Method and system for optimizing cost containing antenna position of Internet of Vehicles based on movable antenna

Abstract

The invention discloses a method and a system for optimizing antenna position cost of Internet of Vehicles based on a movable antenna, and the method comprises the steps: constructing an Internet of Vehicles model of a privacy protection federated learning architecture based on the assistance of an intelligent movable antenna, and enabling a base station to give corresponding compensation according to the total cost of vehicle users in a signal coverage range of the base station, establishing a target function for minimizing the compensation cost of the base station; further decomposing the objective function into two sub-problems, respectively adopting a proposed greedy vehicle user selection algorithm and an adaptive particle swarm optimization algorithm to solve and optimize parameters of the objective function, and in each round of federated communication, the base station obtains a selected vehicle user set and schedules the position of the intelligent movable antenna, so that the intelligent movable antenna is obtained. And the compensation cost of the base station is minimized. According to the method, the compensation cost of the base station is modeled, optimization of the compensation cost of the base station in the Internet of Vehicles is realized, the communication rate between the vehicle user and the base station is improved by scheduling the position of the intelligent movable antenna, and the communication time delay is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of vehicle networking, and specifically relates to an optimization method and system for vehicle networking with a movable antenna, including the cost of antenna position. Background Art

[0002] The problem of optimizing the base station compensation cost in vehicle networking is a popular research direction. Because of the heterogeneous resources of vehicle users, such as computing resources, communication resources, etc., the mobility of vehicle users, the sensitivity of vehicle users to local data, the energy consumed by vehicle users to update the local model, the energy consumed by vehicle users to add noise to the updated local model, and the energy consumed by vehicle users to upload the noisy local model, all have a certain degree of influence on the decision-making of the base station to select vehicle users. At the same time, the intelligent movable antenna deployed at the base station can improve the communication environment of vehicle users and reduce the communication delay by optimizing its position. Because there are many influencing factors, it is difficult to give an optimal method through theoretical derivation. It is necessary to propose a selection strategy for the optimal vehicle users of the base station and a method for scheduling intelligent movable antennas.

[0003] Compared with the technical content of the application document CN118333191A "A Vehicle Selection and Resource Optimization Method and System for Federated Learning in Vehicle Networking", there are the following differences: I. Model Construction and Objective Function; The application document CN118333191A only starts from the vehicle computing power, transmission capacity and fairness constraints, establishes the optimization objective of the federated learning system, and transforms the time-coupled offline problem into an online problem through Lyapunov optimization; the system also uses linear regression to predict the vehicle inherent parameters, and jointly optimizes resource allocation and vehicle selection based on reinforcement learning to minimize the overall system cost and improve the model accuracy and training efficiency.

[0004] This application, under the privacy-preserving federated learning framework, introduces four major components: communication cost, computing cost, privacy cost, and vehicle mobility model, and constructs an objective function to minimize the base station compensation cost for the compensation requirements of vehicle users within the signal coverage of the base station. At the same time, we introduce the scheduling of intelligent movable antennas, incorporate communication rate and delay into the modeling, and achieve double optimization of base station compensation and user experience.

[0005] II. Algorithm Design and Solving Method; In the application document CN118333191A, linear regression and reinforcement learning are respectively used for parameter prediction and online resource scheduling. The algorithm has a high coupling degree but requires training a large number of policy models, and does not consider the physical layer gain of the time-varying communication link.

[0006] This application splits the objective function into two sub - problems: "vehicle user selection" and "antenna position scheduling". For the former, a greedy vehicle user selection algorithm is designed to quickly screen high - value participants; for the latter, an adaptive Particle Swarm Optimization (PSO) algorithm is adopted to dynamically adjust the antenna position to maximize communication quality and minimize compensation cost. This combined method takes into account both computational efficiency and convergence performance, and synchronizes the user set and antenna layout in real - time through each round of federated communication.

[0007] Compared with the technical solution of the patent document CN114051222B, "A Wireless Resource Allocation and Communication Optimization Method Based on Federated Learning in a Vehicular Network Environment", there are the following differences: I. System Model and Optimization Objectives; The patent document CN114051222B constructs a wireless resource allocation and communication optimization model based on the interaction between the base station and vehicle - mounted devices. First, the Jonker–Volgenant algorithm is used to perform the best match according to the channel state and data volume. Then, a traversal algorithm is used to select high - performance vehicles. Finally, the random gradient quantization is used to compress the uplink model parameters to save bandwidth and improve the convergence accuracy.

[0008] In addition to the above communication and resource framework, this application introduces the modeling dimensions of privacy cost and compensation cost, and incorporates intelligent movable antennas into the overall architecture: in each round of federated communication, the base station not only selects participating vehicles but also schedules the antenna position in real - time to minimize the compensation overhead and communication delay in a multi - objective manner, improving the system flexibility and user experience.

[0009] II. Algorithm Process and Performance Advantages; The method of the patent document CN114051222B sequentially executes matching → user screening → quantization compression. Although it optimizes resource allocation and bandwidth utilization, it does not jointly consider user incentives and the dynamic changes in the communication environment.

[0010] This application optimizes the coordination of minimizing compensation cost and maximizing communication rate through the idea of "divide and conquer": the greedy algorithm ensures the low complexity of vehicle selection, and the adaptive PSO efficiently converges to the optimal antenna layout in the continuous space; the two are coupled to achieve unified control of user incentives, communication quality, and delay. Summary of the Invention

[0011] To address the above problems, the present invention provides an optimization method and system for a vehicular network with movable antennas and antenna position costs. By selecting vehicle users and optimizing the position of intelligent movable antennas, the optimization of the base station compensation cost in the vehicular network is achieved, the transmission rate of vehicle users is increased, and the transmission delay is reduced.

[0012] To achieve the above-mentioned invention objectives, based on an optimization method for a vehicle-to-everything (V2X) network with movable antennas considering the cost of antenna positions, the method includes the following steps: S1: Construct a V2X network model based on a privacy-preserving federated learning architecture with intelligent movable antennas. According to the total cost of vehicle users, which consists of three parts: communication cost, computing cost, and privacy cost, the base station provides corresponding compensation, and determines the vehicle users within the signal coverage range of the base station according to the vehicle mobility model, and establishes an objective function to minimize the total cost of the base station; S2: Decompose the target problem into two sub-problems and use the greedy vehicle user selection algorithm and the adaptive particle swarm optimization algorithm to solve the sub-problems respectively, to obtain the set of vehicle users selected by the base station and the positions of the intelligent movable antennas; S3: Obtain the set of vehicle users selected by the federated communication base station in each round and the positions of the antennas after scheduling, calculate the compensation cost of the base station in the current round, and output the optimized base station compensation cost.

[0013] As a further improvement of the present invention, the establishment of the objective function to minimize the total cost of the base station in step S1 includes: The establishment of the objective function to minimize the total cost of the base station includes three parts of costs: communication cost, computing cost, and privacy cost. The optimization problem is as follows: ; ; Among them, represents the set of selected vehicle users, represents the set of intelligent movable antennas; represents the total set of vehicle users; represents a vehicle user, represents whether the vehicle user is selected; represents the total compensation of the base station to the vehicle user; represents the number of intelligent movable antennas; respectively represent the th and th antennas. A represents the antenna system, which consists of a two-dimensional coordinate system. and respectively represent the positions of the th antenna and the th antenna in the two-dimensional coordinate system of the antenna system A. represents the minimum distance between two adjacent antennas; represents the total delay of the vehicle user, is the maximum delay tolerable by the base station; represents the distance traveled by the vehicle user, represents the radius of the signal coverage range of the base station. represents the cosine angle between the base station coverage area and the road, represents the distance between the vehicle user and the entrance of the base station coverage area; is the size of the noise variance added by the vehicle user to the updated local model, represents the total noise variance size threshold of the base station for the selected set of vehicle users.

[0014] As a further improvement of the present invention, the communication cost of the objective function for establishing the minimization of the total base station cost is expressed as: ; is the transmission rate of the vehicle user, is the transmission power of the vehicle user, is the hybrid vector received by the base station from the vehicle user, is the Gaussian white noise in the channel transmission process; is the transmission delay of the vehicle user, is the size of the data volume uploaded by the vehicle user; is the transmission energy consumption of the vehicle user; is the transmission cost of the vehicle user, is the unit cost of the transmission energy consumption.

[0015] As a further improvement of the present invention, the calculation cost of the objective function for establishing the minimization of the total base station cost is expressed as: ; is the calculation delay of the vehicle user, is the number of CPU cycles required for the vehicle user to process a single sample, is the size of the vehicle user's local dataset, is the CPU frequency of the vehicle user's local calculation; is the calculation energy consumption of the vehicle user, is the CPU energy consumption of the vehicle user's local calculation; is the calculation cost of the vehicle user, is the unit cost of the calculation energy consumption.

[0016] As a further improvement of the present invention, the privacy cost of the objective function for establishing the minimization of the total base station cost is expressed as: ; represents the privacy cost of the vehicle user, is the sensitivity of two adjacent datasets in differential privacy, represents the unit cost of the noise energy consumption added by the vehicle user to the local model, Represents the privacy budget of the vehicle user.

[0017] As a further improvement of the present invention, the vehicle mobility model in step S1 is expressed as: ; Represents the distance traveled by the vehicle user, is the driving speed of the vehicle user, Represents the transmission delay of the vehicle user, Represents the computing delay of the vehicle user, Represents the distance traveled by the vehicle user after uploading the updated local model, R Represents the radius of the base station signal coverage area, Represents the cosine angle between the base station coverage area and the road, Represents the distance between the vehicle user and the entrance of the base station coverage area, Represents the distance between the vehicle user and the base station.

[0018] As a further improvement of the present invention, the greedy vehicle user selection algorithm in decomposing the target problem into two sub-problems in step S2 is expressed as: ; ; Among them, Represents the set of selected vehicle users; V represents the total set of vehicle users; Represents the vehicle user, Represents whether the vehicle user is selected; Represents the total compensation of the base station to the vehicle user; Z Represents the number of intelligent movable antennas; Represents the total delay of the vehicle user, is the maximum tolerable delay of the base station; Represents the distance traveled by the vehicle user, R Represents the radius of the base station signal coverage area, Represents the cosine angle between the base station coverage area and the road, Represents the distance between the vehicle user and the entrance of the base station coverage area; is the size of the noise variance added by the vehicle user to the updated local model, Represents the threshold of the total noise variance size of the base station for the selected set of vehicle users; Solved by the greedy vehicle user selection algorithm, and the specific steps include; First, the base station determines the set of vehicle users that meet the delay constraint and are within the signal coverage of the base station. Then, all vehicle users in this set calculate their respective communication costs, computing costs, and privacy costs, calculate the total cost, and upload it to the base station. The base station sorts the vehicle users in the set according to the total cost uploaded from low to high. The base station starts selecting from the vehicle user with the lowest total cost until the threshold of the total noise is met. Finally, the base station obtains the subset of vehicle users selected in the current round.

[0019] As a further improvement of the present invention, the adaptive particle swarm optimization algorithm in decomposing the target problem into two sub-problems in step S2 is expressed as: ; ; Among them, represents the set of intelligent movable antennas; C represents the subset of vehicle users obtained by the greedy vehicle user selection algorithm; represents a vehicle user, represents the total compensation of the base station to the vehicle user; represents the number of intelligent movable antennas; respectively represent the th and th antennas. A represents the antenna system, which consists of a two-dimensional coordinate system. and respectively represent the th antenna and the th antenna in the two-dimensional coordinate system of the antenna system A. represents the minimum distance between two adjacent antennas; The sub-problem is solved by the adaptive particle swarm optimization algorithm. The specific steps are as follows: The first step is to initialize the velocity and position of the particles: ; Among them represents the initial position set of the particle swarm, represents the initial position of particle ; represents the initial velocity set of the particle swarm, represents the initial velocity of particle . There are I particles in total. Among them is further expressed as: ; Among them , represent the rd stored by particle The initial position of an antenna in the coordinate system; The feasible region of the intelligent movable antenna is denoted as A and satisfies , so the positions of the particles and the antenna need to satisfy the following: ; Where represents the stored th antenna's position on the axis in the coordinate system, represents the stored th antenna's position on the axis in the coordinate system, H represents the radius of the antenna system; The second step is the update of the particle velocity and position: During the process of the adaptive particle swarm optimization algorithm, in each iteration, the particle swarm updates its respective position and velocity, which are represented as follows: ; Where represents the current iteration number of the particle swarm algorithm, represents the global optimal position of the individual particle swarm, represents the global optimal position of the particle swarm, represents the th round of particle 's velocity, represents the th round's particle 's position, represents the th round's particle 's position, is the inertial factor of the particle, and are the learning factors of the th round of the particle, and are random numbers that satisfy ; The third step is the adaptive update of the inertial factor: Currently, a relatively popular way to update the inertial factor is to adaptively update the inertial factor in each iteration, which is represented as follows: ; Where is the inertial factor of the particle, represents the maximum inertial weight of the particle, represents the minimum inertial weight of the particle, represents the current iteration number of the particle swarm algorithm, Mrepresents the total number of iterations of the adaptive particle swarm algorithm; The fourth step is to adaptively update the learning factors: The update of the learning factors is based on the distance of the current particle from its individual particle swarm's global optimal position and the global optimal position of the particle swarm The influence is that particles farther from the two optimal positions should obtain larger learning factors in order to approach their optimal positions faster. The adaptive weighted update rule is expressed as follows: ; where and are the learning factors of the -th round of particles. The function F represents the adaptive weighted update function. and respectively represent the distance of the particle from its individual optimal position and the distance from the global optimal position of the population in the -th round. Specifically, and are expressed as follows: ; A piecewise linear function is used as the adaptive weighted update function, which is specifically expressed as follows: ; where and respectively represent the distance of the particle from its individual optimal position and the distance from the global optimal position of the population in the -th round. represents the global optimal position of the individual particle swarm. represents the global optimal position of the particle swarm. represents the -th round. The position of the particle . and are parameters that control the learning factors and satisfy ; represents the distance between the particle and the optimal position. represents the threshold of the current particle's distance from the optimal position; The fifth step is to set the penalty function and the fitness function: Considering the constraints of this sub-problem, a penalty function is set to punish particles that violate the constraints. The penalty function is specifically expressed as follows: ; where is the penalty factor. The penalty function representing the particles is the number of particles violating the constraints; Combined with the penalty function, the fitness function of the particles is expressed as follows: ; where represents the fitness function of the particles, C represents the subset of vehicle users obtained by the greedy vehicle user selection algorithm; represents the vehicle users; represents the total compensation of the base station to the vehicle users; represents the penalty function.

[0020] The present invention provides an optimization system for a vehicle-to-everything network with antenna position cost based on a movable antenna, and the system includes: A construction module for constructing a vehicle-to-everything network model based on a privacy-protected federated learning architecture assisted by an intelligent movable antenna, and establishing an objective function for minimizing the base station compensation cost according to the historical state data of vehicle users; the historical state data includes the communication cost, local computing cost, privacy cost of vehicle users, and the distance between vehicle users and the base station; A vehicle user selection module for obtaining the vehicle users selected by the base station, solving and parameter optimizing the objective function by using a greedy vehicle user selection algorithm, obtaining a vehicle user selection model, and using the result for the antenna position scheduling module; An antenna position optimization module for obtaining the currently optimized antenna position, solving and parameter optimizing the objective function by using an adaptive particle swarm optimization algorithm and the result of the vehicle user selection module, and obtaining the optimized antenna position.

[0021] Beneficial effects: By constructing a vehicle-to-everything network model based on a privacy-protected federated learning architecture with an intelligent movable antenna, establishing an objective function for minimizing the base station compensation cost according to the communication cost, local computing cost, privacy cost of vehicle users, and the mobility of vehicle users, and by selecting vehicle users and optimizing the position of the intelligent movable antenna, the optimization of the base station compensation cost in the vehicle-to-everything network is realized, the transmission rate of vehicle users is improved, and the transmission delay is reduced. Brief Description of the Drawings

[0022] Figure 1 is a flowchart of a method for optimizing cost and antenna position in a vehicle-to-everything network based on a privacy-protected federated learning architecture with an intelligent movable antenna; Figure 2 is a schematic diagram of a system for optimizing cost and antenna position in a vehicle-to-everything network based on a privacy-protected federated learning architecture with an intelligent movable antenna; Figure 3It is a comparison chart of base station compensation costs; Figure 4 It is a comparison chart of the adaptive particle swarm optimization algorithm. Specific implementation manners

[0023] The following combines the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

[0024] Embodiment 1;

[0025] Figure 1 It is a flowchart of a method for optimizing cost and antenna position in a vehicle-to-everything network based on a privacy protection federated learning architecture with intelligent movable antennas. As Figure 1 shown, this embodiment provides a method for optimizing cost and antenna position in a vehicle-to-everything network based on a privacy protection federated learning architecture with intelligent movable antennas. The method includes the following steps: S1: Construct a vehicle-to-everything network model based on a privacy protection federated learning architecture with intelligent movable antennas. According to the total cost of vehicle users within the signal coverage range of the base station, the base station gives corresponding compensation, and establish an objective function for minimizing the total cost of the base station; S2: Decompose the target problem into two sub-problems and use the greedy vehicle user selection algorithm and the adaptive particle swarm optimization algorithm to solve the sub-problems respectively, and obtain the set of vehicle users selected by the base station and the position of the intelligent movable antenna; S3: Obtain the set of vehicle users selected by the base station in each round of federated communication and the position of the antenna after scheduling, calculate the compensation cost of the base station in the current round, and output the optimized base station compensation cost.

[0026] The method of this embodiment realizes the minimization of the base station compensation cost in the vehicle-to-everything network by optimizing the base station compensation cost and the antenna position, and improves the transmission rate of vehicle users and reduces the transmission delay.

[0027] In step S1, the establishment of the objective function for minimizing the base station compensation cost includes: Establish the objective function of the base station compensation cost, and the optimization problem is as follows: ; ; Among them, represents the set of vehicle user selections, represents the set of intelligent movable antennas; represents the total set of vehicle users; Denote the vehicle user, indicate whether the vehicle user is selected; represent the total compensation of the base station to the vehicle user; represent the number of intelligent movable antennas; respectively represent the and the th antennas. A represents the antenna system, which consists of a two-dimensional coordinate system, and respectively represent the th antenna and the th antenna's positions in the two-dimensional coordinate system of the antenna system A, represent the minimum distance between two adjacent antennas; represent the total time delay of the vehicle user, is the maximum time delay tolerated by the base station; represent the distance traveled by the vehicle user, represent the radius of the base station signal coverage area, represent the cosine angle between the base station coverage area and the road, represent the distance between the vehicle user and the entrance of the base station coverage area; is the magnitude of the noise variance added by the vehicle user to the updated local model, represent the threshold of the total noise variance magnitude of the base station for the selected set of vehicle users.

[0028] In step S1, under the vehicle networking framework based on intelligent movable antennas, the communication cost of the vehicle user can be expressed as: ; is the transmission rate of the vehicle user, is the transmission power of the vehicle user, is the mixed vector received by the base station from the vehicle user, is the Gaussian white noise in the channel transmission process; is the transmission time delay of the vehicle user, is the size of the data volume uploaded by the vehicle user; is the transmission energy consumption of the vehicle user; is the transmission cost of the vehicle user, is the unit cost of the transmission energy consumption.

[0029] Specifically, construct a network that includes a base station, with an edge server and an intelligent movable antenna system deployed inside the base station, and N vehicle users.

[0030] The transmission time delay of the vehicle user is expressed as: ; Among them, is the data volume size uploaded by the vehicle user, is the transmission rate of the vehicle user; The required transmission energy consumption is: ;

[0031] Among them, is the transmission power of the vehicle user.

[0032] The transmission cost caused by the vehicle user is expressed as: ; Among them, is the unit cost of transmission energy consumption.

[0033] In step S1, under the vehicle networking framework based on the intelligent movable antenna, the computing cost of the vehicle user can be expressed as: ; is the computing delay of the vehicle user, is the number of CPU cycles required for the vehicle user to process a single sample, is the size of the vehicle user's local dataset, is the CPU frequency of the vehicle user's local computing; is the computing energy consumption of the vehicle user, is the CPU energy consumption of the vehicle user's local computing; is the computing cost of the vehicle user, is the unit cost of computing energy consumption.

[0034] Specifically, the vehicle user's local computing delay is: ; Among them, is the number of CPU cycles required for the vehicle user to process a single sample, is the size of the vehicle user's local dataset, is the CPU frequency of the vehicle user's local computing; The required computing energy consumption is: ; Among them, is the computing power of the vehicle user.

[0035] The computing cost caused by the vehicle user is expressed as: ; Among them, is the unit cost of computing energy consumption.

[0036] In step S1, the privacy cost model of the vehicle user can be expressed as: ; represents the privacy cost of vehicle users, is the sensitivity of two adjacent data sets in differential privacy, represents the unit cost of the noise energy consumption added by vehicle users to the local model, represents the privacy budget of vehicle users.

[0037] Specifically, the total sum of the three parts of the cost for vehicle users is expressed as

[0038] ; In step S2, the greedy vehicle user selection algorithm and the adaptive particle swarm optimization algorithm are used to solve the objective function. Specifically, first, the greedy vehicle user selection algorithm is used to determine the set of vehicle users selected by the base station, and then, according to the determined set of vehicle users, the adaptive particle swarm optimization algorithm is used to solve the antenna position. The specific process is as follows: 1. The first sub-problem after the decomposition of the target problem is expressed as: ; ; Among them, represents the set of vehicle users selected; V represents the total set of vehicle users; represents a vehicle user, represents whether the vehicle user is selected; represents the total compensation of the base station to the vehicle user; Z represents the number of intelligent movable antennas; represents the total delay of vehicle users, is the maximum tolerable delay of the base station; represents the distance traveled by the vehicle user, R represents the radius of the signal coverage range of the base station, represents the cosine angle between the base station coverage range and the road, represents the distance between the vehicle user and the entrance of the base station coverage range; is the size of the noise variance added by the vehicle user to the updated local model, represents the total noise variance size threshold of the base station for the selected set of vehicle users; The greedy vehicle user selection algorithm is used to solve this sub-problem. The specific steps are as follows: First, the base station determines the set of vehicle users that satisfy the latency constraint and are within the signal coverage of the base station. Then, all the vehicle users in this set calculate their respective communication costs, computing costs, and privacy costs, calculate the total cost, and upload it to the base station. The base station sorts the vehicle users in the set according to the total cost they upload in ascending order. The base station starts selecting from the vehicle user with the lowest total cost until the threshold of the total noise is met. Finally, the base station obtains the subset of vehicle users selected in the current round.

[0039] 2. The second sub-problem after the target problem decomposition is expressed as: Solve it through the greedy vehicle user selection algorithm. The specific steps are as follows: First, the base station determines the set of vehicle users that satisfy the latency constraint and are within the signal coverage of the base station. Then, all the vehicle users in this set calculate their respective communication costs, computing costs, and privacy costs, calculate the total cost, and upload it to the base station. The base station sorts the vehicle users in the set according to the total cost they upload in ascending order. The base station starts selecting from the vehicle user with the lowest total cost until the threshold of the total noise is met. Finally, the base station obtains the subset of vehicle users selected in the current round.

[0040] ; ; Among them, represents the set of intelligent movable antennas; C represents the subset of vehicle users obtained by the greedy vehicle user selection algorithm; represents the vehicle user, represents the total compensation of the base station to the vehicle user; represents the number of intelligent movable antennas; respectively represent the th and th antennas. A represents the antenna system, which consists of a two-dimensional coordinate system. and respectively represent the th antenna and the th antenna's positions in the two-dimensional coordinate system of the antenna system A. represents the minimum distance between two adjacent antennas; Solve this sub-problem using the adaptive particle swarm optimization algorithm. The specific steps are as follows: Initialize the velocity and position of the particles: ; Among them represents the initial position set of the particle swarm, represents the particle 's initial position; Represents the initial velocity set of the particle swarm, Represents the particle The initial velocity, and there are a total of I Particles, among which Is further expressed as: ; Among them 、 Represents the particle The initial position of the stored The th antenna in the coordinate system; The feasible region of the intelligent movable antenna is denoted as A, satisfying , so the positions of the particles and the antenna need to satisfy the following: ; Among them Represents the particle The stored The th antenna in the coordinate system Axis position, Represents the particle The stored The th antenna in the coordinate system Axis position, H Represents the radius of the antenna system; Particle velocity and position update: During the process of the adaptive particle swarm optimization algorithm, in each round of iteration, the particle swarm will update its respective position and velocity, which are respectively expressed as follows: ; Among them Represents the current iteration round of the particle swarm algorithm, Represents the global optimal position of the individual particle swarm, Represents the global optimal position of the particle swarm, Represents the Round particle Velocity, Represents the Round when the particle Position, Represents the Round when the particle Position, Is the inertia factor of the particle, And Is the Round particle learning factor, And Are random numbers satisfying ; Adaptive update of the inertia factor: Currently, the more popular way to update the inertia factor is to adaptively update the inertia factor for each iteration, which is expressed as follows: ; where is the inertia factor of the particle, represents the maximum inertia weight of the particle, represents the minimum inertia weight of the particle, represents the current iteration number of the particle swarm algorithm, M represents the total number of iterations of the adaptive particle swarm algorithm; Adaptive update of the learning factor: The adaptive particle swarm optimization algorithm has a faster convergence speed and better convergence performance compared to the traditional particle swarm optimization algorithm. The update of the learning factor is based on the distance of the current particle from its individual optimal position and the global optimal position of the particle swarm . Particles that are farther from the two optimal positions should obtain a larger learning factor in order to approach their optimal positions faster. The adaptive weighted update rule is expressed as follows: ; where and are the learning factors of the particle in the th iteration. The function F represents the adaptive weighted update function, and represent the distances of the particle from its individual optimal position and the global optimal position of the swarm in the th iteration, respectively. Specifically, and are expressed as follows: ; A piecewise linear function is used as the adaptive weighted update function, which is specifically expressed as follows: ; where and represent the distances of the particle from its individual optimal position and the global optimal position of the swarm in the th iteration, respectively, represents the global optimal position of the individual particle swarm, represents the global optimal position of the particle swarm, represents the th iteration, is the position of the particle and are parameters that control the learning factor and satisfy ; Represents the distance between the particle and the optimal position, Indicates the threshold of the current particle's distance from the optimal position; Set the penalty function and the fitness function: Considering the constraints of this sub - problem, we set a penalty function to punish the particles that violate the constraints. The penalty function Is specifically expressed as follows: ; Where Is the penalty factor, Represents the penalty function of the particle, Is the number of particles that violate the constraints; Combined with the penalty function, the fitness function of the particle Is expressed as follows: ; Where Represents the fitness function of the particle, C Represents the subset of vehicle users obtained by the greedy vehicle - user selection algorithm; Represents the vehicle user; Represents the total compensation of the base station to the vehicle user; Represents the penalty function.

[0041] Figure 3 Is a comparison graph of the base - station compensation cost of the algorithm proposed in this patent, the random vehicle - user selection algorithm, and the all - vehicle - user selection algorithm. Figure 4 Is a comparison graph of the adaptive particle swarm optimization algorithm proposed in this patent and the traditional particle swarm optimization algorithm. It can be seen from Figure 3 That the algorithm proposed in this patent has a lower base - station compensation cost compared with other algorithms. It can be seen from Figure 4 That the adaptive particle swarm optimization algorithm proposed in this patent has a faster convergence speed and better convergence performance compared with the traditional particle swarm optimization algorithm.

[0042] In this embodiment, a vehicle - to - everything (V2X) network model based on a privacy - protected federated learning architecture with intelligent movable antennas is constructed. According to the communication cost, local computing cost, privacy cost of vehicle users, and the mobility of vehicle users, an objective function for minimizing the base - station compensation cost is established. By selecting vehicle users and optimizing the position of intelligent movable antennas, the optimization of the base - station compensation cost in the V2X network is achieved, and the transmission rate of vehicle users is improved while the transmission delay is reduced.

[0043] Embodiment 2:;

[0044] Figure 2It is a schematic diagram of a cost and antenna position optimization system in a vehicle-to-everything network based on a privacy-preserving federated learning architecture with intelligent movable antennas. As Figure 2 shown, this embodiment provides a cost and antenna position optimization system in a vehicle-to-everything network based on a privacy-preserving federated learning architecture with intelligent movable antennas. The system includes: A construction module 201, configured to construct a vehicle-to-everything network model based on a privacy-preserving federated learning architecture with intelligent movable antennas, and establish an objective function for minimizing the base station compensation cost according to the historical status data of vehicle users; the historical status data includes the communication cost, local computing cost, privacy cost of vehicle users, and the distance between vehicle users and the base station; A vehicle user selection module 202, configured to obtain the vehicle users selected by the base station, solve and optimize the parameters of the objective function by using a greedy vehicle user selection algorithm, obtain a vehicle user selection model, and use the result for the antenna position scheduling module; An antenna position optimization module 203, configured to obtain the currently optimized antenna position, solve and optimize the parameters of the objective function by using an adaptive particle swarm optimization algorithm and the result of the vehicle user selection module, and obtain the optimized antenna position.

[0045] Preferably, the construction of the objective function for minimizing the base station compensation cost by the construction module 201 includes: Establish the objective function of the base station compensation cost, and the optimization problem is as follows: ; ; Preferably, when constructing the vehicle user selection module, the objective function will be modified to: ; ; Solve the optimization problem through a greedy vehicle user selection algorithm, and obtain a vehicle user selection model.

[0046] Preferably, when constructing the antenna position optimization module, the objective function will be modified to: ; ; Solve the optimization problem through an adaptive particle swarm optimization algorithm, and obtain an optimized antenna position model.

[0047] The specific implementation process of the functions implemented by each module in this embodiment 2 is the same as that in embodiment 1, and will not be elaborated here.

[0048] The experimental results show that the algorithm in this paper exhibits excellent performance in terms of both the base station compensation cost and the average transmission delay metric.

Claims

1. Optimization method for vehicle-to-everything (V2X) with movable antennas considering the cost of antenna positions, characterized in that The method includes the following steps: S1: Construct a vehicle-to-everything (V2X) network model based on a privacy-preserving federated learning architecture with intelligent movable antennas. According to the total cost of vehicle users, which consists of three parts: communication cost, computing cost, and privacy cost, the base station gives corresponding compensation, and determines the vehicle users within the signal coverage of the base station according to the vehicle mobility model, and establishes an objective function to minimize the total cost of the base station; S2: Decompose the target problem into two sub-problems and use the greedy vehicle user selection algorithm and the adaptive particle swarm optimization algorithm to solve the sub-problems respectively, and obtain the set of vehicle users selected by the base station and the positions of the intelligent movable antennas; S3: Obtain the set of vehicle users selected by the federated communication base station in each round and the positions of the antennas after scheduling, calculate the compensation cost of the base station in the current round, and output the optimized compensation cost of the base station.

2. The optimization method based on the vehicle-to-everything (V2X) with movable antennas and including antenna position cost according to claim 1, wherein: The establishment of the objective function to minimize the total cost of the base station in step S1 includes: The establishment of the objective function to minimize the total cost of the base station includes three parts of costs, namely communication cost, computing cost, and privacy cost. The optimization problem is as follows: ； ； Among them, represents the set of vehicle users selected, represents the set of intelligent movable antennas; represents the total set of vehicle users; represents a vehicle user, represents whether the vehicle user is selected; represents the total compensation of the base station to the vehicle users; represents the number of intelligent movable antennas; respectively represent the th and th antennas. A represents the antenna system, which consists of a two-dimensional coordinate system, and respectively represent the th antenna and the th antenna's positions in the two-dimensional coordinate system of the antenna system A, represents the minimum distance between two adjacent antennas; represents the total time delay of the vehicle users, is the maximum time delay that the base station can tolerate; represents the distance traveled by the vehicle user, represents the radius of the base station signal coverage range, represents the cosine angle between the base station coverage range and the road, represents the distance between the vehicle user and the entrance of the base station coverage range; is the magnitude of the noise variance added by the vehicle user to the updated local model, represents the threshold of the total noise variance magnitude of the base station for the selected set of vehicle users.

3. The optimization method based on a movable antenna vehicle network with antenna position cost according to claim 2, wherein: The communication cost of the objective function to minimize the total cost of the base station is expressed as: ； is the transmission rate of the vehicle user, is the transmission power of the vehicle user, is the hybrid vector received by the base station from the vehicle user, is the Gaussian white noise in the channel transmission process; is the transmission delay of the vehicle user, is the size of the data volume uploaded by the vehicle user; is the transmission energy consumption of the vehicle user, is the transmission cost of the vehicle user, is the unit cost of the transmission energy consumption.

4. The optimization method based on the vehicle networking with movable antennas and including the antenna position cost according to claim 2, wherein: The computing cost of the objective function to minimize the total cost of the base station is expressed as: ； The computing delay for vehicle users, The number of CPU cycles required for vehicle users to process a single sample, The size of the local dataset of vehicle users, The CPU frequency of local computing for vehicle users; The computing energy consumption of vehicle users, The CPU energy consumption of local computing for vehicle users; The computing cost of vehicle users, The unit cost of computing energy consumption.

5. The optimization method based on a vehicle-to-everything network with a movable antenna and including the cost of antenna position according to claim 2, wherein: The privacy cost of the objective function to minimize the total cost of the base station is expressed as: ； represents the privacy cost of vehicle users, is the sensitivity of two adjacent data sets in differential privacy, represents the unit cost of the noise energy consumption added by vehicle users to the local model, represents the privacy budget of vehicle users.

6. The optimization method based on the vehicle-to-everything (V2X) with movable antennas and including antenna position cost according to claim 1, wherein: The vehicle mobility model in step S1 is expressed as: ； Represents the distance traveled by the vehicle user, is the driving speed of the vehicle user, Represents the transmission delay of the vehicle user, Represents the computing delay of the vehicle user, Represents the distance traveled by the vehicle user after uploading the updated local model, R Represents the radius of the base station signal coverage area, Represents the cosine angle between the base station coverage area and the road, Represents the distance between the vehicle user and the entrance of the base station coverage area, Represents the distance between the vehicle user and the base station.

7. The optimization method based on a vehicle-to-everything (V2X) network with a movable antenna and including the cost of antenna position according to claim 1, wherein: The greedy vehicle user selection algorithm in the two sub-problems obtained by decomposing the target problem in step S2 is expressed as: ； ； Among them, represents the set of vehicle users selected; V represents the total set of vehicle users; represents a vehicle user, indicates whether the vehicle user is selected; represents the total compensation of the base station to the vehicle users; Z represents the number of intelligent movable antennas; represents the total time delay of the vehicle users, is the maximum time delay tolerated by the base station; represents the distance traveled by the vehicle user, R represents the radius of the signal coverage range of the base station, represents the cosine angle between the base station coverage range and the road, represents the distance between the vehicle user and the entrance of the base station coverage range; is the magnitude of the noise variance added by the vehicle user to the updated local model, represents the threshold of the total noise variance magnitude of the base station for the selected set of vehicle users; Solve it through the greedy vehicle user selection algorithm. The specific steps include: First, the base station determines the set of vehicle users that meet the delay constraint and are within the signal coverage of the base station. Then, all vehicle users in this set calculate their respective communication costs, computing costs, and privacy costs, calculate the total cost, and upload it to the base station. The base station sorts the vehicle users in the set according to the total cost uploaded from low to high. The base station starts selecting from the vehicle user with the lowest total cost until the threshold of the sum of noises is met. Finally, the base station obtains the subset of vehicle users selected in the current round.

8. The optimization method based on a vehicle-to-everything network with a movable antenna and including the cost of antenna position according to claim 1, wherein: The adaptive particle swarm optimization algorithm in the two sub-problems obtained by decomposing the target problem in step S2 is expressed as: ； ； Among them, represents the set of intelligent movable antennas; C represents the subset of vehicle users obtained by the greedy vehicle user selection algorithm; represents the vehicle user, represents the total compensation of the base station to the vehicle users; represents the number of intelligent movable antennas; respectively represent the th and th antennas. A represents the antenna system, which consists of a two-dimensional coordinate system, and respectively represent the th antenna and the th antenna positions in the two-dimensional coordinate system of the antenna system A, represents the minimum distance between two adjacent antennas; Solve this sub-problem through the adaptive particle swarm optimization algorithm. The specific steps are as follows: The first step is to initialize the velocity and position of the particles: ； Among them represents the set of initial positions of the particle swarm, represents a particle 's initial position; represents the set of initial velocities of the particle swarm, represents a particle 's initial velocity. There are altogether I particles, among which is further expressed as: ； Among them and represent the initial positions of the stored particles in the coordinate system of the n-th antenna The feasible region of the intelligent movable antenna is denoted as A, satisfying , so the positions of the particles and the antenna need to satisfy the following: ； Among them represents the position of the th antenna stored by the particle on the axis in the coordinate system, represents the position of the th antenna stored by the particle on the axis in the coordinate system, H represents the radius of the antenna system; The second step is to update the particle velocity and position: During the process of the adaptive particle swarm optimization algorithm, in each round of iteration, the particle swarm will update their respective positions and velocities, which are expressed as follows: ； Among them represents the current iteration round of the particle swarm algorithm, represents the global optimal position of the individual particle swarm, represents the global optimal position of the particle swarm, represents the round of particles speed, represents the round when the particle position, represents the round when the particle position, is the inertia factor of the particle, and is the learning factor of the round of particles, and is a random number satisfying ; The third step is to adaptively update the inertia factor: Currently, the more popular way to update the inertia factor is to adaptively update the inertia factor in each round of iteration, which is expressed as follows: ； wherein is the inertia factor of the particle, represents the maximum inertia weight of the particle, represents the minimum inertia weight of the particle, represents the current iteration round of the particle swarm algorithm, M represents the total iteration rounds of the adaptive particle swarm algorithm; The fourth step is to adaptively update the learning factor: The update of the learning factor is based on the distance of the current particle from the global optimal position of its individual particle swarm and the global optimal position of the particle swarm The particles farther away from the two optimal positions should obtain a larger learning factor to approach their optimal positions faster. The adaptive weighted update rule is expressed as follows: ； Among them and are the learning factors of the round of particles. The function F represents the adaptive weighted update function. and respectively represent the distances of the particle in the round from its individual optimal position and from the global optimal position. Specifically, and are expressed as follows: ； Use a piecewise linear function as the adaptive weighted update function, which is specifically expressed as follows: ； wherein and respectively represent the distances of the particle in the th round from its individual optimal position and from the global optimal position of the population, represents the global optimal position of the individual particle swarm, represents the global optimal position of the particle swarm, represents the th round, and represents the position of the particle and are parameters for controlling the learning factor and satisfy ; represents the distance between the particle and the optimal position, represents the threshold of the current particle's distance from the optimal position; The fifth step is to set the penalty function and the fitness function: Considering the constraints of this sub-problem, a penalty function is set up to penalize the particles that violate the constraints. The penalty function is specifically expressed as follows: ； wherein is the penalty factor, represents the penalty function of the particle, is the number of particles violating the constraints; Combined with the penalty function, the fitness function of the particle is expressed as follows: ； wherein represents the fitness function of the particles C represents the subset of vehicle users obtained by the greedy vehicle user selection algorithm represents the vehicle users represents the total compensation of the base station to the vehicle users represents the penalty function 9. A system using the optimization method for a vehicle-to-everything (V2X) network with movable antennas and including antenna position costs according to any one of claims 1 to 8, characterized in that: The system includes: A building module for building a vehicle-to-everything (V2X) network model based on an intelligent movable antenna-assisted privacy-preserving federated learning architecture, and establishing an objective function for minimizing the base station compensation cost according to the historical status data of vehicle users; the historical status data includes the communication cost, local computing cost, privacy cost of vehicle users, and the distance between vehicle users and the base station. A vehicle user selection module for obtaining the vehicle users selected by the base station, solving and optimizing the parameters of the objective function by using a greedy vehicle user selection algorithm, obtaining a vehicle user selection model, and using the result for the antenna position scheduling module. An antenna position optimization module for obtaining the currently optimized antenna position, solving and optimizing the parameters of the objective function by using an adaptive particle swarm optimization algorithm and the result of the vehicle user selection module, and obtaining the optimized antenna position.

Images