CN-121567700-B - Communication load self-adaptive scheduling method for intelligent lamp post edge node

CN121567700BCN 121567700 BCN121567700 BCN 121567700BCN-121567700-B

Abstract

The embodiment of the application provides a communication load self-adaptive scheduling method of an intelligent lamp post edge node, which is applied to the technical field of edge calculation and Internet of things communication and comprises the steps of collecting electrical, network, queue and environment multidimensional data of the node, and forming a unified node running state data set through filtering, alignment and fusion processing; the method comprises the steps of calculating instant communication load through a multi-factor fusion model, generating a load state mapping table by utilizing a dynamic threshold, intelligently decomposing tasks of high-load nodes according to the mapping table, carrying out double-factor matching decision by combining a real-time link state and node resources, realizing accurate migration from task segments to low-load nodes, monitoring in real time and dynamically adjusting in a migration process, and optimizing weight parameters of a load calculation model based on load reports after each round of scheduling in a self-adaptive feedback mode. The application realizes accurate assessment and forward-looking early warning of node states, completes efficient and reliable refined task scheduling, and forms a closed-loop self-optimizing intelligent system.

Inventors

ZHANG YUEGANG
LI YAN
BAO GE
LI HAIFEI
CHEN BAOWEI
CHEN RONG

Assignees

上海茗岳信息科技有限公司

Dates

Publication Date: 20260508
Application Date: 20260121

Claims (9)

1. A method for adaptively scheduling communication load of an edge node of a smart lamp post, the method comprising: in an intelligent lamp post network, acquiring current, voltage, processing queue length, communication delay, packet loss rate, ambient light, wind speed and temperature of each edge node to form a node running state data set; Concurrently extracting an operation parameter subset for load calculation according to the node identification, wherein the operation parameter subset at least comprises an electrical parameter subset representing the energy supply state of the node, a network parameter subset representing the communication state and an environment disturbance field; calculating the power consumption value of the node in the current scheduling period according to a node energy supply model based on the electric parameter subset in the operation parameter subset, and calculating the network congestion value reflecting the communication congestion degree based on the network parameter subset in the operation parameter subset; Respectively executing normalization and scale mapping processing on the power consumption value and the network congestion value, and converting the power consumption value and the network congestion value into a standardized power consumption factor and a standardized congestion factor which can be compared; according to a preset weight relation, performing weighted fusion calculation on the standardized power consumption factor, the standardized congestion factor and the environment disturbance field, taking a calculation result as an instant communication load of a corresponding edge node, writing the instant communication load into a load field of the node running state data set, and storing a load value of each node and an identifier of whether the load value exceeds a preset threshold value in an associated manner to form a load state mapping table; Screening nodes marked as the super threshold value according to the load state mapping table to form a high load node set, decomposing the communication task of the high load node into migratable data fragments, and generating a scheduling candidate task list; Matching target low-load nodes for each migratable data segment in the list based on the scheduling candidate task list and combining with the communication link state between the nodes, forwarding the migratable data segments to the corresponding target low-load nodes according to the matching result, and generating a task migration execution log; Collecting communication delay, packet loss rate and processing queue length of each node in the migration process, and combining the collected result with the task migration execution log to form load adjustment data, wherein the load adjustment data is used for adjusting the transmission sequence and migration path of unfinished tasks; According to the load adjustment data, dispatching execution instructions are issued to the high-load nodes and the receiving nodes, task migration and load distribution are carried out on the communication tasks, and a final node load state report is generated; And carrying out self-adaptive feedback adjustment on the calculation parameters of the subsequent scheduling process based on the final node load state report.
2. The method of claim 1, wherein the obtaining the current, the voltage, the processing queue length, the communication delay, the packet loss rate, and the ambient light, the wind speed, and the temperature of each edge node to form the node operation state data set comprises: acquiring the electrical, network and environment original data of a node to form a multi-source original data set, wherein the multi-source original data set comprises original electrical data, original queue data, original network data and original environment data; Aiming at the original electrical data in the multi-source original data set, performing self-adaptive filtering and abnormal suppression processing by combining the fluctuation characteristic of the node power supply and the sampling period characteristic, eliminating the influence of instantaneous interference and random noise, and obtaining clean electrical parameters capable of stably representing the change trend of the node energy supply state; performing timestamp reconstruction and alignment processing on the original queue data in the multi-source original data set by taking the clean electrical parameters as time references, and controlling a node to calculate a corresponding relation between a load state and a node energy supply state under a unified time axis to generate time synchronization node core operation data reflecting the node core operation state; Performing cross-domain fusion on the time synchronization node core operation data, the original network data and the original environment data in the multi-source original data set according to the node unique identification, constructing a node-level multi-dimensional state description structure, and performing association packaging on the node internal operation state and the external environment state; And outputting the node-level multidimensional state description structure which is packaged by the association as a node running state data set for subsequent scheduling and analysis.
3. The method of claim 2, further comprising, after outputting the node operational state data set: Extracting historical illumination intensity data and historical temperature data from the node running state data set, and extracting a real-time illumination intensity value and a real-time temperature value at the current moment; Performing statistical modeling based on the time distribution characteristics of the historical illumination intensity data, calculating an illumination intensity reference value representing the environmental stability level of the node, and calculating a corresponding temperature reference value based on the periodic fluctuation characteristics of the historical temperature data; Respectively performing deviation calculation on the real-time illumination intensity value and the real-time temperature value by taking the illumination intensity reference value and the temperature reference value as references to obtain an illumination disturbance value and a temperature disturbance value representing the instantaneous change degree of the environment; performing joint modeling on the illumination disturbance value and the temperature disturbance value, and generating a single environment disturbance quantization parameter through weighted fusion and scale unified processing; And writing the environment disturbance quantization parameter serving as an environment disturbance field into the node running state data set, and establishing field association with the electrical parameter and the network parameter of the corresponding node to form an updated node running state data set.
4. A method according to claim 3, wherein calculating the instant messaging load of each edge node based on the node operation status data set, storing the load value of each node in association with an identification of whether a preset threshold is exceeded, and forming a load status mapping table includes: converging the instant messaging loads of all edge nodes to form a whole network load set covering the whole intelligent lamp pole network; Performing statistical analysis based on the whole network load set, and calculating to obtain a whole network average load value reflecting the whole load level and a load fluctuation range value representing the load discrete degree; Taking the whole network average load value as a reference, and combining the dynamic calculation of the load fluctuation range value to generate a high load judgment threshold value for identifying the load abnormal node; Comparing the instant communication load value of each node with the high load judgment threshold value one by one, and screening out nodes exceeding the high load judgment threshold value; according to the amplitude interval that the instant messaging load value of each node exceeds the high load judging threshold value, carrying out grading treatment on the node load state; and carrying out structural association on the node identifier, the instant messaging load value and the corresponding load grading result to generate a load state mapping table.
5. The method of claim 1, wherein the filtering nodes identified as super-threshold according to the load status mapping table to form a high load node set, and decomposing the communication task of the high load node into migratable data segments, and generating the scheduling candidate task list includes: determining nodes in a high-load classification section based on a load classification result in the load state mapping table, and extracting communication task information to be processed in the high-load nodes; analyzing the communication task information, identifying the task type of each communication task, and matching corresponding task decomposition strategies according to the task type; splitting the communication task into a plurality of task units which have associated identifications and can be processed in parallel based on the task decomposition strategy; Encapsulating processing control information comprising a sequence number, a task identifier, a data size and an expected processing time calculated based on the data size and the task type for each task unit to form a migratable data segment; And calculating the priority of the movable data segments based on the expected processing time, and collecting and sequencing the movable data segments carrying the priority and the processing control information to generate a scheduling candidate task list.
6. The method of claim 5, wherein the method further comprises: Analyzing the processing control information of each migratable data segment in the scheduling candidate task list, and extracting a corresponding task identifier; Inquiring a preset task dependency relation library based on the task identification, and acquiring an original dependency relation description describing a constraint relation among tasks; constructing a task dependency graph representing the execution constraint relation among the migratable data fragments according to the original dependency description; Performing topology analysis on the task dependency graph, and deducing the global execution sequence of each movable data segment under the condition that the dependency constraint condition is met; and writing the deduced global execution sequence result into the scheduling candidate task list, and reordering the migratable data fragments to generate an updated scheduling candidate task list.
7. The method of claim 6, wherein the matching the target low-load nodes for each of the migratable data segments in the list based on the list of scheduling candidate tasks in combination with the inter-node communication link state, and forwarding the migratable data segments to their respective target low-load nodes based on the matching results, comprises: Based on the current intelligent lamp post network topology structure, link quality data representing the communication quality between nodes and resource state data of candidate low-load nodes are obtained concurrently; Calculating a corresponding transmission efficiency score for a communication path from each high-load node to each candidate low-load node by using the link quality data, and calculating a processing capacity matching score of the candidate low-load node to the data segment by combining the resource state data and the feature of the movable data segment extracted from the scheduling candidate task list; Fusing the transmission efficiency scores corresponding to the same candidate low-load node with the processing capacity matching scores to generate a comprehensive forwarding evaluation result for the candidate low-load node; and selecting a corresponding target forwarding node for each migratable data segment according to the comprehensive forwarding evaluation result.
8. The method of claim 1, wherein issuing scheduling execution instructions to the high load node and the receiving node based on the load adjustment data comprises: Analyzing the load adjustment data, and determining a new target node address corresponding to the movable data segment to be adjusted and an updated sending sequence; Generating scheduling control instructions respectively oriented to the high-load node and the receiving node based on the new target node address and the updated sending sequence; Sending a scheduling control instruction to the high-load node, and controlling the high-load node to forward the migratable data fragments to the corresponding new target node addresses according to the updated sending sequence; Sending a scheduling control instruction to a receiving node corresponding to the new target node address, and triggering the receiving and processing of the transferable data segment; And acquiring the execution state information of the scheduling control instruction, and writing the execution state information into the task migration execution log.
9. The method of claim 1, wherein the adaptively feedback adjusting the calculation parameters of the subsequent scheduling process based on the final node load status report comprises: extracting instant messaging load values of all edge nodes after dispatching is completed from the final node load state report; calculating a load balance index reflecting the distribution characteristics of the current dispatching result based on the instant messaging load value; comparing and analyzing the load balance index with a prestored historical load balance index to generate a scheduling deviation analysis result; Updating weight parameters used in the instant messaging load numerical calculation process according to the scheduling deviation analysis result; and applying the updated weight parameters to a load calculation process of the next scheduling period.

Description

Communication load self-adaptive scheduling method for intelligent lamp post edge node Technical Field The application relates to the technical field of edge calculation and Internet of things communication, in particular to a communication load self-adaptive scheduling method of intelligent lamp pole edge nodes. Background In the edge computing network formed by intelligent lamp poles, the overload of nodes is a complex state of multidimensional coupling, which can be characterized by far from a single index. The traditional scheduling method generally depends on isolated indexes such as CPU utilization rate, memory occupation or network queue length to judge, in an outdoor severe environment, node states are affected by multiple factors such as power supply fluctuation, environment temperature and humidity, wireless channel quality and the like, for example, voltage sag can lead to the frequency reduction of a computing unit, further processing queue accumulation is caused, and high-temperature environment can aggravate chip thermal noise, so that communication error rate is increased. The prior art lacks the capability of synchronously acquiring, relating and analyzing and fusion modeling the cross-domain parameters, cannot prospectively capture performance bottlenecks or fault precursors caused by coupling effects, causes load assessment distortion, and buries hidden dangers for subsequent scheduling decisions. The existing load scheduling system has the core parameters such as load threshold value and evaluation weight which are mostly static configuration or are set based on limited experience, and lacks self-learning and self-adapting capability, the intelligent lamp pole network faces highly dynamic service loads such as security video flow tides of the morning and evening peaks and time-varying external environments, the static strategy can misjudge a plurality of nodes in the whole high load or react slowly in the service valley, the optimal global resource allocation cannot be realized, and the fixed strategy can be rapidly invalid along with equipment aging and service mode evolution. Disclosure of Invention The embodiment of the application provides a communication load self-adaptive scheduling method for intelligent lamp pole edge nodes, which realizes the basic promotion of load balancing and precision, resource utilization rate maximization and operation and maintenance cost minimization in a complex outdoor environment, and adopts the following technical scheme: a method for adaptively scheduling communication load of an edge node of a smart lamp post, the method comprising: in an intelligent lamp post network, acquiring current, voltage, processing queue length, communication delay, packet loss rate, ambient light, wind speed and temperature of each edge node to form a node running state data set; Based on the node running state data set, calculating the instant communication load of each edge node, and storing the load value of each node and the identification of whether the load value exceeds a preset threshold value in an associated mode to form a load state mapping table; Screening nodes marked as the super threshold value according to the load state mapping table to form a high load node set, decomposing the communication task of the high load node into migratable data fragments, and generating a scheduling candidate task list; Matching target low-load nodes for each migratable data segment in the list based on the scheduling candidate task list and combining with the communication link state between the nodes, forwarding the migratable data segments to the corresponding target low-load nodes according to the matching result, and generating a task migration execution log; Collecting communication delay, packet loss rate and processing queue length of each node in the migration process, and combining the collected result with the task migration execution log to form load adjustment data, wherein the load adjustment data is used for adjusting the transmission sequence and migration path of unfinished tasks; According to the load adjustment data, dispatching execution instructions are issued to the high-load nodes and the receiving nodes, task migration and load distribution are carried out on the communication tasks, and a final node load state report is generated; And carrying out self-adaptive feedback adjustment on the calculation parameters of the subsequent scheduling process based on the final node load state report. In some possible implementations, the obtaining the current, the voltage, the processing queue length, the communication delay, the packet loss rate, the ambient light, the wind speed, and the temperature of each edge node to form a node operation state data set includes: acquiring the electrical, network and environment original data of a node to form a multi-source original data set, wherein the multi-source original data set comprises original electrical data, orig