Routing optimization method based on big data

Abstract

The invention relates to the technical field of Internet of Things routing management, in particular to a big data-based routing optimization method, which comprises the following steps of: acquiring spectrum data of an Internet of Things routing management area, performing correlation analysis on a spectrum state corresponding to a historical success rate and retransmission times of communication between nodes, and quantifying a probability of successful communication at a future target time. According to the method, the multi-band historical communication success rate and retransmission frequency data are dynamically acquired, the interference risk of the target time period is analyzed and predicted in combination with the time sequence, an interference avoidance strategy based on static channel assessment is upgraded to a probability-driven pre-judgment mechanism, and the anti-interference capability of a transmission path in a complex electromagnetic environment is improved. According to multi-protocol link characteristics of Bluetooth, WiFi, 2.4 G and the like, parameters such as delay, throughput, packet loss rate, flow cost and the like of a single path are synchronously quantified, a unified performance cost file of a heterogeneous network link is established, and end-to-end performance modeling of cross-protocol path combination is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet of Things routing management, and particularly to a routing optimization method based on big data. Background Art

[0002] Internet of Things routing management is a core technical field that supports the interconnection of a large number of intelligent devices, mainly solving the problems of efficient transmission of data packets and resource optimization in heterogeneous network environments. This field focuses on scenarios such as wireless sensor networks, low-power wide-area networks (LPWANs), and mobile ad-hoc networks (Ad-hoc), and needs to address challenges such as limited node energy, dynamic channel changes, coexistence of multiple protocols, and differentiated quality of service (QoS) requirements.

[0003] In existing Internet of Things routing management technologies, such as RPL, which only selects paths based on hop count or remaining energy, it is difficult to quantitatively evaluate the impact of sudden interference on the transmission success rate, resulting in frequent retransmissions and a sharp increase in energy consumption in high-interference scenarios. At the same time, when the throughput of Bluetooth and WiFi links decreases due to channel competition, algorithms based on single-protocol evaluation are difficult to switch to Zigbee low-frequency band links in a timely manner, leading to data transmission bottlenecks. Therefore, improvements are needed. Summary of the Invention

[0004] The purpose of the present invention is to solve the deficiencies in the prior art and propose a routing optimization method based on big data.

[0005] To achieve the above purpose, the present invention adopts the following technical solution, a routing optimization method based on big data, including the following steps: Collect spectrum data in the Internet of Things routing management area, conduct correlation analysis on the historical success rate of communication between nodes and the spectrum status corresponding to the number of retransmissions, quantify the probability of successful communication at a future target time, and obtain a set of predicted interference risk indicators; Collect and process network performance data of Bluetooth, WiFi, 2.4G, Zigbee, LoRa, and NB-IoT, combine the traffic cost and the load status of the access point, evaluate the transmission performance of a single network technology link, obtain a single-segment path performance cost profile, and based on the single-segment path performance cost profile, calculate the expected comprehensive performance of the end-to-end transmission path across different network technologies and access points in combination, and establish a quantified value of the end-to-end expected path performance cost; Based on the set of predicted interference risk indicators and the quantified value of the end-to-end expected path performance cost, associate and match the probability of successful communication of the frequency bands used by each alternative transmission path with the corresponding expected performance, generate a path risk and performance cost mapping table, and based on the path risk and performance cost mapping table, conduct a quantified combined calculation to obtain the comprehensive ranking basis for each path and establish a path evaluation result; Based on the path evaluation results, compare with the communication service quality requirements proposed by the application layer, filter the alternative transmission paths that meet all the constraint conditions, obtain the candidate path list that meets the constraints, and based on the candidate path list that meets the constraints, select the path with the optimal path evaluation result as the current routing decision, and output the preferred routing instruction.

[0006] Preferably, the obtaining step of the predicted interference risk index set is as follows: Collect the spectrum data of all nodes in the Internet of Things routing management area, associate the historical communication success rate data of each node with the number of data packet retransmissions within the corresponding time period, establish the relationship between the node spectrum state and communication behavior, and generate the spectrum state matrix; Based on the spectrum state matrix, calculate the probability of successful communication for each frequency band within the future target time window; According to the probability of successful communication, aggregate the probability distributions of all nodes to obtain the predicted interference risk index.

[0007] Preferably, the obtaining step of the single-segment path performance cost profile is as follows: Collect the historical and real-time transmission data of six network technologies including Bluetooth, WiFi, 2.4G, Zigbee, LoRa, and NB-IoT, analyze the end-to-end delay value, throughput value, stability value, and packet loss rate value, and generate a network performance index set; Based on the network performance index set, synchronously obtain the corresponding traffic cost value and the real-time load status value of the access point for each network technology, and classify and associate the traffic cost value and the load status value with the network performance index set according to the network technology to form a performance cost parameter group; According to the performance cost parameter group, analyze the influence weights of the delay value, throughput value, and packet loss rate value of each network technology on the transmission efficiency, and combine the fluctuation ranges of the stability value and the load status value to calculate the transmission efficiency per unit traffic cost, and generate a single-segment path performance cost profile.

[0008] Preferably, the obtaining step of the end-to-end expected path performance cost quantization value is as follows: Based on the single-segment path performance cost profile, traverse all end-to-end transmission path combinations across network technologies and access points, and associate the single-segment path performance cost profiles corresponding to each segment in each path to form an end-to-end transmission path combination set; According to the end-to-end transmission path combination set, calculate the end-to-end expected path performance cost quantization value for each path.

[0009] Preferably, the obtaining step of the path risk and performance cost mapping table is as follows: Extract the predicted interference risk index set of all alternative transmission paths and the quantification value of the end-to-end expected path performance cost, associate the frequency band identifier of each path with the successful communication probability of the corresponding frequency band, match the expected performance cost value of each path with the successful communication probability, and generate a preliminary mapping relationship between the frequency band and performance cost of the alternative paths; Based on the preliminary mapping relationship between the frequency band and performance cost of the alternative paths, traverse the successful communication probability of each path, and in combination with the quantification value of the end-to-end expected path performance cost, dynamically adjust the matching degree threshold between the successful communication probability and the quantification value of the end-to-end expected path performance cost, and filter out the frequency band-performance cost association data that meet the threshold conditions to form effective frequency band-performance cost mapping data; According to the effective frequency band-performance cost mapping data, arrange the path risk levels in ascending order of the successful communication probability, and at the same time arrange the path cost performance priorities in descending order of the quantification value of the end-to-end expected path performance cost, classify and integrate the cross-correlation relationship between the risk level and the cost performance, and generate a path risk and performance cost mapping table.

[0010] Preferably, the steps for obtaining the path evaluation result are as follows: Extract the predicted interference risk index set of each alternative path and the quantification value of the end-to-end expected path performance cost in the path risk and performance cost mapping table, associate the path identifier with the corresponding interference probability-performance cost data pair, and generate a path risk-performance cost association data set; Based on the path risk-performance cost association data set, calculate the comprehensive sorting score of each path; According to the comprehensive sorting score, arrange the comprehensive sorting scores of all paths in descending order, select the top K paths with the highest comprehensive sorting scores as the preferred routing options, and establish a path evaluation result.

[0011] Preferably, the steps for obtaining the list of candidate paths that meet the constraints are as follows: Extract the end-to-end delay value, stability value and quantification value of the end-to-end expected path performance cost of each alternative transmission path recorded in the path evaluation result, synchronously obtain the delay upper limit threshold, stability lower limit threshold and cost tolerance threshold defined in the application layer service quality requirements, and bind the path parameters and service quality thresholds item by item according to the path identifier to generate an initial matching data set of the path evaluation result and the service quality; Based on the initial matching data set of the path evaluation result and the service quality, traverse the comparison between the delay value of each alternative transmission path and the delay upper limit threshold, the comparison between the stability value and the stability lower limit threshold, and the comparison between the performance cost quantification value and the cost tolerance threshold, and set the path validity flag bit that simultaneously meets the conditions of delay value ≤ delay upper limit threshold, stability value ≥ stability lower limit threshold, and performance cost quantification value ≤ cost tolerance threshold to form a path screening data set with a validity flag bit; Filter the data set according to the path with a validity flag bit, filter all alternative transmission paths with the validity flag bit being true, and sort them in descending order according to the comprehensive sorting score in the path evaluation result to generate a candidate path list that meets the constraints.

[0012] Preferably, the step of obtaining the preferred routing instruction is as follows: Extract the path identifier, comprehensive sorting score value, end-to-end delay value, and end-to-end expected path performance cost quantization value of each path in the candidate path list that meets the constraints, associate the path identifier with the corresponding sorting score - performance parameter data pair, and generate a candidate path full-parameter data set; Based on the candidate path full-parameter data set, sort the candidate paths in descending order according to the comprehensive sorting score value. If there are paths with the same comprehensive sorting score value, compare the priorities of the end-to-end delay value and the performance cost quantization value in turn to generate a candidate path priority list with sorting levels; According to the candidate path priority list with sorting levels, select the transmission path corresponding to the path identifier with a sorting level of 1 to generate a preferred routing instruction.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are as follows: The present invention dynamically collects multi-band historical communication success rate and retransmission times data, combines time series analysis to predict the interference risk in the target period, upgrades the interference avoidance strategy based on static channel evaluation to a probability-driven pre-judgment mechanism, and improves the anti-interference ability of the transmission path in a complex electromagnetic environment. For the characteristics of multi-protocol links such as Bluetooth, WiFi, and 2.4G, synchronously quantify parameters such as the delay, throughput, packet loss rate, and traffic cost of a single segment path, establish a unified performance cost profile for heterogeneous network links, and realize the end-to-end performance modeling of cross-protocol path combinations. Further integrate the interference risk and performance cost parameters, construct a quantitative mapping relationship between path risk and performance cost, sort the candidate paths in multiple dimensions, and ensure the global optimality of the routing decision under constraints such as delay, reliability, and cost. Based on the application layer service quality requirements, screen and sort the candidate paths in real time, and dynamically generate routing instructions through a priority matching mechanism, replacing the fixed strategy of manually preset paths, and enhancing the adaptability of the system to network load fluctuations and business demand changes. For example, in the industrial Internet of Things scenario, this method can identify the trend of the LoRa link stability decline during high-interference periods and automatically switch to the NB-IoT redundant path to avoid data loss caused by link interruption. Brief Description of the Drawings

[0014] Figure 1 It is a step schematic diagram of the present invention. Detailed Embodiment

[0015] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0016] Please refer to Figure 1 , the present invention provides a technical solution, a routing optimization method based on big data, including the following steps: Collect spectrum data in the Internet of Things routing management area, perform correlation analysis on the historical success rate of communication between nodes and the spectrum status corresponding to the number of retransmissions, quantify the probability of successful communication at the future target time, and obtain a predicted interference risk index set; Collect and process network performance data of Bluetooth, WiFi, 2.4G, Zigbee, LoRa, and NB-IoT, combine the traffic cost and the access point load status, evaluate the transmission performance of a single network technology link, obtain a single-segment path performance cost profile, and based on the single-segment path performance cost profile, calculate the expected comprehensive performance of the end-to-end transmission path across different network technologies and access points in combination, and establish a quantified value of the end-to-end expected path performance cost; Based on the predicted interference risk index set and the quantified value of the end-to-end expected path performance cost, associate and match the probability of successful communication of the frequency bands used by each alternative transmission path with the corresponding expected performance, generate a path risk and performance cost mapping table, and based on the path risk and performance cost mapping table, perform a quantified combined calculation to obtain the comprehensive ranking basis for each path and establish a path evaluation result; Based on the path evaluation result, compare the communication service quality requirements proposed by the application layer, screen the alternative transmission paths that meet all the constraint conditions, obtain a list of candidate paths that meet the constraints, and based on the list of candidate paths that meet the constraints, select the path with the best path evaluation result as the current routing decision and output an optimized routing instruction.

[0017] The steps for obtaining the predicted interference risk index set are as follows: Collect spectrum data of all nodes in the Internet of Things routing management area, correlate the historical communication success rate data of each node with the number of data packet retransmissions within the corresponding time period, establish the relationship between the node spectrum status and communication behavior, and generate a spectrum status matrix; Based on the spectrum status matrix, calculate the probability of successful communication of each frequency band within the future target time window. The calculation formula is: ; Among them, is the probability of successful communication, is the historical communication success rate of the node, is the average number of retransmissions of the node within the time window, is the time difference between the target time and the current time; Aggregate the probability distributions of all nodes according to the probability of successful communication to obtain a predicted interference risk indicator.

[0018] Specifically, the original spectrum data of each node in the IoT routing management area collected includes indicators such as the received signal strength indication (RSSI), signal-to-noise ratio (SNR), and channel occupancy rate of each frequency band at a specific timestamp. These data are obtained through periodic scanning by the spectrum sensing module built into the node or an external spectrum analyzer in a preset frequency band (for example, the 2.4GHz ISM band is divided into 13 channels, and the commonly used 433MHz, 868MHz, and 915MHz bands for LoRa). The scanning period is set according to the network dynamics. For example, a full-band scan is performed every 5 minutes. At the same time, retrieve the communication logs of each node in the past period (such as the past 24 hours) from the network management database or the node local log, extract the records of the success or failure of each communication session and the number of packet retransmissions in the failed sessions, align the timestamps of these communication behavior data with the spectrum data collected during the corresponding period. Then, use association rule mining algorithms, such as the Apriori algorithm or the FP-Growth algorithm, to analyze the frequent item sets and association rules between the spectrum index status (for example, divide RSSI into three levels: strong > -65dBm, medium -80dBm to -65dBm, weak < -80dBm; divide SNR into three levels: high > 20dB, medium 10dB to 20dB, low < 10dB; divide the channel occupancy rate by percentage: low < 20%, medium 20% - 60%, high > 60%) and the communication results (high / low success rate, many / few retransmission times). Set the minimum support threshold to 0.1 and the minimum confidence threshold to 0.7. The setting of the threshold is based on historical data statistical analysis to ensure that the selected rules are universal and have a high degree of credibility. For example, it is found through calculation that the support and confidence of the rule "when the channel occupancy rate is high and the SNR is low, the communication success rate is less than 50%" both meet the requirements. Structure the association rules of these quantified spectrum states and communication behaviors, and finally construct a multi-dimensional spectrum state matrix. This matrix is indexed by node ID and timestamp, records the quantified levels of spectrum indicators in different frequency bands and the associated historical communication performance (such as average success rate, average number of retransmissions), and generates a spectrum state matrix.

[0019] Formula: , The benefit of the formula is that it comprehensively considers the long-term historical performance of the node ( ), the recent degree of transmission struggle ( ), and the time difference between the prediction time and the current moment ( ), through the exponential decay function to significantly penalize the situation with a large number of retransmissions, and through the square root function to gently reduce the credibility of long-term predictions, thereby more accurately quantifying the likelihood of communication success at a specific future time point, avoiding the one-sidedness of relying solely on historical averages or instantaneous states for prediction, and providing a forward-looking risk assessment basis for routing decisions.

[0020] (Obtaining steps of the node's historical communication success rate): This parameter represents the probability that a node has successfully completed communication transmissions within a past period of time. When obtaining it, first determine a historical time window, such as the past 24 hours. Then, from the node's communication logs or the historical records of the network management system, count the total number of all communication transmissions that the node has attempted to initiate within this time window and the number of communication transmissions that have been successfully completed . The node's historical communication success rate The calculation formula is: . For example, within the past 24 hours, node A has attempted to send a total of 1500 data packets, and 1380 of them have been confirmed to be successfully received. Then its historical communication success rate is: .

[0021] (Obtaining steps of the average number of retransmissions of the node within the time window): This parameter represents the number of times of retransmission that the node needs to perform on average each time it attempts to send a data packet within a specific time window, reflecting the instability and interference degree of the recent link. When obtaining it, also based on the node's communication logs or the records of the network management system, select a time window related to the prediction target (such as the most recent 1 hour, or the previous corresponding period of the upcoming prediction time window). Count the total number of retransmissions of all data packets that the node has attempted to send within this time window (including the first transmission and all retransmissions) and the total number of data packets that have been attempted to be sent for the first time . The average number of retransmissions of the node within the time window The calculation formula is: . For example, within the most recent 1 hour, node A has attempted to send 120 data packets for the first time, and a total of 36 retransmissions have occurred during this period. Then its average number of retransmissions is: .

[0022] (Obtaining steps of the time difference between the target time and the current time): This parameter represents the duration from the current time to the predicted target time point, used to adjust the credibility of the prediction result. The greater the time difference, the higher the uncertainty. When obtaining it, first determine the current system time and the target time point for which the probability of communication success needs to be predicted 。Time difference That is the difference between the two. The calculation formula is as follows: 。For example, if the current time is 10:00 AM and the communication success probability at 11:30 AM needs to be predicted, then the time difference is 1.5 hours, that is 。

[0023] Calculation process: Substitute the parameter values obtained previously: the historical communication success rate of the node and the average number of retransmissions of the node within the time window as well as the time difference between the target time and the current time hours.

[0024] The calculation process is as follows: ; ; ; ; ; This result indicates that the calculated successful communication probability is approximately 0.431. This value means that after comprehensively considering the historical performance of node A (success rate 0.92), the recent communication difficulties encountered (average retransmissions 0.3 times), and the uncertainty at the predicted time point (1.5 hours later), the estimated probability of successfully completing a communication transmission 1.5 hours later is approximately 43.1%. This probability value is relatively low, lower than its historical average success rate, mainly affected by the recent high average number of retransmissions and the predicted time span. This value is one of the basic data for aggregating the predicted interference risk index in the subsequent steps. A lower value means a higher interference risk or a poorer expected link quality. If the value is greater than a preset reliable communication threshold (for example, 0.8 set according to the quality of service requirements), then the communication risk of this frequency band at this time point is considered low; if it is less than the threshold, the risk is high. The setting basis of this threshold (0.8) is: Through statistical analysis of historical data, it is found that in this Internet of Things application scenario, when the single - time communication success probability is lower than 80%, the end - to - end quality of service (such as delay, packet loss rate) usually cannot meet the application layer requirements. Therefore, 0.8 is selected as the dividing line for judging the high or low communication risk.

[0025] Based on the successful communication probabilities of each node in each frequency band within the future target time window calculated in the previous steps Values, these discrete probability information needs to be aggregated to form an indicator that can macroscopically reflect the potential interference level of the entire routing management area or a specific path. First, determine the aggregation scope, whether to aggregate all nodes in the entire management area or nodes on a specific alternative path, and then select a suitable aggregation method. For example, calculate the arithmetic mean of all relevant node values to obtain a global or path-averaged successful communication probability , where is the total number of nodes, is the successful communication probability of the th node, or calculate the weighted average, and the weights can be set according to the importance of the node, service priority, or expected traffic to be carried. For example , where is the weight of the th node, and the weights can be set according to the network hierarchy of the node (for example, the weight of the backbone node is set to 3, the aggregation node is set to 2, and the end node is set to 1) or the historical traffic contribution (for example, nodes with a traffic proportion exceeding 10% in the past week have a weight of 2.5, and the rest are 1). Another method is to construct a probability distribution and count the number or proportion of nodes whose values fall into different intervals. For example, count the proportion of nodes with , , , or define a risk function to map the value to a risk index. For example, define the risk index , and then aggregate the risk indices, such as calculating the average risk index . A risk threshold can also be set, for example , count the proportion or number of nodes whose values are lower than this threshold as an indicator of the severity of regional interference, that is, when it is lower than this value, the network service quality usually does not meet the standard. Integrate the aggregation results (such as average success probability, weighted average success probability, proportion of low-probability nodes, average risk index, etc.) calculated by the selected aggregation method to finally obtain the predicted interference risk index set.

[0026] The steps to obtain the performance cost profile of a single-segment path are as follows: Collect historical and real-time transmission data of six network technologies, namely Bluetooth, WiFi, 2.4G, Zigbee, LoRa, and NB-IoT, parse the end-to-end delay values, throughput values, stability values, and packet loss rate values, and generate a network performance index set; Based on the network performance metric set, synchronously obtain the traffic cost values corresponding to each network technology and the real-time load status values of the access points, and classify and associate the traffic cost values and load status values with the network performance metric set according to the network technology to form a performance cost parameter group. According to the performance cost parameter group, analyze the influence weights of the delay value, throughput value, and packet loss rate value of each network technology on the transmission efficiency, and combine the stability value and the fluctuation range of the load status value to calculate the transmission efficiency under the unit traffic cost, and generate a single-segment path performance cost profile.

[0027] Specifically, collect the historical and real-time transmission data of six network technologies, namely Bluetooth, WiFi (distinguishing 2.4GHz and 5GHz frequency bands), general 2.4G (such as private protocols), Zigbee, LoRa (distinguishing different regional frequency bands such as EU868, US915), and NB-IoT. The specific operation is to deploy monitoring agent programs at nodes and access points in the network, or use the API interfaces provided by the network management system (NMS) and the Internet of Things platform to periodically (for example, once a minute for high-speed networks such as WiFi and Bluetooth, and once every 15 minutes for low-speed networks such as LoRa and NB-IoT) record and obtain various raw parameters during the actual data transmission process. For example, for the end-to-end delay value, measure the round-trip time (RTT) by sending ICMP Echo requests or specific probe packets, and record its average value, maximum value, and minimum value, or parse the timestamp information in the communication protocol stack to calculate the one-way or two-way delay. For the throughput value, calculate the actual rate (bps or kbps) by counting the number of data bytes successfully transmitted within a unit time (such as 1 second or 10 seconds). For the stability value, calculate the average delay jitter (Jitter) within a continuous monitoring period (such as the past 1 hour), that is, the standard deviation of the RTT, or count the number of link interruptions and the total interruption duration to calculate the link availability percentage. For the packet loss rate value, calculate the percentage of packet loss by comparing the number of packets sent by the sender with the number of packets successfully received and confirmed by the receiver (or detect the lost packets through sequence numbers). Store these parsed specific values (for example, WiFi connection A: average delay 25ms, average throughput 54Mbps, average jitter 8ms, packet loss rate 0.1%; LoRa connection B: average delay 1500ms, average throughput 5kbps, link availability 99.5%, packet loss rate 2%) in a structured manner according to the network technology type, link identifier (such as the Bluetooth connection from node X to node Y, the connection from node Z to WiFi access point P), and timestamp to generate a network performance metric set.

[0028] Based on the network performance metric set containing detailed performance parameters of each network technology link generated in the previous stage, start the data acquisition process in parallel to collect cost and load information. Specifically, for the traffic cost value, query the pre-configured cost database or call the billing system of the operator / service provider through the API to obtain the currency cost of transmitting unit data (such as per MB or per KB) for each network technology (especially cellular technologies such as NB-IoT). For technologies such as Bluetooth, WiFi, Zigbee, LoRa, etc., which usually have no direct traffic fees, their costs can be quantified as the energy consumption cost per unit data, and estimated based on the hardware power consumption model of the node (for example, query the device specification sheet to obtain the current consumption and voltage in the transmit / receive mode, and calculate joules / byte in combination with the transmission rate and duration) or actual measurement values. For example, the NB-IoT cost is 0.05 yuan / MB, and the WiFi energy consumption cost is estimated to be yuan / KB (calculated based on the module power consumption and electricity cost). For the real-time load status value of the access point, query the current number of connected users, CPU utilization, memory occupancy, channel utilization (ChannelUtilization) of the WiFi access point through the network management protocol (such as SNMP), or query the uplink load (such as duty cycle of slot occupancy) of the LoRa gateway, the length of the downlink message queue through the LoRaWAN network server (LNS), or query the NB-IoT base station controller to obtain metrics such as the cell resource block (RB) utilization rate and the number of active users. Match the obtained traffic cost value (such as 0.05 yuan / MB) and the real-time load status value (such as the WiFi AP channel utilization rate of 65%) with the records in the network performance metric set. According to the network technology type and the specific link identifier (ensuring that the cost and load data are associated with the correct network link or its endpoint device / access point), append these cost and load data to the corresponding performance metric records to form a performance cost parameter group.

[0029] According to the performance-cost parameter group integrated in the previous stage, next, it is necessary to evaluate the "cost performance" of each single-segment path, that is, how much transmission efficiency can be obtained per unit cost. First, it is necessary to determine which of the various performance indicators (mainly latency, throughput, packet loss rate) is more important for the current application and allocate appropriate importance weights to them. For example, real-time video may place more emphasis on low latency and low packet loss, so the weights will tilt towards these two items, while file download places more emphasis on high throughput, so the weight of throughput will be higher. At the same time, the stability of the link (such as whether the connection often fluctuates or is interrupted) and the current busy degree (load status) of the access point also need to be considered. For a less stable or overly crowded link, its actual efficiency needs to be discounted. When calculating specifically, it can be imagined that each performance indicator is converted into a "performance score" (such as 0 to 100 points), and the conversion rule is that the higher the score, the better the performance. This means that the original latency value and packet loss rate value need to be scored in reverse (the lower the value, the higher the score). Then, based on the previously determined importance weights, these performance scores are weighted and averaged to obtain a comprehensive performance score. This score also needs to be appropriately reduced according to the stability and load conditions (the reduction amplitude is larger for poor stability or high load). Finally, this performance score after comprehensive evaluation and adjustment is divided by the cost required for this link to transmit a unit data volume (such as 1MB) (traffic fee or equivalent energy consumption cost. If the cost is zero, a very, very small symbolic cost value can be used instead to avoid calculation problems), so as to obtain the transmission efficiency per unit cost. This value intuitively reflects whether this single-segment path is cost-effective. Among them, transmission efficiency per unit cost ≈ (comprehensive performance score × stability adjustment factor × load adjustment factor) / (unit traffic cost + minimum value). The calculated transmission efficiency per unit cost, together with the key information such as the original performance indicators, costs, and loads of this link, is saved together to form the performance-cost file of this single-segment path.

[0030] The steps to obtain the quantization value of the end-to-end expected path performance cost are as follows: Based on the single-segment path performance-cost file, traverse all end-to-end transmission path combinations across network technologies and access points, and associate the single-segment path performance-cost files corresponding to each segment in each path to form a set of end-to-end transmission path combinations; According to the set of end-to-end transmission path combinations, calculate the quantization value of the performance cost of each end-to-end expected path. The calculation formula is: ; Among them, is the quantization value of the performance cost of the end-to-end expected path, is the throughput value of the k-th segment path, is the maximum throughput value of all paths in the system, is the latency value of the k-th segment path, is the maximum delay threshold allowed by the system, is the stability value of the k-th path segment, is the packet loss rate value of the k-th path segment, is the traffic cost value of the k-th path segment, is the maximum traffic cost value of all paths in the system, is the load status value of the access point of the k-th path segment, is the total number of segments included in the end-to-end path, is a small positive number to prevent division-by-zero errors.

[0031] Specifically, based on the single-segment path performance cost profiles generated in the previous step, which contain the detailed performance metrics, traffic costs, and access point load status of single-segment links of various types of network technologies (Bluetooth, WiFi, 2.4G, Zigbee, LoRa, NB-IoT) (for example, direct links between specific nodes or links between nodes and access points), start the path discovery program. This program models the network topology of the entire Internet of Things routing management area as a hybrid graph, where nodes represent Internet of Things devices, relay nodes, gateways, or access points, and edges represent single-segment communication links composed of different network technologies. Using graph traversal algorithms, such as depth-first search (DFS) or breadth-first search (BFS), starting from the specified source node, explore all potential transmission paths that can reach the target node. During the traversal process, it is necessary to cross different network technology domains. For example, a path may connect from the Zigbee link of the source node to the aggregation node, then connect to the gateway through the WiFi link of this node, and finally reach the cloud platform (target node) through the NB-IoT link of the gateway. When traversing, it is necessary to set path search constraints. For example, limit the maximum number of hops (number of segments ), set a maximum number of hops threshold, which is set according to the network scale and application delay requirements. For example, set it to 5 hops to avoid high latency and low reliability caused by overly long paths. At the same time, check and exclude paths that contain loops. For each valid end-to-end path found (for example, path P1: source node A -> Zigbee -> relay B -> WiFi -> gateway C -> NB-IoT -> target D), record the ordered list of segment links that make up this path ([A - B(Zigbee), B - C(WiFi), C - D(NB-IoT)]), and from the single-segment path performance cost profiles, accurately extract and associate the complete performance cost data corresponding to each segment according to the identifiers of each segment link (such as link type, start node, end node). Gather all the discovered valid end-to-end paths and their associated detailed profile information of each segment to form a set of end-to-end transmission path combinations.

[0032] Formula: ; The advantages of the formula are as follows: The formula provides a quantitative method for comprehensively evaluating the quality of end-to-end heterogeneous network paths. It not only considers the throughput of each segment on the path ( ), latency ( ), stability ( ), packet loss rate ( ), and other core performance indicators, but also incorporates two key cost and resource consumption factors, namely traffic cost ( ) and access point load ( ). By aggregating the performance factors of each segment in a multiplicative form, it reflects the characteristic that the path performance is limited by the bottleneck segment (barrel effect). At the same time, through normalization (dividing by ), indicators with different dimensions can be uniformly compared. The denominator part accumulates the cost and load impacts of each segment, so that the higher the total cost and load of the path, the lower the score. The finally calculated value, as a comprehensive performance-cost quantification indicator, can intuitively reflect the overall "cost performance" of the path. The higher the value, the better the balance achieved between performance and cost, providing an effective sorting basis for routing decisions.

[0033] (The throughput value of the k-th segment path) is obtained as follows: This parameter represents the actual data transmission rate of the -th segment link that constitutes the end-to-end path. This value is directly extracted from the "Single-segment Path Performance Cost Archive" associated with this segment. For example, for the second segment of path P1 (the WiFi link from relay B to gateway C), its average throughput is found to be Mbps from the archive.

[0034] (The maximum throughput value of all single-segment paths in the system) is obtained as follows: This parameter serves as the normalization benchmark for throughput and represents the highest throughput that all available single-segment links in the current network environment can provide. For example, by checking the throughput records of all WiFi, NB-IoT, LoRa, etc. links in the archive, it is found that the peak throughput of a certain high-speed WiFi link reaches 120 Mbps, then Mbps is set.

[0035] (The latency value of the k-th segment path) is obtained as follows: This parameter represents the time it takes for a data packet to pass through the The average time required for each segmented link is also directly extracted from the "Single-segment Path Performance Cost Profile" associated with this segment. The delay value in this profile is the average end-to-end delay calculated by measuring the round-trip time (RTT) of the probe packet or analyzing the protocol timestamp in the previous step. For example, the average delay of the second segment of the WiFi link on path P1 is ms.

[0036] (The acquisition step for the maximum single-segment delay threshold allowed by the system) is as follows: This parameter serves as the normalization benchmark for delay or a reference for an acceptable upper limit of single-segment delay. Its setting should be based on the specific requirements for real-time performance in the application scenario. For example, for real-time control applications with low-latency requirements, it may be set to 100 ms; for data acquisition applications that can tolerate higher delays, it can be set to 2000 ms. This threshold needs to be predefined by the system administrator or according to the service level agreement (SLA). For example, for the current hybrid application scenario, a general maximum single-segment tolerance delay is set to ms.

[0037] (The acquisition step for the stability value of the k-th segment path) is as follows: This parameter quantifies the transmission stability of the -th segmented link. This value is also extracted from the "Single-segment Path Performance Cost Profile". For example, the availability of the second segment of the WiFi link on path P1 is (i.e., 99.5%).

[0038] (The acquisition step for the packet loss rate value of the k-th segment path) is as follows: This parameter represents the probability that a data packet is lost when passing through the -th segmented link. It is directly extracted from the "Single-segment Path Performance Cost Profile" associated with this segment. For example, the packet loss rate of the second segment of the WiFi link on path P1 is (i.e., 0.1%).

[0039] (The acquisition step for a small positive number to prevent division-by-zero errors) is as follows: This is a very small positive constant introduced to avoid division-by-zero errors when equals 0. Its value should be much smaller than the smallest non-zero packet loss rate that may occur under normal circumstances. It is set according to experience. For example, set .

[0040] (The acquisition step for the traffic cost value of the k-th segment path) is as follows: This parameter represents the traffic cost of passing through the The cost (monetary cost or energy consumption cost) required to transmit a unit of data (such as MB or KB) for each segmented link is directly extracted from the "Single-segment Path Performance Cost Profile" associated with that segment. The cost values in this profile are calculated based on the operator's billing standard or the device energy consumption model. For example, the estimated energy consumption cost of the second segment of the WiFi link on path P1 is yuan / KB. To facilitate subsequent calculations and unify the unit to yuan / MB, then yuan / MB.

[0041] The steps to obtain (the maximum flow cost value of all single-segment paths in the system) are as follows: This parameter serves as the normalization benchmark for the flow cost and represents the highest unit data transmission cost of all single-segment links in the network. For example, upon checking the profile, it is found that the cost of a certain NB-IoT link is the highest, at 0.15 yuan / MB, then set yuan / MB.

[0042] The steps to obtain (the load status value of the access point at the k-th segment path) are as follows: This parameter quantifies the current load level of the access point (such as WiFi AP, LoRa gateway, NB-IoT base station) connected to the th segmented link. For example, define . For example, for gateway C connected to the second segment of the WiFi link on path P1, the current load status of its WiFi interface is 65%, then .

[0043] The steps to obtain (the total number of segments included in the end-to-end path) are as follows: This parameter indicates how many single-segment links the currently calculated end-to-end path consists of. After determining a path during the path traversal phase, simply count the number of edges included in this path. For example, if path P1 consists of 3 segments (A - B, B - C, C - D), then .

[0044] Calculation process: Select path P1 obtained from the "Set of End-to-End Transmission Path Combinations" for calculation. This path contains 3 segments ( ). For example, the parameters of each segment and the global maximum are as follows: Segment 1 (A - B, Zigbee): Mbps, ms, , , yuan / MB, (associated node B load), Segment 2 (B - C, WiFi): Mbps, ms, , , Yuan / MB, (Associated node C load), segment 3 (C-D, NB-IoT): Mbps, ms, , , Yuan / MB, (Associated node D or base station load), global parameters: Mbps, ms, Yuan / MB, .

[0045] Calculation process: First, calculate the performance factor of each segment : .

[0046] Then, calculate the product of the total performance factors : .

[0047] Next, calculate the cost factor of each segment : .

[0048] Then, calculate the accumulation of the total cost factors : .

[0049] Finally, calculate the end-to-end expected path performance cost quantification value : ; This result indicates that: The calculated end-to-end path P1's expected performance cost quantification value is approximately 18.268. This value itself is a relative, dimensionless comprehensive score. It represents the overall effectiveness and cost-effectiveness of this path after comprehensively considering multiple dimensional factors such as throughput, latency, stability, packet loss rate, traffic cost, and access point load. The higher the value, the better the overall performance of the path usually indicates, that is, it provides better transmission performance under acceptable costs and loads. This value will be used as an important basis for subsequent path evaluation and selection. For example, among multiple candidate paths, the path with the highest value will be given priority. If a benchmark value is set (for example, the average value obtained by analyzing a large amount of historical path data, or the value inversely deduced according to the minimum requirements of the application), then the path with a value higher than this can be regarded as a better path, while the path with a value lower than this may have poor performance cost-effectiveness.

[0050] The steps for obtaining the path risk and performance cost mapping table are as follows: Extract the predicted interference risk index set and the end-to-end expected path performance cost quantization value of all alternative transmission paths, associate the frequency band identifier of each path with the successful communication probability of the corresponding frequency band, match the expected performance cost value and the successful communication probability of each path, and generate a preliminary mapping relationship between the frequency band and performance cost of the alternative paths; Based on the preliminary mapping relationship between the frequency band and performance cost of the alternative paths, traverse the successful communication probability of each path, combine it with the end-to-end expected path performance cost quantization value, dynamically adjust the matching degree threshold of the successful communication probability and the end-to-end expected path performance cost quantization value, and screen the frequency band-performance cost association data that meet the threshold conditions to form effective frequency band-performance cost mapping data; According to the effective frequency band-performance cost mapping data, arrange the path risk levels in ascending order of the successful communication probability, and at the same time arrange the path cost performance priorities in descending order of the end-to-end expected path performance cost quantization value, classify and integrate the cross-correlation relationship between the risk level and the cost performance, and generate a path risk and performance cost mapping table.

[0051] Specifically, extract the predicted interference risk index set calculated in the previous steps for all alternative transmission paths in the system (especially the future target time successful communication probability for each path or its key frequency band value) and the corresponding end-to-end expected path performance cost quantization value , for each alternative transmission path, clarify the main wireless communication frequency band identifier it uses (for example, for a path containing WiFi and LoRa segments, record its used WiFi channel number and LoRa center frequency), find and retrieve the successful communication probability corresponding to this path (or its bottleneck frequency band) from the predicted interference risk index set value, and at the same time extract the end-to-end expected path performance cost quantization value of this path , structurally associate the four key pieces of information of the path identifier, frequency band identifier, successful communication probability and performance cost quantization value , for example, create a record for each alternative path, containing fields {Path_ID, Freq_ID, Q_value, E_value}, such as {Path 12, WiFi_Ch6, 0.85, 22.5}, {Path 15, LoRa_868MHz, 0.92, 15.8}, summarize such records of all alternative paths, and generate a preliminary mapping relationship between the frequency band and performance cost of the alternative paths.

[0052] Based on the preliminary mapping relationship between alternative path frequency bands and performance costs generated in the previous step, traverse and check each alternative path record contained therein, and extract the successful communication probability of each path value and the end-to-end expected path performance cost quantification value , set a matching degree evaluation function to judge the risk of the path (low value represents high risk) and performance cost (high value represents high cost performance) whether it reaches an acceptable matching level. For example, a basic threshold pair can be set: the lowest acceptable successful communication probability and the lowest acceptable performance cost quantification value . The setting of these two thresholds is based on the bottom-line requirements of the quality of service in the actual application scenario. By analyzing historical operation data, determine the and boundary values that meet the basic requirements of 95% of the application scenarios. For example, set , . During the traversal process, check whether each path meets both and conditions, and only retain the path records that meet this dual condition. Or, adopt a dynamically adjusted matching degree threshold strategy. First, calculate the values and value distribution of all alternative paths (such as calculating the mean, standard deviation, percentile). According to the current overall network condition and application priority, dynamically set the screening criteria. For example, if the network load is generally high, it may be necessary to increase the requirement for the value. Pool all the paths that pass the screening and their associated frequency band-performance cost data to form effective frequency band-performance cost mapping data.

[0053] According to the effective frequency band-performance cost mapping data screened in the previous step, classify and sort these data. First, divide the paths into risk levels according to the value of the successful communication probability . Define the risk level criteria. For example: the low risk level (LowRisk) is set to , the medium risk level (MediumRisk) is set to , the high risk level (HighRisk) is set to . These thresholds (0.90, 0.75) are based on the statistical analysis of historical network operation data to determine the actual service interruption or performance degradation probability corresponding to different value intervals. For example, when , the service failure rate exceeds the acceptable 5% red line. Secondly, according to the end-to-end expected path performance cost quantification value Divide the paths according to the cost performance priority, define the priority criteria. For example, set the high priority (HighPriority) as , set the medium priority (MediumPriority) as , set the low priority (LowPriority) as . These thresholds (20, 10) are based on the analysis of the value distribution of the valid paths. For example, by using the K-means clustering or equal-frequency binning method, divide the paths into three performance-cost-benefit groups with significant distinguishability. Then, sort the valid frequency band-performance cost mapping data according to the risk level (based on the value in ascending order, that is, high risk first) and the cost performance priority (based on the value in descending order, that is, high priority first) in multiple levels. Finally, integrate the sorting and classification results to construct a structured table or data structure, clearly showing the identifier of each valid path, the used frequency band, the specific value and value, the corresponding risk level (high, medium, low), and the cost performance priority (high, medium, low), forming a path risk-performance cost mapping table.

[0054] The steps to obtain the path evaluation results are as follows: Extract the predicted interference risk index set and the end-to-end expected path performance cost quantization value of each alternative path in the path risk-performance cost mapping table, associate the path identifier with the corresponding interference probability-performance cost data pair, and generate a path risk-performance cost association data set; Based on the path risk-performance cost association data set, calculate the comprehensive sorting score of each path. The calculation formula is: ; Among them, is the comprehensive sorting score, is the probability of successful communication of the jth path, is the end-to-end expected path performance cost quantization value of the jth path; According to the comprehensive sorting score, sort the comprehensive sorting scores of all paths in descending order, select the top K paths with the highest comprehensive sorting scores as the preferred routing options, and establish the path evaluation results.

[0055] Specifically, read the path risk-performance cost mapping table generated in the previous step. This table records the alternative transmission paths screened and classified preliminarily and their key attributes. Traverse each row record in the table, extract the unique identifier (PathID) of each alternative path, and the core index in the predicted interference risk index set associated with it, that is, the probability of successful communication value, and at the same time extract the end-to-end expected path performance cost quantization value corresponding to this path , pair the identifier of each extracted path with its corresponding and numerical value to form a structured data pair (for example, {Path_ID: 12, Q: 0.85, E: 22.5}, {Path_ID: 15, Q: 0.92, E: 15.8}, {Path_ID: 21, Q: 0.78, E: 35.1}). Collect and organize such data pairs of all alternative paths, ignoring other auxiliary information in the mapping table (such as risk level, priority text description), and generate a data list or set that only contains path identifiers, successful communication probability, and performance cost quantization value. This is the path risk-performance cost association data set.

[0056] Formula: , the benefit of the formula is that it is designed to perform a final comprehensive ranking of alternative paths. It cleverly combines the reliability of the path (reflected by the successful communication probability ) and its performance cost-benefit (reflected by the end-to-end expected path performance cost quantization value ). It uses the natural logarithm to process the success probability, amplifying the negative impact of low probability values on the score, reflecting a high emphasis on reliability. Dividing by introduces the performance cost factor and tends to select higher paths, but the use of the square root avoids the problem of excessive score improvement caused by being too high. An exponential penalty term is introduced, and this term imposes a significant penalty on the combination of "high risk (low ) and high cost / performance (high ), effectively avoiding the selection of paths that have bright performance data but doubtful reliability. The final comprehensive ranking score can more comprehensively and robustly reflect the comprehensive value of the path, providing a refined ranking basis for routing decisions.

[0057] (The probability of successful communication of the j-th path) is obtained as follows: This parameter represents the predicted probability that the th alternative transmission path successfully completes communication within the future target time, reflecting the inherent reliability of the path and the expected interference level. This value directly comes from the value associated with the path in the "path risk-performance cost association data set" generated in the previous step. For example, extract the value corresponding to path 12 from the data set .

[0058] (Steps for obtaining the end-to-end expected path performance cost quantization value for the j-th path): This parameter is the result of quantitatively evaluating the comprehensive performance (throughput, latency, stability, packet loss rate) and cost (traffic cost, load) of the th alternative path, reflecting the overall "cost performance" of the path. This value was calculated in previous steps by aggregating the performance cost data for each segment of the path and applying the evaluation formula. For example, extract the corresponding to path 12 from the data set.

[0059] Calculation process: 1. Calculate the natural logarithm of : . Since , this value is less than or equal to 0.

[0060] 2. Calculate the square root of : . It is required that .

[0061] 3. Calculate the first part: .

[0062] 4. Calculate the failure probability: .

[0063] 5. Calculate the exponential part of the exponential penalty term: .

[0064] 6. Calculate the exponential penalty term: . This value is in the interval (0, 1].

[0065] 7. Multiply the first part by the exponential penalty term to get the final score .

[0066] Specific calculation example: Using the data for path 12: , .

[0067] Calculation process: ; ; ; ; ; ; ; This result shows that the calculated comprehensive ranking score for path 12 is approximately -0.00117. The closer the value is to 0, the higher the comprehensive evaluation of the path. This score comprehensively reflects the reliability of the path and the performance-cost benefit balance, and particularly considers the penalty for the combination of risk and performance cost. This score will be used to sort all alternative paths. A path with a higher value (i.e., a negative number closer to 0) will be considered better. For example, if the score of another path 15 is -0.0025, then path 12 (-0.00117) is better than path 15.

[0068] Based on the comprehensive sorting scores of all alternative paths calculated in the previous step values, perform a sorting operation, call a standard sorting function or algorithm, such as quicksort or mergesort, to sort all path records (including PathID, Q, E, R) in the "Path Risk-Performance Cost Association Data Set" according to the score values in descending order. Since , the values are non-positive, sorting in descending order means arranging the score values from the one closer to 0 (optimal) to the more negative (sub-optimal). After sorting, an ordered list of paths is obtained. Then, determine the number of preferred routing option selections required according to the system configuration or current application requirements The value is usually a fixed small positive integer, for example , indicating that the system always selects the 3 paths with the highest scores as candidates. The setting of the value can be based on considerations of redundancy, load balancing strategy, or the number of available paths. For example, if the system requires at least 2 backup paths, then should be greater than or equal to 3. If the total number of valid paths is less than , then select all valid paths. Select the highest-ranked paths from the sorted path list. Record the identifiers of these paths and their corresponding comprehensive sorting scores and values (which may also include to form the final path evaluation result. This result is an ordered list containing the optimal

[0069] The steps to obtain the list of candidate paths that meet the constraints are as follows: Extract the end-to-end delay value, stability value, and end-to-end expected path performance cost quantification value of each alternative transmission path recorded in the path evaluation result. Synchronously obtain the delay upper limit threshold, stability lower limit threshold, and cost tolerance threshold defined in the application layer service quality requirements. Bind the path parameters and service quality thresholds item by item according to the path identifier to generate an initial matching data set of the path evaluation result and service quality. Based on the initial matching data set of the path evaluation result and service quality, traverse the comparison between the delay value of each alternative transmission path and the delay upper limit threshold, the comparison between the stability value and the stability lower limit threshold, and the comparison between the performance cost quantification value and the cost tolerance threshold. Set the path validity flag bit that simultaneously satisfies the delay value ≤ delay upper limit threshold, stability value ≥ stability lower limit threshold, and performance cost quantification value ≤ cost tolerance threshold to form a path screening data set with validity flag bits. According to the path screening data set with validity flag bits, filter all alternative transmission paths with the validity flag bit being true, and sort them in descending order according to the comprehensive sorting score in the path evaluation result to generate a candidate path list that meets the constraints.

[0070] Specifically, read the path evaluation result generated in the previous step, which includes the list of the top K alternative transmission paths with the highest comprehensive sorting score and their scores. , for each alternative transmission path in the list, it is necessary to trace back its composition information, and obtain or recalculate the end-to-end cumulative delay value and end-to-end overall stability value of this path from the previous steps (such as when generating the set of end-to-end transmission path combinations or calculating values), and at the same time extract the end-to-end expected path performance cost quantification value that has been calculated for this path. , in parallel, access the system configuration or query the application layer interface to obtain the service quality (QoS) requirements corresponding to the current communication task, specifically including the maximum end-to-end delay upper limit threshold that the application can tolerate (for example, for real-time voice transmission, set to 150 milliseconds), the required minimum end-to-end stability lower limit threshold (for example, for critical data reporting, set to 99.9% end-to-end availability, that is ), and a cost tolerance threshold, which represents an upper limit constraint on the comprehensive performance cost index (for example, ), to prevent the selection of paths with performance or resource consumption far exceeding the requirements. Bind the identifier, calculated end-to-end delay , end-to-end stability , and end-to-end expected path performance cost quantification value of each alternative path , with the unified QoS requirements (delay upper limit threshold, stability lower limit threshold, cost tolerance threshold Perform binding association to form an initial matching dataset of path evaluation results and service quality.

[0071] Based on the initial matching dataset of path evaluation results and service quality generated in the previous step, process each alternative transmission path record in the dataset one by one, and perform compliance checks on the constraint conditions. Specifically, compare the actual end-to-end delay value obtained in this path record with the delay upper limit threshold in the service quality requirements of the bound application layer to check if it is satisfied the delay upper limit threshold. Then, compare the actual end-to-end stability value of the path (such as end-to-end availability) with the stability lower limit threshold it is bound to to check if it is satisfied the stability lower limit threshold. Then, compare the end-to-end expected path performance cost quantification value of the path with the cost tolerance threshold it is bound to to check if it is satisfied . Set a boolean-type validity flag for each path. Only when all three above comparison conditions (delay constraint, stability constraint, performance cost upper limit constraint) are true (that is, the path satisfies all three QoS requirements simultaneously), set the validity flag of this path to true (TRUE). If any one condition is not satisfied, set its validity flag to false (FALSE). Integrate all path records with this newly calculated validity flag to form a path screening dataset with validity flags.

[0072] According to the path screening dataset with validity flags formed in the previous step, perform a filtering operation. Traverse all path records in the dataset and only select those alternative transmission paths whose validity flag values are true (TRUE). These paths are the ones that simultaneously satisfy all mandatory service quality constraints (delay, stability, cost tolerance) of the application layer. For all the selected valid paths, based on their comprehensive sorting scores (this score comprehensively considers risk and performance cost) obtained in the path evaluation result stage, perform a final sorting. Call the sorting program and sort in descending order according to the score value (that is, the path with a score value closer to 0 is ranked higher). If there are paths with the same score, they can be further sorted according to the secondary sorting rule (such as preferentially selecting the path with a lower end-to-end delay). Organize the list of paths that have been filtered by QoS constraints and sorted according to the comprehensive score to generate a final list of candidate paths that meet the constraints. This list contains all paths that meet the conditions and are sorted according to the optimization degree, and can be directly used for routing decisions.

[0073] The steps for obtaining the optimal routing instruction are as follows: Extract the path identifier, comprehensive sorting score value, end-to-end delay value, and end-to-end expected path performance cost quantization value of each path in the candidate path list that meets the constraints, associate the path identifier with the corresponding sorting score - performance parameter data pair, and generate a full parameter data set of candidate paths; Based on the full parameter data set of candidate paths, sort the candidate paths in descending order according to the comprehensive sorting score value. If there are paths with the same comprehensive sorting score value, compare the priorities of the end-to-end delay value and the performance cost quantization value in turn to generate a candidate path priority list with sorting levels; According to the candidate path priority list with sorting levels, select the transmission path corresponding to the path identifier with a sorting level of 1 to generate an optimal routing instruction.

[0074] Specifically, access the candidate path list that meets the constraints generated in the previous step. This list already contains paths that have passed the quality of service (QoS) test and are sorted according to the preliminary comprehensive sorting score Sort the paths, traverse each candidate path entry in this list, and extract the unique identifier (PathID) of this path, its corresponding comprehensive sorting score value , the end-to-end delay value of this path ( , for example, 120 milliseconds, which has been calculated or stored in the previous step), and the end-to-end expected path performance cost quantization value of this path ( , for example, 30.5, which has also been calculated or stored in the previous step). Organize these extracted key parameters (path identifier, , , ) into a data record, for example, {Path_ID:A, R_j:-0.00117, D_total:120, E_j:30.5}. Repeat this operation for all paths in the list, gather the complete parameter records of all paths, and generate a full parameter data set of candidate paths. This data set contains all the information required for the final precise sorting and selection.

[0075] Based on the full parameter data set of candidate paths generated in the previous step, perform the final precise sorting process. Although the input source of this data set (the candidate path list that meets the constraints) is already roughly sorted in descending order according to the comprehensive sorting score , it is necessary to handle the case where the scores are exactly the same. Set a multi-level sorting rule: the primary sorting key is the comprehensive sorting score value , sorted in descending order (the closer the value is to 0, the more preferred). When it is detected that there are multiple paths with exactly the same When evaluating score values, enable the first-level tie-breaker and compare the end-to-end delay values of these paths with the same scores. , preferentially select the path with a lower value (i.e., smaller delay), and set the delay priority principle. This is because when the comprehensive scores are the same, faster transmission is usually better. If there are still and paths with the same values, then enable the second-level tie-breaker and compare the end-to-end expected path performance cost quantification values of these paths. , preferentially select the path with a higher value (i.e., better performance cost-benefit), and set the performance cost priority principle. Apply this composite sorting logic that includes the primary sorting key ( descending order) and two levels of tie-breakers ( ascending order, descending order) to fully sort the candidate path full-parameter data set, assign an increasing sorting rank (Rank) starting from 1 to each path in the sorted list, and generate a candidate path priority list with sorting ranks.

[0076] According to the candidate path priority list with sorting ranks generated in the previous step, this list accurately reflects the final order of all paths that meet the constraint conditions arranged by optimality (comprehensive score, delay, performance cost). Directly locate and select the path record with the sorting rank (Rank) assigned as 1 in the list. This path is the absolute optimal choice determined after all evaluation, screening, sorting, and tie-breaking processes. Extract its path identifier (PathID) from this path record ranked first, such as Path A. According to this optimal path identifier, construct a specific routing instruction. The format of this instruction needs to be compatible with the target network device or routing control protocol, and at least includes the selected path identifier, and may also include detailed hop information of this path (for example, node sequence: source -> node B -> node C -> destination, and the technologies used in each segment: Zigbee, WiFi), or directly includes parameters such as interfaces and next-hop addresses required to configure this path. Finally, generate a formatted preferred routing instruction. For example, generate an instruction with the content "SELECT_OPTIMAL_ROUTE: PathID=A, Hops=[(Source, B, Zigbee), (B, C, WiFi), (C, Dest, LoRa)]" and send it to the relevant network nodes for execution.

[0077] The above are only the preferred embodiments of the present invention, and do not limit the present invention in other forms. Any person skilled in the art may use the technical content disclosed above to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A routing optimization method based on big data, characterized in that: The following steps are involved: Collect spectrum data in the IoT routing management area, correlate and analyze the historical success rate of inter-node communication and the spectrum status corresponding to the number of retransmissions, quantify the probability of successful communication at the future target time, and obtain a set of predicted interference risk indicators; Collect and process network performance data of Bluetooth, WiFi, 2.4G, Zigbee, LoRa and NB-IoT, combine traffic cost and access point load status, evaluate the transmission performance of a single network technology link, obtain a single-segment path performance cost profile, and based on the single-segment path performance cost profile, combine and calculate the expected comprehensive performance of the end-to-end transmission path across different network technologies and access points, and establish a quantitative value of the end-to-end expected path performance cost; Based on the predicted interference risk indicator set and the end-to-end expected path performance cost quantified value, the probability of successful communication of the frequency band used by each candidate transmission path is associated and matched with the corresponding expected performance, and a path risk and performance cost mapping table is generated. Based on the path risk and performance cost mapping table, a quantitative combined calculation is performed to obtain a comprehensive ranking basis for each path, and a path evaluation result is established; Based on the path evaluation result, the communication service quality requirements proposed by the application layer are compared, and alternative transmission paths that meet all constraints are screened to obtain a list of candidate paths that meet the constraints. Based on the list of candidate paths that meet the constraints, the path with the best path evaluation result is selected as the current routing decision, and the preferred routing instruction is output.

2. The routing optimization method based on big data according to claim 1, characterized in that: The steps for obtaining the prediction interference risk indicator set are: Collect spectrum data of all nodes in the IoT routing management area, associate the historical communication success rate data of each node with the number of data packet retransmissions in the corresponding time period, establish the relationship between the node spectrum state and communication behavior, and generate a spectrum state matrix; Based on the spectrum state matrix, calculating the probability of successful communication in each frequency band within a future target time window; According to the probability of successful communication, the probability distribution of all nodes is aggregated to obtain the predicted interference risk index.

3. The routing optimization method based on big data according to claim 1, characterized in that: The steps for obtaining the single-segment path performance cost file are as follows: Collect historical and real-time transmission data of six network technologies, including Bluetooth, WiFi, 2.4G, Zigbee, LoRa and NB-IoT, analyze end-to-end delay values, throughput values, stability values ​​and packet loss rate values, and generate a set of network performance indicators; Based on the network performance indicator set, synchronously obtain the flow cost value corresponding to each network technology and the real-time load status value of the access point, and associate the flow cost value and the load status value with the network performance indicator set according to the network technology classification to form a performance cost parameter group; According to the performance cost parameter group, the influence weight of the delay value, throughput value and packet loss rate value of each network technology on the transmission efficiency is analyzed, and the transmission performance under unit traffic cost is calculated by combining the fluctuation range of the stability value and the load status value to generate a single-segment path performance cost file.

4. The routing optimization method based on big data according to claim 1, characterized in that: The step of obtaining the quantized value of the end-to-end expected path performance cost is as follows: Based on the single-segment path performance cost profile, traverse all end-to-end transmission path combinations across network technologies and access points, associate the single-segment path performance cost profiles corresponding to each segment in each path, and form an end-to-end transmission path combination set; According to the end-to-end transmission path combination set, a quantized value of the performance cost of each end-to-end expected path is calculated.

5. The routing optimization method based on big data according to claim 1, characterized in that: The steps for obtaining the path risk and performance cost mapping table are as follows: Extract the predicted interference risk indicator set and the quantified value of the end-to-end expected path performance cost of all alternative transmission paths, associate the frequency band identifier of each path with the probability of successful communication of the corresponding frequency band, match the expected performance cost value of each path with the probability of successful communication, and generate a preliminary mapping relationship between the frequency band and performance cost of the alternative path; Based on the preliminary mapping relationship between the frequency band and performance cost of the alternative paths, the successful communication probability of each path is traversed, and combined with the quantified value of the end-to-end expected path performance cost, the matching threshold between the successful communication probability and the quantified value of the end-to-end expected path performance cost is dynamically adjusted, and the frequency band-performance cost association data that meets the threshold condition is screened to form effective frequency band-performance cost mapping data; According to the effective frequency band-performance cost mapping data, the path risk level is arranged in ascending order according to the probability of successful communication, and the path cost-effectiveness priority is arranged in descending order according to the quantified value of the end-to-end expected path performance cost. The cross-correlation between risk level and cost-effectiveness is classified and integrated to generate a path risk and performance cost mapping table.

6. The routing optimization method based on big data according to claim 1, characterized in that: The steps for obtaining the path evaluation result are: Extract the predicted interference risk indicator set and the end-to-end expected path performance cost quantized value of each alternative path in the path risk and performance cost mapping table, associate the path identifier with the corresponding interference probability-performance cost data pair, and generate a path risk-performance cost associated data set; Based on the path risk-performance cost association data set, a comprehensive ranking score for each path is calculated; According to the comprehensive ranking score, the comprehensive ranking scores of all paths are arranged in descending order, and the first K paths with the highest comprehensive ranking scores are selected as priority routing options to establish a path evaluation result.

7. The routing optimization method based on big data according to claim 1, characterized in that: The steps for obtaining the candidate path list that meets the constraints are: Extract the end-to-end delay value, stability value and end-to-end expected path performance cost quantization value of each candidate transmission path recorded in the path evaluation result, synchronously obtain the delay upper limit threshold, stability lower limit threshold and cost tolerance threshold defined in the application layer service quality requirements, bind the path parameters and the service quality threshold item by item according to the path identifier, and generate an initial matching data set of the path evaluation result and the service quality; Based on the path evaluation results and the initial matching data set of the service quality, the delay value of each candidate transmission path is compared with the delay upper threshold, the stability value is compared with the stability lower threshold, and the performance cost quantization value is compared with the cost tolerance threshold. The path validity flag that satisfies the conditions that the delay value ≤ the delay upper threshold, the stability value ≥ the stability lower threshold, and the performance cost quantization value ≤ the cost tolerance threshold is set to form a path screening data set with validity flags. According to the path screening data set with validity flag, all candidate transmission paths with validity flag set to true are filtered, and the paths are arranged in descending order according to the comprehensive ranking scores in the path evaluation results to generate a list of candidate paths that meet the constraints.

8. The routing optimization method based on big data according to claim 1, characterized in that: The steps of obtaining the preferred routing instruction are: Extracting the path identifier, comprehensive ranking score value, end-to-end delay value and end-to-end expected path performance cost quantized value of each path in the candidate path list that meets the constraints, associating the path identifier with the corresponding ranking score-performance parameter data pair, and generating a candidate path full parameter data set; Based on the candidate path full parameter data set, the candidate paths are sorted in descending order according to the comprehensive ranking score values. If there are paths with the same comprehensive ranking score values, the priorities of the end-to-end delay values ​​and the performance cost quantization values ​​are compared in turn to generate a candidate path priority list with sorting levels; According to the candidate path priority list with sorting levels, a transmission path corresponding to a path identifier with a sorting level of 1 is selected to generate a preferred routing instruction.

Images