Dynamic routing metric method based on nonlinear fusion of multi-dimensional heterogeneous features
By employing a dynamic routing metric method that nonlinearly fuses multidimensional heterogeneous features, the problems of perception sluggishness and convergence oscillation in routing protocols in weak network environments are solved, enabling rapid response and reliable transmission in highly dynamic topology environments.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHEJIANG UNIV
- Filing Date
- 2026-05-15
- Publication Date
- 2026-06-19
AI Technical Summary
Existing distributed routing protocols suffer from problems such as linear metric model distortion, neglect of key stability indicators, and contradiction between sensitivity and stability in weak network environments. This results in routing protocols being slow to perceive in highly dynamic topology environments, having a single metric dimension, and experiencing convergence oscillations.
A dynamic routing metric method based on nonlinear fusion of multidimensional heterogeneous features is adopted. Multidimensional heterogeneous parameters are collected in real time through cross-layer probes, and a second-order exponential weighted moving average algorithm is used for filtering and smoothing. A hybrid operation model is constructed, which includes a linear benchmark and a nonlinear penalty part, to achieve millisecond-level perception of network weakness and energy bottlenecks.
It significantly improves the network's lifespan and data transmission reliability under strong interference environments, ensures the loop-free convergence characteristics of the routing algorithm, quickly triggers route switching, avoids selecting unstable or power-off nodes, and improves the network's anti-interference capability.
Smart Images
Figure CN122247910A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of computer network communication technology, specifically involving a dynamic routing metric method based on nonlinear fusion of multidimensional heterogeneous features. It is applicable to the generation and routing of multi-hop network routing metrics in weak signal and high dynamic environments such as tactical mobile ad hoc networks (MANET), satellite communication networks, underground mine communication, industrial Internet of Things (IIoT), and emergency rescue networks. Background Technology
[0002] In modern complex communication scenarios, the characteristics of weak network environments are becoming increasingly prominent. Such environments typically exhibit the following characteristics: limited and highly fluctuating channel bandwidth, high bit error rate, frequent topology reconfiguration due to node movement, and nodes being limited by battery power.
[0003] Traditional distributed routing protocols typically use relatively simple linear metrics when calculating path quality. Existing technical solutions have the following main limitations: Distortion of linear measurement models: Existing protocols mostly use a linear accumulation method of "reciprocal of bandwidth + latency". However, in weak networks, the deterioration of link quality often exhibits non-linear characteristics (i.e., the "avalanche effect"). For example, when the link packet loss rate rises from 0% to 5%, the service can still be maintained; but when the packet loss rate rises from 15% to 20%, the TCP throughput may instantly drop to zero. Linear models cannot simulate this physical layer collapse characteristic, causing routing protocols to mistakenly select links that are "theoretically feasible but actually paralyzed".
[0004] The lack of key stability metrics: Traditional metrics often overlook "jitter" and "node energy." Jitter is a core indicator of link stability, while energy is crucial in determining network lifespan. Ignoring these two factors can lead to data flows flooding unstable links or low-power nodes, causing routing oscillations or network fragmentation.
[0005] The contradiction between sensitivity and stability: In order to prevent frequent route switching, traditional protocols usually set a large hysteresis threshold, which makes the protocol react to link failures extremely slowly (on the order of seconds or even minutes), and cannot adapt to the ever-changing battlefield of weak networks.
[0006] Therefore, there is an urgent need for a dynamic routing metric method that can integrate multi-dimensional network characteristics, has a non-linear penalty mechanism, and is mathematically rigorous (strictly satisfying the monotonicity of the metric to prevent loops). Summary of the Invention
[0007] To overcome the shortcomings of existing technologies, this invention aims to provide a dynamic routing metric method based on the nonlinear fusion of multidimensional heterogeneous features. This addresses the problems of sluggish perception, single metric dimension, and convergence oscillations in existing distance vector routing protocols in weak network environments with low signal-to-noise ratio and highly dynamic topologies. The method of this invention uses cross-layer probes to collect multidimensional heterogeneous parameters in real time, including effective bandwidth, cumulative delay, signal jitter variance, link packet loss rate, and node remaining energy at the physical and link layers. Time-domain smoothing is performed using a double exponential weighted moving average (DWMA) algorithm. Based on this, a hybrid computational model is constructed, comprising a linear baseline component and a nonlinear penalty component. Exponential and power functions are used to nonlinearly amplify the degradation indicators, achieving millisecond-level perception of subtle network degradation and energy bottlenecks.
[0008] To achieve the above objectives, the present invention may adopt the following specific technical solutions: The dynamic routing metric method based on nonlinear fusion of multidimensional heterogeneous features includes the following steps: Step 1: Establish a multidimensional heterogeneous feature space and obtain the original feature vectors of network nodes; Step 2: Based on the multidimensional heterogeneous feature space established in Step 1, the obtained original feature vectors are filtered and smoothed to output the smoothed feature vectors. Step 3: Using the smoothed feature vector output in Step 2 as input data, combine the calculated linear baseline term with the nonlinear penalty term to obtain the single-hop composite cost of the current link. Step 4: Based on the single-hop composite cost output in Step 3, aggregate the total path metric and use the distance vector algorithm to select the local optimal route and construct the routing table; Step 5: Continuously monitor the change in the single-hop composite cost generated in Step 3. When a preset deterioration condition or periodic condition is met, trigger the broadcast of the updated routing status from Step 4 to the network. The periodic condition refers to the inability to receive periodic heartbeat packets from the link normally, with a preset time interval; the deterioration condition refers to the link quality deteriorating to the point where the route cannot be guaranteed to be loop-free when the link interface has link awareness capability.
[0009] Furthermore, in step 1, the process of establishing the multidimensional heterogeneous feature space is as follows: First, initialize the reference bandwidth constant, smooth forgetting factor, and weight parameters of each feature of the routing protocol; second, deploy state monitoring probes between the network layer and the link layer to periodically collect six-dimensional parameters to form the original feature vector; the six-dimensional parameters include bottleneck bandwidth representing transmission rate, cumulative delay representing transmission delay, link load rate representing interface throughput, reliability coefficient representing transmission success rate, jitter variance representing channel stability, and energy normalization factor representing the remaining power of the node.
[0010] Furthermore, when the status monitoring probe collects the status of the underlying physical link, it uses a second-order exponentially weighted moving average (EWMA) filtering algorithm to preprocess the original sampled values to filter out instantaneous noise interference. Combined with a set hysteresis threshold mechanism, it determines whether to trigger the update calculation of the upper-layer routing table.
[0011] Furthermore, in step 3, the specific process for calculating the single-hop composite cost includes: Step 3.1 Calculate the linear benchmark term: Convert the smoothed jitter variance into equivalent delay according to the preset proportional coefficient, and add it to the smoothed cumulative delay to obtain the comprehensive delay; divide the reference bandwidth constant by the smoothed bottleneck bandwidth to obtain the bandwidth cost; multiply the comprehensive delay and bandwidth cost by the corresponding weight coefficients respectively and sum them to generate the linear benchmark term of the measurement base. Step 3.2: Calculate the nonlinear penalty term: Use a hyperbolic model to perform a high-sensitivity mapping on the smoothed link load rate to obtain the congestion penalty function; use a natural exponential model to map the smoothed reliability coefficient to obtain the reliability penalty function; use a power-law model to map the smoothed energy normalization factor to obtain the energy penalty function; multiply the congestion penalty function, reliability penalty function, and energy penalty function together to generate the nonlinear penalty term. Step 3.3, Composite Cost Output: Multiply the linear baseline term by the nonlinear penalty term to output the final single-hop composite cost of the current link.
[0012] Specifically, in the nonlinear mapping model, when constructing the congestion penalty function, reliability penalty function, and energy penalty function, the sensitivity parameters and weight indices included can be dynamically configured through the software control plane to adapt to the different tolerances of different network service types for latency, packet loss, or energy consumption.
[0013] Furthermore, in step 4, the specific content of constructing and updating the routing table includes: Path metric aggregation: The total metric value of the path is obtained by summing the single-hop composite cost of all single-hop links traversed sequentially on the path from the source node to the destination node. Acyclic monotonicity check: The single-hop composite cost of each hop is always greater than zero, and the total metric is strictly monotonically increasing as the path extends; Optimal path selection and table entry: When a network node receives a routing update message from a neighboring node, it combines the single-hop composite cost calculated locally to obtain the total metric value of multiple candidate paths leading to the same destination node; compares the total metric values of each candidate path, selects the path with the smallest total metric value as the optimal route, and updates the destination address, next-hop node, and total metric value of the optimal route in the local routing table.
[0014] Furthermore, in step 5, the update triggering mechanism includes periodic updates and event-driven fast-trigger updates: Periodic updates: Nodes periodically broadcast the total metric value information in their local routing table to neighboring nodes according to a preset timer; Event-driven rapid update: During periodic link probing, when link congestion, increased packet loss rate, or a sudden drop in node power causes the nonlinear penalty term in step 3 to grow exponentially or exponentially, resulting in the change in the latest calculated single-hop composite cost exceeding a preset hysteresis threshold, or when a physical disconnection of the underlying link is detected, the system ignores the periodic timer and immediately triggers the routing update mechanism. It proactively sends routing update messages containing cost deterioration information to neighboring nodes, prompting rapid convergence of the entire network and isolation of inferior links. Typically, the hysteresis threshold is set as a fixed percentage of the average link cost over a certain period, for example, 20%.
[0015] Compared with the prior art, the present invention has the following advantages: The single-hop composite cost calculation model proposed in step 3 of this invention strictly adheres to the principle of monotonically increasing path cost, ensuring the loop-free convergence characteristic of the routing algorithm and significantly improving the network's lifespan and data transmission reliability under strong interference environments. This invention uses a nonlinear mathematical model to amplify minor link quality degradation into significant metric increases, thereby quickly triggering route switching; it introduces jitter and energy parameters to avoid selecting nodes with high bandwidth but extreme instability or those about to lose power; and it ensures that complex metric calculations ultimately conform to the monotonicity constraints of the distance vector protocol. Attached Figure Description
[0016] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation
[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0018] Example 1
[0019] like Figure 1 As shown, this invention provides a dynamic routing metric method based on nonlinear fusion of multidimensional heterogeneous features, comprising the following steps: (a) Step 1: Establish a multidimensional heterogeneous feature space and obtain the original feature vectors of network nodes.
[0020] The process of establishing a multidimensional heterogeneous feature space is as follows: First, initialize the reference bandwidth constant, smoothing forgetting factor, and weight parameters of each feature for the routing protocol; second, deploy state monitoring probes between the network layer and the link layer to periodically collect six-dimensional parameters to form the original feature vector Vraw. The six-dimensional parameters include bottleneck bandwidth representing transmission rate, cumulative latency representing transmission delay, link load rate representing interface throughput, reliability coefficient representing transmission success rate, jitter variance representing channel stability, and energy normalization factor representing the remaining power of the node.
[0021] Bottleneck bandwidth (Bw): The minimum available transmission rate on the path (unit: Kbps).
[0022] Cumulative delay (Dly): includes physical propagation delay, interface processing delay, and queuing delay (unit: microseconds).
[0023] Link load rate (Load): The proportion of the current throughput of the interface to the maximum bandwidth, quantized as an integer from 0 to 255.
[0024] Reliability coefficient (Rel): The success rate of transmission based on the link layer ACK mechanism, normalized to 0.0 to 1.0.
[0025] Jitter variance (Jit): The rate of change of the interval between consecutive data packet arrivals, used to characterize channel stability.
[0026] Energy normalization factor (Eres): The ratio of a node's remaining charge to its full charge, ranging from 0.0 to 1.0.
[0027] (ii) Step 2: Based on the multidimensional heterogeneous feature space established in step 1, the obtained original feature vector is filtered and smoothed to output the smoothed feature vector.
[0028] To eliminate transient noise and preserve long-term trends, and to prevent minor fluctuations at the lower level from causing frequent recalculations in the upper-layer routing model (i.e., the routing decision and path calculation modules of the upper-layer routing protocol), the raw data is filtered using an exponentially weighted moving average (EWMA) algorithm. Specifically, a second-order smoothing algorithm is employed to enhance noise resistance for high-frequency fluctuating parameters such as jitter variance (Jit).
[0029] First-order smoothness: ; Second-order smoothing: ; in, The original sampled value at the current moment. The final output is a smoothed feature value used for metric calculation. The forgetting factor is between 0 and 1.
[0030] (III) Step 3: Using the smoothed feature vector output in step 2 as input data, combine the calculated linear benchmark term with the nonlinear penalty term to obtain the single-hop composite cost of the current link.
[0031] Specifically, the process for calculating the single-hop composite cost includes: Step 3.1: Calculate the linear baseline term: Convert the smoothed jitter variance into equivalent delay and add it to the smoothed cumulative delay to obtain the comprehensive delay; divide the reference bandwidth constant by the smoothed bottleneck bandwidth to obtain the bandwidth cost; multiply the comprehensive delay and bandwidth cost by the corresponding weight coefficients and sum them to generate the linear baseline term of the metric base.
[0032] Specifically, calculate the linear reference term. Based on bandwidth ,Delay and shaking Calculate the baseline cost. The formula is: ,in, This is a configurable weighting factor.
[0033] The linear baseline (Mbase) retains the linear consideration of bandwidth and latency in classic routing protocols, serving as the physical basis for the metric. It uses the reference bandwidth constant Cref divided by the bottleneck bandwidth Bw to ensure that larger bandwidth results in lower costs. The jitter variance Jit is converted to equivalent latency using a coefficient μ (e.g., treating 1ms of jitter as a fixed 10ms delay), significantly penalizing unstable links. Subsequently, the equivalent latency is combined with the cumulative latency Dly and linearly summed using configurable weight coefficients K1 and K2.
[0034] Step 3.2: Calculate the nonlinear penalty term: Use a hyperbolic model to perform a high-sensitivity mapping on the smoothed link load rate to obtain the congestion penalty function; use a natural exponential model to map the smoothed reliability coefficient to obtain the reliability penalty function; use a power-law model to map the smoothed energy normalization factor to obtain the energy penalty function; multiply the congestion penalty function, reliability penalty function, and energy penalty function together to generate the nonlinear penalty term.
[0035] The nonlinear penalty term utilizes a nonlinear mathematical function to perform a highly sensitive mapping of load, reliability, and energy. Based on load and reliability... and remaining energy Construct a nonlinear penalty function. The formula is: .
[0036] Congestion penalty function (Ψcong): Employs a hyperbolic model. Congestion penalty term. When the load approaches the physical limit of 255 (full load), the congestion penalty value approaches infinity, which conforms to the asymptote characteristic of network queuing theory.
[0037] The reliability penalty function (Ψrel) employs a natural exponential model and includes a sensitivity constant β. Even a small increase in packet loss rate (e.g., from 0% to 10% and then to 30%) will cause the penalty factor to grow exponentially, thus quickly identifying and isolating "gray links." Reliability penalty term. λ is the reliability sensitivity constant.
[0038] Energy penalty function (Ψeng): Employs a power-law model and includes an energy weight γ. When energy is abundant, this term is close to 1; when energy is depleted (Eres→0), this term increases exponentially, forcing routes to avoid low-energy nodes. Energy penalty term , where γ is the energy penalty exponent.
[0039] Step 3.3, Composite Cost Output: Multiply the linear base term (Mbase) and the nonlinear penalty term (Ppenalty) to output the final single-hop composite cost of the current link.
[0040] Specifically, compound costs Output: Multiply the linear baseline term by the nonlinear penalty term to output the final single-hop composite cost of the current link.
[0041] (iv) Step 4: Based on the single-hop composite cost output in step 3, perform path total metric aggregation, and use the distance vector algorithm to complete the selection of the local optimal route and the construction of the routing table.
[0042] The total metric (Metricpath) of the path P from source node S to destination node D is defined as the sum of the single-hop link costs (Costlink) on this path. Monotonicity proof and loop-free guarantee: Since bandwidth (Bw), delay (Dly), and the non-linear penalty term (Ppenalty) are all positive, and the reference bandwidth constant (Cref) > 0, the single-hop cost (Costlink) is always > 0. Therefore, for any node on the path, the total metric of the entire path is strictly greater than the metric of any local segment on that path. This monotonically increasing mathematical property fundamentally guarantees that the purely mathematical analytical algorithm of this invention can run seamlessly on traditional routing protocols based on the Bellman-Ford or DUAL algorithms without ever generating routing loops.
[0043] The specific content of building and updating the routing table includes: ① Path metric aggregation: The single-hop composite cost of all single-hop links traversed sequentially on the path from the source node to the destination node is accumulated to obtain the total metric value of the path.
[0044] ② Acyclic monotonicity check: The single-hop composite cost of each hop is always greater than zero, and the total metric is strictly monotonically increasing as the path extends.
[0045] ③ Optimal Path Selection and Table Entry: When a network node receives a routing update message from a neighboring node, it combines the locally calculated single-hop composite cost to obtain the total metric value of multiple candidate paths leading to the same destination node. The total metric value of each candidate path is compared, and the path with the smallest total metric value is selected as the optimal route. The destination address, next-hop node, and total metric value of the optimal route are then updated in the local routing table.
[0046] Step 5: Continuously monitor the change in the single-hop composite cost generated in Step 3. When a preset deterioration condition or periodic condition is met, trigger the broadcast of the updated routing status from Step 4 to the network. The triggering update mechanism includes periodic updates and event-driven fast-trigger updates. The periodic condition refers to the link's periodic heartbeat packets failing to be received normally, with a preset time interval. The deterioration condition refers to the link quality deteriorating to the point where, even when the link interface has link awareness capabilities, the route cannot be guaranteed to be loop-free.
[0047] Periodic updates: Nodes periodically broadcast the total metric value information in their local routing table to neighboring nodes according to a preset timer.
[0048] Event-driven rapid update: During periodic link probing, when link congestion, increased packet loss rate, or a sudden drop in node power causes the nonlinear penalty term in step 3 to grow exponentially or exponentially, resulting in the change in the latest calculated single-hop composite cost exceeding a preset hysteresis threshold, or when a physical disconnection of the underlying link is detected, the system ignores the periodic timer and immediately triggers the routing update mechanism. It proactively sends routing update messages containing cost deterioration information to neighboring nodes, prompting rapid convergence of the entire network and isolation of inferior links. Typically, the hysteresis threshold is set as a fixed percentage of the average link cost over a certain period, for example, 20%.
[0049] Example 2 Suppose that in a weak network environment, node A needs to transmit data to node B.
[0050] The following is the link characteristic data collected by the monitoring probe at a certain moment and after smoothing: bandwidth ,Delay Shaking Link load (Relative value, within the range of 0-255), reliability (Packet loss rate 20%), remaining energy ratio .
[0051] Parameter settings: Reference bandwidth Weighting factor , , nonlinear constants , .
[0052] The calculation results are as follows:
[0053] Linear reference term .
[0054] Penalty Factor , , .
[0055] Comprehensive punishment .
[0056] Calculate the final single-hop composite cost .
[0057] If in the subsequent process If it rises to 240, then calculate... It will reach 17, and then The cost increment will exceed 1200. At this point, the cost increment far exceeds the set latency threshold, and the system will immediately trigger route reselection, successfully avoiding the poorly performing links that are about to become congested. This demonstrates the high sensitivity and effectiveness of the metric model in weak network environments.
[0058] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A dynamic routing metric method based on nonlinear fusion of multidimensional heterogeneous features, characterized in that, Includes the following steps: Step 1: Establish a multidimensional heterogeneous feature space and obtain the original feature vectors of network nodes; Step 2: Based on the multidimensional heterogeneous feature space established in Step 1, filter and smooth the obtained original feature vectors to output the smoothed feature vectors. Step 3: Using the smoothed feature vector output in Step 2 as input data, combine the calculated linear baseline term with the nonlinear penalty term to obtain the single-hop composite cost of the current link. Step 4: Based on the single-hop composite cost output in Step 3, aggregate the total path metric and use the distance vector algorithm to select the local optimal route and construct the routing table; Step 5: Continuously monitor the change in the single-hop composite cost generated in Step 3. When the preset deterioration condition or periodic condition is met, trigger the broadcast of the updated routing status in Step 4 to the network. The periodic condition refers to the inability to receive periodic heartbeat packets of the link normally, with a time interval of a preset value. The deterioration condition refers to the state where the link quality deteriorates to the point that the route cannot be guaranteed to be a loop-free route when the link interface has link awareness capability.
2. The dynamic routing metric method based on nonlinear fusion of multidimensional heterogeneous features according to claim 1, characterized in that, In step 1, the process of establishing the multidimensional heterogeneous feature space is as follows: First, initialize the reference bandwidth constant, smooth forgetting factor, and weight parameters of each feature of the routing protocol; second, deploy state monitoring probes between the network layer and the link layer to periodically collect six-dimensional parameters to form the original feature vector; the six-dimensional parameters include bottleneck bandwidth representing transmission rate, cumulative delay representing transmission delay, link load rate representing interface throughput, reliability coefficient representing transmission success rate, jitter variance representing channel stability, and energy normalization factor representing the remaining power of the node.
3. The dynamic routing metric method based on nonlinear fusion of multidimensional heterogeneous features according to claim 2, characterized in that, When collecting the status of the underlying physical link, the status monitoring probe uses a second-order exponentially weighted moving average filtering algorithm to preprocess the original sampled values to filter out instantaneous noise interference. Combined with a set hysteresis threshold mechanism, it determines whether to trigger the update calculation of the upper-layer routing table.
4. The dynamic routing metric method based on nonlinear fusion of multidimensional heterogeneous features according to claim 2, characterized in that, In step 3, the specific process for calculating the single-hop composite cost includes: Step 3.1 Calculate the linear benchmark term: Convert the smoothed jitter variance into equivalent delay according to the preset proportional coefficient, and add it to the smoothed cumulative delay to obtain the comprehensive delay; divide the reference bandwidth constant by the smoothed bottleneck bandwidth to obtain the bandwidth cost; multiply the comprehensive delay and bandwidth cost by the corresponding weight coefficients respectively and sum them to generate the linear benchmark term of the measurement base. Step 3.2: Calculate the nonlinear penalty term: Use a hyperbolic model to perform a high-sensitivity mapping on the smoothed link load rate to obtain the congestion penalty function; use a natural exponential model to map the smoothed reliability coefficient to obtain the reliability penalty function; use a power-law model to map the smoothed energy normalization factor to obtain the energy penalty function; multiply the congestion penalty function, reliability penalty function, and energy penalty function together to generate the nonlinear penalty term. Step 3.3, Composite Cost Output: Multiply the linear baseline term by the nonlinear penalty term to output the final single-hop composite cost of the current link.
5. The dynamic routing metric method based on nonlinear fusion of multidimensional heterogeneous features according to claim 4, characterized in that, In the aforementioned nonlinear mapping model, when constructing the congestion penalty function, reliability penalty function, and energy penalty function, the sensitivity parameters and weight indices included are dynamically configured through the software control plane to adapt to the different tolerances of different network service types for latency, packet loss, or energy consumption.
6. The dynamic routing metric method based on nonlinear fusion of multidimensional heterogeneous features according to claim 1, characterized in that, In step 4, the specific content of constructing and updating the routing table includes: Path metric aggregation: The total metric value of the path is obtained by summing the single-hop composite cost of all single-hop links traversed sequentially on the path from the source node to the destination node. Acyclic monotonicity check: The single-hop composite cost of each hop is always greater than zero, and the total metric is strictly monotonically increasing as the path extends; Optimal path selection and table entry: When a network node receives a routing update message from a neighboring node, it combines the single-hop composite cost calculated locally to obtain the total metric value of multiple candidate paths leading to the same destination node; compares the total metric values of each candidate path, selects the path with the smallest total metric value as the optimal route, and updates the destination address, next-hop node, and total metric value of the optimal route in the local routing table.
7. The dynamic routing metric method based on nonlinear fusion of multidimensional heterogeneous features according to claim 1, characterized in that, In step 5, the update triggering mechanism includes periodic updates and event-driven fast-trigger updates: Periodic updates: Nodes periodically broadcast the total metric value information in their local routing table to neighboring nodes according to a preset timer; Event-driven rapid update: During periodic link probing, when the nonlinear penalty term in step 3 grows exponentially or exponentially due to link congestion, increased packet loss rate, or sudden drop in node power, causing the change in the latest calculated single-hop composite cost to exceed the preset hysteresis threshold, or when a physical disconnection of the underlying link is detected, the system ignores the periodic timer and immediately triggers the routing update mechanism, actively sending routing update messages containing cost deterioration information to neighboring nodes, prompting the entire network to converge quickly and isolate inferior links.