A multi-modal network-oriented adaptive elastic routing method

Abstract

The application discloses a kind of adaptive flexible routing methods for multi-modal network, belong to network communication security technical field, including by the service demand and each network modal service capability is abstracted as the multidimensional vector of uniform format, the optimal bearing mode is selected for service using TOPSIS algorithm;In the selected mode, according to four optimization goals, generate the pareto optimal candidate path set;In the data packet transmission stage, according to the minimum reliability requirement set by user dynamically adjusts the number of redundant paths in mode, and when single mode is damaged as a whole, cross-modal dispersion service bearing is formed, and the double-layer redundancy structure of path domain and mode domain is formed;Through phased adaptive switching mechanism, phased progressive response is carried out according to fault severity;The application uses the above method, can effectively deal with persistent network attack and complex fault scene under multi-modal heterogeneous network environment, significantly improve the reliability and transmission performance of end-to-end communication.

Description

Technical Field

[0001] This invention relates to the field of network communication security technology, and in particular to an adaptive resilient routing method for multimodal networks. Background Technology

[0002] With the rapid evolution of network technology, application scenarios such as smart cities, edge computing, and large-scale Internet of Things (IoT) are placing increasingly higher demands on network architecture. Traditional single-mode IP networks are becoming increasingly inadequate in terms of scalability, management flexibility, and service carrying capacity. Therefore, the existing technology report "On the Revolution of the Information Network Development Paradigm" points out that the development paradigm of information networks is undergoing a new round of revolutionary changes, urgently requiring a completely new network architecture to meet these needs.

[0003] Against this backdrop, existing technologies have proposed the PINet multimodal intelligent network architecture. Based on the design principle of "fully definable," PINet allows multiple heterogeneous network modes, such as IPv4, IPv6, GeoIP, MPLS, and SCION, to coexist and operate under unified coordination. This significantly improves the network's service carrying capacity and supports the continuous evolution of network functions without affecting existing deployments. However, while this multimodal convergence and open architecture brings flexibility, it also greatly expands the network's potential attack surface, posing a serious challenge to the reliability and security of communication.

[0004] Regarding cybersecurity threats, existing research has accumulated a great deal of results on traditional single-mode network architectures and proposed the SCION architecture. This architecture embeds encrypted authentication into the path construction and forwarding process, the label-switched forwarding mechanism in MPLS networks, and points out that while inheriting IP network security vulnerabilities, it also introduces attack surfaces such as label injection and the risk of traffic hijacking caused by malicious tampering of geolocation information in GeoIP networks.

[0005] In terms of resilient routing mechanisms, existing methods can be broadly categorized into three types. The first type is a passive response method based on multipath, which relies on pre-computed backup paths to trigger rapid rerouting to restore connectivity after a failure. The second type is a measurement-driven method based on real-time detection, which senses network changes and dynamically adjusts routes by continuously monitoring queue statistics or actively probing. The third type is a proactive method based on prediction, which uses machine learning models to infer future network conditions from historical traffic and topology characteristics, and adjusts routing decisions in advance to avoid potential outages.

[0006] However, all the above methods share a fundamental flaw: they are designed for single-modal networks and are difficult to apply directly to heterogeneous multimodal environments like PINet. Specifically, the topology modeling, protocol semantics, and threat assumptions of existing routing mechanisms are tightly bound to specific network modes, failing to effectively handle cross-modal collaborative interactions. Furthermore, due to the lack of a unified cross-modal threat model, existing solutions cannot systematically analyze attack propagation behavior in multimodal environments, nor can they formulate targeted routing protection strategies accordingly. Faced with the continuously evolving and unknown attacks in PINet, defense methods relying solely on preset rules or historical data are insufficient to guarantee the reliability and continuity of end-to-end communication. Summary of the Invention

[0007] The purpose of this invention is to provide an adaptive resilient routing method for multimodal networks, in order to solve the problems of resilient routing technology in multimodal heterogeneous network environments as described in the background.

[0008] To achieve the above objectives, this invention provides an adaptive resilient routing method for multimodal networks, comprising the following steps: S1. Collect network status information; S2. Model the demand for new service traffic entering the network and match the optimal bearer mode for the current service; S3. In the selected optimal bearer mode, with transmission reliability, end-to-end delay, maximum link utilization and path jitter as optimization objectives, a set of Pareto optimal candidate paths is generated by using a hybrid multi-objective genetic-ant colony cooperative optimization algorithm. S4. During the data packet transmission phase, the number of redundant paths within a mode is dynamically adjusted according to the minimum reliability requirements set by the user, and the service is distributed across modes when a single mode is damaged, forming a dual-layer redundancy structure of path domain and mode domain. S5. The system continuously monitors the link status of each path and responds progressively through a phased adaptive switching mechanism based on the severity of the fault.

[0009] Preferably, step S1 specifically involves: the controller periodically discovering the link connection relationships in the network through the link layer discovery protocol, and generating a global network topology map. ,in, For a set of network nodes, The controller sends feature request messages and port statistics messages to each switching node to obtain the port status information of each node, and calculates the real-time available bandwidth and traffic load of each link to form a network status view for subsequent path calculation.

[0010] Preferably, step S2 specifically includes: When new service traffic enters the network, the service... The requirements are abstracted into a multi-dimensional requirement vector: ,in, This indicates latency and jitter tolerance. Indicates the expected bandwidth level. This represents the reliability requirement, with each component representing a value relative to the original requirement. The normalization results; the network modes Service capabilities are also abstracted into capability vectors. Weighting coefficients are introduced based on the business's emphasis on each performance dimension. ,satisfy ,in, Indicates latency service capability. Indicates bandwidth service capacity. Indicates reliability service capability. This indicates a preference for weighting the latency dimension. This indicates a preference for weighting the bandwidth dimension. This indicates a preference for weighting the reliability dimension. After formalizing the business requirement vector and modal capabilities, modality selection is performed based on the TOPSIS algorithm, and the weighted modal capability vector is denoted as... The ideal solution is the weighted business demand vector. The negative ideal solution is defined by the zero performance state. Calculate the Euclidean distances between each mode and the ideal solution and the negative ideal solution: , The proximity coefficients for each mode are expressed as follows: , ;right Sort the modes in ascending order and select the mode with the smallest proximity coefficient as the optimal mode for carrying the current service. .

[0011] Preferably, in step S3, in the selected optimal load-bearing mode Within the corresponding network domain, the Multipath Resilient Routing Optimization Problem (MRRP) is defined, with four specific optimization objectives: Transmission reliability : Set up a link Availability is This represents the probability that the link is in normal working condition per unit of time. Transmission reliability is defined as the path... Availability of each link The product of and reflects the ability of the path to remain connected without interruption during the transmission cycle, and is expressed as: ; End-to-end delay Defined as a path The sum of the latency of each link is used to directly reflect the path's ability to support real-time services. Let the path be... It consists of several links, and the average delay of each link is denoted as . ,but ;in, Indicates the first in the path The sequence number of the link; Representing a path Total number of links; Maximum link utilization Defined as the maximum percentage of bandwidth utilization across all links during a network's operational cycle, it measures the data transmission capacity a path can provide under the current network conditions. For paths consisting of multi-hop links... Its overall bandwidth capacity is limited by the link with the smallest bandwidth on the path. The bottleneck bandwidth model is used to define the path bandwidth performance: ;in, Indicates a specific moment in time; This represents a time set representing a network's operating cycle; Indicates the source node; Indicates the destination node; Indicates time The set of links; Indicates the first Path; Indicates from node To the node Traffic; Indicates link Total bandwidth traffic; Indicates from node To the node The business traffic requirements; Path jitter Used to measure path The degree of fluctuation in end-to-end delay during transmission, assuming the path... The continuous time delay sample sequence measured above is The jitter is then defined as the standard deviation of the time delay sequence: ;in, Indicates the number of samples in the time-delayed sample sequence; This represents the mean of the time-delayed sample sequence; The optimization objective of MRRP is to minimize the above four optimization objectives simultaneously under conditions of node failure, link failure, and attack interference: .

[0012] Preferably, step S3 uses a hybrid multi-objective genetic-ant colony cooperative optimization algorithm to generate a Pareto optimal candidate path set, specifically including: Individual coding employs a path representation method based on node sequences, treating each node as a gene locus, and representing the entire individual as an ordered sequence from the source node to the destination node. There must be a valid link connection between adjacent nodes; initial population The population is generated by randomly traversing feasible paths in the topology graph, while adhering to topological reachability and maximum hop count constraints, ensuring that the population has a certain degree of diversity and feasibility. For a particular individual in the population Calculate its transmission reliability End-to-end delay Maximum link utilization and path jitter The fitness values ​​on the four objective functions form the multi-objective performance evaluation vector for each candidate path; Based on the above fitness assessment results, the NSGA-II non-dominated ordination mechanism was used to perform Pareto stratification on all individuals in the population: if the path The path is not inferior to the target path on all objectives. And is strictly superior to at least one objective. Then it is called Dominate The algorithm assigns all paths not dominated by any other individual to the first Pareto level, which serves as the set of Pareto optimal candidate paths for the current generation. The remaining individuals are stratified according to their dominance relationship. Within the same Pareto level, the crowding distance is further introduced to sort the individuals, giving priority to retaining sparsely distributed individuals in the target space to maintain the diversity and uniform coverage of the solution set. During the intersection phase, a bridging point identification mechanism is introduced to ensure path validity: given two parent paths... and Seeking satisfaction and between, and There are pairs of bridging nodes with valid links between them, and these are used as the intersection points. Prefix segment and The suffix segments are spliced ​​to generate reasonable crossover solutions. If there are multiple bridging point pairs, one is randomly selected to maintain population diversity. During the mutation phase, a path suffix reconstruction strategy is adopted: randomly select an intermediate node that is not the beginning or end of the path, and replace the sub-path segment from that node to the destination node with a feasible sub-path replanned based on the current topology. If the reconstructed path meets the feasibility requirements, it is accepted as a new individual; otherwise, it is rolled back and a new point is selected for execution. Simultaneously, an ant colony pheromone mechanism is introduced during the crossover and mutation phases, and the crossover, mutation, and pheromone update processes are iterated until the preset maximum number of iterations is reached or the solution set converges. The optimal carrying mode selected in step S2 is then determined. Within the corresponding network domain, multi-objective optimization is performed on the candidate transmission paths to ultimately output a Pareto optimal candidate path set. It is used for subsequent redundant transmission phases.

[0013] Preferably, the ant colony pheromone mechanism is as follows: Set link In the The historical attractiveness of each iteration is determined by the pheromone matrix. Heuristic information Reflects the real-time performance characteristics of the link; at the slave node Select the next hop node When the transition probability is defined as: ; in, Represents a node The reachable adjacency set; To control the relative weight of historical experience and real-time link performance, Represents a node Adjacent links, Indicates time link In the The pheromone concentration in each iteration Indicates link The heuristic, real-time quality information is updated by pheromones according to the following rules: ; in, This is a pheromone volatile factor used to prevent premature convergence. Indicates the first Only one ant checked the link in this iteration. The increase in pheromones; Finally, the pheromone concentration of each link is used as a regulating factor for the crossover probability and mutation probability, enabling the genetic operation to adaptively adjust the operator behavior based on historical search experience: ; ; in, Indicates link Adaptive crossover probability; Indicates the basic crossover probability; Indicates link Current pheromone concentration; This represents the maximum value of the pheromone concentration across all links in the current iteration; Indicates link The adaptive mutation probability, Represents the basic mutation probability. The pheromone regulation coefficient represents the probability of mutation.

[0014] Preferably, in step S4, the user side sets the minimum reliability requirement parameters according to the application scenario. This indicates the user's minimum acceptable reliability requirement for the current transmission task, and is transmitted to the network side via a reliability flag appended to the packet header. The network side then selects paths sequentially from the candidate path set. Given a redundant path, assuming that each path failure event is independent, and that successful data reception from any path by the system receiver is considered a successful overall transmission, the probability model for the joint transmission success of the system is as follows: ;in, Indicates the transmission reliability of redundant paths. This indicates the first path selected from the candidate path set. There are redundant paths, and the product term represents the probability that all paths fail simultaneously. If the current path selection meets the reliability requirements, the path selection process terminates; otherwise, redundant paths are expanded until the threshold condition is met or the maximum number of paths is reached.

[0015] Preferably, in step S4, when network attacks, interference, or strategic congestion cause a synchronous decline in the overall transmission reliability of a certain mode, adding redundant paths within the mode can no longer restore service availability. The system will select a backup mode that is independent of the current mode in terms of control domain, routing system, or link physical characteristics from other modes according to the minimum service reliability requirements, and perform redundant bearing and traffic migration between multiple modes. By distributing service transmission among different modes, the system bypasses the attacked or damaged mode, forming a dual-layer redundancy structure of path domain and mode domain, thereby maintaining service continuity and improving the overall resilience of the network in complex adversarial environments.

[0016] Preferably, in step S5, the system continuously monitors the link status of each path and responds in three progressive stages according to the severity of the fault, based on two mechanisms: periodic triggering and conditional triggering. Specifically: The first phase, path-level fine-tuning: When the link monitoring module detects a single data packet loss on the currently active path, the system determines it as an early signal of potential path performance degradation and adjusts the path based on the current optimal candidate path set. The existing Pareto front candidate path set is sorted by path reliability, bottleneck bandwidth, and time delay stability to obtain a priority sequence. When the path When a single packet loss occurs, the system immediately switches to the next highest priority path. To replace the current path and take over the service transmission; In the second stage, path set recalculation: when the same path becomes unreachable in multiple consecutive detection cycles, the system determines that the fault has been upgraded from instantaneous disturbance to stability degradation or physical failure. At this time, the hybrid multi-objective genetic-ant colony algorithm in step S4 is triggered to regenerate the candidate path set in the current mode to replace the failed path and restore transmission. The third stage, mode-level switching: When unreachability occurs in a single mode and the data reachability of that mode decreases synchronously, and latency and jitter show a systematic deterioration trend, the system determines that the mode-level availability has decreased. This usually stems from a targeted attack against a specific mode. The system will directly perform mode switching on the current transmission link, migrating the service traffic to another mode that is independent of the current mode in terms of routing domain, physical link, or control mechanism. During the switching process, the system will simultaneously consider the mode switching overhead, encapsulation / protocol mapping cost, and the availability and stability of the target mode path, and select the most suitable alternative mode for carrying according to service priority and security requirements.

[0017] Therefore, the above-mentioned adaptive resilient routing method for multimodal networks, adopted in this invention, has the following beneficial effects: (1) In terms of routing robustness, through systematic cross-modal threat analysis and threat-driven routing design, it can maintain network reachability significantly higher than traditional routing protocols (OSPF, ISIS) in both link failure and switch failure scenarios. It can also maintain near-full reachability performance under high failure rate, and has good fault tolerance and cross-topology generalization ability. (2) In terms of fault recovery speed, relying on the phased adaptive switching mechanism, millisecond-level route recovery can be achieved in both link failure and switch failure scenarios. Compared with the second-level reconvergence delay of traditional protocols and the tens to hundreds of millisecond-level recovery delay of MCF, it has stronger real-time response capability, and the recovery time remains highly stable as the failure ratio changes. (3) In terms of overall transmission performance, a reasonable balance is achieved between the three indicators of reachability, recovery speed and end-to-end latency. It can ensure the continuity of critical services while keeping the additional overhead introduced by redundant transmission within a reasonable range. It can still maintain stable service quality in the face of unknown attacks and continuous evolution.

[0018] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating an adaptive resilient routing method for multimodal networks according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the multimodal network system model of the present invention; Figure 3 This is a diagram of the AMRES multipath resilient routing framework of the present invention; Figure 4 The following are comparison results of routing performance under fault scenarios in embodiments of the present invention: (a) is the packet loss rate result under the link failure ratio; (b) is the end-to-end latency result under the link failure ratio; (c) is the routing recovery time result under the link failure ratio; (d) is the packet loss rate result under the switch failure ratio; (e) is the end-to-end latency result under the switch failure ratio; and (f) is the routing recovery time result under the switch failure ratio. Detailed Implementation

[0020] The following detailed description of embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0021] Please see Figures 1-3 An adaptive resilient routing method for multimodal networks, comprising two main aspects: Firstly, a modality-aware path computation method for multimodal networks is provided.

[0022] In multimodal networks, different network modes such as IPv4, IPv6, GeoIP, MPLS, and SCION each have their own advantages and limitations in terms of latency, bandwidth, reliability, and security. When new service traffic enters the network, the most suitable bearer mode needs to be selected based on the service's performance requirements. To this end, this method abstracts the service requirements and the service capabilities of each mode into multidimensional vectors in a unified format, and uses the TOPSIS algorithm (Approximation of Ideal Solution Ranking) for mode matching. By calculating the degree of closeness of each mode to the ideal performance state, the bearer mode with the best overall performance for the current service is selected.

[0023] After mode selection, candidate transmission paths are further calculated within the network domain of the selected mode. Unlike traditional path calculation methods that only aim for the shortest path, this method simultaneously optimizes four metrics: transmission reliability, end-to-end delay, maximum link utilization, and path jitter. A hybrid multi-objective genetic-ant colony collaborative optimization algorithm is employed for path search. This algorithm leverages the global exploration capability of the genetic algorithm to ensure the diversity of candidate paths, while introducing an ant colony pheromone mechanism to empirically reinforce historically high-performing paths, guiding the search towards a high-quality path set. In the genetic operation, a bridging point identification mechanism and a path suffix reconstruction strategy are designed for the special encoding form of path representation as a node sequence. These are used in the crossover and mutation stages, respectively, to ensure that the paths generated during evolution always satisfy network topology connectivity constraints. The final output is a Pareto optimal candidate path set covering multiple objective trade-offs.

[0024] Secondly, a phased adaptive path switching method for multimodal networks is provided.

[0025] In terms of redundant transmission strategy, the user side sets a minimum reliability requirement threshold based on the actual business scenario. The network side dynamically selects several redundant paths from the candidate path set to carry services in parallel. By modeling the success probability of multi-path joint transmission, the number of redundant paths is continuously expanded until the user-defined reliability requirements are met. When the overall availability of a single mode decreases, the system further distributes service carrying across modes by introducing backup paths from other modes that are independent of each other in the routing domain, physical link, or control mechanism, forming a two-layer redundancy structure in the path domain and mode domain.

[0026] In terms of the dynamic switching mechanism, the system continuously monitors the link status of each path and responds in three progressive stages based on the severity of the fault. When a single packet loss is detected, it is determined to be an early sign of path performance degradation, and the system directly switches from the current Pareto front candidate path to the next highest priority path for rapid fine-tuning. When the same path remains unreachable for multiple consecutive probing cycles, it is determined to be a path stability failure, triggering a recalculation of the candidate path set to generate a new optimized path set within the current mode. When unreachability is concentrated in a single mode and shows a systemic deterioration trend, it is determined to be a mode-level availability degradation (usually stemming from a targeted attack against a specific mode). At this point, mode switching is directly executed, migrating the service to the optimal alternative mode based on comprehensive evaluation. This phased mechanism ensures service continuity while keeping the control overhead of route recalculation to a minimum necessary level.

[0027] The specific steps are as follows: S1. Collect network status information: The controller periodically discovers link connections in the network using the Link Layer Discovery Protocol (LLDP) to generate a global network topology map. ,in, It is a set of network nodes (including forwarding devices such as routers and switches). The controller sends feature request messages and port statistics messages to each switching node to obtain the port status information of each node, and calculates the real-time available bandwidth and traffic load of each link to form a network status view for subsequent path calculation.

[0028] S2. Business Requirements Modeling and Modality Matching: When new service traffic enters the network, the service... The requirements are abstracted into a multi-dimensional requirement vector: ,in, This indicates latency and jitter tolerance. Indicates the expected bandwidth level. This represents the reliability requirement, with each component representing a value relative to the original requirement. The normalization results; the network modes The service capabilities (including IPv4, IPv6, GeoIP, MPLS, and SCION) are also abstracted into capability vectors. Weighting coefficients are introduced based on the business's emphasis on each performance dimension. ,satisfy ,in, Indicates latency service capability. Indicates bandwidth service capacity. Indicates reliability service capability. This indicates a preference for weighting the latency dimension. This indicates a preference for weighting the bandwidth dimension. This indicates a preference for weighting the reliability dimension. After formalizing the business requirement vector and modal capabilities, modality selection is performed based on the TOPSIS algorithm, and the weighted modal capability vector is denoted as... The ideal solution is the weighted business demand vector. The negative ideal solution is defined by the zero performance state. Calculate the Euclidean distances between each mode and the ideal solution and the negative ideal solution: , The closer a mode is to the ideal solution, the closer its overall performance is to the business expectation; the similarity coefficient of each mode is expressed as: , ;right Sort the modes in ascending order and select the mode with the smallest proximity coefficient as the optimal mode for carrying the current service. .

[0029] S3. Define a multi-objective optimization problem: In the selected optimal load-bearing mode Within the corresponding network domain, the Multi-Path Resilient Routing Optimization Problem (MRRP) is defined, with four metrics as optimization objectives: Transmission reliability : Set up a link Availability is This represents the probability that the link is in normal working condition per unit of time. Transmission reliability is defined as the path... Availability of each link The product of and reflects the ability of the path to remain connected without interruption during the transmission cycle, and is expressed as: ; End-to-end delay Defined as a path The sum of the latency of each link is used to directly reflect the path's ability to support real-time services. Let the path be... It consists of several links, and the average delay of each link is denoted as . ,but ;in, Indicates the first in the path The sequence number of the link; Representing a path Total number of links; Maximum link utilization Defined as the maximum percentage of bandwidth utilization across all links during a network's operational cycle, it measures the data transmission capacity a path can provide under the current network conditions. For paths consisting of multi-hop links... Its overall bandwidth capacity is limited by the link with the smallest bandwidth on the path. The bottleneck bandwidth model is used to define the path bandwidth performance: ;in, Indicates a specific moment in time; This represents a time set representing a network's operating cycle; Indicates the source node; Indicates the destination node; Indicates time The set of links; Indicates the first Path; Indicates from node To the node Traffic; Indicates link Total bandwidth traffic; Indicates from node To the node The business traffic requirements; Path jitter Used to measure path The degree of fluctuation in end-to-end delay during transmission, assuming the path... The continuous time delay sample sequence measured above is The jitter is then defined as the standard deviation of the time delay sequence: ;in, Indicates the number of samples in the time-delayed sample sequence; This represents the mean of the time-delayed sample sequence; The optimization objective of MRRP is to minimize the above four optimization objectives simultaneously under conditions of node failure, link failure, and attack interference: .

[0030] S4. Hybrid Multi-Objective Genetic-Ant Colony Cooperative Optimization Algorithm: Individual coding employs a path representation method based on node sequences, treating each node as a gene locus, and representing the entire individual as an ordered sequence from the source node to the destination node. There must be a valid link connection between adjacent nodes; initial population By randomly traversing feasible paths in the topology graph, and adhering to topological reachability and maximum hop count constraints, the population is ensured to have a certain degree of diversity and feasibility.

[0031] For a particular individual in the population Calculate its transmission reliability End-to-end delay Maximum link utilization and path jitter The fitness values ​​on the four objective functions form the multi-objective performance evaluation vector for each candidate path.

[0032] Based on the above fitness assessment results, the NSGA-II non-dominated ordination mechanism was used to perform Pareto stratification on all individuals in the population: if the path The path is not inferior to the target path on all objectives. And is strictly superior to at least one objective. Then it is called Dominate The algorithm assigns all paths not dominated by any other individual to the first Pareto level, which serves as the set of Pareto optimal candidate paths for the current generation. The remaining individuals are stratified according to their dominance relationship. Within the same Pareto level, the crowding distance is further introduced to sort the individuals, giving priority to retaining sparsely distributed individuals in the target space to maintain the diversity and uniform coverage of the solution set.

[0033] During the intersection phase, a bridging point identification mechanism is introduced to ensure path validity: given two parent paths... and Seeking satisfaction and between, and There are pairs of bridging nodes with valid links between them, and these are used as the intersection points. Prefix segment and The suffix segments are spliced ​​to generate reasonable crossover solutions. If there are multiple bridging point pairs, one is randomly selected to maintain population diversity.

[0034] During the mutation phase, a path suffix reconstruction strategy is adopted: a non-starting and non-ending intermediate node is randomly selected in the path, and the sub-path segment from the node to the destination node is replaced with a feasible sub-path replanned based on the current topology. If the reconstructed path meets the feasibility requirements, it is accepted as a new individual; otherwise, it is rolled back and a new point is selected for execution.

[0035] Simultaneously, the ant colony pheromone mechanism is introduced during the crossover and mutation phases: Set link In the The historical attractiveness of each iteration is determined by the pheromone matrix. This indicates the frequency and quality of access to a given path in historical iterations; heuristic information. It reflects the real-time performance characteristics of the link, and is used to represent the real-time quality of the link under the current network conditions; in the slave node Select the next hop node When the transition probability is defined as: ; in, Represents a node The reachable adjacency set; To control the relative weights of historical experience (pheromones) and immediate link performance (heuristic information), Represents a node Adjacent links, Indicates time link In the The pheromone concentration in each iteration Indicates link The heuristic, real-time quality information is updated by pheromones according to the following rules: ; in, This is a pheromone volatile factor used to prevent premature convergence. Indicates the first Only one ant checked the link in this iteration. The increase in pheromones.

[0036] Finally, the pheromone concentration of each link is used as a regulating factor for the crossover probability and mutation probability, enabling the genetic operation to adaptively adjust the operator behavior based on historical search experience: ; ; in, Indicates link Adaptive crossover probability; Indicates the basic crossover probability; Indicates link Current pheromone concentration; This represents the maximum value of the pheromone concentration across all links in the current iteration; Indicates link The adaptive mutation probability, Represents the basic mutation probability. The pheromone regulation coefficient represents the probability of mutation.

[0037] The process iterates through the crossover phase, mutation phase, and pheromone update until the preset maximum number of iterations is reached or the solution set converges. The optimal carrying mode selected in step S2 is then... Within the corresponding network domain, multi-objective optimization is performed on the candidate transmission paths to ultimately output a Pareto optimal candidate path set. It is used for subsequent redundant transmission phases.

[0038] S5. Formulate intra-modal redundancy transmission strategy: The user sets minimum reliability requirements based on the application scenario. This indicates the user's minimum acceptable reliability requirement for the current transmission task, and is transmitted to the network side via a reliability flag appended to the packet header. The network side then selects paths sequentially from the candidate path set. Given a redundant path, assuming that each path failure event is independent, and that successful data reception from any path by the system receiver is considered a successful overall transmission, the probability model for the joint transmission success of the system is as follows: ;in, Indicates the transmission reliability of redundant paths. This indicates the first path selected from the candidate path set. There are redundant paths, and the product term represents the probability that all paths fail simultaneously. If the current path selection meets the reliability requirements, the path selection process terminates; otherwise, redundant paths are expanded until the threshold condition is met or the maximum number of paths is reached.

[0039] S6. Formulate inter-modal transmission strategy: When network attacks, interference, or strategic congestion cause a synchronous decline in the overall transmission reliability of a certain mode, adding redundant paths within the mode can no longer restore service availability. The system will select a backup mode that is independent of the current mode in terms of control domain, routing system, or link physical characteristics from other modes according to the minimum reliability requirements of the service. Redundant bearing and traffic migration will be carried out between multiple modes. By distributing service transmission among different modes, the system bypasses the attacked or damaged mode, forming a two-layer redundancy structure of path domain and mode domain, thereby maintaining service continuity and improving the overall resilience of the network in complex adversarial environments.

[0040] S7. Phased adaptive path switching: The system continuously monitors the link status of each path and responds in three progressive stages based on the severity of the fault, using two mechanisms: periodic triggering and conditional triggering. The first phase, path-level fine-tuning: When the link monitoring module detects a single data packet loss on the currently active path, the system determines it as an early signal of potential path performance degradation and adjusts the path based on the current optimal candidate path set. The existing Pareto front candidate path set is sorted by path reliability, bottleneck bandwidth, and time delay stability to obtain a priority sequence. When the path When a single packet loss occurs, the system immediately switches to the next highest priority path. To replace the current path and take over the service transmission; In the second stage, path set recalculation: when the same path becomes unreachable in multiple consecutive probe cycles (e.g., three consecutive probe packets fail to return), the system determines that the fault has escalated from a momentary disturbance to stability degradation or physical failure. At this time, the hybrid multi-objective genetic-ant colony algorithm in step S4 is triggered to regenerate the candidate path set in the current mode to replace the failed path and restore transmission. The third stage, mode-level switching: When unreachability occurs in a single mode and the data reachability of that mode decreases synchronously, and latency and jitter show a systematic deterioration trend, the system determines that the mode-level availability has decreased. This usually stems from a targeted attack against a specific mode. The system will directly perform mode switching on the current transmission link, migrating the service traffic to another mode that is independent of the current mode in terms of routing domain, physical link, or control mechanism. During the switching process, the system will simultaneously consider the mode switching overhead, encapsulation / protocol mapping cost, and the availability and stability of the target mode path, and select the most suitable alternative mode for carrying according to service priority and security requirements.

[0041] In this embodiment, under an IPv4 network environment, the performance of ISIS, MCF, and the proposed method was compared in terms of packet loss rate, end-to-end latency, and route recovery time when the link / switch failure rates were 5%, 10%, 15%, and 20%, respectively. Figure 4 As shown.

[0042] Therefore, this invention adopts the above-mentioned adaptive resilient routing method for multimodal networks, which achieves high reliability and high performance transmission guarantee for end-to-end communication in multimodal network environments through systematic cross-modal threat modeling, hybrid multi-objective path calculation and adaptive dynamic transmission mechanism.

[0043] 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 preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. An adaptive resilient routing method for multimodal networks, characterized in that, Includes the following steps: S1. Collect network status information; S2. Model the demand for new service traffic entering the network and match the optimal bearer mode for the current service; S3. In the selected optimal bearer mode, with transmission reliability, end-to-end delay, maximum link utilization and path jitter as optimization objectives, a set of Pareto optimal candidate paths is generated by using a hybrid multi-objective genetic-ant colony cooperative optimization algorithm. S4. During the data packet transmission phase, the number of redundant paths within a mode is dynamically adjusted according to the minimum reliability requirements set by the user, and the service is distributed across modes when a single mode is damaged, forming a dual-layer redundancy structure of path domain and mode domain. S5. The system continuously monitors the link status of each path and responds progressively through a phased adaptive switching mechanism based on the severity of the fault. Step S2 is as follows: When new service traffic enters the network, the service... The requirements are abstracted into a multi-dimensional requirement vector: ,in, This indicates latency and jitter tolerance. Indicates the expected bandwidth level. This represents the reliability requirement, with each component representing a value relative to the original requirement. The normalization results; the network modes Service capabilities are also abstracted into capability vectors. Weighting coefficients are introduced based on the business's emphasis on each performance dimension. ,satisfy ,in, Indicates latency service capability. Indicates bandwidth service capacity. Indicates reliability service capability. This indicates a preference for weighting the latency dimension. This indicates a preference for weighting the bandwidth dimension. This indicates a preference for weighting the reliability dimension. After formalizing the business requirement vector and modal capabilities, modality selection is performed based on the TOPSIS algorithm, and the weighted modal capability vector is denoted as... The ideal solution is the weighted business demand vector. The negative ideal solution is defined by the zero performance state. Calculate the Euclidean distances between each mode and the ideal solution and the negative ideal solution: , The proximity coefficients for each mode are expressed as follows: , ;right Sort the modes in ascending order and select the mode with the smallest proximity coefficient as the optimal mode for carrying the current service. ; In step S5, the system continuously monitors the link status of each path and responds in three progressive stages based on the severity of the fault, according to two mechanisms: periodic triggering and conditional triggering. The first phase, path-level fine-tuning: When the link monitoring module detects a single data packet loss on the currently active path, the system determines it as an early signal of potential path performance degradation and adjusts the path based on the current optimal candidate path set. The existing Pareto front candidate path set is sorted by path reliability, bottleneck bandwidth, and time delay stability to obtain a priority sequence. When the path When a single packet loss occurs, the system immediately switches to the next highest priority path. To replace the current path and take over the service transmission; In the second stage, path set recalculation: when the same path becomes unreachable in multiple consecutive detection cycles, the system determines that the fault has been upgraded from instantaneous disturbance to stability degradation or physical failure. At this time, the hybrid multi-objective genetic-ant colony algorithm in step S4 is triggered to regenerate the candidate path set in the current mode to replace the failed path and restore transmission. The third stage is mode-level switching: When unreachability occurs in a single mode and the data reachability of that mode decreases synchronously, and latency and jitter show a systematic deterioration trend, the system determines that the mode-level availability has decreased. This usually stems from a targeted attack against a specific mode. The system will directly perform mode switching on the current transmission link, migrating the service traffic to another mode that is independent of the current mode in terms of routing domain, physical link, or control mechanism. During the switching process, the system will simultaneously consider the mode switching overhead, encapsulation / protocol mapping cost, and the availability and stability of the target mode path, and select the most suitable alternative mode for carrying according to service priority and security requirements.

2. The adaptive resilient routing method for multimodal networks according to claim 1, characterized in that, Step S1 specifically involves the controller periodically discovering link connections in the network through a link-layer discovery protocol, thereby generating a global network topology map. ,in, For a set of network nodes, The controller sends feature request messages and port statistics messages to each switching node to obtain the port status information of each node, and calculates the real-time available bandwidth and traffic load of each link to form a network status view for subsequent path calculation.

3. The adaptive resilient routing method for multimodal networks according to claim 1, characterized in that, In step S3, in the selected optimal load-bearing mode Within the corresponding network domain, the Multipath Resilient Routing Optimization Problem (MRRP) is defined, with four specific optimization objectives: Transmission reliability : Set up a link Availability is This represents the probability that the link is in normal working condition per unit of time. Transmission reliability is defined as the path... Availability of each link The product of and reflects the ability of the path to remain connected without interruption during the transmission cycle, and is expressed as: ; End-to-end delay Defined as a path The sum of the latency of each link is used to directly reflect the path's ability to support real-time services. Let the path be... It consists of several links, and the average delay of each link is denoted as . ,but ;in, Indicates the first in the path The sequence number of the link; Representing a path Total number of links; Maximum link utilization Defined as the maximum percentage of bandwidth utilization across all links during a network's operational cycle, it measures the data transmission capacity a path can provide under the current network conditions. For paths consisting of multi-hop links... Its overall bandwidth capacity is limited by the link with the smallest bandwidth on the path. The bottleneck bandwidth model is used to define the path bandwidth performance: ;in, Indicates a specific moment in time; This represents a time set representing a network's operating cycle; Indicates the source node; Indicates the destination node; Indicates time The set of links; Indicates the first Path; Indicates from node To the node Traffic; Indicates link Total bandwidth traffic; Indicates from node To the node The business traffic requirements; Path jitter Used to measure path The degree of fluctuation in end-to-end delay during transmission is denoted as the path... The measured continuous time delay sample sequence is as follows The jitter is then defined as the standard deviation of the time-delayed sample sequence: ;in, Indicates the number of samples in the time-delayed sample sequence; This represents the mean of the time-delayed sample sequence; The optimization objective of MRRP is to minimize the above four optimization objectives simultaneously under conditions of node failure, link failure, and attack interference: 。 4. The adaptive resilient routing method for multimodal networks according to claim 3, characterized in that, Step S3 uses a hybrid multi-objective genetic-ant colony cooperative optimization algorithm to generate a Pareto optimal candidate path set, specifically including: Individual coding employs a path representation method based on node sequences, treating each node as a gene locus, and representing the entire individual as an ordered sequence from the source node to the destination node. There must be a valid link connection between adjacent nodes; initial population The population is generated by randomly traversing feasible paths in the topology graph, while adhering to topological reachability and maximum hop count constraints, ensuring that the population has a certain degree of diversity and feasibility. For a particular individual in the population Calculate its transmission reliability End-to-end delay Maximum link utilization and path jitter The fitness values ​​on the four objective functions form the multi-objective performance evaluation vector for each candidate path; Based on the above fitness assessment results, the NSGA-II non-dominated ordination mechanism was used to perform Pareto stratification on all individuals in the population: if the path The path is not inferior to the target path on all objectives. And is strictly superior to at least one objective. Then it is called Dominate The algorithm assigns all paths not dominated by any other individual to the first Pareto level, which serves as the set of Pareto optimal candidate paths for the current generation. The remaining individuals are stratified according to their dominance relationship. Within the same Pareto level, the crowding distance is further introduced to sort the individuals, giving priority to retaining sparsely distributed individuals in the target space to maintain the diversity and uniform coverage of the solution set. During the intersection phase, a bridging point identification mechanism is introduced to ensure path validity: given two parent paths... and Seeking satisfaction and between, and There are pairs of bridging nodes with valid links between them, and these are used as the intersection points. Prefix segment and The suffix segments are spliced ​​to generate reasonable crossover solutions. If there are multiple bridging point pairs, one is randomly selected to maintain population diversity. During the mutation phase, a path suffix reconstruction strategy is adopted: randomly select an intermediate node that is not the beginning or end of the path, and replace the sub-path segment from that node to the destination node with a feasible sub-path replanned based on the current topology. If the reconstructed path meets the feasibility requirements, it is accepted as a new individual; otherwise, it is rolled back and a new point is selected for execution. Simultaneously, an ant colony pheromone mechanism is introduced during the crossover and mutation phases, and the crossover, mutation, and pheromone update processes are iterated until the preset maximum number of iterations is reached or the solution set converges. The optimal carrying mode selected in step S2 is then determined. Within the corresponding network domain, multi-objective optimization is performed on the candidate transmission paths to ultimately output a Pareto optimal candidate path set. It is used for subsequent redundant transmission phases.

5. The adaptive resilient routing method for multimodal networks according to claim 4, characterized in that, The ant colony pheromone mechanism is as follows: Set link In the The historical attractiveness of each iteration is determined by the pheromone matrix. Heuristic information Reflects the real-time performance characteristics of the link; From the node Select the next hop node When the transition probability is defined as: ； in, Represents a node The reachable adjacency set; To control the relative weight of historical experience and real-time link performance, Represents a node Adjacent links, Indicates time link In the The pheromone concentration in each iteration Indicates link The heuristic, real-time quality information is updated by pheromones according to the following rules: ； in, This is a pheromone volatile factor used to prevent premature convergence. Indicates the first Only one ant checked the link in this iteration. The increase in pheromones; Finally, the pheromone concentration of each link is used as a regulating factor for the crossover probability and mutation probability, enabling the genetic operation to adaptively adjust the operator behavior based on historical search experience: ； ； in, Indicates link Adaptive crossover probability; Indicates the basic crossover probability; Indicates link Current pheromone concentration; This represents the maximum value of the pheromone concentration across all links in the current iteration; Indicates link The adaptive mutation probability, Represents the basic mutation probability. The pheromone regulation coefficient represents the probability of mutation.

6. The adaptive resilient routing method for multimodal networks according to claim 5, characterized in that, In step S4, the user side sets the minimum reliability requirement parameters according to the application scenario. This indicates the user's minimum acceptable reliability requirement for the current transmission task, and is transmitted to the network side via a reliability flag appended to the packet header. The network side then selects paths sequentially from the candidate path set. Given a redundant path, assuming that each path failure event is independent, and that successful data reception from any path by the system receiver is considered a successful overall transmission, the probability model for the joint transmission success of the system is as follows: ;in, Indicates the transmission reliability of redundant paths. This indicates the first path selected from the candidate path set. There are redundant paths, and the product term represents the probability that all paths fail simultaneously. If the current path selection meets the reliability requirements, the path selection process terminates; otherwise, redundant paths are expanded until the threshold condition is met or the maximum number of paths is reached.

7. The adaptive resilient routing method for multimodal networks according to claim 6, characterized in that: In step S4, when network attacks, interference, or strategic congestion cause a synchronous decline in the overall transmission reliability of a certain mode, adding redundant paths within the mode can no longer restore service availability. The system will select a backup mode that is independent of the current mode in terms of control domain, routing architecture, or link physical characteristics from other modes according to the minimum service reliability requirements. Redundant bearing and traffic migration will be performed between multiple modes. By distributing service transmission among different modes, the system bypasses the attacked or damaged mode, forming a dual-layer redundancy structure of path domain and mode domain, thereby maintaining service continuity and improving the overall resilience of the network in complex adversarial environments.

Images