Relay selection and resource allocation method in network connection vehicle cooperation system

Abstract

The invention belongs to the technical field of Internet of Vehicles communication, and particularly relates to a joint resource allocation and relay selection optimization method of a networked vehicle cooperation system. The method comprises the following steps: firstly, constructing a double-hop cooperative relay model from a wireless access point to a vehicle and from the vehicle to the vehicle; secondly, considering constraints such as time delay of each link, and establishing an optimization model taking minimization of end-to-end time delay as a target; then, a double-layer optimization method combining resource allocation and relay selection is provided; the inner layer decouples an original problem into two sub-problems under any given relay strategy and solves a resource allocation problem by using a convex optimization method; the outer layer encodes each chromosome into a relay selection strategy through an improved genetic algorithm, and the output of the inner layer serves as a fitness value to solve the relay selection problem. According to the method, the end-to-end delay performance of the system is guaranteed, and meanwhile, the influence of the delay performance of each link on the system is considered, so that the reliability and the robustness of the network connection vehicle cooperation system in the urban dense scene are effectively improved.

Description

Technical Field

[0001] This invention belongs to the field of vehicle-to-everything (V2X) communication technology, and specifically relates to a resource allocation and relay selection method in a connected vehicle cooperative system. Background Technology

[0002] Intelligent Connected Vehicles (ICVs), as core units of future intelligent transportation systems, rely on efficient and reliable communication capabilities as a crucial foundation for achieving collaborative perception, decision-making, and control among vehicles. Vehicle-to-Everything (V2X) communication supports information interaction between intelligent connected vehicles and the outside world. V2I (Vehicle-to-Infrastructure) communication enables vehicles to access roadside units or cellular base stations to obtain network information and services, while V2V (Vehicle-to-Vehicle) communication directly enables information sharing between vehicles. However, the transmission environment is complex and variable, with issues such as non-line-of-sight propagation due to building obstruction and communication link instability. [2] This severely restricts the performance of traditional vehicle-to-everything (V2X) communication, threatening driving safety and system reliability. Traditional point-to-point communication methods are insufficient to meet the communication needs of V2X systems. Relay cooperation technology, however, can effectively overcome non-line-of-sight propagation limitations, extend communication range, improve link reliability and system capacity, and is a key approach to solving the challenges of V2X communication in densely populated cities. Furthermore, in V2X relay cooperation systems, the optimal selection of relay nodes is a prerequisite for building efficient and reliable communication links. Due to factors such as real-time changes in wireless channel conditions, selecting the optimal relay to maximize system performance becomes a critical issue.

[0003] Existing research has proposed diverse relay selection strategies for different performance goals and constraints. These include focusing on the reliability, real-time performance, and capacity performance of communication systems, aiming to improve related system performance through different relay selection schemes. Furthermore, many studies investigate the impact of resource allocation and relay selection on relay system performance, including using various techniques such as deep learning algorithms, machine learning algorithms, and convex optimization methods to jointly optimize system performance through resource allocation and relay selection.

[0004] Existing research mainly focuses on optimizing the relay selection strategy itself or the overall system delay performance, but it does not consider the impact of the performance of each link on the overall system end-to-end delay. This is because in relay systems with strong fading channels, multiple links with poor delay performance often affect the system's performance. Furthermore, traditional single relay selection methods or resource allocation methods cannot sufficiently improve the system's end-to-end delay performance.

[0005] In summary, researching end-to-end latency optimization in relay cooperative networks through joint resource allocation and relay selection is of great significance. The core idea is to use dynamic resource allocation methods, such as power control and bandwidth allocation, to prioritize the allocation of bandwidth or power resources to vehicle links with poor latency performance, thereby reducing their link latency and ultimately lowering the system's end-to-end latency, while meeting resource constraints and link latency requirements. Furthermore, by employing a reasonable relay selection method, the target vehicle is matched with the relay vehicle that minimizes end-to-end latency, resulting in a globally minimized system end-to-end latency. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention proposes a resource allocation and relay selection method for connected vehicle cooperative systems, comprising:

[0007] S1. Construct a two-hop V2I-V2V cooperative relay model from access point to vehicle and from vehicle to vehicle;

[0008] S2. Taking into account the latency requirements of each link, relay selection strategy and resource constraints, an optimization model with end-to-end latency as the objective was established.

[0009] S3. The problem is modeled as a mixed integer nonlinear programming problem. By analyzing the coupling relationship of variables, the problem is decomposed into two sub-problems. A two-layer genetic algorithm that combines resource allocation and relay selection is proposed.

[0010] S4. The inner layer solves the resource allocation problem. Under any given relay strategy, the convex optimization method is used to find the minimum end-to-end delay and the corresponding resource allocation scheme under the strategy.

[0011] S5. The outer layer solves the relay selection problem by encoding each chromosome as a relay selection strategy through an improved genetic algorithm, and uses the output of the inner layer as the fitness value. Finally, it outputs the optimal relay selection strategy and the global minimum of end-to-end delay under the optimal relay selection.

[0012] Furthermore, in step S1, the model building and link modeling are carried out in the following specific steps:

[0013] S11. To address the issues of strong fading and coverage blind spots in the communication link between the wireless access point and the vehicle, a two-hop relay transmission model is constructed. The transmission model includes M target vehicle nodes and K relay vehicle nodes, with the target vehicle set represented by [missing information]. This indicates that a connection to an Access Point (AP) is required to obtain communication information. However, due to obstruction between the target vehicle and the AP, resulting in poor signal quality, relay vehicles are needed to cooperate in acquiring information for each target vehicle. Simultaneously, the relay vehicles also need to receive information sent to them by the AP. The AP needs to transmit data belonging to the target vehicle through the selected relay vehicles. The relay vehicle set is used for... This indicates that M < K.

[0014] S12. Since large-scale fading components generally depend on vehicle location and undergo slow scale changes, the base station can acquire large-scale fading information for all communication links. Assuming that the access point and base station can share information through an interface, for small-scale fading of different links, the resource scheduling cycle is less than the channel coherence time caused by vehicle mobility, i.e., the impact of vehicle mobility is ignored. Therefore, the channel gain between the k-th relay vehicle and the m-th target vehicle is modeled as follows:

[0015]

[0016] in h represents the large-scale slow fading channel gain. k,m Let κ represent the small-scale fast fading power component, and D represent the path loss constant. k,m The distance between the k-th relay vehicle and the m-th target vehicle is represented by ε, where ε represents the channel fading index. Indicates that it has a standard deviation The log-normal shaded fading random variable. The entire communication transmission is divided into two links and uses FDMA (Frequency Division Multiple Access) to complete the communication between the AP and the target vehicle. Specifically, the V2I link channel gain from the AP to the kth relay vehicle is denoted by H. k H represents the channel gain of the V2V link from the k-th relay vehicle to the m-th target vehicle. k,mThe relay vehicle uses a decoding-forwarding protocol for data forwarding. The delays of the two links are the V2I link delay and the V2V link delay, denoted by τ1 and τ2 respectively. Their physical meanings are the maximum transmission delay between all APs and relay vehicles and the maximum transmission delay between the relay vehicle and the target vehicle, respectively. Furthermore, considering the current shortage of spectrum resources in dense areas, spectrum reuse technology is introduced in the relay transmission of the second link to improve spectrum resource utilization. Since uplink resources are used less densely and interference at the base station is easier to manage, the orthogonally allocated uplink spectrum of cellular user equipment (CUE) is reused by all relay V2V links, which will introduce co-channel interference. The channel gain between the k-th relay vehicle and the base station when the k-th relay vehicle interferes with the CUE is H. k,B The channel gain between the CUE and the base station is H. c The channel gain between CUE and the m-th target vehicle is H. m .

[0017] S13. To ensure that each target vehicle obtains AP information by selecting relay vehicles, a relay selection strategy matrix x is defined. This matrix is ​​a K×M binary matrix, where each element x... k,m ∈{0,1} is the relay selection factor, representing the service relationship between the relay vehicle and the target vehicle, defined as x k,m =1 indicates that the k-th relay vehicle serves the m-th target vehicle. Considering that each relay vehicle can serve one and at most one target vehicle, the following condition must be met:

[0018]

[0019] Assuming each target vehicle has exactly one relay vehicle serving it, the following conditions must be met:

[0020]

[0021] S14. Consider the downlink of access point information based on FDMA for V2I links. Given a relay selection strategy x, let k... m (x) represents the relay vehicle selected by the m-th target vehicle. The achievable downlink data rate is then given by the following formula.

[0022]

[0023] in This represents the transmit power allocated by the AP to the m-th target vehicle, and N0 represents the power spectral density of the noise. This represents the bandwidth allocated by the AP to the m-th target vehicle.

[0024] S15. If the relay selection factor x is not given, the achievable downlink data rate is given by the following formula.

[0025]

[0026] S16. Let... This represents the amount of data that the AP sends to the k-th relay vehicle. This represents the amount of data that the AP sends to the m-th target vehicle. Let the total amount of data to be sent be... Considering the poor performance of V2I links, the transmission delay between each AP and the relay vehicle is specified to have an upper limit τ1, which must satisfy...

[0027]

[0028] S17. Next, consider FDMA-based relay transmission on V2V links. If, under a given relay selection strategy x, from the k-th... m The achievable V2V link data rate from (x) relay vehicles to the m-th target vehicle is given by the following formula.

[0029]

[0030] in This indicates the interference of other relay vehicles reusing the CUE channel with the communication of the m-th target vehicle. This indicates that the CUE interferes with the same lane of the m-th target vehicle. This indicates the bandwidth of the CUE that has been reused. This represents the transmit power allocated by the relay vehicle to the m-th target vehicle. Furthermore, given the maximum interference power threshold I of the V2V link to the CUE uplink... th,max The minimum signal-to-interference-plus-noise ratio (SINNR) γ0 of the uplink can be used to calculate the CUE's transmit power to satisfy the given condition.

[0031]

[0032] S18. If the relay selection factor x is not given, the achievable V2V link data rate is given by the following formula.

[0033]

[0034] in

[0035] S19. Considering the data transmission loss during decoding and forwarding, let... This represents the amount of data that needs to be transmitted to the m-th relay vehicle. Considering the poor performance of V2V links, the transmission delay between each relay vehicle and the target vehicle is defined to have an upper bound τ2, which must satisfy...

[0036]

[0037] Preferably, in step S2, the optimization objective is to minimize the end-to-end delay T. E2E Defined as the sum of V2I link delay τ1 and V2V link delay τ2, it needs to be optimized through joint optimization of the relay selection strategy matrix x and the V2I link transmission bandwidth allocation vector B. (1) Send power allocation vector P (1) V2V link transmit power allocation vector P (2) And the time delay vector τ. Where B (1) Represents a set of K×1 dimensional variables P (1) Represents a K×1 set of variables P (2) Represents an M×1 set of variables And τ represents a 2×1 set of variables {τ1, τ2}. And define... This is the maximum bandwidth that the AP can allocate to the cooperative relay system. This is the AP's maximum transmit power. Let m be the maximum transmission power of the m-th relay vehicle. Let be the minimum transmission rate of the V2V link. The optimization problem is formulated as a resource allocation and relay selection problem that minimizes end-to-end latency, as described below:

[0038]

[0039] Preferably, in step S3, under any given relay selection strategy x, the original problem can be equivalently decoupled into two parallel subproblems: joint bandwidth allocation and power control for V2I links, and power allocation and interference management for V2V links, as shown below.

[0040]

[0041] C4,C5

[0042]

[0043] C7,C9

[0044]

[0045] In the inner layer, for any fixed relay selection strategy x, by solving subproblems P2 and P3, the minimum delay of the V2I link under the specific relay selection strategy x can be obtained as follows: And the minimum latency of the V2V link is Furthermore, suppose that, given x, the optimal solutions for P2 and P3 are... Then the minimum end-to-end latency that the system can achieve is This minimum value is relative to all possible resource allocation schemes for a fixed x. The goal of the outer improved genetic algorithm is to search for the globally optimal relay selection strategy x in all possible solution spaces of x. * This ensures that the minimum end-to-end latency generated by it, after optimal resource allocation within the inner layer, reaches the global minimum, i.e. and It is the global minimum end-to-end delay that problem P1 requires.

[0046] Preferably, in step S4, for subproblem P2, based on the determined relay selection factor x, the optimal solution to the resource allocation problem can be obtained as follows: and And the minimum latency of the V2I link is For subproblem P3, based on a determined relay selection factor x, the optimal solution to the resource allocation problem can be obtained as follows: The minimum latency of a V2V link is The minimum end-to-end delay can then be obtained as follows:

[0047] Preferably, in step S5, based on the inner optimization framework established in S4, the core task of the outer algorithm is to efficiently search for the optimal relay selection strategy that minimizes end-to-end latency in the discrete relay selection strategy space. Since the relay selection variables are discrete, and the strategy space grows exponentially with the number of vehicles and relay nodes, traditional exhaustive search or gradient methods are difficult to apply. Therefore, this invention uses an improved genetic algorithm as the outer optimization engine, which has the advantages of parallel search, avoiding getting trapped in local optima, and effectively handling combinatorial optimization problems.

[0048] The beneficial effects of this invention are:

[0049] Unlike traditional methods that minimize the total transmission latency for all users, this invention focuses on end-to-end latency, which is obtained by summing the V2V and V2I link delays and is constrained by the worst-case link latency. The proposed method minimizes the end-to-end latency under any relay selection strategy through a fair resource allocation approach, avoiding the unfair phenomenon in traditional methods where resource allocation tends to favor the central vehicle in order to minimize the total latency for all users. Simultaneously, this invention reuses the uplink spectrum resources of cellular users for V2V links, improving system resource utilization. Overall, this invention effectively improves the reliability and real-time performance of the system. Attached Figure Description

[0050] Figure 1 This is the V2I-V2V cooperative relay communication model in this embodiment of the invention;

[0051] Figure 2 This is a flowchart illustrating the steps in an embodiment of the present invention;

[0052] Figure 3 This is a flowchart of the joint resource allocation and relay selection method in an embodiment of the present invention. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] This invention provides a joint optimization method for resource allocation and relay selection in V2I-V2V two-hop relay cooperative networks. The invention defines the target scenario, models the communication links between entities, lists and solves the optimization problem, and presents a reliable and real-time joint optimization method for resource allocation and relay selection. The process specifically includes the following steps:

[0055] S1. Construct a two-hop V2I-V2V cooperative relay model from access point to vehicle and from vehicle to vehicle;

[0056] S2. Taking into account the latency requirements of each link, relay selection strategy and resource constraints, an optimization model with end-to-end latency as the objective was established.

[0057] S3. The problem is modeled as a mixed integer nonlinear programming problem. By analyzing the coupling relationship of variables, the problem is decomposed into two sub-problems. A two-layer genetic algorithm that combines resource allocation and relay selection is proposed.

[0058] S4. The inner layer solves the resource allocation problem. Under any given relay strategy, the convex optimization method is used to find the minimum end-to-end delay and the corresponding resource allocation scheme under the strategy.

[0059] S5. The outer layer solves the relay selection problem by encoding each chromosome as a relay selection strategy through an improved genetic algorithm, and uses the output of the inner layer as the fitness value. Finally, it outputs the optimal relay selection strategy and the global minimum of end-to-end delay under the optimal relay selection.

[0060] In the embodiments provided by this invention, considering the relay cooperative transmission in the vehicle-to-everything (V2I) system, a V2I-V2V two-hop relay cooperative system model is constructed, as shown in the attached figure. Figure 1The communication entities constructed in this method include an access point (AP), a relay vehicle, a target vehicle, a cellular user, and a base station. The communication flow direction is from the AP to the relay vehicle, and then from the relay vehicle to the target vehicle. A relay selection strategy is required to select the relay vehicle, and the spectrum resources of the link between the relay vehicle and the target vehicle reuse the uplink from the cellular user to the base station. Therefore, communication between the relay vehicle and the target vehicle will interfere with communication between the cellular user and the base station, and the uplink signal of the cellular user will also interfere with communication between the relay vehicle and the target vehicle. The communication links involved in this embodiment and their physical meanings include:

[0061] The V2I link between the access point and the relay vehicle carries information sent from the access point to the target vehicle and the relay vehicle.

[0062] The V2V link between the relay vehicle and the target vehicle carries information forwarded by the relay vehicle to the target vehicle.

[0063] The uplink between cellular users and base stations carries communication information between them.

[0064] Interference links between cellular users and target vehicles represent interference from cellular users to target vehicles caused by spectrum reuse.

[0065] Interference links between relay vehicles and base stations represent interference caused by spectrum reuse on the base station side by relay vehicles to cellular users.

[0066] The model building and link modeling in step S1 are as follows:

[0067] S11. To address the issues of strong fading and coverage blind spots in the communication link between the wireless access point and the vehicle, a two-hop relay transmission model is constructed (see Appendix). Figure 1 The transmission model has M target vehicle nodes and K relay vehicle nodes. The set of target vehicles is represented by... This indicates that a connection to a wireless access point (AP) is required to obtain communication information. However, due to obstruction between the target vehicle and the AP, resulting in poor signal quality, relay vehicles are needed to cooperate in acquiring information for each target vehicle. Simultaneously, the relay vehicles also need to receive information sent to them by the AP. The AP needs to transmit data belonging to the target vehicles through the selected relay vehicles. The relay vehicle set is used for... This indicates that M < K.

[0068] S12. Since large-scale fading components generally depend on vehicle location and undergo slow scale changes, the base station can acquire large-scale fading information for all communication links. Assuming that the access point and base station can share information through an interface, for small-scale fading of different links, the resource scheduling cycle is less than the channel coherence time caused by vehicle mobility, i.e., the impact of vehicle mobility is ignored. Therefore, the channel gain between the k-th relay vehicle and the m-th target vehicle is modeled as follows:

[0069]

[0070] in h represents the large-scale slow fading channel gain. k,m Let κ represent the small-scale fast fading power component, and D represent the path loss constant. k,m The distance between the k-th relay vehicle and the m-th target vehicle is represented by ε, where ε represents the channel fading index. Indicates that it has a standard deviation The log-normal shaded fading random variable. The entire communication transmission is divided into two links and uses FDMA (Frequency Division Multiple Access) to complete the communication between the AP and the target vehicle. Specifically, the V2I link channel gain from the AP to the kth relay vehicle is denoted by H. k H represents the channel gain of the V2V link from the k-th relay vehicle to the m-th target vehicle. k,m The relay vehicle uses a decoding-forwarding protocol for data forwarding. The delays of the two links are the V2I link delay and the V2V link delay, denoted by τ1 and τ2 respectively. Their physical meanings are the maximum transmission delay between all APs and relay vehicles and the maximum transmission delay between the relay vehicle and the target vehicle, respectively. Furthermore, considering the current shortage of spectrum resources in dense areas, spectrum reuse technology is introduced in the relay transmission of the second link to improve spectrum resource utilization. Since uplink resources are used less densely and interference at the base station is easier to manage, the orthogonally allocated uplink spectrum of cellular user equipment (CUE) is reused by all relay V2V links, which will introduce co-channel interference. The channel gain between the k-th relay vehicle and the base station when the k-th relay vehicle interferes with the CUE is H. k,B The channel gain between the CUE and the base station is H. c The channel gain between CUE and the m-th target vehicle is H. m .

[0071] S13. To ensure that each target vehicle obtains AP information by selecting relay vehicles, a relay selection strategy matrix x is defined. This matrix is ​​a K×M binary matrix, where each element x... k,m∈{0,1} is the relay selection factor, representing the service relationship between the relay vehicle and the target vehicle, defined as x k,m =1 indicates that the k-th relay vehicle serves the m-th target vehicle. Considering that each relay vehicle can serve one and at most one target vehicle, the following condition must be met:

[0072]

[0073] Assuming each target vehicle has exactly one relay vehicle serving it, the following conditions must be met:

[0074]

[0075] S14. Consider the downlink of access point information based on FDMA for V2I links. Given a relay selection strategy x, let k... m (x) represents the relay vehicle selected by the m-th target vehicle. The achievable downlink data rate is then given by the following formula.

[0076]

[0077] in This represents the transmit power allocated by the AP to the m-th target vehicle, and N0 represents the power spectral density of the noise. This represents the bandwidth allocated by the AP to the m-th target vehicle.

[0078] S15. If the relay selection factor x is not given, the achievable downlink data rate is given by the following formula.

[0079]

[0080] S16. Let... This represents the amount of data that the AP sends to the k-th relay vehicle. This represents the amount of data that the AP sends to the m-th target vehicle. Let the total amount of data to be sent be... Considering the poor performance of V2I links, the transmission delay between each AP and the relay vehicle is specified to have an upper limit τ1, which must satisfy...

[0081]

[0082] S17. Next, consider FDMA-based relay transmission on V2V links. If, under a given relay selection strategy x, from the k-th... m The achievable V2V link data rate from (x) relay vehicles to the m-th target vehicle is given by the following formula.

[0083]

[0084] in This indicates the interference of other relay vehicles reusing the CUE channel with the communication of the m-th target vehicle. This indicates that the CUE interferes with the same lane of the m-th target vehicle. This indicates the bandwidth of the CUE that has been reused. This represents the transmit power allocated by the relay vehicle to the m-th target vehicle. Furthermore, given the maximum interference power threshold I of the V2V link to the CUE uplink... th,max The minimum signal-to-interference-plus-noise ratio (SINNR) γ0 of the uplink can be used to calculate the CUE's transmit power to satisfy the given condition.

[0085]

[0086] S18. If the relay selection factor x is not given, the achievable V2V link data rate is given by the following formula.

[0087]

[0088] in

[0089] S19. Considering the data transmission loss during decoding and forwarding, let... This represents the amount of data that needs to be transmitted to the m-th relay vehicle. Considering the poor performance of V2V links, the transmission delay between each relay vehicle and the target vehicle is defined to have an upper bound τ2, which must satisfy...

[0090]

[0091] The optimization problem and mathematical model establishment in step 2 are as follows:

[0092] S21. By integrating the above two links, the end-to-end transmission delay for completing all information transmission is given by the following formula.

[0093] T E2E =τ1+τ2

[0094] S22. The optimization objective is to minimize the end-to-end delay T. E2E This requires joint optimization of the relay selection strategy matrix x and the V2I link transmission bandwidth allocation vector B. (1) Send power allocation vector P (1) V2V link transmit power allocation vector P (2) And the time delay vector τ. Where B (1) Represents a set of K×1 dimensional variables

[0095] P (1) Represents a K×1 set of variables P (2) Represents an M×1 set of variables And τ represents a 2×1 set of variables {τ1, τ2}. The optimization problem is formulated as a resource allocation and relay selection problem that minimizes end-to-end latency, as described below:

[0096]

[0097] S23. Definition This is the maximum bandwidth that the AP can allocate to the cooperative relay system. This is the AP's maximum transmit power. Let m be the maximum transmission power of the m-th relay vehicle. The minimum transmission rate for the V2V link is defined by constraints C1 and C2, which are two relay selection factor constraints. Constraint C3 states that the transmission delay between each vehicle and the AP in the V2I link should not exceed the V2I link delay τ1. Constraint C6 represents that the transmission delay between each vehicle in the V2V link should not exceed the V2V link delay τ2. Combining this with the optimization objective of minimizing the sum of τ1 and τ2, this is essentially a minimax problem coupled through the relay selection strategy x. These two constraints represent the goal of minimizing the maximum delays τ1 and τ2 of each link through fair resource allocation across different links, even if it requires sacrificing local performance by transferring resources to higher-latency links to achieve delay balance. Constraints C4, C5, and C7 are resource total constraints, while constraints C8 and C9 are reliability constraints. This means that the interference levels caused by spectrum reuse, whether for the uplink of cellular users or relay links, must be within certain limits. Cellular user reliability is ensured through interference limiting at the base station, while relay link rate constraints reduce interference from cellular users to target vehicles.

[0098] The problem decomposition, joint resource allocation, and relay selection methods in step 3 are described in detail below:

[0099] S31. From the optimization problem, it can be seen that V2I link delay and V2V link delay are coupled together through the relay selection strategy x. This coupling makes the relay selection and resource allocation problems more difficult to handle. Furthermore, due to the binary constraints in the problem, it is a combinatorial mixed-integer nonlinear problem, and is non-convex NP-hard. Therefore, this invention proposes an improved two-layer distributed genetic algorithm. In this method, B is first calculated for any given x. (1) P (1) P (2) And τ, and then, instead of exhaustively listing all possible cases of x, this invention chooses to use an improved genetic algorithm to globally optimize x.

[0100] S32. Specifically, under any given relay selection strategy x, the original problem can be equivalently decoupled into two parallel subproblems: joint bandwidth allocation and power control for V2I links, and power allocation and interference management for V2V links, as shown below.

[0101]

[0102] C4,C5

[0103]

[0104] C7,C9

[0105]

[0106] S33. In the inner layer, for any fixed relay selection strategy x, by solving subproblems P2 and P3, the minimum delay of the V2I link under the specific relay selection strategy x can be obtained as follows: And the minimum latency of the V2V link is Furthermore, suppose that, given x, the optimal solutions for P2 and P3 are... Then the minimum end-to-end latency that the system can achieve is This minimum value is relative to all possible resource allocation schemes for a fixed x. The goal of the outer improved genetic algorithm is to search for the globally optimal relay selection strategy x in all possible solution spaces of x. * This ensures that the minimum end-to-end latency generated by it, after optimal resource allocation within the inner layer, reaches the global minimum, i.e. and This refers to the globally minimum end-to-end latency required by problem P1. See the attached flowchart for the method. Figure 3 .

[0107] The solution for the inner layer fitness value in step 4 is as follows:

[0108] S41. Solve problem P2 using convex optimization methods.

[0109] S411. Since constraints C4 and C5 in problem P2 are both linear constraints, they are also convex constraints. For constraint C3, generally, it is observed that... It is about The concave function, and It is the perspective function of g(·). From this, we can derive... It is about and The concave function.

[0110] S412. Problem P2 is a convex problem that can be solved using the Lagrange dual method. To solve the given convex optimization problem P2 using the Lagrange dual method, the augmented Lagrangian function is constructed as follows:

[0111]

[0112] Where λ m μ, ν are non-negative Lagrange multipliers. The dual problem of problem P2 is:

[0113]

[0114] The dual function is:

[0115]

[0116] S413. Since problem P2 satisfies the Slater condition, P2 and its dual problem P2.1 have strong duality. To solve P2, this invention first uses a given λ m Solve P2.2 to obtain D({λ, μ, ν). m},μ,ν), then through the condition {λ m To maximize D({λ ≥ 0}, μ ≥ 0, and ν ≥ 0 m},μ,ν).

[0117] First, given {λ m Given that μ≥0 and ν≥0, problem P2.2 can be divided into the following subproblems.

[0118]

[0119] S414. For subproblem P2.3, it is easy to obtain the optimal solution for end-to-end delay as follows:

[0120]

[0121] S415. For subproblem P2.4, this problem is convex and satisfies the Slater condition. Therefore, it can be solved by applying the Karush-Kuhn-Tucker condition. The Lagrange equation for the m-th subproblem is given below.

[0122]

[0123] L can be obtained m about and The first derivative is shown in the following equation.

[0124]

[0125] From the above equation, we can obtain that given {λ} m Under the conditions ≥0}, μ≥0, and ν≥0, we can solve for...

[0126]

[0127] Where, [x] + =max{0,x}, which can be used to modify problem P2.4.

[0128]

[0129] It can be seen that subproblem P2.5 is about The linear programming problem is solved to obtain... The range of values ​​is

[0130]

[0131] in Then, based on the optimal solution and Solve problem P2.1. This problem is convex but not differentiable, and can be solved using the elliptic method.

[0132] Finally, the optimal solution to the dual problem P2.1 is found. μ * and ν * Down, This is the optimal solution to problem P2. Therefore, at the optimal solution, the following equation must satisfy...

[0133]

[0134] The optimal solutions for the two resource allocation variables are obtained respectively.

[0135]

[0136] S416. Therefore, for subproblem P2, based on the determined relay selection factor x, the optimal solution to the resource allocation problem can be obtained as follows: and And the minimum latency of the V2I link is

[0137] S42. Solve problem P3 using convex optimization methods.

[0138] S421. Next, solve subproblem P3, rewriting problem P3 as follows:

[0139]

[0140] C7,C8

[0141]

[0142] Constraint C9 is derived through linear transformation, and constraint C6 is derived through substitution of the following formula.

[0143]

[0144] S422. This change makes constraint C9 a linear constraint, while constraint C6 remains a non-convex constraint. Therefore, we first use the continuous convex approximation to transform the above subproblem into a convex optimization problem. Assume a new variable... and therefore Rewritten as

[0145]

[0146] Question P3.1 has been rewritten as

[0147]

[0148] S423. Use the continuous convex approximation method to apply the constraints. Transformed into a convex constraint. Observation. Inequality can be obtained

[0149]

[0150] Where X m and Y m Defined respectively and If and only if and When the above inequality is equal, that is, R m Obtain the lower bound. When As it approaches positive infinity, it exists (X) m ,Y m () = (1, 0). Therefore, the problem is rewritten as

[0151]

[0152] S424. At this point, all constraints are non-convex, and the entire problem P3 is transformed into a convex problem, which can be solved using the Lagrange dual method. To solve the given convex optimization problem using the Lagrange dual method, Lagrange multipliers are introduced, and the augmented Lagrange function is constructed as follows:

[0153]

[0154] Where α m θ m Given that β is a nonnegative Lagrange multiplier, the dual problem is expressed as:

[0155] Ρ3.4:maxD(α,β,θ)

[0156] stα≥0,β≥0,θ≥0

[0157] The dual function is:

[0158]

[0159] S425. As can be seen from problems P3.4 and P3.5, dual decomposition transforms the original problem into a two-layer optimization problem. The inner loop solves for the optimal power. Given the minimum end-to-end delay τ², the outer layer iteratively updates the Lagrange multipliers. Based on the KKT conditions, L can be obtained regarding... The first derivative of τ2 is shown in the following equation.

[0160] Based on the above equation, it is easy to obtain the optimal solutions for the end-to-end delay and power allocation of the dual problem P3.5 as follows:

[0161] S426. According to the gradient descent method, the Lagrange multipliers are updated as follows, where η is a sufficiently small positive step size.

[0162]

[0163]

[0164] S427. Where η is the number of iterations not yet completed, the minimum V2V link delay for problem P3 is finally obtained under the given relay selection strategy x. Optimal power allocation is

[0165] S43. For any given relay selection strategy x, the minimum end-to-end delay achievable by the system under this strategy can be calculated through the above inner convex optimization process. The calculated latency value represents the optimal end-to-end latency performance achievable through optimized resource allocation under the current relay selection scheme x. More importantly, The fitness value will be directly used as the core indicator for evaluating the quality of the chromosome in the outer genetic algorithm. The inner optimization enables the outer algorithm to effectively search and evolve within the vast space of discrete relay selection strategies in order to find the globally optimal relay selection scheme.

[0166] The outer discrete variable solution and strategy output in step 5 are as follows:

[0167] S51. Based on the inner-layer optimization framework established in S4, the core task of the outer-layer algorithm is to efficiently search for the optimal relay selection strategy that minimizes end-to-end latency within the discrete relay selection strategy space. Since the relay selection variables are discrete, and the strategy space grows exponentially with the number of vehicles and relay nodes, traditional exhaustive search or gradient methods are difficult to apply. Therefore, this invention employs an improved genetic algorithm as the outer-layer optimization engine. Its advantages lie in its ability to perform parallel searches, avoid getting trapped in local optima, and effectively handle combinatorial optimization problems. The specific steps are described below.

[0168] S52. Population Initialization. The genetic algorithm introduces the concept of chromosomes based on natural selection. The relay selection strategy is used to construct the chromosome information of each individual. An initial population is generated according to the preset population size. Each chromosome uses a one-dimensional permutation encoding to represent the relay allocation scheme. The encoding length is the number of target vehicles M, and each gene position... Let represent the relay vehicle number selected by the target vehicle m in the i-th chromosome, and satisfy . To ensure a one-to-one mapping between relay vehicles and target vehicles, the following requirements must be met. Therefore, for the i-th chromosome, the chromosome information can be represented as:

[0169]

[0170] The initialization phase of a genetic algorithm directly impacts search efficiency and global optimality. To improve the quality of the initial population, this invention designs a heuristic initialization strategy based on channel quality and interference level to generate high-quality initial solutions. According to prior knowledge, the target vehicle tends to choose vehicles with better channel quality between it and other relay vehicles, while the channel quality between it and other relay vehicles is relatively worse. This is because the interference level from other relay vehicles is related to the channel quality between them. This invention proposes a heuristic evaluation function to quantify the initialization strategy, defining the fitness of target vehicle m in selecting relay vehicle j as...

[0171]

[0172] In the formula and H is a weighting factor. B,j H represents the channel gain between the base station and relay vehicle j. j,m This represents the channel quality between the target vehicle m and the relay vehicle j. Let represent the average channel gain from other relays to the target vehicle m. Each target vehicle selects a relay vehicle based on its fitness value. While introducing high-quality solutions, this invention carefully controls the proportion of heuristic solutions to prevent premature loss of population diversity and getting trapped in local optima. A hybrid initialization strategy is adopted, setting the ratio of greedy solutions to random solutions to ω.

[0173] S53. Fitness Function Design. The design of the fitness function is the core of the genetic algorithm's performance. Essentially, it transforms the inner-layer optimization results into a quantitative indicator for evaluating the merits of the relay selection strategy. Based on the inner-layer optimization framework established in step 4, the process involving the fitness function is described in detail.

[0174] S531. Regarding chromosome l i Define the equivalent relay selection strategy matrix as x i Among them, easy to obtain

[0175] S532. x i The fixed relay selection strategy is passed as input to the inner optimization module;

[0176] S533. Solve for x according to S4. i Optimal resource allocation scheme and minimum V2I link latency

[0177] S534. Solve for x according to S4. i Optimal interference management scheme and minimum V2V link latency

[0178] S535. Calculate the current policy x based on the inner optimization output. i End-to-end delay

[0179] S536. Define the fitness function as follows:

[0180]

[0181] A higher fitness value indicates a lower cost of the objective function, suggesting a better relay selection strategy.

[0182] S54. Selection Operation. In order to preserve better chromosomes (solutions) to produce better offspring, sort the population individuals in ascending order of fitness, set the number of elite individuals, copy these optimal individuals to the new generation population, randomly select 3 individuals, select the 2 with the best fitness as parents, place them in the mating pool to reproduce them, and perform the following crossover steps.

[0183] S55. Crossover Operation. To ensure that chromosomes cannot contain duplicate genes, sequential crossover is performed on selected parent individuals to generate offspring individuals. This operation preserves some sequence information of the parent individuals. Input two parent individuals, p1 and p2. Randomly select two cut points, with only one gene between the two cut points. Copy the gene segment located between the cut points from parent p1 to the same position in the offspring. Extract genes not present in the cut point from parent p2 in sequence, forming the remaining gene sequence. Fill the non-gene segment positions in the offspring with the remaining gene sequence in sequence (preserving the order of parent p2).

[0184] S56. Mutation Operation. Mutation is performed on offspring individuals with a certain probability, i.e., the relay assignment numbers at two positions in the offspring are randomly swapped. Mutation is performed with a certain probability: two gene positions are randomly selected and their gene values ​​are swapped. If no mutation is triggered, the original gene sequence is maintained.

[0185] S57. The algorithm repeats steps 1 to 5 until the number of generations meets the specified number or the relative change in optimal fitness over 15 consecutive generations is less than a threshold. Then, the chromosome with the minimum fitness value is selected. The gene value of this chromosome is chosen as the optimal relay selection scheme.

[0186] S58. Relay selection and resource allocation are modeled as mixed-integer nonlinear programming problems, and an optimization method for joint resource allocation and relay selection is proposed. (Flowchart attached) Figure 3 The outer layer uses an improved genetic algorithm to solve for the relay selection strategy x. The inner layer decomposes the original problem into two subproblems and uses convex optimization to solve the resource allocation problem, which serves as the fitness function for the outer genetic algorithm. Finally, the globally optimal end-to-end delay and the corresponding optimal relay selection strategy are obtained through iterative solutions using the genetic algorithm.

[0187] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for resource allocation and relay selection in a connected vehicle cooperative relay system, characterized in that... This invention provides a joint optimization method for resource allocation and relay selection in two-hop relay cooperative networks. Addressing the problem of high end-to-end latency caused by poor latency performance of one or more links in intelligent connected vehicle relay systems, this invention defines the target scenario, models the communication of the wireless access point to vehicle V2I link and the vehicle to vehicle V2V link, lists and solves the optimization problem, and presents a reliable and real-time joint optimization method for resource allocation and relay selection. The process specifically includes the following steps: S1. Construct a two-hop V2I-V2V cooperative relay model from access point to vehicle and from vehicle to vehicle; S2. Taking into account the latency requirements of each link, relay selection strategy and resource constraints, an optimization model with end-to-end latency as the objective was established. S3. The problem is modeled as a mixed integer nonlinear programming problem. By analyzing the coupling relationship of variables, the problem is decomposed into two sub-problems. A two-layer genetic algorithm that combines resource allocation and relay selection is proposed. S4. The inner layer solves the resource allocation problem. Under any given relay strategy, the convex optimization method is used to find the minimum end-to-end delay and the corresponding resource allocation scheme under the strategy. S5. The outer layer solves the relay selection problem by encoding each chromosome as a relay selection strategy through an improved genetic algorithm, and uses the output of the inner layer as the fitness value. Finally, it outputs the optimal relay selection strategy and the global minimum of end-to-end delay under the optimal relay selection. V2I link refers to the communication link between a vehicle and a wireless access point, while V2V link refers to the communication link between vehicles.

2. The resource allocation and relay selection method in a connected vehicle cooperative system according to claim 1, characterized in that, Step S1 constructs a two-hop communication architecture from the access point in densely populated urban areas to the relay vehicle and then to the target vehicle. Channel modeling is performed on all communication links, and expressions for metrics such as data rate and latency are analyzed and listed. Specifically: S11. To address the issues of strong fading and coverage blind spots in the communication link between the wireless access point and the vehicle, a two-hop relay transmission model is constructed. The transmission model includes M target vehicle nodes and K relay vehicle nodes, with the target vehicle set represented by [missing information]. This indicates that a connection to a wireless access point (AP) is required to obtain communication information. However, due to obstruction between the target vehicle and the AP, resulting in poor signal quality, relay vehicles are needed to cooperate in acquiring information for each target vehicle. Simultaneously, the relay vehicles also need to receive information sent to them by the AP. The AP needs to transmit data belonging to the target vehicles through the selected relay vehicles. The relay vehicle group is used for... This indicates that M < K. S12. Since large-scale fading components generally depend on vehicle location and undergo slow scale changes, the base station can acquire large-scale fading information for all communication links. Assuming that the access point and base station can share information through an interface, for small-scale fading of different links, the resource scheduling cycle is less than the channel coherence time caused by vehicle mobility, i.e., the impact of vehicle mobility is ignored. Therefore, the channel gain between the k-th relay vehicle and the m-th target vehicle is modeled as follows: in h represents the large-scale slow fading channel gain. k,m κ represents the small-scale fast fading power component, d represents the path loss constant, and d represents the path loss constant. k,m The distance between the k-th relay vehicle and the m-th target vehicle is represented by ε, where ε represents the channel fading index. Indicates that it has a standard deviation The log-normal shaded fading random variable. The entire communication transmission is divided into two links and communication between the wireless access point (AP) and the target vehicle is completed through frequency division multiple access (FDMA). Specifically, the channel gain of the V2I link from the wireless access point (AP) to the k-th relay vehicle is represented by H. k H represents the channel gain of the V2V link from the k-th relay vehicle to the m-th target vehicle. k,m The relay vehicle uses a decoding-forwarding protocol for data forwarding. The delays of the two links are the V2I link delay and the V2V link delay, denoted by τ1 and τ2 respectively. Their physical meanings are the maximum transmission delay between all APs and relay vehicles and the maximum transmission delay between the relay vehicle and the target vehicle, respectively. Furthermore, considering the current shortage of spectrum resources in dense areas, spectrum reuse technology is introduced in the relay transmission of the second link to improve spectrum resource utilization. Since uplink resources are used less densely and interference at the base station is easier to manage, the orthogonally allocated uplink spectrum of the cellular user (CUE) is reused by all relay V2V links, which will introduce co-channel interference. The channel gain between the k-th relay vehicle and the base station when the k-th relay vehicle interferes with the CUE is H. k,B The channel gain between the CUE and the base station is H. c The channel gain between CUE and the m-th target vehicle is H. m . S13. To ensure that each target vehicle obtains AP information by selecting relay vehicles, a relay selection strategy matrix x is defined. This matrix is ​​a K×M binary matrix, where each element x... k,m ∈{0,1} is the relay selection factor, representing the service relationship between the relay vehicle and the target vehicle, defined as x k,m =1 indicates that the k-th relay vehicle serves the m-th target vehicle. Considering that each relay vehicle can serve one and at most one target vehicle, the following condition must be met: Assuming each target vehicle has exactly one relay vehicle serving it, the following conditions must be met: S14. Consider the downlink of access point information based on FDMA for V2I links. Given a relay selection strategy x, let k... m (x) represents the relay vehicle selected by the m-th target vehicle. The achievable downlink data rate is then given by the following formula. in This represents the transmit power allocated by the AP to the m-th target vehicle, and N0 represents the power spectral density of the noise. This represents the bandwidth allocated by the AP to the m-th target vehicle. S15. If the relay selection factor x is not given, the achievable downlink data rate is given by the following formula. S16. Let... This represents the amount of data that the AP sends to the k-th relay vehicle. This represents the amount of data that the AP sends to the m-th target vehicle. Let the total amount of data to be sent be... Considering the poor performance of V2I links, the transmission delay between each AP and the relay vehicle is specified to have an upper limit τ1, which must satisfy... S17. Next, consider FDMA-based relay transmission on V2V links. If, under a given relay selection strategy x, from the k-th... m The achievable V2V link data rate from (x) relay vehicles to the m-th target vehicle is given by the following formula. in This indicates the interference of other relay vehicles reusing the CUE channel with the communication of the m-th target vehicle. This indicates that the CUE interferes with the same lane of the m-th target vehicle. This indicates the bandwidth of the CUE that has been reused. This represents the transmit power allocated by the relay vehicle to the m-th target vehicle. Furthermore, given the maximum interference power threshold of the V2V link to the CUE uplink, it is defined as I. th,max The minimum signal-to-interference-plus-noise ratio (SINNR) of the uplink and the uplink is defined as γ0. The transmit power of the CUE can be calculated to satisfy the following condition. S18. If the relay selection factor x is not given, the achievable V2V link data rate is given by the following formula. in S19. Considering the data transmission loss during decoding and forwarding, let... This represents the amount of data that needs to be transmitted to the m-th relay vehicle. Considering the poor performance of V2V links, the transmission delay between each relay vehicle and the target relay vehicle is defined to have an upper bound τ2, which must satisfy... 。 3. The relay selection and resource allocation method in a connected vehicle cooperative system according to claim 2, characterized in that, Step S2 involves establishing a mixed-integer nonlinear programming model for joint resource allocation and relay selection. Specifically: S21. By integrating the above two links, the end-to-end transmission delay T used to complete all information transmission is reduced. E2E Defined as the sum of V2I link delay and V2V link delay, given by the following formula. T E2E =τ1+τ2 S22. The optimization objective is to minimize the end-to-end delay T. E2E This requires joint optimization of the relay selection strategy matrix x and the V2I link transmission bandwidth allocation vector B. (1) Send power allocation vector P (1) V2V link transmit power allocation vector P (2) And the time delay vector τ. Where B (1) Represents a set of K×1 dimensional variables P (1) Represents a K×1 set of variables P (2) Represents an M×1 set of variables And τ represents a 2×1 set of variables {τ1, τ2}. The optimization problem is formulated as a resource allocation and relay selection problem that minimizes end-to-end latency, as described below: S23. Definition This is the maximum bandwidth that the AP can allocate to the cooperative relay system. This is the AP's maximum transmit power. Let m be the maximum transmission power of the m-th relay vehicle. This represents the minimum transmission rate for the V2V link. Constraints C1 and C2 are two relay selection factor constraints. Constraint C3 states that the link delay for each V2I link is no greater than the link delay τ1 for that stage. Constraint C6 represents that the link delay for each V2V link is no greater than the link delay τ2 for that stage. Constraints C4, C5, and C7 are total resource constraints, and constraints C8 and C9 are reliability constraints.

4. The relay selection and resource allocation method in a connected vehicle cooperative system according to claim 3, characterized in that, In step S3, the problem is decoupled by analyzing the coupling relationship of variables, and a two-layer genetic algorithm is proposed. The algorithm is characterized by an inner layer solving for continuous variables related to resource allocation and an outer layer solving for discrete variables related to relay selection. The steps include: S31. From the optimization problem, it can be seen that V2I link delay and V2V link delay are coupled together through the relay selection strategy x. This coupling makes the relay selection and resource allocation problems more difficult to handle. Furthermore, due to the binary constraints in the problem, it is a combinatorial mixed-integer nonlinear problem, and is non-convex NP-hard. Therefore, this invention proposes an improved two-layer distributed genetic algorithm. In this method, B is first calculated for any given x. (1) P (1) P (2) And τ, and then, instead of exhaustively listing all possible cases of x, this invention chooses to use an improved genetic algorithm to globally optimize x. S32. Specifically, under any given relay selection strategy x, the original problem can be equivalently decoupled into two parallel subproblems: joint bandwidth allocation and power control for V2I links, and power allocation and interference management for V2V links, as shown below. C4,C5 C7,C9 S33. In the inner layer, for any fixed relay selection strategy x, by solving subproblems P2 and P3, the minimum delay of the V2I link under the specific relay selection strategy x can be obtained as follows: And the minimum latency of the V2V link is Furthermore, suppose that, given x, the optimal solutions for P2 and P3 are... Then the minimum end-to-end latency that the system can achieve is This minimum value is relative to all possible resource allocation schemes for a fixed x. The goal of the outer improved genetic algorithm is to search for the globally optimal relay selection strategy x in all possible solution spaces of x. * This ensures that the minimum end-to-end latency generated by it, after optimal resource allocation within the inner layer, reaches the global minimum, i.e. and It is the global minimum end-to-end delay that problem P1 requires.

5. The relay selection and resource allocation method in a connected vehicle cooperative system according to claim 4, characterized in that... The inner layer solves two subproblems and outputs the optimal end-to-end delay under a given relay selection strategy. Specifically: S41. Solve problem P2. S411. It can be proven that constraints C4 and C5 in problem P2 are both linear constraints, and constraint C3 has a certain relationship with the variable. exist The upper part is also a convex constraint. S412. Problem P2 is a convex problem, which can be solved using the Lagrange duality method. The augmented Lagrange function is constructed as follows: Where λ m μ, ν are non-negative Lagrange multipliers. The dual problem of problem P2 is: The dual function is: S413. Since problem P2 satisfies the Slater condition, P2 and its dual problem P2.1 have strong duality. To solve P2, we first use the given λ m Solve P2.2 to obtain D({λ, μ, ν). m },μ,ν), then through the condition {λ m To maximize D({λ ≥ 0}, μ ≥ 0, and ν ≥ 0 m },μ,ν). Given {λ m Given ≥0, μ≥0, and ν≥0, the dual problem can be divided into the following subproblems. S414. For subproblem P2.3, it is easy to obtain the optimal solution for end-to-end delay as follows: S415. For subproblem P2.4, each subproblem is convex and satisfies the Slater conditions. Therefore, this problem can be solved by applying the Karouch-Kun-Tucker conditions (KKT conditions). The Lagrange equation for the m-th subproblem is given below. According to L m about and The first derivative, given {λ} m Under the conditions ≥0}, μ≥0, and ν≥0, we can solve for... Where, [x] + =max{0,x}, which can be used to modify problem P2.

4. It can be seen that subproblem P2.5 is about The linear programming problem is solved to obtain... The range of values ​​is in Then, based on the optimal solution and Solve problem P2.

1. This problem is convex but not differentiable, and can be solved using the elliptic method. Finally, the optimal solution to the dual problem P2.1 is found. μ * and ν * Down, This is the optimal solution to problem P2. Therefore, at the optimal solution, the following equation must satisfy... The optimal solutions for the two resource allocation variables are obtained respectively. S416. Therefore, for subproblem P2, based on the determined relay selection factor x, the optimal solution to the resource allocation problem can be obtained as follows: and And the optimal latency of the V2I link is S42. Solve problem P3. S421. Rewrite problem P3 as follows: C7,C8 Constraint C9 is derived through linear transformation, and constraint C6 is derived through substitution of the following formula. S422. This change makes constraint C9 a linear constraint, while constraint C6 remains a non-convex constraint. Therefore, we first use the continuous convex approximation to transform the above subproblem into a convex optimization problem. Assume a new variable... and therefore Rewritten as Question P3.1 has been rewritten as S423. Use the continuous convex approximation method to apply the constraints. Transformed into a convex constraint. Observation. Inequality can be obtained Where X m and Y m Defined respectively and If and only if and When the above inequality is equal, that is, R m Obtain the lower bound. When As it approaches positive infinity, it exists (X) m ,Y m () = (1, 0). Therefore, the problem is rewritten as S424. At this point, all constraints are non-convex, and the entire problem P3 is transformed into a convex problem, which can be solved using the Lagrange dual method. To solve the given convex optimization problem using the Lagrange dual method, Lagrange multipliers are introduced, and the augmented Lagrange function is constructed as follows: Where α m θ m With β being a nonnegative Lagrange multiplier, the dual problem is represented as P3.4: maxD(α,β,θ). stα≥0,β≥0,θ≥0 The dual function is: S425. As can be seen from problems P3.4 and P3.5, dual decomposition transforms the original problem into a two-layer optimization problem. The inner loop solves for the optimal power. Given the minimum end-to-end delay τ², the outer layer iteratively updates the Lagrange multipliers. Based on the Karouch-Kun-Tucker conditions and KKT conditions, L... (2) about By taking the first derivative of τ2, it is easy to obtain the optimal solutions for the end-to-end delay and power allocation of the dual problem P3.5 as follows: S426. According to the gradient descent method, the Lagrange multipliers are updated as follows, where η is a sufficiently small positive step size. S427. Where η is the number of iterations not yet completed, the optimal V2V link delay for problem P3 is finally obtained under the given relay selection strategy x. Optimal power allocation is S43. For any given relay selection strategy x, the minimum end-to-end delay achievable by the system under this strategy can be calculated through the above inner convex optimization process. The calculated latency value represents the optimal end-to-end latency performance achievable through optimized resource allocation under the current relay selection scheme x. More importantly, The fitness value will be directly used as the core indicator for evaluating the quality of the chromosome in the outer genetic algorithm. The inner optimization enables the outer algorithm to effectively search and evolve within the vast space of discrete relay selection strategies in order to find the globally optimal relay selection scheme.

6. A relay selection and resource allocation method in a connected vehicle cooperative system according to claims 4 and 5, characterized in that, In step S5, the outer layer uses a genetic algorithm to find the optimal relay selection strategy that achieves the globally optimal end-to-end delay, using the end-to-end delay obtained in claim 4 as the fitness value. Specifically: S51. The core task of the outer layer algorithm is to efficiently search for the optimal relay selection strategy that minimizes end-to-end delay in the discrete relay selection strategy space. This invention uses an improved genetic algorithm as the outer layer optimization engine, which has the advantages of parallel search, avoiding getting trapped in local optima, and effectively handling combinatorial optimization problems. S52. Population Initialization. The genetic algorithm introduces the concept of chromosomes based on natural selection. The relay selection strategy is used to construct the chromosome information of each individual. An initial population is generated according to the preset population size. Each chromosome uses a one-dimensional permutation encoding to represent the relay allocation scheme. The encoding length is the number of target vehicles M, and each gene position... Let represent the relay vehicle number selected by the target vehicle m in the i-th chromosome, and satisfy . To ensure a one-to-one mapping between relay vehicles and target vehicles, the following requirements must be met. Therefore, for the i-th chromosome, the chromosome information can be represented as: The initialization phase of a genetic algorithm directly impacts search efficiency and global optimality. To improve the quality of the initial population, this invention designs a heuristic initialization strategy based on channel quality and interference level to generate high-quality initial solutions. According to prior knowledge, the target vehicle tends to choose vehicles with better channel quality between it and other relay vehicles, while the channel quality between it and other relay vehicles is relatively worse. This is because the interference level from other relay vehicles is related to the channel quality between them. This invention proposes a heuristic evaluation function to quantify the initialization strategy, defining the fitness of target vehicle m in selecting relay vehicle j as... In the formula and H is a weighting factor. B,j H represents the channel gain between the base station and relay vehicle j. j,m This represents the channel quality between the target vehicle m and the relay vehicle j. Let represent the average channel gain from other relays to the target vehicle m. Each target vehicle selects a relay vehicle based on its fitness value. While introducing high-quality solutions, this invention carefully controls the proportion of heuristic solutions to prevent premature loss of population diversity and getting trapped in local optima. A hybrid initialization strategy is adopted, setting the ratio of greedy solutions to random solutions to ω. S53. Fitness Function Design. The design of the fitness function is the core of the genetic algorithm's performance. Its essence is to transform the inner-layer optimization results into a quantitative indicator for evaluating the merits of the relay selection strategy. Based on the inner layer output obtained in claim 4, the process of solving the fitness function is described in detail. S531. Regarding chromosome l i Define the equivalent relay selection strategy matrix as x i Among them, easy to obtain S532. x i The fixed relay selection strategy is passed as input to the inner optimization module; S533. Solve for x according to S4. i Optimal resource allocation scheme and minimum V2I link latency S534. Solve for x according to S4. i Optimal interference management scheme and minimum V2V link latency S535. Calculate the current policy x based on the inner optimization output. i End-to-end delay S536. Define the fitness function as follows: A higher fitness value indicates a lower cost of the objective function, suggesting a better relay selection strategy. S54. Selection Operation. In order to preserve better chromosomes (solutions) to produce better offspring, sort the population individuals in ascending order of fitness, set the number of elite individuals, copy these optimal individuals to the new generation population, randomly select 3 individuals, select the 2 with the best fitness as parents, place them in the mating pool to reproduce them, and perform the following crossover steps. S55. Crossover Operation. To ensure that chromosomes cannot contain duplicate genes, sequential crossover is performed on selected parent individuals to generate offspring individuals. This operation preserves some sequence information of the parent individuals. Input two parent individuals, p1 and p2. Randomly select two cut points, with only one gene between the two cut points. Copy the gene segment located between the cut points from parent p1 to the same position in the offspring. Extract genes not present in the cut point from parent p2 in sequence, forming the remaining gene sequence. Fill the non-gene segment positions in the offspring with the remaining gene sequence in sequence (preserving the order of parent p2). S56. Mutation Operation. Mutation is performed on offspring individuals with a certain probability, i.e., the relay assignment numbers at two positions in the offspring are randomly swapped. Mutation is performed with a certain probability: two gene positions are randomly selected and their gene values ​​are swapped. If no mutation is triggered, the original gene sequence is maintained. S57. The algorithm repeats steps 1 to 5 until the number of generations meets the specified number or the relative change in optimal fitness over 15 consecutive generations is less than a threshold. Then, the chromosome with the minimum fitness value is selected. The gene value of this chromosome is chosen as the optimal relay selection scheme. S58. This paper models relay selection and resource allocation as a mixed-integer nonlinear programming problem and proposes a joint optimization method for resource allocation and relay selection. The outer layer uses an improved genetic algorithm to solve for the relay selection strategy x. The inner layer decomposes the original problem into two subproblems and uses a convex optimization method to solve the resource allocation problem, which serves as the fitness function for the outer genetic algorithm. Finally, the globally optimal end-to-end delay and the corresponding optimal relay selection strategy are obtained through iterative solutions using the genetic algorithm.

Images