058] 16. MEC (Multi-access Edge Computing): A technology that deploys computing and storage resources at the network edge, used to reduce latency, relieve the pressure on the core network, and support real-time data processing and local services.

[0059] 17. SEPP (Security Edge Protection Proxy): A security edge protection proxy used to audit and security monitor network resource adjustment operations. It ensures that all operations comply with predefined security policies and initiates a rollback mechanism when anomalies are detected to ensure system security.

[0060] In this embodiment, a communication system based on multi-source data collaboration and intelligent decision-making is provided. The overall structure of the communication system based on multi-source data collaboration and intelligent decision-making and the processes executed by each component therein are as Figure 1 shown. Figure 1 In the structure shown, only some components and some or all of the processes they execute are required to achieve flexible resource scheduling.

[0061] Refer to Figure 1, A communication system based on multi-source data collaboration and intelligent decision-making includes functional modules such as a first functional module, a second functional module, a third functional module, a fourth functional module, a fifth functional module, a sixth functional module, a seventh functional module, and an eighth functional module. In this embodiment, the first functional module is the CU, the second functional module is the UPF, the third functional module is the NWDAF, the fourth functional module is the PCF, the fifth functional module is the AMF, the seventh functional module is the NEF, and the eighth functional module is the SEPP as an example for illustration. Among them, the first functional module (CU) is a part of the radio access network. The first functional module (CU) is connected to user equipment such as mobile phones and is also connected to the communication core network; the second functional module (UPF), the third functional module (NWDAF), the fourth functional module (PCF), the fifth functional module (AMF), the seventh functional module (NEF), the eighth functional module (SEPP), etc. are parts of the communication core network.

[0062] Refer to Figure 1 , When the communication system based on multi-source data collaboration and intelligent decision-making runs, each functional module executes corresponding processes. Among them, the processes executed by the first functional module (CU), the second functional module (UPF), the third functional module (NWDAF), and the fourth functional module (PCF) can implement the basic functions of the communication system based on multi-source data collaboration and intelligent decision-making.

[0063] Specifically, refer to Figure 1 , The first functional module (CU) in the radio access network (RAN) starts each execution process. The first functional module (CU) executes Process 1 and collects slice-level underlying performance metric information at a certain period (for example, 50 ms) through a standardized E1 interface. The specific underlying performance metric information to be collected includes physical resource block (PRB) utilization rate, scheduling queue depth, and channel quality indicator (CQI) distribution, etc. Therefore, the underlying performance metric information collected by the first functional module (CU) can be in the form of a time series. The underlying performance metric information can be used to monitor the physical layer performance of the communication system in real time and provide basic data support for subsequent policy generation.

[0064] Refer to Figure 1 , The first functional module (CU) performs lightweight encryption processing on the collected underlying performance metric information through an encryption algorithm (such as AES-128) and transmits it to the fifth functional module (AMF) through the N2 interface of the 5G core network. The fifth functional module (AMF), as a relay node, does not decrypt the data and executes Process 2, directly forwarding the encrypted data to the third functional module (NWDAF), and the third functional module (NWDAF) performs subsequent decryption and processing.

[0065] Meanwhile, the second functional module (UPF) executes Process 3, uses the Packet Forwarding Control Protocol (PFCP) to monitor in real time the transport layer metrics at the traffic flow level of the Radio Access Network (RAN), i.e., the traffic volume metric information, where the traffic volume metric information specifically includes information such as end-to-end delay, packet loss rate, and throughput fluctuation, and the traffic volume metric information can also be in the form of a time series. The second functional module (UPF) reports the traffic volume metric information to the sixth functional module (SMF) through the N4 interface.

[0066] Refer to Figure 1 , to ensure the time synchronization of multi-source data (including underlying performance metric information and traffic volume metric information), the sixth functional module (SMF) deploys a Precision Clock Synchronization Module as the primary clock source. The sixth functional module (SMF) executes Process 4 and Process 5, and performs clock synchronization with the CU and UPF through the IEEE 1588v2 protocol. Specifically, the sixth functional module (SMF) acts as the primary clock, sends a synchronization message (Sync message) with a timestamp to the second functional module (UPF) through the N4 interface, and receives the delay response message (Delay_Resp message) returned by the second functional module (UPF) to calculate the transmission delay and correct the time deviation. Meanwhile, the sixth functional module (SMF) indirectly transmits the time synchronization signal to the first functional module (CU) through the fifth functional module (AMF), ensuring that the physical layer data (underlying performance metric information) collected by the first functional module (CU) and the transport layer data (traffic volume metric information) collected by the second functional module (UPF) are precisely aligned in the time dimension.

[0067] Refer to Figure 1 , the sixth functional module (SMF) forwards the traffic volume metric information sent by the second functional module (UPF) to the third functional module (NWDAF).

[0068] By executing Process 1 - 6, the raw data (underlying performance metric information and traffic volume metric information) from the first functional module (CU) and the second functional module (UPF) are aggregated to the third functional module (NWDAF) through the fifth functional module (AMF) and the sixth functional module (SMF).

[0069] The third functional module (NWDAF) executes Process 7. First, it decrypts the received encrypted data to restore the original content of the data. Then, the third functional module (NWDAF) generates a feature vector based on the decrypted underlying performance metric information and traffic volume metric information.

[0070] Specifically, the third functional module (NWDAF) runs the built-in stream processing engine to perform windowed analysis on the decrypted multi-source data (underlying performance metric information and traffic volume metric information) received. The stream processing engine in the third functional module (NWDAF) can set the size of a time window with a fixed size (e.g., 1 s). The time point of the start (or end) of this time window is the time position of the time window, and the time window can be slid so that the time window reaches multiple time positions. For the underlying performance metric information and traffic volume metric information in the form of a time series, no matter what the time position of the time window is, there is always a part of the information within the time window. Suppose the time window slides to time positions T1, T2, T3... respectively. Then, the parts of the underlying performance metric information and traffic volume metric information corresponding to T1, T2, T3... are obtained respectively. For the underlying performance metric information and traffic volume metric information within the time window at the time position T1, statistical feature calculations are performed to obtain the coefficient of variation of the physical resource block utilization rate corresponding to the time position T1 (by calculating the mean and standard deviation of the PRB utilization rate, the coefficient of variation is obtained, which reflects the fluctuation of the PRB utilization rate), the 95th percentile value of the delay metric (P95; by sorting the delay data and calculating the P95 position, the distribution of high-delay events is obtained), and the autocorrelation parameter of the traffic flow (by calculating the autocorrelation function of the time series, the periodicity and dependence parameters of the traffic flow are obtained). These data form the feature vector corresponding to the time position T1, and its form is:

[0071] [Coefficient of variation of PRB, 95th percentile value of delay, autocorrelation of traffic flow]

[0072] According to the same principle, the feature vectors corresponding to each time position of T2, T3... can be calculated respectively. Multiple feature vectors can also form the form of a time series. These feature vectors can comprehensively reflect the operating state and performance metrics of the communication system, providing rich data support for subsequent intelligent decision-making analysis.

[0073] Refer to Figure 1 , the third functional module (NWDAF) sends the generated feature vectors to the fourth functional module (PCF), and the fourth functional module (PCF) executes processes 8-10 to generate policy instruction information according to the feature vectors.

[0074] Specifically, the fourth functional module (PCF) executes process 8, integrating an intelligent decision-making module based on a bidirectional long short-term memory (BiLSTM) neural network model for in-depth analysis of network performance data and policy generation. BiLSTM is an improved long short-term memory (LSTM) network structure that can be used to process sequential data. The training data of the BiLSTM model comes from historical network performance data and optimized policy samples annotated by experts. The training objective is to minimize the difference between the policy recommendations and the actual optimization effects. Through the bidirectional processing mechanism and attention mechanism, the BiLSTM model can analyze feature vectors in real time and generate targeted policy recommendation information.

[0075] In this embodiment, the content of the policy recommendation information may belong to multiple levels, that is, the content of a specific policy recommendation information output by the BiLSTM model corresponds to one of the multiple levels. Specifically, the content and levels of the policy recommendation information are shown in Table 1.

[0076] Table 1

[0077]

[0078]

[0079] Among them, the larger the number representing the level size, the higher the level.

[0080] For the policy recommendation information output by the BiLSTM model, the fourth functional module (PCF) executes process 9, combining the policy recommendation information with the service level agreement (SLA) template to dynamically generate specific policy instruction information. Specifically,

[0081] Table 2

[0082]

[0083]

[0084] When the fourth functional module (PCF) executes process 9, for a specific policy recommendation information, according to the level size of the policy recommendation information, it selects the corresponding service level agreement (SLA) from Table 2 and performs a judgment, and generates policy instruction information with corresponding content or performs corresponding operations according to the satisfaction of the service level agreement (SLA).

[0085] When the strategic instruction information is generated in the execution process 9 of the fourth functional module (PCF), to ensure the feasibility of the strategic instruction information, the fourth functional module (PCF) can also execute process 10 to conduct a success rate simulation verification on the strategic instruction information through the digital twin engine. Specifically, the digital twin engine constructs a refined simulation model that includes the core network signaling processing delay. This model can highly reproduce the operating state and behavior of the actual network. By applying the Monte Carlo method for a hundred iterations of simulation, the digital twin engine can evaluate the success probability of the strategic instruction information. The Monte Carlo method is a statistical simulation method based on random sampling, which estimates the target statistic or expected value through the simulation results of a large number of random samples.

[0086] The digital twin engine evaluates the success probability of the strategic instruction information generated in the execution process 9 of the fourth functional module (PCF). If the evaluated success execution probability of the strategic instruction information exceeds 98%, the digital twin engine marks the strategic instruction information as passed the verification; otherwise, it is marked as failed the verification, and information on the verification failure is fed back. For example, the fourth functional module (PCF) can combine the latest network performance data, re-evaluate the network state, and adjust the level division and content setting of the strategic recommendation information. According to the new level of the strategic recommendation information and the SLA threshold, new strategic instruction information is regenerated, and the regenerated strategic instruction information is sent to the digital twin engine for verification until the success execution probability of the strategic instruction information exceeds 98%. Finally, the failed verification strategies and their reasons are recorded, and the relevant information is fed back to the neural network model for optimizing the training data and parameters of the neural network model.

[0087] If the execution process 10 of the fourth functional module (PCF) evaluates that the strategic instruction information passes the verification, then the fourth functional module (PCF) executes process 11. After encoding and compressing the strategic instruction information by ASN.1 PER, it is sent to the second functional module (UPF) through the 5QI = 80 high-priority channel. The second functional module (UPF) executes process 12 and performs resource scheduling in response to the strategic instruction information.

[0088] When the second functional module (UPF) executes process 12, it adopts the ASN.1 PER (Packed Encoding Rules) encoding and compression technology to control the size of each strategic instruction information within 500 bytes. This encoding method can significantly reduce the transmission overhead of the message and improve the transmission efficiency. Subsequently, the strategic instruction information is transmitted through the dedicated data bearer channel with 5QI (QoS Indicator) = 80 and sent down to the first functional module (CU). 5QI = 80 indicates that this data bearer has the characteristics of high priority and low latency, ensuring that the strategic instruction can reach the execution node quickly and reliably.

[0089] After the policy instruction information is sent down, the first functional module (CU) adjusts the scheduling algorithm of the MAC layer according to the policy instruction information.

[0090] The specific adjustment methods include:

[0091] 1. Adjust the service priority weights in the weighted fair queue (WFQ): By adjusting the weights in the WFQ, optimize the scheduling order of different services to ensure that high-priority services can obtain more resource allocations, thereby improving the overall network performance and service quality.

[0092] 2. Activate the fast frequency-selective handover mechanism based on the channel state information (CSI) report: The first functional module (CU) dynamically allocates spectrum resources according to the CSI report reported by the user equipment (UE). This mechanism can adjust the spectrum allocation in real time according to the channel quality, improve the transmission efficiency, and reduce the performance degradation caused by poor channel quality.

[0093] The second functional module (UPF) can monitor the execution process of Process 12. If the second functional module (UPF) detects local network congestion between the first functional module (CU) and the user equipment (UE) etc. (specifically, it can be judged by traffic and delay metrics) during real-time monitoring, then the second functional module (UPF) can trigger the access traffic steering, switching, and splitting (ATSSS) mechanism of the sixth functional module (SMF). This mechanism dynamically steers delay-sensitive services to the edge UPF to ensure the smoothness of critical tasks. When the end-to-end delay exceeds 20 milliseconds or the packet loss rate exceeds 1%, the ATSSS mechanism will switch the relevant traffic to the nearest edge UPF instance, and at the same time, through the dynamic weight calculation of DNS redirection, access the computing task to the geographically nearest MEC node. Meanwhile, the core network works in cooperation with the software-defined network (SDN) controller through the border gateway protocol link state (BGP-LS), dynamically adjusts the FlexE channel bandwidth in 100 Mbps steps, and applies differentiated service code point (DSCP) markings for different service flows to ensure the efficient utilization of network bandwidth. The SDN controller dynamically adjusts the network topology according to real-time traffic and policy instructions to optimize the data transmission path.

[0094] In this embodiment, through Figure 1The communication system based on multi-source data collaboration and intelligent decision-making and the processes 1-12 executed thereby, especially the processes executed by the first functional module (CU), the second functional module (UPF), the third functional module (NWDAF), and the fourth functional module (PCF), can integrate multi-source data such as the underlying performance metric information from the physical layer and the traffic metric information from the link layer, generate policy instruction information through an intelligent decision-making mechanism based on a neural network model, achieve more comprehensive and efficient network optimization, and significantly improve data security and policy reliability at the same time; specifically, the communication system based on multi-source data collaboration and intelligent decision-making in this embodiment can monitor and collaboratively optimize the performance of the physical layer and the transport layer of the radio access network in real time, ensure the efficient utilization and dynamic adjustment of network resources, and meet the high-performance requirements in complex network environments.

[0095] In this embodiment, the seventh functional module (NEF) can execute process 13 for edge computing to adjust the policy instruction information generated by the fourth functional module (PCF). Specifically, the seventh functional module (NEF) uses the common API framework (CAPIF) to adjust the policy instruction information. Specifically, the seventh functional module (NEF) can make corresponding adjustments according to Table 3 based on the level of the policy instruction information.

[0096] Table 3

[0097]

[0098] Refer to Figure 1 , after the second functional module (UPF) completes the implementation of the policy instruction information by executing process 12, the fifth functional module (AMF) executes process 14 to collect status feedback information through radio resource control (RRC) measurement reports, focusing on monitoring the reference signal received power (RSRP) fluctuations and the hybrid automatic repeat request (HARQ) retransmission rate. At the same time, the fifth functional module (AMF) uses the minimized drive test (MDT) function to collect deep coverage data, which will form a policy effect evaluation data set. The fifth functional module (AMF) sends the policy effect evaluation data set to the fourth functional module (PCF). Specifically, the fifth functional module (AMF) can send the policy effect evaluation data set to the machine learning module of the fourth functional module (PCF) at a certain period (such as 15 minutes).

[0099] Refer to Figure 1, the fourth functional module (PCF) executes process 15, evaluates the data set according to the policy effect, and dynamically adjusts the parameters of the neural network model. Specifically, the machine learning module of the fourth functional module (PCF) triggers incremental online learning, adopts the online gradient descent algorithm, and dynamically adjusts the parameters of the BiLSTM model according to the policy effect evaluation data. Through continuous learning and updating, the system can quickly adapt to network changes and optimize the service quality.

[0100] In this embodiment, referring to Figure 1 , the eighth functional module (SEPP) executes process 16, monitors the execution process of the policy instruction information, and performs multi-level rollback when detecting abnormal execution of the policy instruction information. Specifically, when the eighth functional module (SEPP) executes process 16, to ensure the security of the communication system, it can audit all resource adjustment operations to ensure compliance with the predefined security policy template; once an abnormal operation is found, SEPP will start a three-level rollback mechanism: first, recover from the session level, if the problem is still not solved, then roll back from the slice level, and finally restore to a safe state at the full communication system level. The trigger conditions for the rollback mechanism include but are not limited to policy execution failure, security audit exception, or serious deviation of network performance indicators from expectations.

[0101] In this embodiment, the communication system based on multi-source data collaboration and intelligent decision-making can achieve the following functional mechanisms by executing Figure 1 all the processes in

[0102] 1. Multi-source data collaborative acquisition mechanism: The CU and UPF respectively collect physical layer and transport layer metrics, use the IEEE1588v2 protocol to achieve cross-layer data time synchronization, and ensure data security through encrypted transmission.

[0103] 2. Multidimensional feature vector construction mechanism: Based on a sliding time window (1 second), calculate the PRB coefficient of variation, the P95 value of delay, and the traffic autocorrelation parameter, and generate a feature vector that fuses physical layer and transport layer metrics as the input for intelligent decision-making.

[0104] 3. Hierarchical policy generation mechanism based on BiLSTM: Adopt a bidirectional LSTM network with an attention mechanism to map the feature vector to five-level policy recommendations, and combine historical data and expert annotations for model training.

[0105] 4. SLA dynamic binding verification mechanism: Dynamically bind the policy recommendation to the preset SLA template, and realize automatic upgrade and downgrade adjustment of the policy level through conditional judgment (delay / throughput threshold).

[0106] 5. Digital twin-driven Monte Carlo verification mechanism: Build a core network simulation model, calculate the policy success rate through hundreds of Monte Carlo simulations, and only allow policies with a success rate > 98% to be issued for execution. The failed policies are fed back to the model for iterative optimization.

[0107] 6. Lightweight Policy Transmission Mechanism: The ASN.1 PER encoding is used to compress the policy instructions within 500 bytes, and the 5QI = 80 high-priority channel is used to ensure low-latency and reliable transmission.

[0108] 7. Dynamic Resource Scheduling and Combination Technology Mechanism: Integrate WFQ weight adjustment, CSI frequency selection switching, ATSSS traffic guidance, and FlexE bandwidth dynamic allocation to achieve multi-dimensional resource optimization.

[0109] 8. CAPIF-Driven Edge Policy Execution Mechanism: Based on the general API framework, realize the dynamic mapping of five types of policies and edge resources, including resource reservation / redistribution, DNS path optimization, and security protocol enhancement.

[0110] 9. Incremental Online Learning Mechanism: Through the MDT data feedback with a 15-minute cycle, the online gradient descent algorithm is used to update the BiLSTM model parameters in real time to achieve adaptive optimization.

[0111] 10. Three-Level Security Rollback Mechanism: The SEPP proxy triggers a progressive rollback from the session level to the slice level and then to the system level based on the policy execution status to ensure fast recovery in abnormal scenarios.

[0112] In this embodiment, each functional module in Figure 1 can be grouped according to its function, thus forming Figure 2 the data acquisition module, data processing module, intelligent decision-making module, policy verification module, policy execution module, and network feedback module shown. Refer to Figure 2 where the functions of each module are as follows:

[0113] Data Acquisition Module: This module is responsible for real-time collection of physical layer and transport layer performance metrics in the radio access network, and performs encryption processing and time synchronization. It includes four sub-modules: physical layer metric collection, transport layer metric collection, data encrypted transmission, and clock synchronization. Physical layer metric collection: Collect physical layer metrics such as physical resource block (PRB) utilization, scheduling queue depth, and channel quality indicator (CQI) distribution through the central unit (CU). Transport layer metric collection: Collect transport layer metrics such as end-to-end latency, packet loss rate, and throughput fluctuation through the UPF. Data encrypted transmission: Use the AES-128 encryption algorithm to encrypt the collected raw data, and transmit it to the AMF through the N2 interface of the 5G core network, and then forward it to the NWDAF. Clock synchronization: Deploy the IEEE 1588v2 protocol through the SMF to achieve clock synchronization between the CU and the UPF, ensuring accurate alignment of multi-source data in the time dimension.

[0114] Data Processing Module: This module is responsible for decrypting the received encrypted data, extracting features, and generating feature vectors to provide data support for subsequent analysis. It includes four sub-modules: data aggregation, data decryption, feature extraction, and feature vector generation. Data Aggregation: Aggregate the encrypted data collected by CU and UPF to NWDAF through AMF. Data Decryption: NWDAF decrypts the encrypted data using the key to restore the original content of the data. Feature Extraction: Adopt the sliding time window technology to calculate the coefficient of variation of PRB utilization, the 95th percentile value (P95) of latency, and the autocorrelation parameter of traffic flow. Feature Vector Generation: Package the extracted features into feature vectors to represent the state of each time window for subsequent intelligent decision-making analysis.

[0115] Intelligent Decision-making Module: This module generates targeted optimization strategies based on feature vectors and dynamically generates control instructions in combination with the Service Level Agreement (SLA). It includes four sub-modules: model training, real-time analysis, policy classification, and SLA verification. Model Training: Use historical network performance data and expert-annotated optimization strategy samples to train a Bidirectional Long Short-Term Memory (BiLSTM) neural network model to minimize the difference between policy recommendations and actual optimization effects. Real-time Analysis: The BiLSTM model analyzes feature vectors in real time to generate policy recommendation levels (such as conventional optimization, resource reconstruction, security policy adjustment, etc.). Policy Classification: Provide differentiated control policies for different network performance problems according to the policy recommendation levels. SLA Verification: Combine the preset SLA template to verify whether the policy meets the performance requirements and dynamically adjust the policy recommendation levels.

[0116] Policy Verification Module: This module virtually verifies the generated optimization strategies through a digital twin engine to ensure the feasibility and effectiveness of the strategies. It includes four sub-modules: policy reception, digital twin simulation, Monte Carlo evaluation, and verification feedback. Policy Reception: Receive the optimization strategies generated by PCF and conduct preliminary verification. Digital Twin Simulation: Build a refined network simulation model to simulate the execution effects of the strategies. Monte Carlo Evaluation: Conduct multiple iterative simulations through the Monte Carlo method to evaluate the success probability of the strategies. Verification Feedback: According to the evaluation results, feedback whether the policy passes the verification. If it fails, adjust the policy and verify it again.

[0117] Policy Execution Module: This module is responsible for distributing the verified policies to the network execution nodes and performing resource adjustment and optimization. It includes four sub-modules: instruction distribution, resource adjustment, MAC layer scheduling optimization, and edge computing coordination. Instruction Distribution: The policy instructions are distributed to the UPF through the Npcf interface, and the UPF executes the specific policies. Resource Adjustment: According to the policy instructions, the network resource allocation is adjusted, such as PRB allocation and traffic guidance. MAC layer Scheduling Optimization: The weights of the weighted fair queue (WFQ) are adjusted, the frequency-selective handover mechanism based on the channel state information (CSI) is activated, and the scheduling efficiency is optimized. Edge Computing Coordination: It coordinates with the NEF and adjusts the edge computing resource allocation according to the policy to optimize the computing and security resources of the edge nodes.

[0118] Network Feedback Module: This module is responsible for collecting network status feedback data, evaluating the policy effects, and dynamically updating the model parameters. It includes four sub-modules: status monitoring, in-depth data collection, model update, and security auditing. Status Monitoring: The network status feedback is collected through the AMF, such as the reference signal received power (RSRP) fluctuation and the hybrid automatic repeat request (HARQ) retransmission rate. In-depth Data Collection: The in-depth coverage data is collected by using the minimized drive test (MDT) function to form a policy effect evaluation data set. Model Update: The machine learning module of the PCF dynamically adjusts the BiLSTM model parameters according to the feedback data to optimize the model performance. Security Auditing: The security edge protection proxy (SEPP) audits the resource adjustment operations to ensure compliance with the predefined security policies and starts the rollback mechanism in case of anomalies.

[0119] Figure 2 The system shown in Figure 1 The communication system based on multi-source data collaboration and intelligent decision-making shown in

[0120] In this embodiment, by running Figure 1 The communication system based on multi-source data collaboration and intelligent decision-making shown in Figure 3 , the communication method based on multi-source data collaboration and intelligent decision-making can be executed. Referring to

[0121] S1. Collect the underlying performance metric information from the radio access network;

[0122] S2. Collect the traffic metric information from the radio access network;

[0123] S3. Generate a feature vector according to the underlying performance metric information and the traffic metric information;

[0124] S4. Run a neural network model to process the feature vector to obtain the policy instruction information;

[0125] S5. Perform resource scheduling in response to the policy instruction information.

[0126] Among them, step S1 can be executed by the first functional module (CU), step S2 and S5 can be executed by the second functional module (UPF), step S3 can be executed by the third functional module (NWDAF), and step S4 can be executed by the fourth functional module (PCF).

[0127] By executing Figure 3 the communication method based on multi-source data collaboration and intelligent decision-making shown, it is possible to achieve the same technical effect as Figure 1 the communication system based on multi-source data collaboration and intelligent decision-making shown.

[0128] It is possible to write a computer program that executes the communication method based on multi-source data collaboration and intelligent decision-making in this embodiment, write this computer program into a computer device or a storage medium. When the computer program is read and run, the communication method based on multi-source data collaboration and intelligent decision-making in this embodiment is executed, so as to achieve the same technical effect as the communication method based on multi-source data collaboration and intelligent decision-making in the embodiment.

[0129] It should be noted that, unless otherwise specified, when a certain feature is referred to as "fixed" or "connected" to another feature, it can be directly fixed or connected to another feature, or indirectly fixed or connected to another feature. In addition, the up, down, left, right, etc. descriptions used in this disclosure are only relative to the mutual positional relationship of the various components of this disclosure in the drawings. The singular forms "a", "an", and "the" used in this disclosure are also intended to include the plural forms, unless the context clearly indicates otherwise. In addition, unless otherwise defined, all the technical and scientific terms used in this embodiment have the same meaning as those commonly understood by those skilled in the art of this technology field. The terms used in the description of this embodiment of the specification are only for describing specific embodiments, rather than for limiting the present invention. The term "and / or" used in this embodiment includes any combination of one or more of the related listed items.

[0130] It should be understood that although terms such as first, second, and third may be used in this disclosure to describe various elements, these elements should not be limited to these terms. These terms are only used to distinguish elements of the same type from each other. For example, without departing from the scope of this disclosure, the first element may also be referred to as the second element, and similarly, the second element may also be referred to as the first element. The use of any and all examples or exemplary languages ("for example", "such as", etc.) provided in this embodiment is only intended to better illustrate the embodiments of the present invention, and will not impose a limitation on the scope of the present invention unless otherwise required.

[0131] It should be recognized that embodiments of the present invention can be implemented or carried out by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable memory. The methods can be implemented in a computer program using standard programming techniques - including a non-transitory computer-readable storage medium configured with the computer program, wherein the storage medium so configured causes the computer to operate in a specific and predefined manner - according to the methods and drawings described in the specific embodiments. Each program can be implemented in a high-level procedural or object-oriented programming language to communicate with the computer system. However, if desired, the program can be implemented in assembly or machine language. In any case, the language can be a compiled or interpreted language. In addition, for this purpose the program is capable of running on a programmed application-specific integrated circuit.

[0132] In addition, the operations of the processes described in this embodiment can be performed in any suitable order, unless this embodiment otherwise indicates or is otherwise detected to be clearly inconsistent with the context. The processes described in this embodiment (or variations and / or combinations thereof) can be executed under the control of one or more computer systems configured with executable instructions, and can be implemented as code (e.g., executable instructions, one or more computer programs, or one or more applications) executed commonly on one or more processors, by hardware, or a combination thereof. The computer program includes multiple instructions executable by one or more processors.

[0133] Furthermore, the method can be implemented in any type of computing platform operably connected, including but not limited to personal computers, minicomputers, mainframes, workstations, network or distributed computing environments, separate or integrated computer platforms, or communicating with charged particle tools or other imaging devices, etc. Aspects of the present invention can be implemented in machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into the computing platform, such as a hard disk, optical read and / or write storage medium, RAM, ROM, etc., such that it can be read by a programmable computer and, when the storage medium or device is read by the computer, can be used to configure and operate the computer to perform the processes described herein. In addition, the machine-readable code, or portions thereof, can be transmitted via a wired or wireless network. When such media include instructions or programs that implement the above steps in conjunction with a microprocessor or other data processor, the invention of this embodiment includes these and other different types of non-transitory computer-readable storage media. When programmed according to the methods and techniques of the present invention, the present invention also includes the computer itself.

[0134] A computer program can be applied to input data to perform the functions of this embodiment, thereby converting the input data to generate output data stored in a non-volatile memory. The output information can also be applied to one or more output devices such as a display. In a preferred embodiment of the present invention, the converted data represents physical and tangible objects, including specific visual depictions of physical and tangible objects generated on the display.

[0135] The above are only the preferred embodiments of the present invention. The present invention is not limited to the above-mentioned embodiments. As long as it achieves the technical effects of the present invention by the same means, any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the scope of protection of the present invention. Within the scope of protection of the present invention, its technical solutions and / or implementation manners can have various different modifications and changes.

Claims

1. A communication system based on multi-source data collaboration and intelligent decision-making, characterized in that The communication system based on multi-source data collaboration and intelligent decision-making includes: A first functional module for collecting underlying performance metric information from a radio access network; A second functional module for collecting traffic metric information from the radio access network; A third functional module for generating a feature vector based on the underlying performance metric information and the traffic metric information; A fourth functional module for processing the feature vector by running a neural network model to obtain policy instruction information; The second functional module is used to perform resource scheduling in response to the policy instruction information.

2. The communication system based on multi-source data collaboration and intelligent decision-making according to claim 1, characterized in that, The generating of the feature vector based on the underlying performance metric information and the traffic metric information includes: Setting the size of a time window; Sliding the time window so that the time window reaches multiple time positions; For any determined time position of the time window, performing statistical feature calculations on the underlying performance metric information and the traffic metric information in the time window to obtain the coefficient of variation of the physical resource block utilization rate, the 95th percentile value of the delay metric, and the autocorrelation parameter of the traffic volume corresponding to the time position, and forming the feature vector corresponding to the time position with the coefficient of variation of the physical resource block utilization rate, the 95th percentile value of the delay metric, and the autocorrelation parameter of the traffic volume.

3. The communication system based on multi-source data collaboration and intelligent decision-making according to claim 1, characterized in that The running of the neural network model to process the feature vector to obtain policy instruction information includes: Inputting the feature vector into the running neural network model for processing; Obtaining the policy recommendation information output by the running neural network model; the content of the policy recommendation information belongs to one of multiple levels; Determining a service level agreement according to the level corresponding to the policy recommendation information; Using the service level agreement to verify the policy recommendation information; Generating the policy instruction information according to the verification result.

4. The communication system based on multi-source data collaboration and intelligent decision-making according to claim 1, wherein: The fourth functional module is used to perform success rate simulation verification on the policy instruction information through the digital twin engine. When the verification is passed, the policy instruction information is sent to the second functional module. When the verification fails, the regeneration of the policy instruction information is performed; The fourth functional module is used to encode and compress the policy instruction information by ASN.1 PER and send it to the second functional module through a 5QI = 80 high-priority channel.

5. The communication system based on multi-source data collaboration and intelligent decision-making according to any one of claims 1-4, characterized in that, The communication system based on multi-source data collaboration and intelligent decision-making further includes a fifth functional module; The fifth functional module is used to encrypt and forward the underlying performance metric information collected by the first functional module to the third functional module, collect status feedback information through a radio resource control measurement report, form a policy effect evaluation data set, and send the policy effect evaluation data set to the fourth functional module; The fourth functional module is used to dynamically adjust the parameters of the neural network model according to the policy effect evaluation data set.

6. The communication system based on multi-source data collaboration and intelligent decision-making according to claim 5, characterized in that, The communication system based on multi-source data collaboration and intelligent decision-making further includes a sixth functional module, a seventh functional module, and an eighth functional module; The sixth functional module is used to maintain clock synchronization between the first functional module and the second functional module; The seventh functional module is used to perform edge computing and adjust the policy instruction information; The eighth functional module is used to monitor the execution process of the policy instruction information. When the execution of the policy instruction information is detected to be abnormal, multi-level rollback is performed.

7. The communication system based on multi-source data collaboration and intelligent decision-making according to claim 6, wherein: The first functional module is a CU, the second functional module is a UPF, the third functional module is a NWDAF, the fourth functional module is a PCF, the fifth functional module is an AMF, the sixth functional module is an SMF, the seventh functional module is a NEF, and the eighth functional module is a SEPP.

8. A communication method based on multi-source data collaboration and intelligent decision-making, characterized in that, The communication method based on multi-source data collaboration and intelligent decision-making includes: Collecting underlying performance metric information from a radio access network; Collecting traffic metric information from the radio access network; Generating a feature vector according to the underlying performance metric information and the traffic metric information; Running a neural network model to process the feature vector to obtain policy instruction information; Responding to the policy instruction information and performing resource scheduling.

9. A computer device, characterized in that, It includes a memory and a processor. The memory is used to store at least one program, and the processor is used to load at least one program to execute the communication method based on multi-source data collaboration and intelligent decision-making according to claim 8.

10. A computer-readable storage medium storing a program executable by a processor, characterized in that, The program executable by the processor, when executed by the processor, is used to execute the communication method based on multi-source data collaboration and intelligent decision-making according to claim 8.

Images