An intelligent networking method for high-dynamic large-scale mobile ad hoc networks

Abstract

The present invention belongs to the technical field of intelligent mobile ad hoc networks, and discloses an intelligent networking method for high-dynamic large-scale mobile ad hoc networks, including the following steps: S1, establishing a networking model for high-dynamic large-scale mobile ad hoc networks; S2, constructing an intelligent routing decision-making method for high-dynamic large-scale mobile ad hoc networks according to communication requirements; S3, designing a security authentication strategy for node organization and routing transmission. By adopting the above-mentioned intelligent networking method for high-dynamic large-scale mobile ad hoc networks, the present invention realizes the organization and management of a large number of network nodes, solves the routing decision-making problem in various situations, ensures the security of network node access and transmission processes, and is applicable to new network application scenarios with high dynamics and large scale.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent mobile ad hoc networks, and in particular to an intelligent networking method for high-dynamic large-scale mobile ad hoc networks. Background Art

[0002] In today's network technology field, due to the advantages of mobile ad hoc networks such as not relying on infrastructure, flexible deployment, and strong anti-destruction ability, it has become one of the most active research and application directions in the network technology field, and with the increasing popularity of intelligent portable devices, it has penetrated into various industrial fields such as detection and inspection, emergency disaster relief, battlefield communication, and aerospace in an explosive manner. However, with the popularization of the application of new intelligent nodes such as unmanned aerial vehicles and high-speed vehicles, the demand for high-dynamic large-scale new mobile networking has emerged.

[0003] For traditional mobile networking, problems such as instantaneous changes in network topology, frequent connection interruptions, and rapid network domain switching caused by high dynamics, as well as the existence of infrastructure such as available fixed base stations, mobile base stations, and satellite base stations, are likely to have a subversive impact on the routing decision-making of ad hoc networks; a large number of network nodes have caused a sharp increase in network data volume, an increase in calculation or decision-making time, and the network transmission requirements caused by specific target tasks of mobile ad hoc networks are sudden, variable, and homogeneous, so it is difficult to quickly organize and converge network resources according to application requirements; in addition, there are various external threats in specific mobile ad hoc networks, such as node forgery, identity forgery, address forgery, etc., which will cause problems such as high possibility of privacy leakage and poor connection security controllability.

[0004] Therefore, the intelligent networking method has become a bottleneck for new network applications with high dynamics and large scale, and there is an urgent need for an intelligent mobile ad hoc network suitable for new network applications with high dynamics and large scale. Summary of the Invention

[0005] The purpose of the present invention is to provide an intelligent networking method for high-dynamic large-scale mobile ad hoc networks, which realizes the organization and management of a large number of network nodes, solves the routing decision-making problems in various situations, ensures the security of network node access and transmission processes, and is applicable to new network application scenarios with high dynamics and large scale.

[0006] To achieve the above purpose, the present invention provides an intelligent networking method for high-dynamic large-scale mobile ad hoc networks, including the following steps:

[0007] S1. Establish a networking model for high-dynamic large-scale mobile ad hoc networks: establish an organization model of guarantee nodes and functional nodes from a logical perspective, and establish an intelligent network domain division model from a physical perspective;

[0008] S2. According to the communication requirements, construct an intelligent routing decision method for high-dynamic large-scale mobile ad hoc networks, which is specifically divided into three cases: proactive intelligent routing decision, reactive intelligent routing decision, and opportunistic routing decision;

[0009] S3. Design a security authentication strategy for node organization and routing transmission, including two parts: node secure access authentication and transmission connection secure authentication.

[0010] Preferably, in step S1 of establishing the networking model, network nodes are divided into guarantee nodes and functional nodes according to their functions; and according to the realization of the networking function, it is divided into four functional modules: node organization engine, network domain division engine, routing decision engine, and security authentication engine;

[0011] All engine modules are deployed on network guarantee nodes, and the functions of each engine are completed through distributed cooperation; routing decision engine modules and security authentication engine modules are deployed on network functional nodes.

[0012] Preferably, the specific steps of step S1 for establishing the ad hoc network architecture model are as follows:

[0013] S111. Use mathematical optimization to construct the generalized network utility maximization problem of the ad hoc network architecture as follows:

[0014]

[0015] subject to Rx≤c

[0016] where, represents the guarantee node vector; represents the functional node vector; represents the transmission rate of the source node q in the network domain o; identifies the network utility function;

[0017] S112. Use mathematical optimization, causal reasoning, and reinforcement learning to model the Markov decision process and solve the generalized network utility maximization problem as follows:

[0018]

[0019] where, represents the state space; represents the action space; represents the state transition probability; represents the reward function;

[0020] S113. Obtain the optimal optimization algorithm through reinforcement learning, and solve to obtain the best routing strategy R * and transmission strategy x * .

[0021] Preferably, the specific steps of the node organization in step S1 are as follows:

[0022] S121. Node virtual organization:

[0023] Virtualize the guarantee nodes according to the logical connectivity to form a virtual mapping of the guarantee nodes.

[0024] Establish virtual mappings of multiple functional nodes according to the different types of functional nodes.

[0025] Connect the virtual mapping of the functional nodes and the virtual mapping of the guarantee nodes based on the physical or logical relationship to form a distributed virtual mapping organization of the network nodes.

[0026] S122. Optimized deployment of guarantee nodes:

[0027] First, establish a distributed structural causal model for the process of continuously and dynamically adjusting the guarantee node i according to the network environment changes.

[0028] Then, according to the conversion relationship between the causal model and the Markov decision process, represent the structural causal model as a multi-agent interference Markov decision process; find the optimal deployment strategy π by maximizing the long-term cumulative utility function of the network * ; obtain the optimal deployment strategy of the guarantee nodes to achieve efficient coverage of the target area by the guarantee nodes.

[0029] Preferably, in step S1, the specific process of the network domain division method is as follows:

[0030] S131. Scenario awareness:

[0031] Utilize the existing network, environment perception technology and prior knowledge to establish a scenario awareness model based on multi-modal meta-transfer learning; perceive the network scenario within the entire network domain through node-to-node interaction.

[0032] S132. Network domain division:

[0033] Based on the network scenario awareness result, establish a mathematical optimization model for network domain division; automatically solve the optimization problem based on distributed causal reinforcement learning to obtain the optimal routing strategy and transmission strategy to implement an optimized network management model.

[0034] Preferably, the specific process of constructing the reactive intelligent routing decision method in step S2 is as follows:

[0035] S211. Based on the node organization model and the network domain division model, construct a graph network model as follows:

[0036] G=(V, E, x V )

[0037] Among them, V represents the set of nodes; represents the set of links between nodes; x V represents the characteristic information of the nodes;

[0038] S212. Use the conditional random field model to establish a joint distribution for predicting the connectivity between nodes as follows:

[0039] p φ (y V |x V , E)

[0040] Among them, φ represents the model parameters; directly maximize the lower bound of log p φ (y L |x V , E) and optimize it as follows:

[0041]

[0042] Among them, y L represents the connectivity of some nodes q θ (y U |x V ) represents the connectivity distribution of the node set U;

[0043] S213. Use the variational expectation maximization framework EM to optimize the lower bound in step S212 to obtain the model q θ :

[0044] In the M-phase, fix q θ , and update p φ by maximizing the likelihood function as follows:

[0045]

[0046] In the E-phase, update q θ by optimizing the objective function as follows:

[0047]

[0048] Among them, NB(i) is the neighbor of node i;

[0049] S214. Alternately iterate and train p φ and q θ until convergence to obtain the node connectivity prediction vector y V ;

[0050] S215. Based on the connectivity vector y V, node priority, and target tasks to establish a dynamically updated reactive routing strategy, laying a routing foundation for real-time communication between nodes and between high-level nodes.

[0051] Preferably, the specific process of constructing the reactive intelligent routing decision method in step S2 is as follows:

[0052] S221. Based on the routing discovery process, establish a target-conditioned Markov decision process as follows:

[0053]

[0054] Among them, represents the network state space; represents the action space for selecting the next-hop routing node; represents the transition probability of the network state; G represents the target routing space; represents the network revenue function; γ represents the discount factor;

[0055] S222. Establish a set of Markov decision processes as follows:

[0056]

[0057] Among them, p k represents the causal relationship between the network state and the routing selection action;

[0058] S223. Randomly select one environment from K network environments, and use the interaction strategy π I to generate a network state transition sequence τ = {(s1, a1), (s2, a2),...}, and use the inductive model F to construct a causal model

[0059] S224. Obtain the causal model and the routing strategy π G through learning, and generate a real-time strategy between communication nodes.

[0060] Preferably, the specific process of constructing the opportunistic routing decision method in step S2 is as follows:

[0061] S231. Based on the node dynamic adjustment process, establish a Markov decision imitation learning model as follows:

[0062]

[0063] Among them, is the network state space, is the action space for routing selection; Denote the transition probability of the network state; r represents the set of immediate network revenue functions; T represents the number of transitions of the network state;

[0064] S232. To obtain the opportunistic routing decision between nodes, construct a demonstration dataset as follows:

[0065]

[0066] where represents the trajectory of the potentially available routes of the node; c i represents the routing demand;

[0067] S233. Use imitation learning to construct a node opportunistic routing decision model, that is, find a potential routing policy π(a|s) from a set Π of available routing policies of a class of networks, and simulate the optimal routing policy of the demonstration dataset

[0068] Define the loss function:

[0069] where is the indicator function;

[0070] Expected loss function:

[0071] Expected loss function for T steps:

[0072] where ρ π represents the distribution of the network state;

[0073] S234. Obtain the opportunistic routing decision by minimizing the loss function as follows:

[0074]

[0075] Use the adaptive gradient descent algorithm to solve this optimization problem and obtain the delay-tolerant opportunistic routing policy between nodes.

[0076] Preferably, in the process of network node security access authentication in step S3, it is as follows:

[0077] S311. Construct the public identity RID and the initial stealth identity IID of the access node, then each node to be accessed has a triple <RID, IID, Skey>; where Skey is the security private key for data encryption and decryption;

[0078] ​S312. When node A intends to access the network, it sends an access request carrying the stealth identifier of A to the authentication server; the authentication server verifies A based on the known information of legitimate access devices; if the verification passes, a new stealth identifier of A is generated using the dual-identifier resolution mapping, and an access success confirmation carrying the new stealth identifier of A is sent to A.

[0079] Construct the dual-identifier resolution mapping as follows:

[0080] [z a (X) RID →ψ[z a (X) IID

[0081] Among them, X represents the access node, a represents the access request, and z a (X) RID represents the access identifier, and z a (X) IID represents the connection identifier;

[0082] S313. After node A receives the access success confirmation, it updates the stealth identifier.

[0083] Preferably, in the process of network transmission connection security authentication in step S3, it is as follows:

[0084] S321. When node A needs to send information to node B, it first sends a transmission application carrying the identity identifier RID A |IID A of node A and the public identifier RID B of node B to the authentication server;

[0085] S322. The authentication server verifies the identity information of node A and B. If the identity verification passes, a set of encryption and decryption keys is generated using the dual identifiers of both parties as follows:

[0086] (K A , K B )←Extract(pp, Skey, RID A |IID A , RID B |IID B )

[0087] Among them, pp represents the public parameters, which are sent to node A and node B in ciphertext form;

[0088] ​S323. Node A encrypts and encapsulates the data using the key obtained from the authentication server and its own stealth identifier, and sends it to Node B. When Node B receives the data packet, it verifies and decrypts the data packet using the key obtained from the authentication server. If the verification passes, it sends a reception success confirmation to Node A.

[0089] S324. After receiving the confirmation information from Node B, Node A sends a transmission end confirmation to the authentication server. The authentication server regenerates the new stealth identifiers of Node A and Node B and sends them to A and B respectively. Node A and Node B update their new stealth identifiers for the next information transmission.

[0090] Therefore, the present invention adopts the above-mentioned intelligent networking method for a high-dynamic large-scale mobile ad hoc network, realizes the organization and management of a large number of network nodes, solves the routing decision problem in various situations, ensures the security during the access and transmission of network nodes, and is applicable to new network application scenarios with high dynamics and large scale.

[0091] The technical solution of the present invention will be further described in detail below through the accompanying drawings and embodiments. Description of the Drawings

[0092] Figure 1 is the intelligent identifier self-organizing network architecture diagram of the intelligent networking method for a high-dynamic large-scale mobile ad hoc network of the present invention;

[0093] Figure 2 is the schematic diagram of node organization of distributed virtual mapping of the intelligent networking method for a high-dynamic large-scale mobile ad hoc network of the present invention;

[0094] Figure 3 is the "send - authenticate - receive" three-party collaborative trusted transmission control process of the intelligent networking method for a high-dynamic large-scale mobile ad hoc network of the present invention. Detailed Embodiments

[0095] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the embodiments of 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 embodiments of the present invention, and are not used to limit the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in this application without creative efforts shall fall within the scope of protection of this application.

[0096] As Figure 1 shown, an intelligent networking method for a high-dynamic large-scale mobile ad hoc network mainly includes three parts: an intelligent networking model, a routing decision method, and a security authentication method.

[0097] S1. Combine the actual application situation and network characteristics of high-dynamic large-scale mobile ad hoc networks, establish an intelligent networking architecture model, establish an organizational model for guarantee nodes and functional nodes from a logical perspective, and establish an intelligent network domain division model from a physical perspective.

[0098] S2. For the communication requirements of guarantee nodes, random communication requirements, and communication requirements without available links, establish a proactive intelligent routing decision-making method based on graph Markov neural networks, a reactive routing decision-making method based on causal reasoning, and a delay-tolerant opportunistic routing decision-making method based on imitation learning respectively.

[0099] S3. For the security issues of node access and transmission connections, design a node security authentication method based on "public - stealth" dual identities and a three-party collaborative transmission security authentication method of "send - authenticate - receive".

[0100] Embodiment

[0101] S1. Establish an intelligent networking architecture model;

[0102] The intelligent networking model mainly includes three parts: ad hoc network architecture, node organization, and network domain division.

[0103] S11. Ad hoc network architecture model;

[0104] In the application requirement environment of high-dynamic large-scale networking, the formed ad hoc network will cover multiple different network domains.

[0105] According to functions, network nodes are divided into two categories: one is guarantee nodes, which are network nodes specifically used for routing relay, such as communication satellites, operator fixed base stations, UAV / unmanned vehicle mobile base stations, etc.; the other is functional nodes, which are network nodes that complete established target tasks and have a certain relay function at the same time, such as reconnaissance UAVs, unmanned mine sweepers, etc.

[0106] According to the functions of realizing networking, it is divided into four modules: node organization engine, network domain division engine, routing decision engine, and security authentication engine.

[0107] All engine modules are deployed on network guarantee nodes, and the functions of each engine are completed through distributed cooperation; on network functional nodes, mainly the routing decision engine and security authentication engine are deployed.

[0108] The node organization engine reasonably organizes all nodes accessing the network, which is divided into guarantee node organization and functional node organization. The two are mutually mapped and associated to achieve efficient management of network nodes without topology.

[0109] The network domain division engine combines node organization and scenario awareness to appropriately divide the network domain and optimize the deployment of guarantee nodes.

[0110] The routing decision engine maintains intelligent routing in various situations according to network status, node level, task level, etc.

[0111] Based on existing encryption and decryption technologies, the security authentication engine uses "public - stealth" dual identification to achieve secure management of node access and transmission connections.

[0112] S111. Use mathematical optimization to model the high - dynamic large - scale ad - hoc network as a generalized network utility maximization problem.

[0113] First, define the set of guarantee nodes and the set of functional nodes as NID c and NID nc . The set of available paths from the source node q to the destination node through guarantee nodes or functional nodes is the adjacency matrix H q , then the available path matrix can be expressed as H = [H1,..., H S ; where S represents the number of source nodes.

[0114] Then, let w q，j represent the traffic of the source node q on the j - th path, and satisfy w q，j ≥0 and where, m represents the number of available paths from the source node q to the destination node; then the routing matrix of the high - dynamic large - scale ad - hoc network is:

[0115] R = HW

[0116] where W is the diagonal matrix composed of the vectors w q , q = 1,..., S.

[0117] Finally, let represent the transmission rate of the source node q in the network domain o, c l represent the capacity of the l - th path (such as link bandwidth, node processing capacity, node energy, etc.), and the generalized network utility maximization problem is as follows:

[0118]

[0119] subject to Rx ≤ c

[0120] where, represents the guarantee node vector, which is composed of the location, resources, authentication capabilities, etc. of the guarantee nodes; represents the functional node vector, which is composed of the location, tasks, and association relationships with other nodes of the functional nodes.

[0121] S112. The characteristics of high - dynamics and large - scale of the network lead to the network utility function There are various forms, and the routing matrix and path capacity change rapidly. Therefore, to solve this optimization problem using mathematical optimization, causal reasoning, and reinforcement learning, the specific process is as follows:

[0122] Introduce a causal model to learn the adjacency matrix H q , and then obtain the routing matrix R, and model the causal model as a Markov decision process:

[0123]

[0124] Among them, represents the state space, which consists of network scenarios, path capacity c, optimization problem and its dual form, etc.; represents the action space, which consists of optimization algorithms; represents the state transition probability; represents the reward function.

[0125] S113. Find the optimal optimization algorithm through reinforcement learning, and solve for the best routing strategy R * and transmission strategy x * .

[0126] Therefore, through the solution of the data model of the high-dynamic large-scale ad hoc network architecture, efficient management of the entire network can be achieved.

[0127] S12. Node organization;

[0128] S121. Node virtual organization;

[0129] Under the application requirements of high-dynamic large-scale networking, traditional node management methods that rely on network topologies are costly or ineffective. Distributed virtual mapping is an effective way to achieve topology-free node management.

[0130] First, virtualize the guarantee nodes based on logical connectivity to form a virtual mapping of the guarantee nodes in the high-dynamic large-scale ad hoc network;

[0131] Then, establish virtual mappings of multiple functional nodes according to different types of functional nodes, such as clustering, peer-to-peer, grouping, etc.;

[0132] Finally, connect the virtual mapping of the functional nodes and the virtual mapping of the guarantee nodes according to physical or logical relationships to form a distributed virtual mapping organization of network nodes, as Figure 2 shown.

[0133] As Figure 2 shown, based on the routing strategy R * and transmission strategy x *, the high-dynamic large-scale ad hoc network mathematical model is decomposed into 2 interrelated sub-optimization problems as follows:

[0134] First, using the Lagrange multiplier method, the generalized network utility maximization problem is transformed into the following optimization problem:

[0135]

[0136] where r q represents the q-th column vector of matrix R, and R lq represents the element in the l-th row and q-th column of matrix R, represents the set of all column vectors of H q , and λ represents the Lagrange multiplier vector.

[0137] Through the known routing strategy R * and transmission strategy x * , the optimal λ * can be obtained; using the non-linear function approximation algorithm, the function is decomposed as follows:

[0138]

[0139] Then, the sub-optimization problems are solved separately:

[0140]

[0141]

[0142] The guaranteed node vector and the functional node vector

[0143] are obtained. and Based on the obtained and the optimized generalized network utility maximization problem, the routing strategy and the transmission strategy are obtained, and then the equilibrium solution

[0144] Finally, the difference degree between the solution of the overall optimization problem and the sub-optimization solution is evaluated, and the definition of the difference degree is as follows:

[0145]

[0146] By iteratively optimizing the difference degree, the best decomposition of the overall problem is obtained, thereby improving the overall performance of the network.

[0147] S122. Optimized deployment of guaranteed nodes;

[0148] Combined with the ad hoc network architecture, the security node organization, and the functional node organization, a causal reinforcement learning is introduced to establish an optimized deployment method for security nodes, so as to achieve efficient coverage of the target network area, as follows:

[0149] First, the process of continuously and dynamically adjusting the security node i according to the network environment changes is modeled as a distributed structural causal model, as follows:

[0150]

[0151] Among them, U i represents exogenous variables, such as weather changes, geographical environment, human interference, etc.; V i represents endogenous variables, such as network status, node deployment strategy, network utility function, etc.; F i represents structural functions, such as state transition, node deployment strategy, network utility function, etc.; P i represents the distribution of all exogenous variables, such as Gaussian distribution, etc.; n is the number of security nodes.

[0152] Secondly, according to the conversion relationship between the causal model and the Markov decision process, the structural causal model is characterized as a multi-agent interference Markov decision process.

[0153] Then, to achieve efficient coverage of the overall network by security nodes, an optimal deployment strategy π is found by maximizing the long-term cumulative utility function of the network * .

[0154] Finally, the best deployment strategy of security nodes is obtained to achieve efficient coverage of the target area by security nodes.

[0155] S13, Network domain division method;

[0156] S131, Scenario awareness;

[0157] Using existing networks, environment perception technologies, prior knowledge, etc., a scenario awareness method based on multi-modal meta-transfer learning is established, as follows:

[0158] First, the perception information such as location, resources, status, environment, meteorology, etc. obtained by node i is represented as a set Among them, the vector s i，j represents the j-th type of perception information of node i, and m represents the number of types of perception information.

[0159] Secondly, a network scenario dataset is obtained from the network scenario perception information set S using the multi-modal feature learning method i The feature extractor Θ of the network scenario is initialized and trained using a deep neural network model and the classifier θ i and the classifier θ i, obtain Θ according to the stochastic gradient descent learning algorithm i , as follows:

[0160]

[0161] where α represents the learning rate, represents the empirical loss function.

[0162] Then, for a specific network scenario awareness task train the network scenario classifier θ′ i , as follows:

[0163]

[0164] where β represents the learning rate, represents the data training set, and represent different meta-operations respectively, such as scaling and translation of the network scenario. Each meta-operation has the following learning rule:

[0165]

[0166] where η represents the learning rate, represents the test data set. At the same time, the basic classifier θ of the network scenario i , is updated as follows:

[0167]

[0168] In this way, through training, the basic perception model θ of the network scenario can be obtained i , which contains information such as the spatial location of the network, network resources, environmental meteorology, etc. Using the basic perception model θ i , through the classifier θ′ i perceive a specific network scenario awareness and update the model θ i , and the network scenario where node i is located can be quickly identified through iterative loops.

[0169] Finally, perceive the network scenarios within the entire network domain through interactions between nodes.

[0170] S132, Network domain division;

[0171] Under the high-dynamic large-scale network application requirements, based on the network scenario recognition results of step S131, establish a network domain division mathematical optimization model, which can achieve efficient network management in combination with the actual scenario. The specific process is as follows:

[0172] First, use the Lagrange multiplier method to transform the generalized network utility maximization problem into the following optimization problem:

[0173]

[0174] Next, decompose the above optimization problem into n sub-problems. The O-th sub-optimization problem is as follows:

[0175]

[0176] According to the optimal safeguard node vector and the functional node vector The O-th sub-optimization problem is as follows:

[0177]

[0178] To solve each sub-optimization problem, establish an idea of automatically solving sub-optimization problems based on distributed causal reinforcement learning:

[0179] First, for each network domain o ∈ {1,..., n}, construct a multi-network domain causal model as follows:

[0180]

[0181] Among them, is composed of the optimal safeguard node vector, the optimal functional node vector, network domain O, network scenarios, etc.; represents the association relationships between safeguard nodes, the association relationships between functional nodes, and the association relationships between safeguard nodes and functional nodes; represents the probability distribution of; represents the shared information between network domain o and other network domains.

[0182] Then, through distributed causal discovery learning, obtain Characterize as a generalized Markov decision process as follows:

[0183]

[0184] Among them, the state space includes the spatial deployment of safeguard nodes and functional nodes, network scenarios, network domain information, etc.; represents the action space of network domain o, which is composed of a set of distributed optimization algorithms; γ ∈ [0, 1] represents the discount rate.

[0185] Therefore, the solution task for each sub-problem o can be defined as:

[0186]

[0187] Among them, represents the Markov decision process, r oDenote the revenue function.

[0188] Next, use distributed causal reinforcement learning to select the best optimization algorithm to solve each sub-optimization problem, and obtain the optimal routing strategy and transmission strategy

[0189] Finally, based on and optimize the network management model.

[0190] S2. Construct an intelligent routing decision method for high-dynamic large-scale mobile ad hoc networks.

[0191] The routing decision method mainly includes three cases: proactive intelligent routing decision, reactive intelligent routing decision, and opportunistic routing decision.

[0192] S21. Proactive intelligent routing decision method;

[0193] Based on the node organization model, domain division model, node priority, etc., design a method for predicting node connectivity based on graph Markov neural network, and dynamically maintain the proactive routing strategy between nodes and between high-priority nodes with potential transmission tasks, as follows:

[0194] S211. According to the node organization model, domain division model, etc., construct a graph network model, such as:

[0195] G=(V, E, x V )

[0196] where V represents the set of nodes; represents the set of links between nodes; X V represents the characteristic information of the nodes, such as historical trajectory, established tasks, priority, etc.

[0197] If (i, j)∈E, it means that there is one or more links between node i∈V and i∈V, then it is said that node i∈V and i∈V are connected; let the matrix C=[c ij |V|×|V| represent the connectivity matrix, where |V| represents the number of nodes.

[0198] When (i, j)∈E, then c ij =1, otherwise c ij =0; vectorize the connectivity matrix C as

[0199] y V : y V =vec(C)

[0200] S212. Use the conditional random field model to establish a joint distribution to achieve node connectivity prediction, as follows:

[0201] p φ (y V |x V ,E)

[0202] Among them, φ represents the model parameters.

[0203] Since it is difficult to directly maximize log p φ (y L |x V ,E), where y L represents the connectivity of some nodes , so its lower bound is optimized:

[0204]

[0205] Among them, q θ (y U |x V ) represents the connectivity distribution of the node set U.

[0206] S213. Use the variational expectation maximization framework (Expectation Maximization, EM) to optimize the above lower bound to obtain the model q θ .

[0207] In the M-phase, fix q θ , and update p φ by maximizing the following likelihood function:

[0208]

[0209] In the E-phase, update q θ by optimizing the following objective function:

[0210]

[0211] Among them, NB(i) is the neighbor of node i.

[0212] S214. Alternately iterate and train p φ and q θ until convergence to obtain the connectivity prediction vector y V .

[0213] S215. Based on the connectivity vector y V , node priority, target task, etc., establish a dynamically updated reactive routing policy to lay a routing foundation for ensuring real-time communication between nodes and between high-level nodes.

[0214] S22. Reactive intelligent routing decision method;

[0215] Based on the node organization model, domain division model, established tasks of nodes, reactive routing strategies, etc., design a reactive intelligent routing decision-making method based on causal reasoning to maintain the routing strategy between randomly communicating task nodes in real time, as follows:

[0216] S221. Model a target-conditioned Markov decision process based on the routing discovery process, as follows:

[0217]

[0218] The goal is to find a routing strategy π by maximizing the network revenue G .

[0219] Among them, represents the network state space, such as network scenarios, guarantee node deployment, network resources, etc.; represents the action space for selecting the next-hop routing node; represents the transition probability of the network state; G represents the target routing space; represents the network revenue function, that is, the one-step immediate revenue based on the target g ∈ G is r(s, a, g); γ represents the discount factor.

[0220] S222. To improve the generalization ability of the routing strategy, establish a set of Markov decision processes, as follows:

[0221]

[0222] Among them, p k represents the causal relationship between the network state and the routing selection action.

[0223] S223. Randomly select one environment from K network environments, and use the interaction strategy π I to generate a network state transition sequence τ = {(s1, a1), (s2, a2),...}, and use the induction model F to construct a causal model

[0224] S224. Use the causal model to place the routing strategy π G in the context to ensure that the routing selection task can be executed in the new network environment.

[0225] To learn the induction model F, divide the set of all Markov decision processes into two non-overlapping sets and

[0226] In the training stage, use the supervised learning method in ​Learn F, and use the DAgger algorithm to train the learning routing policy π G 。

[0227] In the test phase, evaluate whether a causal model can be constructed through learning F in the new network environment Whether a causal model can be utilized to find the routing policy π G 。

[0228] S225. Obtain the causal model through learning and the routing policy π G 。For specific random communication requirements, based on the routing policy π G , generate real-time policies between communication nodes.

[0229] S23. Opportunistic routing decision method

[0230] When there is no available link, establish an opportunistic routing decision method based on imitation learning, and generate a routing policy under delay tolerance conditions based on the node organization model, the established node tasks, the domain division model, the communication task QoS, the established routing policy, etc., as follows

[0231] S231. According to the node organization model, the established node tasks, the domain division model, etc., model the dynamic adjustment process of nodes as a Markov decision imitation learning model, as follows

[0232]

[0233] where is the network state space, such as network scenario images, node location information, network resources, etc.; is the action space of routing selection; represents the transition probability of the network state; r represents the set of immediate network revenue functions; T represents the number of network state transitions.

[0234] S232. According to the node organization and domain division, simulate the actual network environment, obtain the trajectory set of the potential available routing policies of nodes, and construct a demonstration data set to obtain the opportunistic routing decision between nodes as follows

[0235]

[0236] where represents the trajectory of the potential available routing of nodes; c i represents the routing requirement.

[0237] S233. Build a node opportunistic routing decision model using imitation learning, that is, find a potential routing policy π(a|s) from a set Π of available routing policies for a class of networks, and simulate the optimal routing policy of the demonstration dataset of the optimal routing policy

[0238] Define the loss function:

[0239] where is the indicator function.

[0240] Expected loss function:

[0241] Expected loss function for T steps:

[0242] where ρ π represents the distribution of network states.

[0243] S234. By minimizing the loss function the potential opportunistic routing decision can be obtained as follows:

[0244]

[0245] Use the adaptive gradient descent algorithm to solve this optimization problem and obtain the delay-tolerant opportunistic routing policy between nodes.

[0246] S3. Design a security authentication policy for node organization and routing transmission.

[0247] The security authentication method mainly includes two parts: node secure access authentication and transmission connection security authentication.

[0248] S31. The process of network node secure access authentication is as follows:

[0249] S311. Construct the public identifier RID (RouterID) and the initial stealth identifier RID (RouterID) of the access node; select the organic combination of the access node's ID, product serial number, and other physical information as the public identifier RID; extract the organic combination of the access node's parameter information, buffer information, backup file information, etc. as the stealth identifier IID; then each access node to be connected has a triple <RID, IID, Skey>; where Skey is the security private key for data encryption and decryption.

[0250] S312. When node A intends to access the network, it sends an access request carrying the invisible identifier of A to the authentication server; the authentication server verifies A based on the known information of legitimate access devices; if the verification passes, a new invisible identifier of A is generated using the dual-identifier resolution mapping, and an access success confirmation carrying the new invisible identifier of A is sent to A.

[0251] To achieve the rapid conversion of "public - invisible" identifiers, a dual-identifier resolution mapping is constructed as follows:

[0252] [z a (X) RID →ψ[z a (X) IID

[0253] Among them, X represents the access node, a represents the access request, and z a (X) RID represents the access identifier, and z a (X) IID represents the connection identifier.

[0254] S313. After node A receives the access success confirmation, it updates the invisible identifier.

[0255] S32. The network transmission connection security authentication process is as Figure 3 shown:

[0256] S321. When node A needs to send information to node B, it first sends a transmission request carrying the identity identifier RID A |IID A (RouterIDA, InvisibleIDA) of node A and the public identifier RID B (RouterIDB) of node B to the authentication server;

[0257] S322. The authentication server verifies the identity information of nodes A and B. If the identity verification passes, a set of encryption and decryption keys is generated using the dual identifiers of both parties:

[0258] (K A ,K B )←Extract(pp, Skey, RID A |IID A ,RID B |IID B )

[0259] Among them, pp represents the public parameters and is sent to nodes A and B in ciphertext form.

[0260] ​S323. Node A encrypts and encapsulates the data using the key obtained from the authentication server and its own stealth identifier, and sends it to Node B. When Node B receives the data packet, it verifies and decrypts the data packet using the key obtained from the authentication server. If the verification passes, it sends a reception success confirmation to Node A.

[0261] S324. After receiving the confirmation information from Node B, Node A sends a transmission end confirmation to the authentication server. The authentication server regenerates the new stealth identifiers of Node A and Node B and sends them to A and B respectively. Node A and Node B update their new stealth identifiers for the next information transmission.

[0262] Therefore, the present invention adopts the above intelligent networking method for a high-dynamic large-scale mobile ad hoc network, realizes the organization and management of a large number of network nodes, solves the routing decision problem in various situations, ensures the security of network node access and transmission processes, and is applicable to new network application scenarios with high dynamics and large scale.

[0263] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that they can still modify or equivalently replace the technical solutions of the present invention, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. An intelligent networking method for high-dynamic large-scale mobile ad hoc networks, characterized in that, It includes the following steps: S1. Establish a networking model for a high-dynamic large-scale mobile ad hoc network, including three parts: an ad hoc network architecture, node organization, and network domain division. Among them, establish a guarantee node and functional node organization model from a logical perspective, and establish an intelligent network domain division model from a physical perspective; S11. Ad hoc network architecture model: The network nodes are divided into guarantee nodes and functional nodes according to their functions; they are divided into four functional modules according to the functions of networking: a node organization engine, a network domain division engine, a routing decision engine, and a security authentication engine; The guarantee node is a network node specifically used for routing relay, deploying all engine modules, and completing the functions of each engine through distributed cooperation; the functional node is a network node used to complete a given target task, and at the same time has a relay function, deploying a routing decision engine module and a security authentication engine module; The node organization engine reasonably organizes all the nodes accessing the network, divided into guarantee node organization and functional node organization, which are mutually mapped and associated to achieve efficient management of network nodes without topology; the network domain division engine combines node organization and scenario awareness to divide the network into network domains and optimize the deployment of guarantee nodes; the routing decision engine maintains intelligent routes in various situations according to network status, node level, and task level; the security authentication engine is based on encryption and decryption technologies, using "public - invisible" dual identifiers to achieve secure management of node access and transmission connections; Among them, the specific steps to establish an ad hoc network architecture model are as follows: S111. Use mathematical optimization to construct the generalized network utility maximization problem of the ad hoc network architecture as follows: subject to Rx≤c; Among them, represents the guarantee node vector; represents the functional node vector; represents the transmission rate of the source node q in the network domain o; identifies the network utility function; S112. Use mathematical optimization, causal reasoning, and reinforcement learning to model the Markov decision process and solve the generalized network utility maximization problem as follows: Among them, represents the state space; represents the action space; represents the state transition probability; represents the reward function; S113. Obtain the optimal optimization algorithm through reinforcement learning, and solve to obtain the optimal routing policy R * and transmission policy x * ; S12. The specific steps of node organization are as follows: S121. Node virtual organization; First, virtualize and organize the guarantee nodes according to logical connectivity to form a virtual mapping of guarantee nodes; Then, establish virtual mappings of multiple functional nodes according to different types of functional nodes; Finally, based on physical or logical relationships, connect the virtual mappings of functional nodes with the virtual mappings of guarantee nodes to form a distributed virtual mapping organization of network nodes; S122. Optimized deployment of guarantee nodes; First, establish a distributed structural causal model for the process of continuously and dynamically adjusting guarantee node i according to network environment changes; Then, according to the conversion relationship between the causal model and the Markov decision process, represent the structural causal model as a multi-agent interference Markov decision process; Finally, the optimal deployment strategy π is found by maximizing the long-term cumulative utility function of the network * , and the best deployment strategy of the guarantee nodes is obtained to achieve the efficient coverage of the target area by the guarantee nodes; S13. The specific process of the network domain division method is as follows: S131. Scenario awareness; Use existing network, environment perception technologies, and prior knowledge to establish a scenario awareness model based on multi-modal meta-transfer learning; perceive the network scenarios within the entire network domain through node - to - node interaction; S132. Network domain division; Based on the network scenario recognition results, establish a mathematical optimization model for network domain division; based on distributed causal reinforcement learning, automatically solve the optimization problem to obtain the optimal routing strategy and transmission strategy Realize the optimization of the network management model; S2. According to communication requirements, construct an intelligent routing decision method for a high-dynamic large-scale mobile ad hoc network, specifically divided into three cases: proactive intelligent routing decision, reactive intelligent routing decision, and opportunistic routing decision; S3. Design the security authentication policy for node organization and routing transmission, including two parts: node secure access authentication and transmission connection security authentication.

2. An intelligent networking method for a high-dynamic large-scale mobile ad-hoc network according to claim 1, characterized in that, The specific process of constructing the proactive intelligent routing decision method in step S2 is as follows: S211. Based on the node organization model and the network domain division model, construct a graph network model as follows: G = (V, E, x v ); Among them, V represents the set of nodes; represents the set of links between nodes; x V represents the characteristic information of the nodes; S212. Use the conditional random field model to establish a joint distribution for predicting the connectivity between nodes as follows: p φ (y V |x V ,E); where φ represents the model parameters; optimize the lower bound of directly maximizing logp φ (y L |x V , E) as follows: Among them, y L represents the connectivity of some nodes , and q θ (y U |x V ) represents the connectivity distribution of the node set U; S213. Use the variational expectation-maximization framework EM to optimize the lower bound in step S212 to obtain the model q θ : At the M-phase, fix q θ , update p by maximizing the likelihood function φ , as follows: In the E-phase, update q by optimizing the objective function θ , as follows: Among them, NB(i) is the neighbor of node i; S214. Alternately and iteratively train p φ and q θ until convergence to obtain the node - to - node connectivity prediction vector y V ; S215. Based on the connectivity vector y V and the node priority and the target task, establish a dynamically updated reactive routing policy to lay a routing foundation for real-time communication between nodes and between high-level nodes.

3. An intelligent networking method for a high-dynamic large-scale mobile ad hoc network according to claim 1, characterized in that, The specific process of constructing the reactive intelligent routing decision method in step S2 is as follows: S221. Based on the routing discovery process, establish a target conditional Markov decision process as follows: Among them, represents the network state space; represents the action space for selecting the next-hop routing node; p: represents the transition probability of the network state; G represents the target routing space; r: represents the network revenue function; γ represents the discount factor; S222. Establish a set of Markov decision processes as follows: Among them, p k represents the causal relationship between the network state and the routing selection action; S223. Randomly select one environment from K network environments and use the interaction strategy π I to generate a network state transition sequence τ = {(s1, a1), (s2, a2),...} and use the inductive model F to construct a causal model S224. Obtain a causal model through learning and a routing policy π G to generate a real-time policy between communication nodes.

4. An intelligent networking method for a high-dynamic large-scale mobile ad hoc network according to claim 1, characterized in that, The specific process of constructing the opportunistic routing decision method in step S2 is as follows: S231. Based on the node dynamic adjustment process, establish a Markov decision imitation learning model as follows: Among them, is the network state space, is the action space for routing selection; represents the transition probability of the network state; r represents the set of immediate network revenue functions; T represents the number of network state transitions; S232. To obtain the opportunistic routing decision between nodes, a demonstration data set is constructed as follows: Among them, represents the trajectory of the potentially available routes of the node; c i represents the routing requirement; S233. Build a node opportunistic routing decision model using imitation learning, that is, find a potential routing policy π(a|s) from a set Π of available routing policies for a class of networks, and simulate the optimal routing policy of the demonstration dataset of the optimal routing policy Define the loss function: wherein, is an indicator function; Expected loss function: Expected loss function for step T: where ρ π represents the distribution of network states; S234. Obtain the opportunistic routing decision by minimizing the loss function as follows: Use the adaptive gradient descent algorithm to solve this optimization problem to obtain the delay-tolerant opportunistic routing policy between nodes.

5. An intelligent networking method for a high-dynamic large-scale mobile ad hoc network according to claim 1, characterized in that The process of network node secure access authentication in step S3 is as follows: S311. Construct the public identifier RID and the initial stealth identifier IID of the access node, so that each node to be accessed has a triple <RID, IID, Skey>; where Skey is the security private key for data encryption and decryption. S312. When node A intends to access the network, send an access request carrying the stealth identifier of A to the authentication server; the authentication server verifies A according to the known information of legitimate access devices; if the verification passes, use the dual-identifier resolution mapping to generate a new stealth identifier of A, and send an access success confirmation carrying the new stealth identifier of A to A. Construct the dual-identifier resolution mapping as follows: [z a (X) RID →ψ[z a (X) IID ; Among them, X represents an access node, a represents an access request, and z a (X) RID represents an access identifier, and z a (X) IID represents a connection identifier; S313. When node A receives the access success confirmation, update the stealth identifier.

6. An intelligent networking method for a high-dynamic large-scale mobile ad hoc network according to claim 1, characterized in that, The process of network transmission connection security authentication in step S3 is as follows: S321. When node A needs to send information to node B, it first sends a transmission request with the identity identifier RID of node A A |IID A and the public identifier RID of node B B to the authentication server; S322. The authentication server verifies the identity information of nodes A and B. If the identity verification passes, generate a set of encryption and decryption keys using the dual identifiers of both parties as follows: (K A ,K B ) ← Extract(pp, Skey, RIDA|IIDA, RID B |IID B ); Among them, pp represents the public parameter, which is sent to node A and node B in ciphertext form. S323. Node A encrypts and encapsulates the data using the key obtained from the authentication server and its own stealth identifier, and sends it to node B; when node B receives the data packet, it verifies and decrypts the data packet using the key obtained from the authentication server; if the verification passes, send a reception success confirmation to node A. S324. After node A receives the confirmation information from node B, send a transmission end confirmation to the authentication server; the authentication server regenerates the new stealth identifiers of nodes A and B and sends them to A and B respectively. Nodes A and B update their new stealth identifiers for the next information transmission.

Images