Embodied intelligent operation and maintenance equipment configuration method considering input and output

Abstract

The application belongs to the technical field of communication network operation and maintenance, and discloses a body intelligent operation and maintenance equipment configuration method considering input and output, constructs an input side index set, which is divided into two types of equipment cost and communication resource cost; the equipment cost covers equipment purchase cost, operation energy consumption cost, operation and maintenance labor cost and fault loss cost, and the communication resource cost covers slice resource exclusive cost, shared bandwidth premium cost and port conflict implicit loss cost; a device operation type and a communication service type output side index set are synchronously established; in combination with network node levels and service priorities, an entropy weight method is adopted to distribute the weight of each index, and the implicit cost of adjacent link congestion caused by node configuration is converted; a communication link digital twin mirror image is constructed, the maximum of the whole network input-output ratio is taken as the target, and the initial configuration scheme is solved by taking the device performance and other conditions as the constraints, the scheme is introduced into the twin mirror image, simulation traffic is injected to complete six index verifications, and if the threshold is not met, iteration is performed.

Description

Technical Field

[0001] This invention belongs to the field of communication network operation and maintenance technology, specifically a method for configuring intelligent operation and maintenance equipment that takes into account input and output. Background Technology

[0002] Embodied intelligence technology, leveraging its physical interaction and on-site perception capabilities, is gradually being applied to the inspection and maintenance of network nodes, improving the automation level of physical layer maintenance. However, existing technologies still face the following technical challenges in the deep integration of device configuration decisions and network communication systems:

[0003] Existing equipment configuration methods mostly focus on optimizing equipment hardware performance parameters, without incorporating network communication metrics such as transmission quality, bandwidth usage, latency jitter, and port resource conflicts into the input-output quantitative model. This means that configuration optimization can only achieve local optimization of the equipment, making it difficult to balance end-to-end communication performance with the overall lifecycle cost.

[0004] Intelligent operation and maintenance equipment typically adopts a single-path decision-making architecture with centralized cloud computing and terminal execution. The edge side lacks localized capabilities for rapid verification of input and output and fine-tuning of configuration. When network congestion, slice isolation failure, or partial disconnection occurs, the issuance of configuration commands and status feedback links will be blocked. At the same time, the computing load of centralized scheduling nodes increases non-linearly with the expansion of equipment scale, making it difficult to meet the scalability requirements of large-scale communication networks.

[0005] Existing input-output assessments mostly employ static, periodic accounting mechanisms, which cannot form a closed loop with real-time multimodal perception data from embodied intelligent agents. Configuration adjustments lag significantly behind fluctuations in business load and network status. Furthermore, existing methods lack pre-emptive communication layer feasibility verification of configuration schemes, which not only fails to reduce operation and maintenance costs but also increases the risk of communication network failures and hidden maintenance expenses. Summary of the Invention

[0006] The purpose of this invention is to provide a method for configuring intelligent operation and maintenance equipment that takes into account input and output, so as to solve one or more problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for configuring intelligent operation and maintenance equipment that takes into account input and output, comprising the following specific steps:

[0008] Furthermore, based on the link layer discovery protocol and the simple network management protocol, the topology connection relationship, port status, bandwidth resources and protocol type of the nodes in the target operation and maintenance network are actively scanned, and the network hierarchy is divided and the nodes are marked according to the core layer, aggregation layer and edge layer.

[0009] In-band network telemetry technology is used to perform secondary verification of the scanning results. The actual connectivity of the links and the real-time occupancy status of the ports are collected by hop-by-hop telemetry messages to correct topology deviations.

[0010] Collect static attributes and historical operating data of maintenance equipment to build equipment asset profiles, including information such as hardware procurement costs, energy consumption parameters, maintenance cycles, historical failure frequencies, and load fluctuation curves; establish a communication port resource ledger to record the network slice affiliation, reserved resource ratio, historical conflict frequency, and SLA level of the services carried by each physical port, and finally form a mapping relationship table of maintenance equipment, communication ports, network slices, and service nodes.

[0011] Furthermore, input-side and output-side indicator sets are constructed separately. The input side is divided into equipment costs and communication resource costs. Equipment costs include equipment purchase costs, operating energy consumption costs, maintenance manpower costs, and failure loss costs. Communication resource costs include three categories: exclusive slice resource costs, shared bandwidth premium costs, and port conflict implicit loss costs, which are used to quantify the input of communication resource occupation.

[0012] The output side is divided into equipment operation indicators and communication service indicators. Equipment operation indicators include equipment availability, data transmission success rate, average link latency, latency jitter, fault response time, and service support saturation. Communication service indicators include two core indicators: slice SLA compliance rate and service traffic continuity rate.

[0013] Based on the network node hierarchy, the priority of the carried services, and the SLA level of the corresponding slice, the entropy weight method is used to allocate the weight coefficients of each indicator. The core layer nodes focus on slice reliability, average link latency and latency jitter, and SLA compliance rate indicators, while the edge layer nodes focus on resource cost and energy efficiency indicators. A congestion transmission coefficient is set to calculate the implicit cost of adjacent link congestion caused by the core node configuration scheme.

[0014] Furthermore, the embodied intelligent agents deployed on each network node collect data on the physical operating status of the devices through sensors, and simultaneously use in-band network telemetry technology to collect fine-grained operational indicators of the communication layer, such as real-time traffic, hop-by-hop packet loss rate, latency jitter, and slice resource utilization rate of the corresponding network ports.

[0015] The boundary clock protocol is used to align the timestamps of physical operation status data and communication telemetry data. By combining port ID and spatial location identifier, the spatial dimension mapping of the two types of data is realized, forming a synchronous sensing dataset that corresponds one-to-one with physical device status, communication link status, and slice resource status, and finally generating a link traffic baseline fingerprint.

[0016] Furthermore, based on the topology ledger and sensing data, a digital twin mirror of the communication links of the target operation and maintenance network is constructed;

[0017] With the goal of maximizing the overall network input-output ratio, and with constraints such as equipment performance thresholds, link bandwidth limits, total port resources, routing convergence rules, and slice isolation requirements, an adaptive genetic algorithm is used to solve the initial configuration scheme. The scheme involves equipment deployment location, working parameter levels, communication port allocation, slice resource ratio, data transmission scheduling strategy, and inspection frequency settings.

[0018] The initial configuration scheme is imported into the digital twin image, and simulated traffic matching the characteristics of real business is injected for simulation.

[0019] The system verifies six communication metrics: link load rate, port conflict probability, average link latency, end-to-end latency jitter, slice isolation, and route convergence time. If any metric exceeds a preset threshold, the system returns to the solution stage for re-iteration until the solution passes the communication feasibility verification.

[0020] Furthermore, the configuration scheme is broken down into global scheduling tasks and edge execution sub-tasks according to the network layer. Global core configuration instructions, cross-domain route adjustments and slice resource redistribution are uniformly managed by the cloud, while lightweight configuration tasks that can be executed independently on the edge nodes are pushed down to the corresponding nodes for execution.

[0021] The command issuance is completed using a dual protocol stack consisting of message queue telemetry transmission protocol and network configuration protocol. High-priority global configuration commands are issued through the network configuration protocol signaling channel, while routine operation and maintenance configuration commands are issued through message queue telemetry transmission protocol.

[0022] Redundant communication links are configured for nodes carrying high-priority services. At the same time, a local arbitration module for signaling is configured on the edge nodes to perform port resource pre-occupancy verification on the instructions sent to the local node. If there is no conflict, the instructions are executed directly. If there is a conflict, the local arbitration module performs parameter fine-tuning before sending the instructions for execution.

[0023] Furthermore, the embodied intelligent agent parses the corresponding configuration instructions, performs physical configuration operations locally, and simultaneously completes remote parameter configuration and slice resource adjustment of the corresponding network device through the network configuration protocol;

[0024] During execution, the embodied intelligent agent transmits the operation progress at fixed intervals, and at the same time transmits the hop-by-hop link quality data and port resource usage data of the corresponding link through in-band network telemetry technology. The edge node synchronously completes local input-output verification and resource conflict detection. When an anomaly is detected, the edge node triggers a local rollback mechanism, restores the port and slice configuration through the network configuration protocol, and synchronously starts local traffic bypass scheduling.

[0025] Furthermore, collect equipment operation data and network communication data within the preset operating cycle, calculate the actual input-output ratio and the actual compliance rate of the sliced ​​SLA, and conduct a deviation analysis between the actual value and the expected value of the plan.

[0026] The traffic fingerprinting algorithm is used to locate the source of deviations, distinguish four types of deviation causes: traffic model deviation, routing policy deviation, port configuration deviation, and device performance deviation, and then make corresponding parameter corrections.

[0027] Validated configuration schemes are stored in the policy library of the corresponding network layer. The applicable topology features, traffic models, slice SLA requirements, communication environment constraints and input-output benchmarks are marked. Deviation tracing results are fed back to the digital twin mirror of the communication link to optimize the simulation accuracy of the twin model. The configuration effect of the entire network is summarized regularly, and the adaptive genetic algorithm and weight allocation rules are iteratively optimized.

[0028] The beneficial effects of this invention are as follows:

[0029] 1. This invention constructs an input-side indicator set, divided into two categories: equipment costs and communication resource costs. Equipment costs include equipment purchase costs, operating energy consumption costs, maintenance manpower costs, and failure loss costs. Communication resource costs include the cost of exclusive access to sliced ​​resources, the premium cost of shared bandwidth, and the implicit loss cost of port conflicts. Simultaneously, output-side indicator sets for equipment operation and communication services are established. Combining network node hierarchy and service priority, the entropy weight method is used to allocate the weights of each indicator, calculating the implicit cost of adjacent link congestion caused by node configuration. A digital twin image of the communication link is constructed. With the goal of maximizing the overall network input-output ratio, and constrained by conditions such as equipment performance and link bandwidth, the initial configuration scheme is solved. The scheme is imported into the twin image and simulated traffic is injected to complete the verification of six indicators. If the threshold is not met, iterative recalculation is performed. This scheme achieves global collaborative configuration of physical equipment and communication resources, which can avoid the risk of implicit communication failures and ensure the optimal balance between input costs and service output throughout the network's entire lifecycle.

[0030] 2. This invention, based on the network hierarchy of core layer, aggregation layer, and edge layer, decomposes the complete configuration scheme into two categories: global scheduling tasks and edge execution sub-tasks. Global operations such as global core configuration commands, cross-domain route adjustments, and slice resource reallocation are uniformly managed by the cloud, while lightweight configuration tasks that can be independently executed locally on edge nodes, such as port parameter configuration and inspection strategy adjustments, are offloaded to the corresponding nodes for execution. A dual-protocol stack is composed of a message queue telemetry transmission protocol and a network configuration protocol to complete hierarchical command distribution. High-priority global commands are distributed through the network configuration protocol signaling channel, while routine operation and maintenance configuration commands are distributed through the message queue telemetry transmission protocol. Edge nodes are equipped with a local arbitration module for configuration signaling to perform port resource pre-occupancy verification on local commands. If resource conflicts exist, local parameter fine-tuning is performed. This architecture can alleviate the centralized computing load on the cloud, improve the scalability of large-scale communication networks, enhance the configuration response speed on the edge side, and reduce the risk of configuration command failure when the network is partially disconnected.

[0031] 3. This invention deploys embodied intelligent agents on each network node to parse corresponding configuration instructions, execute physical configuration operations locally, and simultaneously complete remote parameter configuration and slice resource adjustment for corresponding network devices through network configuration protocols. During execution, the embodied intelligent agents transmit operation progress and link status data back at fixed intervals. Edge nodes synchronously perform local input-output verification and resource conflict detection. When an anomaly is detected, a local configuration rollback is triggered, and local traffic bypass scheduling is initiated simultaneously. Device and network data are collected during the operating cycle, actual input-output indicators are calculated, and four types of deviation causes are located and parameters are corrected through a traffic fingerprint matching algorithm. The policy library is updated and the twin model and solution algorithm are optimized. This adapts to real-time fluctuations in business load and network status, continuously improving the accuracy of configuration schemes and reducing the hidden overhead of network operation and maintenance. Attached Figure Description

[0032] Figure 1 This is a flowchart of the intelligent operation and maintenance equipment configuration method of the present invention, which takes into account input and output.

[0033] Figure 2 This is a flowchart of the twin verification sub-configuration scheme of the present invention. Detailed Implementation

[0034] 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.

[0035] like Figures 1 to 2As shown, this embodiment of the invention provides a method for configuring intelligent operation and maintenance equipment that takes into account input and output;

[0036] This embodiment takes the provincial government communication backbone network from the city-level core node to the county-level edge access node as the application scenario. Based on the link layer discovery protocol and simple network management protocol, it actively scans the topology connection relationship, port status, bandwidth resources and protocol type of all network nodes, and completes the network layer division and node marking according to the core layer city-level core router, the aggregation layer county-level aggregation optical transmission equipment, and the edge layer township government access switch.

[0037] In-band network telemetry technology is used to perform secondary verification of the scanning results. Telemetry probe messages are injected path-by-path from the city-level core router to the township-level government access switches. When a message passes through each node, the device identifier, ingress port number, egress port number, and real-time port occupancy data are automatically embedded. The cloud aggregates the path information of all probe messages to generate an actual connectivity topology matrix, which is compared node-by-node with the topology matrix obtained from the static scan. If the number of adjacent nodes or the port correspondence is inconsistent, the topology ledger is updated based on the in-band telemetry data, and the deviation nodes are marked and the deviation type is recorded.

[0038] Collect static attributes and historical operating data of maintenance equipment at all levels to construct equipment asset profiles that include hardware procurement costs, energy consumption parameters, maintenance cycles, historical failure frequencies, and load fluctuation curves; establish a communication port resource ledger to record the network slice affiliation, reserved resource ratio, historical conflict frequency, and SLA level of the services carried by each physical port for the three types of network slices: government office, emergency command, and public services, forming a multi-element mapping relationship table of maintenance equipment, communication ports, network slices, and government business nodes.

[0039] The mapping table uses communication ports as the core associated nodes. A single maintenance device corresponds to multiple physical communication ports, a single physical port corresponds to one or more logical network slices, and a single network slice corresponds to multiple government service nodes. The mapping table is equipped with a dynamic update trigger mechanism. When four types of events occur, such as device addition / removal, port expansion, slice creation / deregistration, and service node migration, the corresponding entries are automatically updated. Among them, device addition / removal and port expansion events synchronously update the associated data in the topology ledger and the port resource ledger, while slice creation / deregistration and service node migration events only update the port resource ledger and the mapping table, ensuring logical consistency of data across multiple ledger types.

[0040] This embodiment constructs an input-side indicator set and an output-side indicator set respectively. The input side is divided into equipment costs and communication resource costs. Equipment costs include equipment purchase costs, operating energy consumption costs, maintenance manpower costs, and failure loss costs. Communication resource costs include three categories: exclusive slice resource costs, shared bandwidth premium costs, and port conflict implicit loss costs, which are used to quantify the input of communication resource occupation.

[0041] The implicit cost of port conflicts is calculated based on a conflict probability model built from historical conflict data for the corresponding port. This model, combined with the port resource utilization rate, SLA level of the service being carried, and the baseline value of a single conflict loss under the current configuration, calculates the expected implicit cost of port conflicts under the current configuration and includes it in the total investment cost of the corresponding node. The baseline value of a single conflict loss is derived from statistics of the duration of historical conflicts and the corresponding service's SLA compensation standard.

[0042] The output side is divided into equipment operation indicators and communication service indicators. Equipment operation indicators include equipment availability, data transmission success rate, average link latency, latency jitter, fault response time, and service support saturation. Communication service indicators include two core indicators: slice SLA compliance rate and service traffic continuity rate.

[0043] The service support saturation is calculated using a single network communication device as the statistical unit. It is the ratio of the peak value of the government service traffic actually carried by the device within the statistical period to the rated capacity threshold of the device. Combined with the ratio of the number of government service nodes simultaneously carried by the device to the rated number of access points, the weighted average of the two ratios is taken as the final saturation value, which represents the degree of service capacity utilization of network communication device resources.

[0044] Based on the network node level, the priority of carrying government services, and the SLA level of the corresponding slice, the entropy weight method is used to allocate the weight coefficients of each indicator. First, the objective basic weight is calculated based on the historical operating data of each indicator using the entropy weight method. Then, the weight correction coefficient is set according to the three dimensions of the network level to which the node belongs, the priority of the carried services, and the SLA level of the slice.

[0045] For core router nodes in prefecture-level cities, the correction coefficient for communication service indicators is higher than that for equipment cost indicators, with an emphasis on slice reliability, average link latency and latency jitter, and SLA compliance rate. For access switch nodes in townships, the correction coefficient for resource cost indicators is higher than that for communication service indicators, with an emphasis on resource cost and energy efficiency indicators.

[0046] The correction coefficients for indicators corresponding to high-priority emergency command operations increase synchronously with the upgrade of the operation level. The basic weights are multiplied by the corresponding correction coefficients and then normalized to obtain the final weight coefficients of each indicator adapted to the current node.

[0047] The congestion propagation coefficient is set to offset the implicit cost of adjacent link congestion caused by the core node configuration scheme. The congestion propagation coefficient is set comprehensively based on the hierarchical position of the core node, the number of adjacent links, and the priority of the services carried. The propagation coefficient of the core router node in the prefecture-level city is higher than that of the aggregation node in the district and county.

[0048] The calculation first calculates the port resource utilization and link load rate of the core node itself based on the initial configuration scheme, obtaining the load rate increment relative to the baseline state. Then, this increment is multiplied by the congestion transmission coefficient of the corresponding adjacent link to obtain the expected load increment of the adjacent link. Combining the total bandwidth resources of the adjacent links, and based on the extent to which the expected load increment exceeds the link's safe load threshold, the calculation matches the collision probability and loss baseline value under historical congestion scenarios, converting it into the corresponding expected implicit congestion loss cost, which is then included in the total investment cost of the core node configuration scheme.

[0049] Formula for calculating the hidden cost of congestion in adjacent links:

[0050]

[0051] The expected load increment of adjacent links satisfies: The probability of a conflict is obtained by matching historical data with the magnitude of the expected load increase exceeding the link security threshold.

[0052] This represents the implicit cost of expected congestion in adjacent links caused by core node configuration, which is included in the total investment cost of the core node configuration scheme.

[0053] This represents the congestion propagation coefficient, which is determined comprehensively based on the core node's hierarchical position, the number of adjacent links, and the priority of the services it carries.

[0054] This represents the increase in the load rate of the core node relative to the baseline state after configuration.

[0055] This represents the expected load increment of adjacent links after conversion by the transmission coefficient;

[0056] This indicates the probability of congestion conflicts occurring on adjacent links, obtained by matching the magnitude of the expected load increase exceeding the safety threshold with historical congestion scenario data;

[0057] This represents the baseline loss value for a single congestion conflict on the corresponding link, which is calculated based on the duration of historical conflicts and the business SLA compensation standard.

[0058] The embodied intelligent agents deployed in core data centers at the city and county levels collect data on device panel indicator light status, device surface temperature, fan speed, and optical module parameters using various sensors, including vision, touch, and RFID. This data, acquired through these sensors, is then uploaded to edge nodes. Simultaneously, the network management and acquisition terminal uses in-band network telemetry technology to collect fine-grained communication layer operational metrics such as real-time traffic, hop-by-hop packet loss rate, latency jitter, and slice resource utilization for the corresponding network ports, and synchronizes these metrics to the edge nodes.

[0059] A boundary clock protocol is used to align the physical operation status data and communication telemetry data with timestamps. By combining the port ID and the rack location identifier in the data center, the spatial dimension mapping of physical data and communication data is completed, forming a synchronous sensing dataset with one-to-one correspondence between physical device status, communication link status, and slice resource status.

[0060] A baseline fingerprint of link traffic is generated for each 10G physical link from a single municipal core router to a district / county aggregation device. Feature data from seven consecutive stable working days in the synchronous sensing dataset is extracted, including four core features: time-series distribution characteristics of government business traffic, port resource occupancy range, latency jitter fluctuation range, and average resource occupancy of three types of slices. After feature standardization, the baseline fingerprint of the corresponding link is formed.

[0061] Baseline fingerprints are fully updated monthly, using feature data from seven consecutive stable working days. When service slice adjustments or equipment expansion / upgrades occur, temporary corrections to the corresponding link's baseline fingerprints are triggered. These temporary baselines are generated using perception data from three consecutive stable working days following the event, and a formal baseline update is completed after a total of seven stable working days. When network topology changes, a new baseline collection process is initiated for newly added links, and the baseline fingerprints for disappeared links are invalidated and archived.

[0062] Based on topology ledgers and synchronous sensing data, a digital twin mirror of the communication links of the provincial government communication backbone network is constructed. This digital twin mirror includes four modules: topology mapping, traffic simulation, constraint verification, and result output. The output of the topology mapping module is connected to the input of the traffic simulation module, and the output of the traffic simulation module is connected to the input of the constraint verification module. The constraint verification module is equipped with a feedback branch. When the verification fails, the deviation result is fed back to the configuration solution stage. When the verification passes, the result is output to the result output module, forming an iterative closed-loop simulation link.

[0063] The topology mapping module outputs a virtual topology structure that corresponds one-to-one with the physical network nodes and links; the traffic simulation module outputs full-link simulation data on link load, port occupancy, and latency jitter; the constraint verification module outputs verification pass indicators and out-of-limit values; and the result output module summarizes all data and outputs a scheme simulation report.

[0064] The objective function is to maximize the overall network input-output ratio, with constraints including device performance thresholds, link bandwidth limits, total port resources, routing convergence rules, and slice isolation requirements. The routing convergence rules limit the maximum number of hops for a single government service transmission path to no more than 8 hops and the number of equivalent detour paths for a single service link to no more than 4. All generated transmission paths must simultaneously meet both the maximum hop count and the number of equivalent detour paths requirements; transmission paths that do not meet these requirements are directly eliminated. Furthermore, the routing convergence time after a topology change is capped at 50ms and is included as a network-wide device-level constraint, incorporated into the overall verification dimension of the configuration scheme. If this constraint is not met, the overall configuration scheme is deemed infeasible, and the process returns to the solution iteration stage.

[0065] The objective function for the overall network's input-output ratio is:

[0066]

[0067] This represents the overall input-output ratio of the entire network, which is the core objective of configuration optimization. The larger the value, the better the input-output efficiency.

[0068] This represents the standardized value of the i-th output-side indicator, covering two categories of indicators: equipment operation and communication services.

[0069] The weight coefficient corresponding to the i-th output-side indicator is obtained by combining the entropy weight method with node level and business priority.

[0070] This indicates the total number of output-side indicators;

[0071] This represents the standardized value of the j-th input-side indicator, covering two categories of indicators: equipment costs and communication resource costs.

[0072] The weight coefficient corresponding to the j-th input-side indicator is obtained by combining the entropy weight method with node level and business priority.

[0073] This indicates the total number of input-side indicators.

[0074] An adaptive genetic algorithm is used to solve for the initial configuration scheme, which includes equipment deployment locations, operating parameter levels, communication port allocation, resource ratios of three types of slices, data transmission scheduling strategies, and inspection frequency settings. The algorithm uses a mixed integer encoding method, sequentially concatenating the equipment deployment location number, port number, slice resource percentage, scheduling strategy number, and inspection frequency value into a single chromosome. After inputting the topology ledger, port resource ledger, indicator weight coefficients, and constraint thresholds, the algorithm outputs the optimal configuration scheme chromosome that satisfies all constraints. Decoding this chromosome yields the complete set of initial configuration parameters.

[0075] The fitness value of the algorithm is the overall network input-output efficiency value of the corresponding configuration scheme. First, all input cost indicators are positively normalized, converting them into standardized scores where higher values ​​indicate lower input costs. Similarly, all output indicators are positively normalized, converting them into standardized scores where higher values ​​indicate better service output. These two scores are then multiplied by their respective weights to obtain a weighted sum of inputs and a weighted sum of outputs. The ratio of the weighted sum of outputs to the weighted sum of inputs is used as the fitness value.

[0076] The formula for calculating the fitness of an adaptive genetic algorithm is as follows:

[0077]

[0078] The fitness value represents the individual fitness value corresponding to a single configuration scheme in a genetic algorithm, and is the core evaluation criterion for iterative optimization of the algorithm;

[0079] This represents the positive standardized value of the i-th output-side indicator; the higher the value, the better the corresponding service output performance.

[0080] This represents the positive normalized value of the j-th input-side cost indicator. After conversion, a higher value indicates a lower corresponding input cost.

[0081] This represents the weight coefficient corresponding to the i-th output-side indicator;

[0082] This represents the weight coefficient corresponding to the j-th input-side indicator;

[0083] This indicates the total number of output-side indicators;

[0084] This indicates the total number of input-side indicators.

[0085] The crossover and mutation probabilities are adjusted using an adaptive rule. When the difference between the average fitness of the population and the fitness of the best individual is greater than a preset threshold, the crossover probability is set to 0.7 and the mutation probability is set to 0.05. When the difference is less than or equal to the preset threshold, the crossover probability decreases linearly to 0.4 with the number of iterations, and the mutation probability increases linearly to 0.1. The algorithm sets a dual iteration termination condition: the rate of change of the best fitness is less than 0.1% for 20 consecutive generations, or the total number of iterations reaches 200. The optimization stops when either condition is met.

[0086] The initial configuration scheme is imported into the digital twin image, and simulated traffic matching the real business characteristics of government office, emergency command, and public service is injected for simulation. The six communication indicators after configuration are verified: link load rate, port conflict probability, average link latency, end-to-end latency jitter, slice isolation, and route convergence time. If any indicator exceeds the preset threshold, the solution is returned to the solution stage for re-iteration until the scheme passes the communication feasibility verification.

[0087] During the slice isolation verification, traffic transmission path, port resource usage, and bandwidth quota data of the three types of government slices are extracted respectively to check whether there are port resource exclusive conflicts or transmission path resource preemption between different slices; the deviation between the actual bandwidth ratio and the quota ratio of each slice is calculated. If the deviation exceeds the preset range or cross-slice resource illegal occupation occurs, it is determined that the slice isolation does not meet the requirements.

[0088] The configuration scheme is broken down into global scheduling tasks and edge execution sub-tasks according to the network layer. The task decomposition follows three judgment criteria: scope of impact, cross-domain resource accessibility, and business priority. Emergency command slice resource reallocation across districts and counties, cross-aggregation layer routing adjustment, and configuration changes affecting more than 3 township nodes are judged as global scheduling tasks and are uniformly managed by the cloud. Port parameter configuration of access switches in a single township, local inspection policy adjustment, and public service slice resource fine-tuning that do not affect adjacent nodes are judged as edge execution sub-tasks and are assigned to the corresponding edge nodes for execution. After the decomposition is completed, a task hierarchy list is generated, marking the responsible node, execution time limit, and priority level of each task.

[0089] The system employs a dual-protocol stack consisting of a message queue telemetry transmission protocol and a network configuration protocol to issue commands. High-priority commands such as global scheduling, emergency fault handling, and emergency command service configuration adjustments are issued through the network configuration protocol signaling channel; routine commands such as daily maintenance and inspection adjustments are issued through the message queue telemetry transmission protocol.

[0090] The two types of protocols are configured with two transmission links, one primary and one backup. Each link has an independent heartbeat detection mechanism and sends a link probe message every 10 seconds. When the primary transmission link of a single protocol loses three consecutive probe messages, it automatically switches the pending instructions of that protocol to its own backup transmission link for transmission, and records the channel abnormal event. After the channel is restored, it switches back to the original protocol channel.

[0091] Redundant communication links are configured for nodes carrying emergency command services. At the same time, local arbitration modules for signaling are set up at the edge nodes of districts and counties to perform port resource pre-occupancy verification on instructions sent to the local area. During the verification, all physical port numbers, slice resource quotas and bandwidth usage values ​​involved in the configuration instructions are extracted and compared with the currently occupied resources in the local port resource ledger. When the cumulative reserved resource ratio of the target port is less than 70% of the safety threshold, it is determined that there is no resource conflict and the request is directly allowed.

[0092] When the cumulative reserved resource ratio exceeds the safety threshold, the arbitration module will first fine-tune the parameters within the locally schedulable elastic bandwidth margin; if the threshold requirement still cannot be met, the resource conflict information will be immediately reported to the cloud management node to apply for global slice resource quota adjustment, and the configuration will be issued after the cloud returns the adjustment instruction.

[0093] The embodied intelligent agent parses the corresponding configuration instructions and first breaks them down into two categories: physical operation items and network configuration items. The two types of operation items are associated with a unified task number. The physical operation items are completed by the mechanical execution unit of the embodied intelligent agent, including on-site physical operations such as optical module plugging and unplugging, port jumper adjustment, and equipment power switch. Once an operation is completed, a corresponding operation completion identifier is sent back.

[0094] The network configuration item is sent to the target device by the embodied intelligent agent communication unit through the network configuration protocol. It completes remote parameter configuration such as port VLAN division and slice bandwidth adjustment. After receiving the configuration success response from the device, it records the corresponding completion flag. After all associated physical operation items and network configuration items return completion flags, it is determined that a single configuration command has been executed and enters the status feedback stage of the next cycle.

[0095] During execution, the embodied intelligent agent transmits the operation progress at a fixed interval of 5 minutes; the hop-by-hop link quality data and port resource usage data of the corresponding link are synchronously collected by the network management and acquisition terminal through in-band network telemetry technology and reported to the edge node.

[0096] Data is categorized into two types based on its urgency: emergency data such as configuration execution anomalies, link failures, and port conflicts are transmitted immediately via the network configuration protocol channel; routine data such as regular operation progress and periodic link quality are transmitted in batches via the message queue telemetry transmission protocol.

[0097] The backhaul links of the two types of protocols are configured with two transmission paths, one primary and one backup. Each link has an independent heartbeat detection mechanism. When the primary backhaul link of a single protocol loses three consecutive probe messages, the backhaul data of that protocol will be automatically switched to its own backup transmission path.

[0098] The district / county edge nodes simultaneously complete local input-output verification and resource conflict detection. Taking the jurisdiction of a single district / county as the boundary, they extract local equipment energy consumption and port resource usage data within the current period as input items, and extract local link availability and government business traffic continuity rate data as output items. They use weight coefficients matched with edge layer nodes to calculate the local input-output ratio. If the ratio is lower than the preset range of the benchmark value, or if the port resource usage exceeds the safety threshold, it is judged as a configuration anomaly.

[0099] Upon detecting an anomaly, edge nodes trigger partial rollback and traffic rerouting commands in parallel, executing them sequentially: First, rollback commands are sent to the corresponding intelligent agents and network devices to restore port configurations and slice resource allocations to their baseline states before the configuration operation; Simultaneously, traffic rerouting requests are reported to the cloud management node, and within the backup access links under the jurisdiction of the district / county, affected business traffic is rerouted locally according to the priority of emergency command, government affairs, and public services from high to low; Rerouting scheduling across districts / counties is uniformly coordinated by the cloud and executed by adjacent nodes, and after the configuration rollback is completed and the link status returns to normal, the rerouting traffic is gradually switched back to the original link for transmission.

[0100] Collect equipment operation data and network communication data within the monthly operation cycle, calculate the actual input-output ratio and the actual compliance rate of the three types of government affairs slice SLA, and conduct deviation analysis between the actual values ​​and the expected values ​​of the plan;

[0101] The source of deviations was located using a link traffic baseline fingerprinting algorithm, identifying four types of deviation causes and correcting the corresponding parameters. These included: traffic model deviations caused by unexpectedly high traffic during peak government service periods, which were addressed by adjusting the simulated traffic distribution parameters and peak percentage within the digital twin mirror; routing strategy deviations caused by excessive hop counts in cross-county routes, which were addressed by adjusting the routing convergence constraints and path selection priorities in the configuration solution stage; port configuration deviations caused by unreasonable slice bandwidth quotas, which were corrected by adjusting the slice resource quotas and reserved bandwidth percentages for the corresponding ports; and equipment performance deviations caused by aging equipment leading to decreased availability, which were addressed by adjusting the equipment operating parameter levels and inspection frequency settings. The corrected parameters were then synchronously updated to the corresponding configuration scheme and the digital twin model parameter library.

[0102] The input to the link traffic baseline fingerprint matching algorithm is the link hop-by-hop traffic time-series data, port occupancy data, latency and latency jitter data collected in the current running cycle, as well as the pre-stored link traffic baseline fingerprint database.

[0103] The algorithm first extracts four types of feature vectors from the current running data: peak traffic periods, traffic fluctuation variance, port conflict period proportion, and latency distribution characteristics. Then, it calculates the similarity with the baseline feature vectors of the corresponding links in the baseline fingerprint database one by one.

[0104] The judgment rules are as follows: When the similarity between the peak traffic period and the fluctuation variance is less than 0.8, it is judged as a traffic model deviation; when the similarity of the delay distribution characteristics is less than 0.8 and the difference between the route hop count and the baseline is greater than 2, it is judged as a routing policy deviation; when the peak traffic characteristics do not deviate significantly from the baseline, but the difference between the peak port occupancy rate and the baseline is greater than 30% and the proportion of conflict periods is more than twice that of the baseline, it is judged as a port configuration deviation; when none of the above characteristics deviate significantly, but the difference between the device availability rate and the baseline is greater than 5%, it is judged as a device performance deviation. The algorithm finally outputs the deviation cause label and the corresponding deviation quantification value.

[0105] Validated configuration schemes are stored in the policy library of the corresponding network layer, and the applicable topology features, government traffic models, slice SLA requirements, communication environment constraints and input-output benchmark values ​​are marked. Deviation tracing results are fed back to the digital twin mirror of the communication link. The loss function is constructed by the difference between the actual operation data and the simulation data. The gradient descent method is used to update the two core parameters of the traffic simulation module, namely the traffic propagation coefficient and the link loss coefficient. The parameter learning rate is fixed at 0.02.

[0106] After acquiring full running data for each single cycle, multiple rounds of iterative optimization are performed based on that batch of data until the average relative error between the simulation output and the actual data is less than 5% or the maximum number of iterations is reached, at which point the current update is stopped, thus optimizing the simulation accuracy of the twin model.

[0107] Each quarter, the overall network configuration effect is summarized, and the adaptive genetic algorithm and weight allocation rules are iteratively optimized. All valid configuration schemes and actual input-output data within the period are summarized, the coding features and constraint parameters of high-fitness schemes are extracted, and the initial population generation rules and crossover mutation parameter boundaries of the adaptive genetic algorithm are updated. Based on the full range of indicator operation data within the period, the entropy weight base value of each indicator is recalculated, and the weight correction coefficient matrix is ​​updated in combination with the changes in node level and government business priority.

[0108] The strategy library is stored independently in a three-tier architecture of core layer, aggregation layer and edge layer. Each level of strategy library only stores the configuration schemes that can be executed at the corresponding level.

[0109] When calling the strategy library, the core parameters of the current network scenario, such as topology, service level, and resource reserves, are first extracted and matched one by one with the marked parameters of the solutions in the strategy library. The solution with the highest matching degree and the best input-output baseline value is selected as the reference baseline for the initial solution.

[0110] Each time a new valid solution is added, it is automatically compared with the old solution under the same applicable conditions. If the new solution has better input and output performance, it will replace the old solution and retain the historical version record. Historical solutions that have been valid for more than 1 year will be cleaned up regularly.

[0111] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[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 equivalents.

Claims

1. A method for configuring personalized intelligent operation and maintenance equipment that takes into account input and output, characterized in that, The specific steps include the following: Scan the topology and port resource status of the target network to complete the network hierarchy division; collect data from maintenance equipment, establish a communication port resource ledger, and form a mapping relationship between maintenance equipment, communication ports, network slices, and service nodes; Construct an input-side indicator set and an output-side indicator set. The input side includes equipment costs and communication resource costs, while the output side includes equipment operation indicators and communication service indicators. Allocate the weight of each indicator based on network node hierarchy and business priority. Collect physical operation data of maintenance equipment and communication operation data of corresponding communication ports, complete the timestamp alignment and spatial mapping of the two types of data, and generate link traffic baseline fingerprint; A digital twin mirror of the communication link is constructed. With the goal of maximizing the input-output ratio of the entire network, and with constraints such as equipment performance, link bandwidth, port resources, routing convergence rules and slice isolation requirements, the initial configuration scheme is obtained by solving the problem. The solution is imported into the digital twin image of the communication link, and simulated traffic is injected to complete the verification. The initial configuration scheme is broken down into global scheduling tasks and edge execution sub-tasks according to the network hierarchy, with the cloud and edge nodes respectively undertaking the management and control responsibilities. A dual-protocol stack is used to complete the hierarchical distribution of instructions; The embodied intelligent agent performs physical configuration and network device parameter configuration, and transmits operation progress and link status data back at fixed intervals; the edge nodes synchronously perform local input-output verification and resource conflict detection, and trigger configuration rollback when an anomaly occurs, and synchronously start local traffic bypass scheduling. Collect equipment and network data during the operation cycle, calculate actual input and output indicators, complete deviation tracing and parameter correction; update the strategy library, and iteratively optimize the digital twin model, configure the solution algorithm and indicator weight allocation rules.

2. The method for configuring personalized intelligent operation and maintenance equipment that takes into account input and output as described in claim 1, characterized in that, Based on the link layer discovery protocol and simple network management protocol, the topology, port status, bandwidth resources and protocol type of network nodes are scanned, and the network layers are divided and nodes are marked according to the core layer, aggregation layer and edge layer. The scanning results are verified twice using in-band network telemetry technology to correct topology deviations; static attributes and historical operating data of the maintenance equipment are collected to construct equipment asset profiles, and a resource ledger for the communication ports is established, recording the slice ownership, reserved resource ratio, historical conflict frequency, and service SLA level of the communication ports, forming a multi-element mapping relationship table of maintenance equipment, communication ports, network slices, and service nodes.

3. The method for configuring personalized intelligent operation and maintenance equipment that takes into account input and output, as described in claim 2, is characterized in that... The input-side indicator set is divided into equipment costs and communication resource costs; the equipment costs include equipment purchase costs, operating energy consumption costs, maintenance manpower costs, and failure loss costs; The communication resource costs include the cost of exclusive access to slice resources, the cost of shared bandwidth premium, and the cost of implicit losses due to port conflicts; the output-side indicator set is divided into equipment operation indicators and communication service indicators. The equipment operation metrics include equipment availability, data transmission success rate, average link latency, latency jitter, fault response time, and service support saturation; the communication service metrics include slice SLA compliance rate and service traffic continuity rate.

4. The method for configuring personalized intelligent operation and maintenance equipment that takes into account input and output, as described in claim 3, is characterized in that... Combining the network node hierarchy, service priority, and slice SLA level, the entropy weight method is used to allocate the weight coefficients of each indicator to calculate the implicit cost of adjacent link congestion caused by node configuration. Core layer nodes focus on slice reliability, average link latency, latency jitter, and SLA compliance rate, while edge layer nodes focus on resource cost and energy efficiency. A congestion propagation coefficient is set to quantify and calculate the implicit cost of congestion in adjacent links caused by the configuration of core nodes.

5. The method for configuring personalized intelligent operation and maintenance equipment that takes into account input and output, as described in claim 4, is characterized in that... Each network node's embodied intelligent agent collects physical operating status data of the device through sensors, and the network management and acquisition terminal simultaneously collects fine-grained operating indicators of the communication port through the in-band network telemetry technology. The two types of data are timestamped using a boundary clock protocol. The spatial mapping is completed by combining the port ID and spatial location identifier to form a synchronous sensing dataset and generate the link traffic baseline fingerprint.

6. The method for configuring personalized intelligent operation and maintenance equipment that takes into account input and output, as described in claim 5, is characterized in that... A digital twin image of the communication link is constructed based on the topology ledger and sensing data; with the goal of maximizing the input-output ratio of the entire network, and with constraints such as device performance threshold, link bandwidth limit, total port resources, routing convergence rules, and slice isolation requirements, an adaptive genetic algorithm is used to solve the initial configuration scheme, which includes device deployment location, working parameter level, port allocation, slice resource ratio, transmission scheduling strategy, and inspection frequency.

7. The method for configuring personalized intelligent operation and maintenance equipment that takes into account input and output as described in claim 6, characterized in that, The initial configuration scheme is imported into the digital twin image of the communication link, and the simulated traffic matching the real service characteristics is injected for simulation. The system verifies six metrics: link load rate, port conflict probability, average link latency, latency jitter, slice isolation, and route convergence time. If any of these metrics exceeds the preset threshold, the system returns to the solution iteration stage until the solution passes the communication feasibility verification.

8. The method for configuring personalized intelligent operation and maintenance equipment that takes into account input and output, as described in claim 7, is characterized in that... Global core configuration, cross-domain route adjustment, and the redistribution of slice resources are uniformly managed by the cloud, while edge local lightweight configuration tasks are devolved to the corresponding edge nodes for execution. The instructions are issued using a dual protocol stack consisting of a message queue telemetry transmission protocol and a network configuration protocol. High-priority global instructions are issued through the network configuration protocol channel, while routine operation and maintenance instructions are issued through the message queue telemetry transmission protocol. Edge nodes perform port resource pre-occupancy verification on local commands, and complete local parameter fine-tuning if conflicts exist; the edge nodes are equipped with a local arbitration module, which is specifically responsible for the execution of port resource pre-occupancy verification and conflict parameter fine-tuning.

9. A method for configuring personalized intelligent operation and maintenance equipment that takes into account input and output, as described in claim 8, is characterized in that, The embodied intelligent agent parses the configuration instructions, executes the physical configuration operation locally, and simultaneously completes the remote parameter configuration of network devices and the adjustment of slice resources through the network configuration protocol. During execution, the embodied intelligent agent returns the operation progress according to the fixed period. The link quality and the port resource usage data are collected and reported synchronously by the in-band network telemetry. The edge node synchronously performs local input-output verification and conflict detection. In case of an anomaly, a local rollback is triggered to restore the port and slice configuration and synchronously start the local traffic bypass scheduling.

10. A method for configuring personalized intelligent operation and maintenance equipment that takes into account input and output, as described in claim 9, is characterized in that, Collect equipment and network communication data within a preset operating cycle, calculate the actual input-output ratio and the SLA compliance rate of the slice, and compare the actual value with the expected value to perform deviation analysis. The link traffic baseline fingerprint matching algorithm is used to locate the source of deviation, distinguish four types of deviation causes, and complete the corresponding parameter correction. Valid solutions are stored in the corresponding level of the strategy library, applicable conditions and input-output benchmark values ​​are marked, and the deviation results are fed back to optimize the simulation accuracy of the digital twin image of the communication link. The configuration solution algorithm and the indicator weight allocation rules are periodically iterated and optimized.

Images