Mobile 5G fused cross-domain resource scheduling system

Abstract

The invention discloses a mobile 5G fused cross-domain resource scheduling system, and belongs to the technical field of cross-domain resource scheduling. Comprising a service feature intelligent sensing module, a multi-domain network topology dynamic sensing module, an intelligent resource slice evolution arrangement module, a cross-domain protocol intelligent adaptation conversion module, a data flow intelligent optimization scheduling module, a network loop risk prediction self-healing module and a security cooperative defense module. According to the method, lightweight probes are deployed at edge computing nodes, link states, loads and delay indexes are collected in real time, occupation of excessive CPU or memory resources is avoided, network nodes and links are represented as graph structures, edge weights are dynamically adjusted, real-time and accurate sensing of the network states is achieved, node overload risks are predicted in combination with an ARIMA model, potential problems are found in advance, and the method has the advantages of being high in reliability and high in reliability. Therefore, timely and accurate network state input is provided for intelligent resource slice arrangement, network changes can be responded more quickly and efficiently, and the overall system performance and stability are improved.

Description

Technical Field

[0001] This invention relates to the field of cross-domain resource scheduling technology, and in particular to a cross-domain resource scheduling system integrating mobile 5G. Background Technology

[0002] With the widespread application of 5G networks and the deepening of digital transformation in vertical industries, cross-domain services have placed higher demands on network resource scheduling. Existing solutions are mostly limited to a single management domain, and when facing cross-domain scenarios, they suffer from problems such as rigid resource allocation, poor protocol compatibility, lagging network status awareness, and weak security protection.

[0003] However, existing cross-domain resource scheduling in mobile 5G still has certain shortcomings. The existing network topology awareness relies on a periodic polling mechanism, which leads to high resource consumption, high perception latency, inability to accurately perceive network status in real time, lack of ability to predict node overload risks, and inability to provide accurate input for intelligent resource slicing orchestration in a timely manner. Resource scheduling adopts a static or simple rule allocation method, which cannot dynamically adjust resource allocation strategies according to service needs and network status, resulting in low service satisfaction rate and network resource utilization, and lacking a balance optimization mechanism between the service side and the network side. Loop risk handling relies on manual intervention and traditional STP / RSTP protocols, which cannot predict loop risks in advance, and the response speed after loop detection is slow and the self-healing time is long, resulting in poor network stability and low availability, which can easily cause service interruption and network paralysis. To address these issues, a cross-domain resource scheduling system integrating mobile 5G is proposed. Summary of the Invention

[0004] The purpose of this invention is to provide a cross-domain resource scheduling system that integrates mobile 5G to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a cross-domain resource scheduling system integrating mobile 5G, comprising a service characteristic intelligent perception module, a multi-domain network topology dynamic perception module, an intelligent resource slicing evolution orchestration module, a cross-domain protocol intelligent adaptation and conversion module, a data flow intelligent optimization scheduling module, a network loop risk prediction and self-healing module, and a security collaborative defense module;

[0006] The business characteristic intelligent perception module identifies business types and demand characteristics based on real-time business behavior data and through AI feature extraction algorithms.

[0007] The multi-domain network topology dynamic perception module collects data in real time through lightweight probes based on 5G signaling and edge node data to dynamically perceive the network topology status.

[0008] The intelligent resource slicing evolution orchestration module performs dynamic orchestration of resource slices based on business characteristics and network topology status through multi-agent reinforcement learning computation.

[0009] The cross-domain protocol intelligent adaptation and conversion module performs conversion based on business characteristics through protocol sandbox technology, enabling seamless cross-domain protocol adaptation.

[0010] The data flow intelligent optimization and scheduling module performs cross-domain data flow scheduling through dynamic routing optimization based on the path instructions arranged by the intelligent resource slices.

[0011] The network loop risk prediction and self-healing module performs loop risk prediction and self-healing operations based on the network topology change trend.

[0012] The security collaborative defense module, based on the communication requirements of the data flow intelligent optimization scheduling module, the cross-domain protocol intelligent adaptation and conversion module, and the network loop risk prediction and self-healing module, establishes an end-to-end secure channel through 5G slicing and quantum keys, and coordinates with the service feature intelligent perception module and the cross-domain protocol intelligent adaptation and conversion module to ensure data security.

[0013] Preferably, the business feature intelligent perception module is wirelessly connected to the intelligent resource slicing evolution orchestration module; the multi-domain network topology dynamic perception module is wirelessly connected to the intelligent resource slicing evolution orchestration module and the network loop risk prediction and self-healing module; the network loop risk prediction and self-healing module is wirelessly connected to the security collaborative defense module; the intelligent resource slicing evolution orchestration module is wirelessly connected to the cross-domain protocol intelligent adaptation and conversion module and the data flow intelligent optimization and scheduling module; and the cross-domain protocol intelligent adaptation and conversion module is wirelessly connected to the data flow intelligent optimization and scheduling module and the security collaborative defense module.

[0014] Preferably, the intelligent perception module for business characteristics collects business behavior data in real time through 5G network signaling, edge node sensors, and business terminal logs, preprocesses the raw data, including cleaning, removing noise and outliers, and segments it according to time windows to form structured input;

[0015] The formula is as follows: Key features are extracted from preprocessed data using AI algorithms.

[0016]

[0017] In the formula, Represents the business feature vector. This represents the raw business behavior data within the time window t, where i represents the feature dimension index. This represents the dynamic weight of the i-th feature, which is adjusted according to business priorities or real-time requirements. This represents the historical mean of the i-th feature. This represents the historical standard deviation of the i-th feature. This represents the rate of change of the i features over the time window t. Indicates the attenuation coefficient. This represents the exponential decay coefficient.

[0018] Preferably, the business feature intelligent perception module identifies business types and demand features based on the business feature vector, using the following formula:

[0019] ,

[0020] In the formula, This represents the probability that a business belongs to the i-th class given a business feature vector. This represents the k-th business type. Let the covariance matrix of the k-th business class be represented. The square of the Mahalanobis distance is used to measure the deviation of the feature vector from the mean of the k-th business class. K represents the total number of business types. Based on real-time changes in business behavior, combined with an adaptive redundant variable mechanism, the threshold and classification boundary of the feature extraction model are dynamically adjusted to balance discriminative power and recall.

[0021] Preferably, the multi-domain network topology dynamic perception module collects signaling interaction information through the base station, deploys lightweight probes on the edge computing node, and collects link status, load, and latency indicators in real time. The probes only collect key fields to avoid consuming too much CPU or memory resources.

[0022] The acquired data is preprocessed to convert it into topology-related features, and the network nodes and links are represented as a graph structure. Where V represents the set of nodes and E represents the set of edges, the edge weights are dynamically adjusted based on the preprocessed link states, and the formula is as follows:

[0023] ,

[0024] In the formula, This represents the dynamic weight of the link, specifically the dynamic weight value of the link between node i and node j at time t. This represents the normalization coefficient, used to balance the influence of different indicators on the weights. This represents the end-to-end transmission delay between node i and node j at time t. This represents the available bandwidth between node i and node j at time t. This represents the packet loss rate between node i and node j at time t; the lower the latency, the lower the packet loss rate. The smaller, The larger the value, the higher the link weight; the higher the bandwidth... The larger the value, the greater its contribution to the link weight; the lower the packet loss rate. The smaller, The larger the value, the greater the weight contribution. Based on the depth-first search of the topology graph, loop paths are detected in real time. Combined with historical load data, the ARIMA model is used to predict node overload risk. The detected anomalies are synchronized to the network loop risk prediction and self-healing module to trigger self-healing operation.

[0025] Preferably, the intelligent resource slicing evolution orchestration module obtains the service type and resource requirements of the perceived service characteristics module, and analyzes the network status provided by the multi-domain network topology dynamic perception module, and dynamically allocates resource slices according to service requirements and network status.

[0026] Constructing the state space and dynamic space:

[0027] The state space S includes:

[0028] Business characteristics and status: including business priorities and resource requirements;

[0029] Network topology status: including link weights, node loads, and total available resources;

[0030] Historical scheduling records: Recent resource allocation strategies;

[0031] Action space A includes:

[0032] Resource allocation action: Assign a specific slice and resource allocation amount to each business;

[0033] Dynamic adjustment action: Adjust allocated resources according to changes in network status;

[0034] Designing reward functions through multi-agent reinforcement learning:

[0035] The business-side reward is implemented using the following formula:

[0036]

[0037] In the formula, This represents the business-side reward value, a quantitative indicator of the degree to which business needs are met. This indicates the latency requirement for the business request. This represents the end-to-end delay in the actual allocated path. This indicates the latency satisfaction rate, representing the degree of deviation between the actual latency and the business requirements. This represents the bandwidth actually allocated to the service. This indicates the bandwidth requirement of the service request. This represents the broadband fulfillment rate, which is the ratio of actual allocated broadband to service demand.

[0038] Preferably, the network-side reward of the intelligent resource slicing evolution orchestration module is implemented using the following formula:

[0039] ,

[0040] In the formula, This represents the network-side reward value. This represents the node load penalty coefficient, used to adjust the negative impact of node load on rewards. This represents the load of node i at time t. This represents the slice overload penalty coefficient. It represents the overload state of slice k at time t; it receives service characteristics and network topology status in real time and generates resource allocation strategies through multi-agent reinforcement learning.

[0041] Preferably, the cross-domain protocol intelligent adaptation and conversion module acquires the protocol type and characteristics of business data, analyzes the communication direction, data encapsulation method, port number, and header fields of the protocol, establishes a protocol feature database, maps the characteristics of different protocols to a unified protocol meta-model (such as defining communication modes, data structures, and state machines), identifies protocol conflict points in cross-domain scenarios, creates an independent virtual running environment for each protocol, prevents resource competition or conflicts between protocol parsers, deploys a protocol parser within the sandbox, supports dynamic loading and unloading, embeds a protocol conversion engine in the sandbox, and converts the data packets of the source protocol into the format of the target protocol; for the WebSocket protocol, it initiates a handshake request through the HTTP protocol upgrade mechanism and completes the handshake process within the sandbox; for other protocols, it uses custom protocol conversion rules, according to the protocol meta-model, to decompose the data packets of the source protocol into a general data structure, and then re-encapsulates them according to the specifications of the target protocol; based on the business priority provided by the business feature intelligent perception module and the link status provided by the multi-domain network topology dynamic perception module, it dynamically selects the optimal protocol adaptation strategy, detects anomalies in the protocol conversion process, triggers the isolation mechanism of the security collaborative defense module, and prevents network failures caused by protocol conflicts.

[0042] Preferably, the intelligent data flow optimization and scheduling module receives path instructions output by the intelligent resource slicing evolution orchestration module, parses the key information in the path instructions, transforms the path instructions into business requirements, calls dynamic routing to generate the optimal routing path based on the business requirement model, dynamically adjusts the routing strategy in conjunction with the link status provided by the multi-domain network topology dynamic perception module, distributes business traffic to multiple paths according to the link load distribution, plans cross-domain transmission paths based on the cross-domain requirements in the path instructions and the topology information of the multi-domain network topology dynamic perception module, prioritizes links with high protocol compatibility, and triggers the cross-domain protocol intelligent adaptation and conversion module to perform dynamic protocol conversion if there is a protocol conflict in the cross-domain links.

[0043] Preferably, the network loop risk prediction and self-healing module acquires real-time network topology information, collects the interface status, port speed, and VLAN configuration parameters of network devices, inputs the current network topology and traffic data into a deep learning model, outputs a loop risk score, triggers an alarm if a high-risk loop is predicted, generates different levels of alarms based on the risk level, prioritizes blocking redundant links forming the loop, switches service traffic to backup paths through dynamic routing optimization to avoid loop areas, and combines path instructions from the data flow intelligent optimization scheduling module to ensure that the quality of service is met after the switch.

[0044] Preferably, the security collaborative defense module receives the security level and service requirements provided by the service characteristic intelligent perception module, dynamically constructs secure network slices based on 5G slicing, allocates independent resources for highly sensitive services and isolates the communication environment, and establishes an end-to-end quantum encryption channel through quantum key distribution within the slice, using quantum mechanics principles to ensure the key distribution is not eavesdroppable and cannot be copied. Simultaneously, it collaborates with the cross-domain protocol intelligent adaptation and conversion module to apply quantum encryption to protect the data stream in real time during protocol conversion, preventing security vulnerabilities caused by man-in-the-middle attacks and protocol conflicts, and monitors the communication status throughout the process, dynamically adjusting the encryption strategy.

[0045] Compared with the prior art, the beneficial effects of the present invention are:

[0046] 1. This invention deploys lightweight probes on edge computing nodes to collect link status, load, and latency metrics in real time, avoiding excessive CPU or memory resource consumption. It represents network nodes and links as a graph structure and dynamically adjusts edge weights to achieve real-time and accurate perception of network status. Combined with the ARIMA model, it predicts node overload risks and identifies potential problems in advance, providing timely and accurate network status input for intelligent resource slicing orchestration. This enables faster and more efficient response to network changes and improves overall system performance and stability.

[0047] 2. This invention constructs a state space and action space through multi-agent reinforcement learning, and designs dual reward functions on the service side and the network side. It realizes dynamic and optimal orchestration of resource slicing, and can adjust the resource allocation strategy in real time according to service needs and network status, which significantly improves service satisfaction rate and network resource utilization. The service-side reward function ensures the service quality of key services, and the network-side reward function avoids network overload, realizing the balance optimization of services and network, and realizing intelligent and adaptive resource allocation.

[0048] 3. This invention uses a deep learning model to analyze network topology and traffic data in real time, predict loop risks and automatically perform self-healing operations, construct a network topology map, identify redundant paths as potential loop risk points, and extract loop risk features by combining historical fault data to achieve early prediction of loop risks. When a high-risk loop is detected, redundant links are blocked first, and business traffic is switched to backup paths through dynamic routing optimization to avoid loop areas, which greatly improves the reliability and availability of the network. Attached Figure Description

[0049] Figure 1 This is a schematic diagram of the structure of a cross-domain resource scheduling system integrating mobile 5G according to the present invention;

[0050] Figure 2 The present invention provides an operational flow of a cross-domain resource scheduling system integrating mobile 5G. Figure 1 ;

[0051] Figure 3 The present invention provides an operational flow of a cross-domain resource scheduling system integrating mobile 5G. Figure 2 . Detailed Implementation

[0052] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0053] Example 1

[0054] Please see Figures 1-3 As shown, the present invention provides a technical solution including a business characteristic intelligent perception module, a multi-domain network topology dynamic perception module, an intelligent resource slicing evolution orchestration module, a cross-domain protocol intelligent adaptation and conversion module, a data flow intelligent optimization scheduling module, a network loop risk prediction and self-healing module, and a security collaborative defense module.

[0055] The business characteristic intelligent perception module identifies business types and demand characteristics based on real-time business behavior data and through AI feature extraction algorithms.

[0056] The multi-domain network topology dynamic perception module collects data in real time through lightweight probes based on 5G signaling and edge node data to dynamically perceive the network topology status.

[0057] The intelligent resource slicing evolution orchestration module performs dynamic orchestration of resource slices based on business characteristics and network topology status through multi-agent reinforcement learning computation.

[0058] The cross-domain protocol intelligent adaptation and conversion module performs conversion based on business characteristics through protocol sandbox technology, enabling seamless cross-domain protocol adaptation.

[0059] The data flow intelligent optimization and scheduling module performs cross-domain data flow scheduling through dynamic routing optimization based on the path instructions arranged by the intelligent resource slices.

[0060] The network loop risk prediction and self-healing module performs loop risk prediction and self-healing operations based on the network topology change trend.

[0061] The security collaborative defense module, based on the communication requirements of the data flow intelligent optimization scheduling module, the cross-domain protocol intelligent adaptation and conversion module, and the network loop risk prediction and self-healing module, establishes an end-to-end secure channel through 5G slicing and quantum keys, and coordinates with the service feature intelligent perception module and the cross-domain protocol intelligent adaptation and conversion module to ensure data security.

[0062] Preferably, the business feature intelligent perception module is wirelessly connected to the intelligent resource slicing evolution orchestration module; the multi-domain network topology dynamic perception module is wirelessly connected to the intelligent resource slicing evolution orchestration module and the network loop risk prediction and self-healing module; the network loop risk prediction and self-healing module is wirelessly connected to the security collaborative defense module; the intelligent resource slicing evolution orchestration module is wirelessly connected to the cross-domain protocol intelligent adaptation and conversion module and the data flow intelligent optimization and scheduling module; and the cross-domain protocol intelligent adaptation and conversion module is wirelessly connected to the data flow intelligent optimization and scheduling module and the security collaborative defense module.

[0063] Preferably, the intelligent perception module for business characteristics collects business behavior data in real time through 5G network signaling, edge node sensors, and business terminal logs, preprocesses the raw data, including cleaning, removing noise and outliers, and segments it according to time windows to form structured input;

[0064] The formula is as follows: Key features are extracted from preprocessed data using AI algorithms.

[0065] ,

[0066] In the formula, Represents the business feature vector. This represents the raw business behavior data within the time window t, where i represents the feature dimension index. This represents the dynamic weight of the i-th feature, which is adjusted according to business priorities or real-time requirements. This represents the historical mean of the i-th feature. This represents the historical standard deviation of the i-th feature. This represents the rate of change of the i features over the time window t. Indicates the attenuation coefficient. This represents the exponential decay coefficient.

[0067] Preferably, the business feature intelligent perception module identifies business types and demand features based on the business feature vector, using the following formula:

[0068] ,

[0069] In the formula, This represents the probability that a business belongs to the i-th class given a business feature vector. This represents the k-th business type. Let the covariance matrix of the k-th business class be represented. The square of the Mahalanobis distance is used to measure the deviation of the feature vector from the mean of the k-th business class. K represents the total number of business types. Based on real-time changes in business behavior, combined with an adaptive redundant variable mechanism, the threshold and classification boundary of the feature extraction model are dynamically adjusted to balance discriminative power and recall.

[0070] Preferably, the multi-domain network topology dynamic perception module collects signaling interaction information through the base station, deploys lightweight probes on the edge computing node, and collects link status, load, and latency indicators in real time. The probes only collect key fields to avoid consuming too much CPU or memory resources.

[0071] The acquired data is preprocessed to convert it into topology-related features, and the network nodes and links are represented as a graph structure. Where V represents the set of nodes and E represents the set of edges, the edge weights are dynamically adjusted based on the preprocessed link states, and the formula is as follows:

[0072] ,

[0073] In the formula, This represents the dynamic weight of the link, specifically the dynamic weight value of the link between node i and node j at time t. This represents the normalization coefficient, used to balance the influence of different indicators on the weights. This represents the end-to-end transmission delay between node i and node j at time t. This represents the available bandwidth between node i and node j at time t. This represents the packet loss rate between node i and node j at time t; the lower the latency, the lower the packet loss rate. The smaller, The larger the value, the higher the link weight; the higher the bandwidth... The larger the value, the greater its contribution to the link weight; the lower the packet loss rate. The smaller, The larger the value, the greater the weight contribution. Based on the depth-first search of the topology graph, loop paths are detected in real time. Combined with historical load data, the ARIMA model is used to predict node overload risk. The detected anomalies are synchronized to the network loop risk prediction and self-healing module to trigger self-healing operation.

[0074] Preferably, the intelligent resource slicing evolution orchestration module obtains the service type and resource requirements of the perceived service characteristics module, and analyzes the network status provided by the multi-domain network topology dynamic perception module, and dynamically allocates resource slices according to service requirements and network status.

[0075] Constructing the state space and dynamic space:

[0076] The state space S includes:

[0077] Business characteristics and status: including business priorities and resource requirements;

[0078] Network topology status: including link weights, node loads, and total available resources;

[0079] Historical scheduling records: Recent resource allocation strategies;

[0080] Action space A includes:

[0081] Resource allocation action: Assign a specific slice and resource allocation amount to each business;

[0082] Dynamic adjustment action: Adjust allocated resources according to changes in network status;

[0083] Designing reward functions through multi-agent reinforcement learning:

[0084] The business-side reward is implemented using the following formula:

[0085] ,

[0086] In the formula, This represents the business-side reward value, a quantitative indicator of the degree to which business needs are met. This indicates the latency requirement for the business request. This represents the end-to-end delay in the actual allocated path. This indicates the latency satisfaction rate, representing the degree of deviation between the actual latency and the business requirements. This represents the bandwidth actually allocated to the service. This indicates the bandwidth requirement of the service request. This represents the broadband fulfillment rate, which is the ratio of actual allocated broadband to service demand.

[0087] Preferably, the network-side reward of the intelligent resource slicing evolution orchestration module is implemented using the following formula:

[0088] ,

[0089] In the formula, This represents the network-side reward value. This represents the node load penalty coefficient, used to adjust the negative impact of node load on rewards. This represents the load of node i at time t. This represents the slice overload penalty coefficient. It represents the overload state of slice k at time t; it receives service characteristics and network topology status in real time and generates resource allocation strategies through multi-agent reinforcement learning.

[0090] Preferably, the cross-domain protocol intelligent adaptation and conversion module acquires the protocol type and characteristics of the business data, analyzes the communication direction, data encapsulation method, port number, and header fields of the protocol, establishes a protocol feature database, maps the features of different protocols to a unified protocol meta-model (such as defining communication mode, data structure, and state machine), and identifies protocol conflict points in cross-domain scenarios.

[0091] Each protocol has an independent virtual runtime environment to prevent resource contention or conflict between protocol parsers. Protocol parsers are deployed within the sandbox, supporting dynamic loading and unloading. A protocol conversion engine is embedded in the sandbox to convert data packets of the source protocol into the format of the target protocol. For the WebSocket protocol, a handshake request is initiated through the HTTP protocol upgrade mechanism, and the handshake process is completed within the sandbox.

[0092] For other protocols, custom protocol conversion rules are used to decompose the data packets of the source protocol into a general data structure based on the protocol meta-model, and then repackage them according to the specifications of the target protocol. Based on the business priority provided by the business characteristic intelligent perception module and the link status provided by the multi-domain network topology dynamic perception module, the optimal protocol adaptation strategy is dynamically selected, anomalies in the protocol conversion process are detected, and the isolation mechanism of the security collaborative defense module is triggered to prevent network failures caused by protocol conflicts.

[0093] Preferably, the intelligent data flow optimization and scheduling module receives path instructions output by the intelligent resource slicing evolution orchestration module, parses the key information in the path instructions, transforms the path instructions into business requirements, calls dynamic routing to generate the optimal routing path based on the business requirement model, dynamically adjusts the routing strategy in conjunction with the link status provided by the multi-domain network topology dynamic perception module, distributes business traffic to multiple paths according to the link load distribution, plans cross-domain transmission paths based on the cross-domain requirements in the path instructions and the topology information of the multi-domain network topology dynamic perception module, prioritizes links with high protocol compatibility, and triggers the cross-domain protocol intelligent adaptation and conversion module to perform dynamic protocol conversion if there is a protocol conflict in the cross-domain links.

[0094] Preferably, the network loop risk prediction and self-healing module acquires real-time network topology information, collects the interface status, port speed, and VLAN configuration parameters of network devices, inputs the current network topology and traffic data into a deep learning model, outputs a loop risk score, and triggers an alarm if a high-risk loop is predicted.

[0095] In this embodiment, alarms of different levels are generated according to the risk level. If a loop is detected, the redundant links forming the loop are blocked first. The service traffic is switched to the backup path through dynamic routing optimization to avoid the loop area. Combined with the path instructions of the data flow intelligent optimization scheduling module, the service quality is ensured after the switch.

[0096] Preferably, the security collaborative defense module receives the security level and business requirements provided by the business characteristic intelligent perception module, dynamically constructs a secure network slice based on 5G slicing, allocates independent resources for highly sensitive services and isolates the communication environment, and establishes an end-to-end quantum encryption channel through quantum key distribution within the slice, using the principles of quantum mechanics to ensure that the key distribution is not eavesdroppable and cannot be copied.

[0097] Simultaneously, the collaborative cross-domain protocol intelligent adaptation and conversion module applies quantum encryption to protect the data stream in real time during the protocol conversion process, preventing security vulnerabilities caused by man-in-the-middle attacks and protocol conflicts, and monitors the communication status throughout the process, dynamically adjusting the encryption strategy.

[0098] Example 2

[0099] In a smart port multi-area equipment collaborative scheduling scenario, this scenario encompasses the port's container terminal operation area, warehousing management area, transportation dispatch center, and remote data monitoring platform. Specifically, the container terminal operation area deploys 5G+Industrial Ethernet to support real-time control of heavy equipment such as gantry cranes and quay cranes, requiring end-to-end latency ≤15ms and bandwidth stability ≥200Mbps. The warehousing management area uses 5G SA networking to handle real-time uploading of cargo inventory data and AGV robot path scheduling, requiring a data transmission packet loss rate ≤0.1%. The transportation dispatch center interacts with the terminal and warehousing areas via the 5G public network to process transactional tasks such as vehicle dispatch instructions and cargo loading plans. The remote data monitoring platform needs to acquire equipment operation data across different areas for fault diagnosis and efficiency analysis, placing extremely high demands on data transmission security to prevent the leakage of sensitive information such as equipment parameters and work plans.

[0100] The current scenario presents significant scheduling challenges: traditional scheduling relies on manual configuration of static resource slices, and real-time control services at the terminal often experience excessive latency due to link congestion, while non-real-time data from the storage area consumes redundant bandwidth; heterogeneous protocols exist across different areas, such as the terminal's PROFINET protocol, the storage area's MQTT protocol, and the monitoring platform's HTTP protocol, resulting in an 8% conflict rate in cross-domain data transmission protocols, leading to data disconnections; network status is polled every 10 seconds, resulting in delayed detection of node overload and link failures, and operational interruptions have occurred due to delays in the transmission of quay crane control commands; loop risks are handled using the STP protocol, with self-healing times exceeding 3 minutes, affecting continuous terminal operations; security protection relies solely on firewalls, leaving remote monitoring data transmission vulnerable to theft.

[0101] The business characteristic intelligent perception module collects three types of business data through 5G signaling and edge sensors. After cleaning, the data is segmented into 2-second time windows, and feature vectors are extracted using AI algorithms. For example, the weight of the delay feature of the dock control business is set to 0.7. The business type is identified by combining the Mahalanobis distance formula, and the three business requirement indicators of real-time control, data upload, and transaction interaction are output.

[0102] The multi-domain network topology dynamic perception module deploys lightweight probes at edge nodes in each region to collect key fields such as link latency, bandwidth, and packet loss rate, constructs a topology graph G(V,E), and dynamically adjusts edge weights according to the link weight formula. For example, when the link latency from the dock to the dispatch center decreases, the weight is increased. Loops are detected through depth-first search, and the overload risk of the gantry crane control node is predicted by combining the ARIMA model. Abnormal information is synchronized to the loop self-healing module.

[0103] The intelligent resource slicing evolution orchestration module constructs a state space and an action space based on business needs and network status, and uses a dual reward function to calculate: the business-side reward ensures the terminal controls the business latency satisfaction rate, and the network-side reward avoids the overload of warehouse nodes, ultimately allocating low-latency slices to the terminal and elastic bandwidth slices to the warehouse.

[0104] The cross-domain protocol intelligent adaptation and conversion module establishes a protocol meta-model, creates independent environments for PROFINET and MQTT protocols in the sandbox, decomposes the dock control data into a general structure and encapsulates it into the HTTP protocol for transmission to the monitoring platform; when an abnormal protocol format is detected, a security isolation mechanism is triggered.

[0105] The data flow intelligent optimization and scheduling module parses the slice orchestration path instructions, selects high-weight links for the terminal control instructions, distributes the warehouse inventory data to three backup paths for transmission, and automatically calls the protocol adaptation module when a protocol conflict is detected.

[0106] The network loop risk prediction and self-healing module collects device interface status and VLAN configuration, inputs them into a deep learning model, and outputs a risk score. When a high risk is predicted in the link between the warehouse area and the dispatch center, it blocks the redundant link and switches to the backup path, with the self-healing time controlled within 1.5 seconds.

[0107] The security collaborative defense module builds a 5G security slice for port control services, establishes an encrypted channel through quantum keys, encrypts data in real time during protocol conversion, and dynamically adjusts the encryption strategy when abnormal access is detected.

[0108] After implementation, the average latency of real-time control services at the port was reduced to 9.2ms, and the latency satisfaction rate increased from 76% to 99.5%; the data packet loss rate in the storage area decreased to 0.03%, and bandwidth utilization increased by 42%; the cross-domain protocol conversion power reached 99.8%, and the conflict problem was completely resolved; the network status awareness latency was shortened to 0.3 seconds, and the node overload prediction accuracy was 91%; the loop self-healing time was reduced to 1.2 seconds, and no more operational interruptions occurred; the encryption rate of sensitive data transmission was 100%, and there were no security leakage incidents.

[0109] This addresses the problems of traditional methods, such as lagging network state perception and high resource consumption, rigid resource allocation and lack of balancing optimization mechanisms, poor cross-domain protocol adaptation and compatibility leading to easy failures, passive handling of loop risks and low self-healing efficiency.

[0110] Working principle: Real-time collection of business behavior data through 5G network signaling, edge node sensors and business terminal logs; cleaning of raw data, removal of noise and outliers, and segmentation into structured inputs according to time windows; extraction of key features from preprocessed data using AI algorithms, calculation of business feature vectors through dynamic weight adjustment mechanism, and identification of business type and demand features by combining Mahalanobis distance; dynamic adjustment of classification boundaries based on real-time changes in business behavior.

[0111] The multi-domain network topology dynamic perception module collects signaling interaction information through base stations and deploys lightweight probes on edge computing nodes to collect link status, load, and latency metrics in real time, avoiding excessive resource consumption. The module represents network nodes and links as a graph structure, dynamically adjusting edge weights based on end-to-end transmission latency, available bandwidth, and packet loss rate. It detects loop paths through depth-first search and predicts node overload risks using an ARIMA model, synchronizing anomaly information to the loop risk prediction module to trigger self-healing operations. The intelligent resource slicing evolutionary orchestration module obtains the service type and resource requirements from the service feature module, analyzes the network status provided by the multi-domain network topology module, and constructs a state space containing service feature status, network topology status, and historical scheduling records. The action space includes resource allocation and dynamic adjustment actions; a dual reward function on the business and network sides is designed through multi-agent reinforcement learning to receive business characteristics and network topology status in real time, generating a dynamic resource slicing orchestration strategy that meets business needs and optimizes network resource utilization; a cross-domain protocol intelligent adaptation and conversion module obtains the protocol type and characteristics of business data, analyzes communication direction, data encapsulation method, port number and header fields, establishes a protocol feature database and maps different protocols to a unified protocol metamodel; an independent virtual running environment is created for each protocol, and a protocol parser and conversion engine are deployed in the sandbox, handling WebSocket handshakes through the HTTP protocol upgrade mechanism, and dynamically selecting the optimal adaptation strategy based on business priority and link status. The system triggers a security isolation mechanism when an anomaly is detected. The intelligent data flow optimization and scheduling module receives path instructions from the intelligent resource slicing orchestration module, parses key information into business requirements, and uses a dynamic routing algorithm to generate the optimal path based on the business requirement model. It also dynamically adjusts routing strategies based on link status provided by the multi-domain network topology awareness module, distributing traffic across multiple paths according to link load distribution, prioritizing links with high protocol compatibility. If a protocol conflict is detected, the protocol adaptation and conversion module is triggered to perform dynamic protocol conversion. The network loop risk prediction and self-healing module acquires real-time network topology information, collects network device interface status, port speeds, and VLAN configuration parameters, and inputs the current network topology and traffic data into a deep learning model to calculate loop risks. The system is structured as follows: A scoring system generates alarms based on risk levels. When a high-risk loop is detected, redundant links are blocked first. Dynamic routing optimization switches service traffic to backup paths, and path instructions from the data flow scheduling module ensure that service quality requirements are met after the switch. The security collaborative defense module receives security levels and service requirements from the service characteristic intelligent perception module. Based on 5G slicing, it dynamically constructs secure network slices, allocating independent resources and isolating communication environments for highly sensitive services. Within the slice, an end-to-end quantum encryption channel is established through quantum key distribution to ensure the key distribution is uneavesdroppable and uncopyable. The collaborative cross-domain protocol intelligent adaptation and conversion module applies quantum encryption to protect data streams in real time during protocol conversion, monitoring communication status throughout and dynamically adjusting encryption strategies.

[0112] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their likenesses.

[0113] This invention provides a cross-domain resource scheduling system integrating mobile 5G. Many methods and approaches exist for implementing this technical solution; the above description is merely a preferred embodiment. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this invention, and these improvements and modifications should also be considered within the scope of protection of this invention. All components not explicitly stated in this embodiment can be implemented using existing technologies.

Claims

1. A cross-domain resource scheduling system integrating mobile 5G, characterized in that, It includes a business characteristic intelligent perception module, a multi-domain network topology dynamic perception module, an intelligent resource slicing evolution orchestration module, a cross-domain protocol intelligent adaptation and conversion module, a data flow intelligent optimization and scheduling module, a network loop risk prediction and self-healing module, and a security collaborative defense module. The business characteristic intelligent perception module identifies business types and demand characteristics based on real-time business behavior data. The identified business characteristics are then used by the intelligent resource slicing evolution orchestration module to dynamically adjust resource allocation. The multi-domain network topology dynamic perception module collects data based on 5G signaling and edge nodes, and collects service behavior data in real time through lightweight probes to obtain the network topology status in real time. The intelligent resource slicing evolution orchestration module performs dynamic orchestration of resource slices and generates path instructions based on business characteristics and network topology status through multi-agent reinforcement learning computation. The cross-domain protocol intelligent adaptation and conversion module receives service characteristics as input, achieves seamless conversion from source protocol to target protocol through protocol sandbox technology and detects anomalies, provides the adapted data stream to the data flow intelligent optimization and scheduling module, and synchronizes the abnormal results to the security collaborative defense module to trigger the isolation mechanism. The data flow intelligent optimization and scheduling module performs cross-domain data flow scheduling through dynamic route optimization based on the path instructions output by the intelligent resource slicing evolution orchestration module and the protocol compatibility status provided by the cross-domain protocol intelligent adaptation and conversion module. The network loop risk prediction and self-healing module performs loop risk prediction and self-healing operations based on the network topology change trend. When a loop risk is detected, it sends an adjustment command to the intelligent resource slice evolution orchestration module. Based on the communication requirements of the intelligent data flow optimization scheduling module, the intelligent cross-domain protocol adaptation and conversion module, and the network loop risk prediction and self-healing module, the security collaborative defense module, during the resource scheduling process, uses 5G slicing and quantum keys to establish an end-to-end secure channel and collaborative defense strategy to ensure the security of data transmission and the stability and reliability of the entire scheduling process.

2. The cross-domain resource scheduling system integrating mobile 5G according to claim 1, characterized in that: The intelligent perception module for business characteristics collects business behavior data in real time through 5G network signaling, edge node sensors, and business terminal logs. It preprocesses the raw data, including cleaning, noise and outlier removal, and segments it by time window to form structured input. Then, it extracts key features from the preprocessed data using AI algorithms, achieving the following formula: ， In the formula, Represents the business feature vector. This represents the raw business behavior data within the time window t, where i represents the feature dimension index. This represents the dynamic weight of the i-th feature. This represents the historical mean of the i-th feature. This represents the historical standard deviation of the i-th feature. This represents the rate of change of the i features over the time window t. This represents an attenuation coefficient of 1, with a value of [0,1]. This represents the exponential decay coefficient.

3. The cross-domain resource scheduling system integrating mobile 5G according to claim 2, characterized in that: The intelligent perception module for business characteristics identifies business types and demand characteristics based on business feature vectors, using the following formula: ， In the formula, This represents the probability that a business belongs to the i-th class given a business feature vector. This represents the k-th business type. Let the covariance matrix of the k-th business class be represented. The square of the Mahalanobis distance is used to measure the degree of deviation of the feature vector from the mean of the k-th business type, where K represents the total number of business types.

4. The cross-domain resource scheduling system integrating mobile 5G according to claim 1, characterized in that, The multi-domain network topology dynamic perception module collects signaling interaction information from base stations and deploys lightweight probes on edge computing nodes to collect link status, load, and latency metrics in real time. The probes only collect key fields. The acquired data is preprocessed and converted into topology-related features, representing the network node and link structure. Where V represents the set of nodes and E represents the set of edges, the edge weights are dynamically adjusted according to the preprocessed link state.

5. A cross-domain resource scheduling system integrating mobile 5G according to claim 1, characterized in that: The intelligent resource slicing evolution orchestration module obtains the service type and resource requirements of the service characteristic perception module, and analyzes the network status provided by the multi-domain network topology dynamic perception module. Based on the service requirements and network status, it dynamically allocates resource slices. A reward function is designed using multi-agent reinforcement learning, which includes both business-side rewards and network-side rewards. The reward function is used to guide multi-agent reinforcement learning to dynamically optimize resource slicing allocation strategies. Specifically, it includes business-side rewards that quantify the degree to which business needs are met, and network-side rewards that assess network load status.

6. A cross-domain resource scheduling system integrating mobile 5G according to claim 5, characterized in that: The network-side reward of the intelligent resource slicing evolution orchestration module is implemented using the following formula: ， In the formula, This represents the network-side reward value. This represents the node load penalty coefficient. This represents the load of node i at time t. This represents the slice overload penalty coefficient. It represents the overload state of slice k at time t; it receives service characteristics and network topology status in real time and generates resource allocation strategies through multi-agent reinforcement learning.

7. A cross-domain resource scheduling system integrating mobile 5G according to claim 1, characterized in that: The cross-domain protocol intelligent adaptation and conversion module acquires the protocol type and characteristics of business data, analyzes the communication direction, data encapsulation method, port number, and header fields of the protocol, establishes a protocol feature database, maps the characteristics of different protocols to a unified protocol meta-model, identifies protocol conflict points in cross-domain scenarios, creates an independent virtual running environment for each protocol, deploys a protocol parser in the sandbox, embeds a protocol conversion engine in the sandbox, and converts the data packets of the source protocol into the format of the target protocol. For the WebSocket protocol, a handshake request is initiated through the HTTP protocol upgrade mechanism, and the handshake process is completed within a sandbox; For other protocols, custom protocol conversion rules are used to decompose the data packets of the source protocol into a general data structure based on the protocol meta-model, and then repackage them according to the specifications of the target protocol. Based on the business priority provided by the business characteristic intelligent perception module and the link status provided by the multi-domain network topology dynamic perception module, the optimal protocol adaptation strategy is dynamically selected, anomalies in the protocol conversion process are detected, and the isolation mechanism of the security collaborative defense module is triggered to prevent network failures caused by protocol conflicts.

8. A cross-domain resource scheduling system integrating mobile 5G according to claim 1, characterized in that: The intelligent data flow optimization and scheduling module receives path instructions output by the intelligent resource slicing evolution orchestration module, parses key information in the path instructions, including business requirement identifiers, path requirement parameters, source and destination node information, priority information, cross-domain requirements, and other constraints, transforms the path instructions into business requirements, calls dynamic routing to generate the optimal routing path based on the business requirement model, and dynamically adjusts the routing strategy in conjunction with the link status provided by the multi-domain network topology dynamic perception module. According to the link load distribution, the business traffic is distributed to multiple paths. Based on the cross-domain requirements in the path instructions and the topology information from the multi-domain network topology dynamic perception module, cross-domain transmission paths are planned, prioritizing links with high protocol compatibility. If there are protocol conflicts in the cross-domain links, the cross-domain protocol intelligent adaptation and conversion module is triggered to perform dynamic protocol conversion.

9. A cross-domain resource scheduling system integrating mobile 5G according to claim 1, characterized in that: The network loop risk prediction and self-healing module acquires real-time network topology status, collects interface status, port speed, and VLAN configuration parameters of network devices, inputs the current network topology and traffic data into a deep learning model, and outputs a loop risk score. If a high-risk loop is predicted, an alarm is triggered, and different levels of alarms are generated according to the risk level. If a loop is detected, redundant links forming the loop are blocked first, and service traffic is switched to backup paths through dynamic routing optimization to avoid the loop area. Combined with the path instructions of the data flow intelligent optimization scheduling module, the service quality is ensured after the switch.

10. A cross-domain resource scheduling system integrating mobile 5G according to claim 1, characterized in that: The security collaborative defense module receives the security level and business requirements provided by the business characteristic intelligent perception module, dynamically constructs a secure network slice based on 5G slicing, establishes an end-to-end quantum encryption channel through quantum key distribution within the slice, and collaborates with the cross-domain protocol intelligent adaptation and conversion module to apply quantum encryption to protect the data stream in real time during the protocol conversion process, preventing security vulnerabilities caused by man-in-the-middle attacks and protocol conflicts, and monitors the communication status throughout the process and dynamically adjusts the encryption strategy.

Images