A Federated Learning Incentive Method and System for Vehicle Networks Based on Nash Game Theory and Coalition Game Theory

Abstract

This invention discloses a federated learning incentive method and system for vehicle-to-everything (V2X) networks based on Nash game theory and coalition game theory. The incentive method includes: initializing a set of vehicle clients in the V2X environment; broadcasting global model parameters to all clients via a server; establishing a two-stage Nash game model, wherein the first stage determines the coalition set within the vehicle client set, and the second stage negotiates and optimizes resource contributions; based on the two-stage Nash game model, the server and the coalition set negotiate and determine the optimal data contribution ratio and the optimal resource contribution ratio; and the server allocates rewards to the coalition set according to the optimal data contribution ratio and the optimal resource contribution ratio. This invention can adapt to the high dynamism of V2X networks and achieve fair and efficient resource allocation.

Description

Technical Field

[0001] This invention belongs to the fields of vehicle networking and distributed machine learning technology, and particularly relates to a vehicle networking federated learning incentive method and system based on Nash game and federated game. Background Technology

[0002] With the rapid development of intelligent transportation systems, modern vehicles are equipped with advanced sensors that generate massive amounts of data. This data contains valuable information that can be used to improve traffic management, enhance autonomous driving capabilities, and analyze road safety. Federated Learning (FL) offers a highly attractive solution for connected vehicle environments by supporting distributed model training while maintaining data locality and privacy.

[0003] However, deploying federated learning in vehicular networks faces numerous challenges: First, the high mobility of vehicles leads to frequent changes in network topology and unpredictable participation patterns, severely interfering with the training process. While existing research recognizes the potential impact of vehicle mobility on federated learning (FL) performance, it fails to fully consider this impact and cannot maintain effective incentives when participants frequently join and leave the network. Second, data heterogeneity weakens the effectiveness of incentive mechanisms. Data in vehicular networks exhibits non-independent identically distributed (non-IID) characteristics, resulting in unsatisfactory convergence speeds. Existing methods, such as introducing proximal regularization terms and training GAN networks, can alleviate the non-IID problem, but they suffer from instability and overhead constraints in the vehicular environment and fail to establish a fair compensation mechanism that considers both the quality and quantity of data contributions. Finally, the strict constraints of vehicle performance and energy, along with significant heterogeneity, make efficient incentive optimization imperative. Some studies have explored client selection strategies based on static resource allocation, but these methods fail to consider the cumulative effects of high mobility and time-varying data quality. Traditional federated learning incentive mechanisms, such as FedAvg and FedProx, primarily focus on static environments and are ill-suited to the dynamic characteristics of vehicular networks. Stackelberg game-based methods require multiple iterations to reach equilibrium, making them inefficient in high-mobility scenarios. While auction mechanisms consider resource heterogeneity, they are prone to participant churn under frequently changing network conditions. Summary of the Invention

[0004] To address the aforementioned technical issues, this invention proposes a federated learning incentive method and system for vehicle-to-everything (V2X) networks based on Nash game theory and coalition game theory, in order to adapt to the high dynamism of V2X networks and achieve fair and efficient resource allocation.

[0005] To achieve the above objectives, this invention provides a federated learning incentive method for vehicle networks based on Nash game theory and coalition game theory, comprising:

[0006] Initialize the set of vehicle clients in the vehicle networking environment, and broadcast global model parameters to all clients;

[0007] A two-stage Nash game model is established, wherein the first stage is used to determine the alliance set in the set of vehicle clients, and the second stage is used to negotiate and optimize resource contributions.

[0008] Based on the server and the alliance set, the optimal data contribution ratio and the optimal resource contribution ratio are determined through negotiation using the two-stage Nash game model.

[0009] The server distributes rewards to the alliance set based on the optimal data contribution ratio and the optimal resource contribution ratio.

[0010] Optionally, the following steps are included before establishing the two-stage Nash game model:

[0011] A stable alliance structure is constructed based on a multi-dimensional scoring function that considers data quality, communication efficiency, and mobility costs.

[0012] Within each alliance in the stable alliance structure, a cluster head and a backup cluster head are selected based on a three-dimensional score of server communication efficiency, location centrality, and link stability.

[0013] Optionally, the multidimensional scoring function is:

[0014]

[0015] Among them, P m (M(t)) is a multidimensional scoring function, Q data (M(t)) represents the data quality score, Q comm (M(t)) represents the communication efficiency score. For mobility cost scoring, μ1 and μ2 are normalization factors.

[0016] Optionally, constructing the stable alliance structure includes: a merging module, a splitting module, and a switching module;

[0017] The merging module is:

[0018] P(M i (t)∪M j (t))≥P(M i (t))+P(M j (t))+τ merege And |M i (t)|+|M j (t)|≤

[0019] l limit ,

[0020] Among them, M i (t) represents the alliance i, M j (t) represents the alliance j, P(M) i (t) represents the multidimensional scoring function value of alliance i, P(M) j (t) represents the multidimensional scoring function value of alliance j, τ merge l is the threshold for the merge operation. limit Limit the size of the alliance;

[0021] The splitting module is:

[0022] P(M i (t))<τ split or |M i (t)|>l limit ,

[0023] And the resulting subset of the alliance {M i1 M i2 ,…M ik The following conditions must be met:

[0024]

[0025] Where, τ split τ is the splitting threshold. gain_split M is the threshold for splitting payoff. ij (t) represents the j-th alliance in the subset of alliances split from alliance i, P(M ij (t) is the multidimensional scoring function value of the j-th alliance in the alliance subset, j is the index of the j-th alliance in the split alliance subset, and k is the number of sets in the split alliance subset;

[0026] The switching module is:

[0027] P m (M j (t)∪{v m})≥P m (M i (t)\{v m})+τ switch ,

[0028] Where, τ switch To switch the revenue threshold, v m Let the set of vehicles be v = {v1, v2, ..., v...} N The m-th client in}.

[0029] Optionally, cluster heads and backup cluster heads can be selected based on a three-dimensional score of server communication efficiency, location centrality, and link stability.

[0030] Optionally, the two-stage Nash game model is as follows:

[0031]

[0032] Where V(y(t),z(t)) is the server utility function, U i (y j (t),z j Let y(t) be the coalition utility function, and y(t) be the optimal coalition structure Π obtained from the Nash game. * z(t) is the solution vector composed of the optimal data contribution ratio of each alliance in (t), and z(t) is the optimal alliance structure Π obtained from the Nash game. * The solution vector y is composed of the optimal computational contribution proportion of each alliance in (t). j (t) represents the optimal alliance structure Π obtained from solving the Nash game. * The optimal data contribution ratio of the j-th alliance in (t), z j (t) represents the optimal alliance structure Π obtained from solving the Nash game. * The optimal computational resource contribution ratio of the j-th alliance in (t), Π * (t) represents the optimal alliance structure.

[0033] Optionally, the negotiation to determine the optimal data contribution ratio and the optimal resource contribution ratio includes: using a multi-starting point sequential quadratic programming algorithm to negotiate and determine the optimal data contribution ratio and the optimal resource contribution ratio;

[0034] Among them, the method of negotiating and determining the optimal data contribution ratio and the optimal resource contribution ratio by using a multi-starting point sequence quadratic programming algorithm includes: using Latin hypercube sampling to generate a preset number of initial points, executing the quadratic programming algorithm in parallel for each initial point to obtain a local optimal solution, and selecting the solution with the largest objective function value among the local optimal solutions as the Nash bargaining solution;

[0035] Based on the Nash bargaining solution, the optimal data contribution ratio and the optimal resource contribution ratio are obtained.

[0036] Optionally, the server may distribute rewards to the alliance set based on the optimal data contribution ratio and the optimal resource contribution ratio, including:

[0037] Based on the optimal data contribution ratio and resource contribution ratio, the local model parameters are updated using the stochastic gradient descent algorithm.

[0038] The updated local model parameters are sent to the cluster head for local aggregation, and the global model parameters are updated by weighted averaging of the aggregated local model parameters using the central aggregation server.

[0039] Rewards are distributed to the alliance set based on the updated local and global model parameters.

[0040] This invention also provides a federated learning incentive system for vehicle networking based on Nash game and coalition game, including: a data initialization module, a coalition formation module, a cluster management module, a bargaining optimization module, and a global aggregation module;

[0041] The data initialization module is used to initialize the set of vehicle clients in the vehicle network environment, and the server broadcasts global model parameters to all clients.

[0042] The alliance formation module is used to construct a stable alliance structure based on a multi-dimensional scoring function of data quality, communication efficiency, and mobility cost.

[0043] The cluster management module is used to select a cluster head and a backup cluster head within each alliance in the stable alliance structure based on a three-dimensional score of server communication efficiency, location centrality, and link stability.

[0044] The negotiation optimization module is used to establish a two-stage Nash game model. In the first stage, it is used to determine the alliance set in the vehicle client set. In the second stage, it is used to negotiate and optimize resource contribution. Based on the two-stage Nash game model, the server and the alliance set negotiate to determine the optimal data contribution ratio and the optimal resource contribution ratio.

[0045] The global aggregation module is used by the server to distribute rewards to the alliance set according to the optimal data contribution ratio and the optimal resource contribution ratio.

[0046] The present invention also provides a storage medium storing a computer program, which, when running, executes a federated learning incentive method for vehicle networking based on Nash game theory and alliance game theory.

[0047] Compared with the prior art, the present invention has the following advantages and technical effects:

[0048] This invention, by introducing coalition game theory, enables vehicles to self-organize into stable learning coalitions, effectively mitigating the instability caused by frequent entry and exit of individual vehicles from the network. It employs a Nash bargaining framework to achieve a fair allocation of resource contributions, and compared to traditional fixed-reward mechanisms, it can dynamically adjust incentive strategies based on real-time network conditions and data quality. Attached Figure Description

[0049] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0050] Figure 1 This is a flowchart of a federated learning incentive method for vehicle networking based on Nash game theory and alliance game theory, according to an embodiment of the present invention. Detailed Implementation

[0051] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0052] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0053] Example 1:

[0054] This embodiment proposes a federated learning incentive method for vehicle networks based on Nash game theory and federated game theory, such as... Figure 1 As shown, the specific steps include:

[0055] Initialize the set of vehicle clients in the vehicle networking environment, and broadcast global model parameters to all clients;

[0056] A two-stage Nash game model is established, in which the first stage is used to determine the alliance set in the vehicle client set, and the second stage is used to negotiate and optimize resource contributions.

[0057] Based on the server-side and the consortium set, a two-stage Nash game model is used to negotiate and determine the optimal data contribution ratio and the optimal resource contribution ratio.

[0058] The server distributes rewards to the alliance based on the optimal data contribution ratio and the optimal resource contribution ratio.

[0059] Furthermore, before establishing the two-stage Nash game model, the following steps are also required:

[0060] A stable alliance structure is constructed based on a multi-dimensional scoring function that considers data quality, communication efficiency, and mobility costs.

[0061] Within each alliance in a stable alliance structure, the cluster head and backup cluster head are selected based on a three-dimensional score of server communication efficiency, location centrality, and link stability.

[0062] Specifically, mobility-aware alliance formation: In a vehicle-to-everything (V2X) environment, consider an edge network comprising a server and a set of N heterogeneous vehicle nodes V = {v1, v2, ..., v...} NThe vehicle-to-everything (V2X) network is modeled as a federated learning system, where the central server acts as a global model aggregator and the vehicles act as distributed learning clients. The learning process consists of T global iterations, denoted as... To achieve an effective and resource-constrained incentive mechanism deployment, this embodiment proposes an adaptive alliance formation mechanism based on game theory. First, a comprehensive evaluation framework is established, considering the client v after joining alliance M(t). m Define a multidimensional scoring function:

[0063]

[0064] Among them, P m (M(t)) is a multidimensional scoring function, Q data (M(t)) represents the data quality score, Q comm (M(t)) represents the communication efficiency score. For mobility cost scoring, μ1 and μ2 are normalization factors.

[0065] More specifically, 0 < μ i <1, the normalization of each sub-item is as follows:

[0066] Data quality score: Measure client v using Generalized Earth Distance (GEMD) m The data heterogeneity improved after joining the alliance.

[0067] Communication efficiency rating: Assess communication capabilities relative to the league average.

[0068] Mobility cost score: Quantify the difference between client speed and the alliance's average speed.

[0069] Where α is the adjustment parameter, r m For client v m transmission rate, S represents the average client transmission rate in consortium i. m For client v m speed, For the average client speed in alliance i, s max This represents the maximum client speed within Alliance i.

[0070] Based on a scoring mechanism, the alliance formation process is modeled as a hedonistic alliance game, where clients make autonomous decisions to maximize individual utility. Client preference relations and Pareto order are defined, and the alliance dynamically adapts through three basic operations (merge, split, switch).

[0071] Furthermore, building a stable alliance structure includes: a merging module, a splitting module, and a switching module;

[0072] The merging module is:

[0073] P(M i (t)∪M j (t))≥P(M i (t))+p(M j (t))+τ merge And |M i (t)|+|M j (t)|≤

[0074] l limit ,

[0075] Among them, M i (t) represents the alliance i, M j (t) represents the alliance j, P(M) i (t) represents the multidimensional scoring function value of alliance i, P(M) j (t) represents the multidimensional scoring function value of alliance j, τ merge l is the threshold for the merge operation. limit Limit the size of the alliance;

[0076] The split module is:

[0077] P(M i (t))<τ split or |M i (t)|>l limit ,

[0078] And the resulting subset of the alliance {M i1 M i2 ,…M ik The following conditions must be met:

[0079]

[0080] Where, τ split τ is the splitting threshold. gain_split M is the threshold for splitting payoff. ij (t) represents the j-th alliance in the subset of alliances split from alliance i, P(M ij (t) is the multidimensional scoring function value of the j-th alliance in the alliance subset, j is the index of the j-th alliance in the split alliance subset, and k is the number of sets in the split alliance subset;

[0081] Switch to the following module:

[0082] P m (M j (t)∪{v m})≥P m (M i(t)\{v m})+τ switch ,

[0083] Where, τ switch To switch the revenue threshold, v m For vehicle collection The m-th client in the process.

[0084] Furthermore, the cluster head and backup cluster head are selected based on a three-dimensional score of server communication efficiency, location centrality, and link stability.

[0085] Specifically, cluster head selection in the consortium: After establishing a federated learning consortium for vehicle-to-everything (V2X) networks, cluster head selection is crucial for optimizing communication efficiency and model performance. This embodiment proposes a cluster head selection algorithm based on multi-dimensional scoring, which evaluates candidate nodes through a three-dimensional index system:

[0086]

[0087]

[0088] Scoring function g(v) m M i (t) Quantify the fitness of each node as a cluster head:

[0089]

[0090] in, This represents the weight of the k-th indicator. The three indicators are:

[0091] Server communication efficiency metrics:

[0092] Location centrality index:

[0093] Link stability metrics:

[0094] in, To be the optimal cluster head in alliance i in round t, Let g be the weight of the k-th indicator. k For the k-th index, For client v m With Alliance M i The average distance of other clients in (t), where ω1 is the normalization factor and ω2 is the normalization factor.

[0095]

[0096] Where, Δs=|s m -s j |For client vm and client v j The absolute velocity difference, where ε is the velocity difference threshold. dir j,m This is a variable that takes the value 0 or 1, used to represent the relative direction of movement between two clients; a value of 1 indicates that they are moving in the same direction. R trans This represents the maximum distance range that clients can communicate over. d(m,j) represents the distance between client v. m and client v j The distance between them.

[0097] Dynamic Alliance Selection: After establishing a stable and efficient alliance network, a dynamic alliance selection strategy is introduced. This strategy can effectively adapt to real-time network stability and evaluate the value of each alliance based on multiple factors. First, the alliance selection ratio is defined as: r(t) = ζ·r base ·R stab Among them, r base Based on the selection ratio, This is the network stability index. Based on the selection ratio r(t), the number of alliances selected in round t is calculated as follows: Then, the overall rating of the alliance is defined as:

[0098] Score(M i (t))=ζ1·Q data (M i (t))+ζ2·Q his (M i (t))

[0099] Where ζ is the influencing factor, λ is the attenuation coefficient, and σ is the speed standard deviation of the vehicle client set v. Let be the average speed of the vehicle client set, I(t) be the number of alliances obtained by the alliance formation algorithm, ζ1 be the influence factor, and ζ2 be the influence factor.

[0100] A candidate list L(t) is formed by sorting the scores in descending order, and the top τ(t) alliances are selected to participate in the FL task.

[0101] Furthermore, the two-stage Nash game model is as follows:

[0102]

[0103] Where V(y(t),z(t)) is the server utility function, U i (y j (t),z j Let y(t) be the coalition utility function, and y(t) be the optimal coalition structure Π obtained from the Nash game. * z(t) is the solution vector composed of the optimal data contribution ratio of each alliance in (t), and z(t) is the optimal alliance structure Π obtained from the Nash game.* The solution vector y is composed of the optimal computational contribution proportion of each alliance in (t). j (t) represents the optimal alliance structure Π obtained from solving the Nash game. * The optimal data contribution ratio of the j-th alliance in (t), z j (t) represents the optimal alliance structure Π obtained from solving the Nash game. * The optimal computational resource contribution ratio of the j-th alliance in (t), Π * (t) represents the optimal alliance structure.

[0104] Specifically, Nash bargaining game: in forming the optimal alliance structure Π * After (t), the cluster head of each selected alliance engages in concurrent bargaining negotiations with the server to determine the optimal contribution ratio and the amount of reward available to each alliance. A multi-party Nash bargaining solution (NBS) is constructed to optimize resource allocation in round t. The payoff for both negotiating parties is zero at the point of divergence, therefore the point of divergence is defined as the origin s0 = 0. According to Nash's theorem, the optimal agreement requires maximizing the weighted Nash product.

[0105] Furthermore, the negotiation to determine the optimal data contribution ratio and the optimal resource contribution ratio includes: using a multi-starting point sequential quadratic programming algorithm to negotiate and determine the optimal data contribution ratio and the optimal resource contribution ratio;

[0106] Among them, the optimal data contribution ratio and optimal resource contribution ratio are determined by negotiating using a multi-starting-point sequence quadratic programming algorithm, which includes: using Latin hypercube sampling to generate a preset number of initial points, executing the quadratic programming algorithm in parallel for each initial point to obtain a local optimal solution, and selecting the solution with the largest objective function value among the local optimal solutions as the Nash bargaining solution;

[0107] Based on the Nash bargaining solution, the optimal data contribution ratio and the optimal resource contribution ratio are obtained.

[0108] Furthermore, the server distributes rewards to the consortium based on the optimal data contribution ratio and the optimal resource contribution ratio, including:

[0109] Based on the optimal data contribution ratio and resource contribution ratio, the stochastic gradient descent algorithm is used to update the local model parameters;

[0110] The updated local model parameters are sent to the cluster head for local aggregation, and the global model parameters are updated by weighted averaging of the aggregated local model parameters using the central aggregation server.

[0111] Rewards are distributed to the alliance collection based on the updated local and global model parameters.

[0112] Specifically, local training and aggregation: each client v m ∈M i(t) Based on its alliance's training strategy, perform H local iterations on the private dataset using the stochastic gradient descent (SGD) algorithm:

[0113]

[0114] Where h = {1, 2, ..., H-1}, η is the learning rate. This is the mini-batch data for the h-th iteration. After local training, the updated local model parameters are sent to the cluster head for local aggregation. The aggregated model parameters from each consortium are efficiently transmitted to the central aggregation server using quantization and compression techniques, where a weighted average is performed to update the global model.

[0115] Incentive Allocation: To ensure long-term, efficient participation, participating teams that complete the training receive rewards based on Nash bargaining results. The reward scheme satisfies incentive compatibility (IC) and individual rationality (IR) constraints.

[0116]

[0117] Where γ1 and γ2 are the reward values ​​for unit effective computational contribution and unit effective data contribution, respectively, and C i (t) and D i (t) represents the effective computational contribution and effective data contribution of the consortium, respectively.

[0118] The following is a detailed description of Example 1:

[0119] like Figure 1 As shown, the specific implementation steps of Example 1 are as follows:

[0120] 1) System Initialization and Parameter Settings: Simulate a 15km two-way highway scenario with three lanes in each direction. The server is located at kilometer 7.5. Vehicle speeds in lanes 0-2 are 40-60, 60-80, and 80-100 km / h, respectively, while speeds in lanes 3-5 (reverse direction) are -40-(-60), -60-(-80), and -80-(-100) km / h, respectively. Set up 100 heterogeneous clients, divided into three categories:

[0121] High-end equipment (30%): 1.5-3.0 GFLOPS (computing power), 15-25MHz (bandwidth), 80-120mW (transmission power);

[0122] Mid-range equipment (50%): 0.8-1.5 GFLOPS, 8-15 MHz, 50-80 mW;

[0123] Low-end equipment (20%): 0.3-0.8 GFLOPS, 3-8 MHz, 20-50 mW;

[0124] 2) Alliance Formation Process: During initialization, each vehicle node forms a separate alliance V = {{v1}, {v2}, ..., {v...} N}}. Calculate the initial preference P m (M(t)), set the invalid operation count no_ops_count = 0. Iterate through the following operations until convergence (no_ops_count < 3):

[0125] Merge operation: Iterate through all alliance pairs (M) i If M(t) is satisfied, then merge them:

[0126] Collective returns exceed the threshold: P(M) i (t)∪M j (t))≥P(M i (t))+P(M j (t))+τ merge ;

[0127] Size constraint satisfies: |M i (t)|+|M j (t)|≤l limit ;

[0128] The merged partitions satisfy the Pareto order:

[0129] Split operation: against Alliance M i (t), if P(M) i (t))<τ split or |M i (t)|>l limit Generate all possible partitioning schemes and select the one that satisfies:

[0130]

[0131] The execution of the plan under the given conditions.

[0132] Toggle operation: For v m ∈M i (t) Switch to M j (t), examine individual benefit improvement and target coalition capacity constraints.

[0133] 3) Cluster head selection and coalition selection: For each formed coalition M i (t)∈Π ini (t), calculate the three-dimensional score of all nodes:

[0134] Calculate server communication efficiency g1, calculate location centrality g2, and calculate link stability g3;

[0135] Select the node with the highest score as the cluster head. The second highest value serves as a backup cluster head. Based on network stability R stab Calculate the selection ratio r(t) to determine the number of participating alliances τ(t). Calculate the overall score (Score) for each alliance. i (t)), select the first τ(t) alliances to form Π * (t).

[0136] 4) Solving Nash bargaining: Constructing an optimization problem:

[0137]

[0138] Generated using Latin hypercube sampling (LHS) There are several initial points. The SLSQP algorithm is executed in parallel at each initial point to obtain local optima. The solution with the largest objective function value is selected as the final Nash bargaining solution (y). * (t),z * (t)).

[0139] 5) Federated learning training execution: The server will globally train the model ω t The data is sent to each cluster head, which then broadcasts it to all members of the consortium. Each client performs H=5 rounds of local training, with a batch size B=32 and a learning rate η=0.01. The cluster heads collect model updates from within the consortium and perform a weighted average.

[0140]

[0141] The server aggregates models from various alliances:

[0142]

[0143] 6) Incentive Allocation and Historical Updates: Based on the negotiation results (y * (t),z * (t) Calculate the rewards for each alliance and update the historical performance set of the alliances. The earnings records of the most recent 5 rounds will be retained for subsequent evaluation.

[0144] The algorithm is shown in Table 1:

[0145] Table 1

[0146]

[0147]

[0148] Example 2:

[0149] This embodiment also provides a vehicle-to-everything (V2X) federated learning incentive system based on Nash game theory and coalition game theory, including: a data initialization module, a coalition formation module, a cluster management module, a bargaining optimization module, and a global aggregation module;

[0150] The data initialization module is used to initialize the set of vehicle clients in the vehicle networking environment. The server broadcasts global model parameters to all clients.

[0151] The alliance formation module is used to construct stable alliance structures based on a multi-dimensional scoring function of data quality, communication efficiency, and mobility costs.

[0152] The cluster management module is used to select the cluster head and backup cluster head within each alliance in a stable alliance structure based on a three-dimensional score of server communication efficiency, location centrality, and link stability.

[0153] The negotiation optimization module is used to establish a two-stage Nash game model. In the first stage, it is used to determine the alliance set in the vehicle client set. In the second stage, it is used to negotiate and optimize resource contributions. Based on the two-stage Nash game model, the server and the alliance set negotiate to determine the optimal data contribution ratio and the optimal resource contribution ratio.

[0154] The global aggregation module is used by the server to distribute rewards to the alliance set based on the optimal data contribution ratio and the optimal resource contribution ratio.

[0155] Example 3:

[0156] This embodiment also provides an electronic device, including: a memory and a processor, wherein the memory stores a computer program that is run by the processor, and the computer program executes the above-mentioned incentivized learning method for vehicle networking based on Nash game and federation game when run by the processor.

[0157] Example 4:

[0158] This embodiment also provides a storage medium on which a computer program is stored. When the computer program is run, it executes the above-described incentivized learning method for vehicle networking based on Nash game theory and federation game theory.

[0159] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A federated learning incentive method for vehicle networking based on Nash game theory and coalition game theory, characterized in that, include: Initialize the set of vehicle clients in the vehicle networking environment, and broadcast global model parameters to all clients; A two-stage Nash game model is established, wherein the first stage is used to determine the alliance set in the set of vehicle clients, and the second stage is used to negotiate and optimize resource contributions. Based on the server and the alliance set, the optimal data contribution ratio and the optimal resource contribution ratio are determined through negotiation using the two-stage Nash game model. The server distributes rewards to the alliance set based on the optimal data contribution ratio and the optimal resource contribution ratio; Before establishing the two-stage Nash game model, the following steps are also required: A stable alliance structure is constructed based on a multi-dimensional scoring function that considers data quality, communication efficiency, and mobility costs. Within each alliance in the stable alliance structure, a cluster head and a backup cluster head are selected based on a three-dimensional score of server communication efficiency, location centrality, and link stability. The stable alliance structure is constructed by including: a merging module, a splitting module, and a switching module; The merging module is: and , in, For the alliance , For the alliance , For the alliance The multidimensional scoring function value, For the alliance The multidimensional scoring function value, The threshold for the merge operation. Limit the size of the alliance; The splitting module is: or , And the subset of the split alliance The following conditions must be met: ， in, The splitting threshold, The threshold for splitting rewards, For the alliance The first in the split alliance subset One alliance, For the first in the subset of the alliance The multidimensional scoring function values ​​of each alliance For the first subset of the split alliance The subscript of the alliance, The number of sets in the resulting union subsets; The switching module is: ， in, To switch the revenue threshold, For vehicle collection The first in One client.

2. The incentive method for federated learning in vehicle networking based on Nash game theory and coalition game theory according to claim 1, characterized in that, The multidimensional scoring function is: in, For multidimensional scoring functions, Score the data quality. Rate communication efficiency. Scoring for mobility costs, and This is the normalization factor.

3. The incentivization method for federated learning in vehicle networking based on Nash game theory and coalition game theory as described in claim 1, characterized in that, The cluster head and backup cluster head are selected based on a three-dimensional score of server communication efficiency, location centrality, and link stability.

4. The incentivizing method for federated learning in vehicle networking based on Nash game theory and coalition game theory as described in claim 1, characterized in that, The two-stage Nash game model is as follows: in, For server utility functions, For the coalition utility function, Optimal Alliance Structure for Nash Game The solution vector is composed of the optimal data contribution ratio for each alliance. Optimal Alliance Structure for Nash Game The solution vector is composed of the optimal computational contribution ratio of each alliance. Optimal Alliance Structure for Nash Game The first in The optimal data contribution ratio for each alliance Optimal Alliance Structure for Nash Game The first in The optimal proportion of computing resources contributed by each alliance This represents the optimal alliance structure.

5. The incentivizing method for federated learning in vehicle networking based on Nash game theory and coalition game theory according to claim 1, characterized in that, The negotiation to determine the optimal data contribution ratio and the optimal resource contribution ratio includes: using a multi-starting-point sequential quadratic programming algorithm to negotiate and determine the optimal data contribution ratio and the optimal resource contribution ratio; Among them, the method of negotiating and determining the optimal data contribution ratio and the optimal resource contribution ratio by using a multi-starting point sequence quadratic programming algorithm includes: using Latin hypercube sampling to generate a preset number of initial points, executing the quadratic programming algorithm in parallel for each initial point to obtain a local optimal solution, and selecting the solution with the largest objective function value among the local optimal solutions as the Nash bargaining solution; Based on the Nash bargaining solution, the optimal data contribution ratio and the optimal resource contribution ratio are obtained.

6. The incentivizing method for federated learning in vehicle networking based on Nash game theory and coalition game theory according to claim 1, characterized in that, The server distributes rewards to the alliance set based on the optimal data contribution ratio and the optimal resource contribution ratio, including: Based on the optimal data contribution ratio and resource contribution ratio, the local model parameters are updated using the stochastic gradient descent algorithm. The updated local model parameters are sent to the cluster head for local aggregation, and the global model parameters are updated by weighted averaging of the aggregated local model parameters using the central aggregation server. Rewards are distributed to the alliance set based on the updated local and global model parameters.

7. A vehicle-to-everything (V2X) federated learning incentive system based on Nash game theory and coalition game theory, used to implement the method as described in any one of claims 1-6, characterized in that, include: The module includes a data initialization module, an alliance formation module, a cluster management module, a negotiation optimization module, and a global aggregation module. The data initialization module is used to initialize the set of vehicle clients in the vehicle network environment, and the server broadcasts global model parameters to all clients. The alliance formation module is used to construct a stable alliance structure based on a multi-dimensional scoring function of data quality, communication efficiency, and mobility cost. The cluster management module is used to select a cluster head and a backup cluster head within each alliance in the stable alliance structure based on a three-dimensional score of server communication efficiency, location centrality, and link stability. The negotiation optimization module is used to establish a two-stage Nash game model. In the first stage, it is used to determine the alliance set in the vehicle client set. In the second stage, it is used to negotiate and optimize resource contribution. Based on the two-stage Nash game model, the server and the alliance set negotiate to determine the optimal data contribution ratio and the optimal resource contribution ratio. The global aggregation module is used by the server to distribute rewards to the alliance set according to the optimal data contribution ratio and the optimal resource contribution ratio.

8. A storage medium, characterized in that, The storage medium stores a computer program, which, when executed, performs the method as described in any one of claims 1-6.

Images