A method for deploying roadside units of urban vehicle-mounted networks

Abstract

This invention provides a method for deploying roadside units (RSUs) in urban vehicle-to-everything (V2X) networks, belonging to the field of intelligent transportation technology. The method includes: 1) gridding the target urban area and establishing a multi-objective optimization model; 2) initializing algorithm parameters and generating a population; 3) using the NSGA-III algorithm in parallel with multiple subpopulations and adaptive parameter adjustments for iterative optimization; 4) selecting the Pareto optimal solution set and generating a deployment strategy; 5) executing the deployment strategy and a data offloading strategy; and 6) verifying the running IBRSG strategy from multiple dimensions, providing guidance for parameter settings in practical applications. This invention, through multiple population strategies and adaptive parameter adjustments, generates more Pareto optimal solutions, allowing decision-makers to flexibly balance latency, cost, and quality of service according to their needs. The adopted IBRSG strategy optimizes vehicle-to-RSU connections through iterative game theory, reducing total latency by 15-20% compared to existing methods and improving load balancing.

Description

Technical Field

[0001] This invention belongs to the field of intelligent transportation technology, and in particular relates to a method for deploying roadside units of urban vehicle-mounted networks. Background Technology

[0002] With the acceleration of urbanization and the increasing complexity of transportation networks, Intelligent Transportation Systems (ITS) have become a core technology for improving traffic efficiency, safety, and service quality. Among these technologies, the Vehicle-to-Infrastructure (V2I) communication framework plays a crucial role in ITS, using Roadside Units (RSUs) as intermediaries to enable data transmission, real-time information sharing, and traffic management optimization between vehicles. The deployment of RSUs directly impacts network coverage, communication latency, and overall system performance. In urban environments, RSUs must address the challenges posed by high-density vehicle traffic, diverse road types, and obstacles such as buildings, gardens, and lakes.

[0003] Existing research primarily focuses on optimizing RSU deployment, including static deployments (such as fixed-location RSUs) and dynamic deployments (such as mobile RSUs). Optimization objectives are typically balancing communication benefits (such as coverage, response latency, and service profitability) and cost factors (such as construction and maintenance costs). While these studies have contributed to technological advancements, significant limitations remain in practical applications, primarily in the following four aspects:

[0004] 1) Limitations of Single-Objective Optimization: Existing methods often focus on a single objective (such as minimizing latency or cost), failing to balance the conflicts between multiple objectives (such as the trade-off between latency and cost), resulting in a lack of flexibility in deployment schemes in practical applications. For example, the lowest-cost solution may not meet the low-latency requirements of high-traffic areas, leading to traffic congestion or security risks. While some optimization algorithms effectively reduce costs, they neglect QoS guarantees in sensitive areas, resulting in uneven overall network performance.

[0005] 2) Ignoring Environmental Constraints: Urban environments contain obstacles such as buildings, gardens, and lakes, limiting the feasible deployment locations of RSUs. Existing methods often ignore these physical constraints and fail to effectively address signal interference caused by excessively close spacing between RSUs, reducing the feasibility of deployment schemes. For example, while Guerna et al.'s genetic algorithm optimizes coverage, it does not consider obstacles, making some schemes unfeasible in real-world maps and increasing the cost of replanning.

[0006] 3) Low efficiency in large-scale optimization: City maps need to be divided into thousands of grid cells (e.g., 2500 decision variables), resulting in a huge decision variable space. Traditional algorithms (such as NSGA-III and MOEA / D) converge slowly and have poor solution set diversity in large-scale spaces, making it difficult to efficiently explore the optimal solution. For example, in high-density urban scenarios, NSGA-II-related algorithms are prone to getting trapped in local optima and cannot generate sufficient Pareto solution sets, affecting decision diversity.

[0007] 4) Insufficient data offloading strategies: Data offloading between vehicles and RSUs is highly dynamic, and existing strategies (such as nearest-distance selection and random allocation) cannot balance latency and load balancing, leading to network congestion or resource waste. For example, traditional methods do not consider vehicle mobility and RSU load, resulting in a surge in latency during high-traffic periods, affecting real-time services such as collision warnings. Summary of the Invention

[0008] The purpose of this invention is to address the limitations of single-objective optimization, neglect of environmental constraints, low efficiency of large-scale optimization, and insufficient data offloading strategies, and to provide a method for deploying roadside units in urban vehicle networks, comprising the following steps:

[0009] S1: The target urban area is gridded, with each grid serving as a potential RSU deployment location. A multi-objective optimization model is then established based on identified obstacles and latency-sensitive areas.

[0010] S2: Configure the parameters required for the algorithm to run, and randomly generate an initial population, and evenly divide the entire initial population into multiple subpopulations;

[0011] S3: Multiple subpopulations run the NSGA-III algorithm synchronously and independently and make adaptive parameter adjustments. After multiple iterations, the subpopulations complete the collaborative optimization between the subpopulations through migration operations until the maximum number of iterations is reached or the convergence condition is met.

[0012] S4: Merge all subpopulations, select all Pareto optimal solutions from the final population through non-dominated sorting, and output a complete set of Pareto optimal deployment schemes that can be used for decision-making based on the optimal solutions;

[0013] S5: Select a deployment plan and run the data offloading strategy;

[0014] S6: Multi-dimensional verification of the IBRSG strategy provides guidance for parameter settings in practical applications.

[0015] Furthermore, in S1, the gridding process specifically includes: dividing the target city area into K grids, where each grid k corresponds to a variable. ,when A value of 1 indicates that an RSU has been deployed; when A value of 0 indicates that no RSU has been deployed.

[0016] The multi-objective optimization model includes the objective function of minimizing total delay, the objective function of minimizing delay in sensitive areas, the objective function of minimizing the number of RSUs, as well as obstacle constraints and minimum spacing constraints.

[0017] Furthermore, total delay The minimization objective function includes transmission delay, processing delay, queue delay, and propagation delay, and is expressed as:

[0018] ;

[0019] in, For vehicles In time The latency is calculated based on vehicle trajectory and RSU coverage, and the transmission rate is estimated using the Shannon formula.

[0020] Minimize the maximum latency of the sensitive area To ensure communication reliability in critical areas, it is represented as:

[0021] ;

[0022] This objective function provides limited QoS guarantees for high-priority areas;

[0023] Number of RSUs The minimization objective function is used to control construction and maintenance costs, and is expressed as:

[0024] ;

[0025] The multi-objective optimization model indirectly minimizes the budget through this objective function, while balancing the conflict with the delay objective;

[0026] Obstacle constraints are implemented using binary variables. The location of the obstacle is represented as, and the constraint is represented as ;

[0027] Minimum spacing constraints are used to prevent RSUs from being too densely packed and to avoid signal interference. This is achieved by setting a minimum RSU spacing. The constraint is 50 meters, and is expressed as follows: ,in, The distance is Euclidean.

[0028] Furthermore, in S2, the parameters that need to be set include: total delay. Maximum delay in sensitive areas Number of RSUs Location of obstacles Distance constraints Population size N is between 360 and 500, number of subpopulations m is between 3 and 5, and maximum number of iterations G is between 50 and 100;

[0029] Randomly generate a population P, divide it into m subpopulations, and evaluate the feasibility of the initial solution and the objective value.

[0030] Furthermore, in S3, the iterative optimization of multiple subpopulations specifically includes the following steps:

[0031] S31: Calculate the total delay using the IBRSG strategy. Maximum delay in sensitive areas and the number of RSUs ;

[0032] S32: Run the NSGA-III algorithm to generate offspring through selection, crossover, and mutation operations;

[0033] S33: Density constraints are handled using a calibration mechanism.

[0034] S34: Dynamically adjust the cross rate and variability Among them, cross rate ; Variation rate If the current migration operation does not improve the subpopulation, then increase the mutation rate. Reduce cross rate If the current migration operation improves the subpopulation, then increase the crossover rate. Reduce mutation rate ;

[0035] S35: Every 5-10 generations, the first 10-20% of individuals in each subpopulation migrate to other subpopulations, replacing the worst 20-30% of individuals in each subpopulation, in order to promote population diversity.

[0036] Furthermore, in S33, the calibration mechanism specifically includes the following steps:

[0037] S331: Calculate the spacing between all RSU pairs and identify violations of the minimum RSU spacing. RSU pair;

[0038] S332: Based on traffic volume within the coverage area, retain the RSU with relatively high traffic volume and remove the other RSU to avoid overly dense deployment;

[0039] S333: Re-evaluate the objective function after calibration to ensure the feasibility of the solution.

[0040] Furthermore, S3 also includes a constraint handling mechanism, which specifically employs: The comparison method handles constraint violation solutions, allowing for comparison of solution quality under constraint relaxation, and is expressed as:

[0041] ;

[0042] in, The dynamic tolerance threshold (a non-negative real number) is adaptively adjusted during algorithm iterations: the initial value is set to a large value to allow exploration of infeasible regions, and then gradually tightened to guide the population to converge to a feasible solution; To constrain the degree of violation, the distance between the solution and the feasible region is quantified (e.g., the deviation value when the minimum spacing constraint is violated in an RSU deployment). Its calculation is based on the sum of the degree of violation of inequality constraints and equality constraints. This indicates that solution A dominates solution B, meaning that all objective function values ​​of A are not inferior to B and are strictly superior to B at least in one objective. The comparison rule covers three cases: when the constraint violation degrees of both solutions are below a threshold, the dominance relationship is compared first; when the violation degrees are equal, the dominance relationship is directly compared; otherwise, the solution with the lower violation degree is selected. This mechanism uses dynamic... Value balancing algorithm exploration and convergence ensures effective maintenance of population diversity in complex urban environments and leads to feasible Pareto solutions.

[0043] Furthermore, S4 specifically involves: after the algorithm terminates its iterations, merging all subpopulations and selecting all Pareto optimal solutions from the final population using non-dominated sorting. These optimal solutions are then decoded into a list of specific RSU location coordinates, and the performance metrics corresponding to each deployment scheme are calculated. Finally, a complete set of Pareto optimal schemes available for decision-making is output.

[0044] Furthermore, in S5, the data offloading strategy includes the following steps:

[0045] S51: Initialize the random policy and short strategy ;

[0046] S52: Update the connectivity policy for each vehicle sequentially and select the RSU that minimizes the total latency;

[0047] S53: Iterate until the policy change is less than the error threshold, i.e., error=0.01, to ensure convergence to a stable Nash equilibrium.

[0048] Furthermore, S6 specifically involves: evaluating the overall performance of the solution by conducting tests in different city scenarios before and after deployment, and comparing it with traditional algorithms. At the same time, it analyzes the impact of key parameters on the results to verify the robustness of the solution and provide guidance for parameter settings in practical applications.

[0049] Compared with the prior art, the beneficial effects of the present invention are mainly reflected in:

[0050] 1. This invention generates more Pareto optimal solutions through multiple group strategies and adaptive parameter adjustments (e.g., AM-NSGA-III-c generates 22.16 solutions in high-density scenarios, while NSGA-III only generates 10.56), allowing decision-makers to flexibly balance latency, cost, and service quality according to their needs.

[0051] 2. This invention integrates obstacle constraint and spacing calibration mechanisms to ensure the feasibility of deployment schemes. The calibration mechanism reduces RSU density (e.g., the number of RSUs deployed in AM-NSGA-III-c is more than 30% less than that in AM-NSGA-III), thereby reducing signal interference and construction costs.

[0052] 3. This invention achieves accelerated convergence (reaching stability within 50 generations) with multiple populations and adaptive strategies, and maintains solution diversity in a large space (2500 decision variables) (IGD reduced to 0.2408, HV increased to 0.8312). Compared with the traditional NSGA-III, the computation time is shortened by 20-30%.

[0053] 4. The IBRSG strategy adopted in this invention optimizes vehicle-RSU connection through iterative game theory, which reduces total latency by 15-20% compared with existing methods (such as nearest distance and genetic algorithm) and improves load balancing (load variance is reduced by 25%).

[0054] 5. The NSGA-III algorithm used in this invention performs excellently in both high- and low-density urban scenarios, and remains feasible even as the sensitive area increases. (Value reduced by 10-15%), suitable for complex urban environments, such as high-density areas in Chengdu or low-density areas in Shanghai.

[0055] 6. The method of this invention minimizes the number of RSUs while ensuring coverage, reducing deployment and maintenance costs (saving 20-40% of the budget), and is highly cost-effective.

[0056] 7. The method of this invention is based on real data (such as DiDi trajectory), the solution is easy to integrate into the ITS platform, supports real-time adjustment, and is highly practical. Attached Figure Description

[0057] Figure 1 This is a flowchart illustrating a method for deploying roadside units in an urban vehicle network according to the present invention.

[0058] Figure 2 Examples of RSU (Roadside Unit) deployment for different algorithms in high-density scenarios.

[0059] Figure 3 Examples of RSU (Roadside Unit) deployment for different algorithms in low-density scenarios. Detailed Implementation

[0060] The following will describe in more detail a method for deploying roadside units of an urban vehicle network according to the present invention, with reference to the schematic diagrams, which illustrate preferred embodiments of the invention. It should be understood that those skilled in the art can modify the invention described herein while still achieving the advantageous effects of the invention. Therefore, the following description should be understood as being of general knowledge to those skilled in the art and is not intended to limit the invention.

[0061] like Figure 1 As shown, a method for deploying roadside units for urban vehicle-mounted networks includes the following steps:

[0062] Step 1: Data source:

[0063] Using the DiDi Gaia open dataset (Chengdu city vehicle trajectories), traffic flow is generated through route planning via the Gaode Map API. Additionally, OpenStreetMap data is integrated to identify obstacles and road layouts.

[0064] Step 2: Problem Modeling and Preprocessing:

[0065] Map processing: The 1 square kilometer area was divided into a 50×50 grid (2500 decision variables), and obstacle locations were identified and marked using OpenStreetMap. Sensitive area S is defined based on historical accident data (e.g., 10% of the grid is high priority), and a multi-objective optimization model is established, including total delay. Minimize the objective function:

[0066] ;

[0067] in, For vehicles In time The latency is calculated based on vehicle trajectory and RSU coverage, and the transmission rate is estimated using the Shannon formula.

[0068] Minimize the maximum latency of the sensitive area :

[0069] ;

[0070] The number of RSUs Minimize the objective function:

[0071] .

[0072] Step 3: Algorithm parameter initialization and population generation:

[0073] Configure the parameters required for the algorithm to run, including: population size N between 360 and 500, number of subpopulations m between 3 and 5, and crossover rate. Variation rate The maximum iteration G is between 50 and 100, and the minimum RSU spacing is... Meters, delay threshold 2 seconds, IBRSG error threshold = 0.01. RSU coverage radius R = 200 meters, transmission power based on IEEE 802.11p standard.

[0074] An initial population is randomly generated, meaning each grid cell is randomly assigned whether to deploy RSUs, and their basic feasibility is checked. Finally, the entire population is evenly divided into multiple subpopulations to prepare for parallel optimization.

[0075] Step 4: Parallel Iterative Optimization of Multiple Subpopulations:

[0076] This stage is the core optimization loop. Each subpopulation undergoes independent evolutionary computation: First, the performance of the current RSU deployment scheme is evaluated using the IBRSG data offloading strategy, and the total latency is calculated. Maximum delay in sensitive areas and the number of RSUs These three objective function values ​​are then used; next, the core operations of the NSGA-III algorithm (selection, crossover, mutation) are run to generate new candidate solutions; for the AM-NSGA-III-c version of the algorithm of this invention, an additional calibration step is performed at this step to identify and merge RSUs that are too close together to avoid signal interference; at the same time, the crossover rate is adaptively adjusted according to the search progress of the subpopulation. and variability To balance global exploration with local development, each subpopulation will undergo a migration operation after operating independently for several generations: the top 10-20% of individuals will migrate to other subpopulations, and the worst 20-30% of individuals in each subpopulation will be replaced. Superior individuals will be exchanged to promote information sharing and maintain population diversity.

[0077] Step 5: Pareto optimal solution set extraction and solution generation:

[0078] The algorithm terminates iteration once the maximum number of iterations is reached or the convergence condition is met. All subpopulations are merged, and all Pareto optimal solutions are selected from the final population using non-dominated sorting. These optimal solutions are decoded into a list of specific RSU location coordinates, and performance metrics (such as expected latency, coverage, cost, etc.) for each deployment scheme are calculated. Finally, a complete set of Pareto optimal schemes available for decision-making is output.

[0079] Step 6: Execution of dynamic data offloading strategy:

[0080] For the selected or ultimately implemented RSU deployment scheme, run the IBRSG (Iterative Best Response Sequence Game) strategy. This strategy dynamically selects the optimal RSU for each vehicle in the network through iterative calculation to minimize the total communication latency of the entire system and ensures that the strategy converges to a stable Nash equilibrium.

[0081] Step 7: Solution Verification and Parameter Tuning

[0082] Before and after deployment, the generated solution was validated in multiple dimensions. Tests were conducted in different urban scenarios, including high and low density, and compared with traditional algorithms to evaluate the overall performance of the solution. Simultaneously, the impact of key parameters (such as population size and minimum RSU spacing) on ​​the results was analyzed to verify the robustness of the solution and provide guidance for parameter settings in practical applications.

[0083] Specifically: such as Figure 2 , Figure 3 As shown, the X and Y axes in the figure constitute a gridded map system based on actual geographic coordinates. The X-axis represents the east-west coordinates (east is the positive direction), and the Y-axis represents the north-south coordinates (north is the positive direction). The coordinate system has its origin (0,0) at the lower left corner of the map area, and each coordinate unit corresponds to an actual distance of 5 meters. This coordinate system was established by uniformly dividing a 1 square kilometer experimental area into 50×50 grids (a total of 2500 decision units). The actual area of ​​each grid is 20 square meters (1000m / 50=20m / grid). Grid index... The relationship with the coordinate position is as follows: ,in , These are grid coordinates.

[0084] Figure 2 (a) NSGA-III algorithm deployment results: The number of RSUs is too high (about 24), the distribution is messy and there are many places where they are too dense (the distance between many pairs of RSUs is <50m, which is obviously a risk of signal interference), the coverage of sensitive areas is uneven, there are coverage gaps at the high-traffic intersection on the right, the total delay (268975.4 seconds) and the maximum delay in sensitive areas (0.747~0.843 seconds) are both too high, and there is no global optimal solution.

[0085] Figure 2 (b) Deployment results of AM-NSGA-III algorithm (with adaptive parameter adjustment, no spacing calibration mechanism): The number of RSUs is moderate (about 20), the coverage of high traffic areas is sufficient, and the total latency is better than NSGA-III (223522.9 seconds), but there are still some local over-density issues (such as the upper right corner), 1-2 pairs of RSUs violate the constraint, and the load balance needs to be improved.

[0086] Figure 2(c) AM-NSGA-III-c (the complete algorithm used in this invention, with multiple population adaptive NSGA-III + spacing calibration mechanism): Optimal number of RSUs (only 18), most reasonable distribution, all RSU spacing ≥ 50m, completely eliminating signal interference risk; accurate coverage of all sensitive areas, lowest maximum latency in sensitive areas (0.735~0.812 seconds), optimal total latency (202880.4 seconds), best load balancing, and significantly better overall performance than AM-NSGA-III-C (the complete algorithm used in this invention, with multiple population adaptive NSGA-III + spacing calibration mechanism): Figure 2 (a) and Figure 2 (b)

[0087] Figure 3 (a) The deployment results of the MOEA / D algorithm are as follows: the number of RSUs is very small (about 9) and they are scattered. There are many coverage vulnerabilities (the upper left corner and the low traffic area below are completely uncovered). The RSUs around the sensitive area are too far apart. The maximum latency in the sensitive area is the highest (f2≈1.62s). There is no globally feasible optimal solution. The concentrated load results in extremely poor load balancing.

[0088] Figure 3 (b) Deployment results of the NSGA-III algorithm: The number of RSUs is too large (about 24), with obvious redundancy and local density (2-3 pairs of RSUs with a spacing of <50m), resulting in serious cost waste. The maximum latency in the sensitive area (0.747~0.843 seconds) is higher than that of the algorithm of this invention.

[0089] Figure 3 (c) is the deployment result of the AM-NSGA-III algorithm (with adaptive parameter adjustment, without spacing calibration mechanism): the number of RSUs is moderate (about 13), the coverage is relatively complete, the total latency is low (10243.95 seconds), and the performance is better than (a)(b), but there is still one RSU spacing problem, and the load variance is slightly high.

[0090] Figure 3 (d) AM-NSGA-III-c (the complete algorithm of this invention, with spacing calibration mechanism): The number of RSUs is optimal (about 10), all of which are precisely deployed on the main road nodes with the highest traffic, the spacing between all RSUs is ≥50m, the coverage of sensitive areas is the most complete, the maximum latency in sensitive areas is the lowest (f2≈0.87s), the total latency is the lowest, the load is the most balanced, and it still maintains extremely high cost performance in low-density scenarios.

[0091] Figure 2 and Figure 3The deployment results of four core algorithms were visually compared using real Chengdu maps (high-density Tianfu Square scenario and low-density Du Fu Thatched Cottage scenario): In the high-density scenario, MOEA / D failed to generate a deployment map due to the lack of feasible solutions, while the AM-NSGA-III-c algorithm of this invention performed best among the other three. In the low-density scenario, all four algorithms generated deployment maps, with AM-NSGA-III-c still outperforming in overall performance. Both scenarios strongly demonstrate that the AM-NSGA-III-c algorithm (multi-population adaptive NSGA-III + spacing calibration mechanism) proposed in this invention achieves the best trade-off between latency, cost, and service quality in complex urban environments. Regardless of whether the scenario is high-density or low-density, this algorithm consistently achieves the lowest number of RSUs (lowest cost), completely eliminates spacing violations (no signal interference risk), accurately covers all latency-sensitive areas (lowest maximum latency), minimizes total latency, and has the most balanced load, significantly outperforming traditional NSGA-III, MOEA / D, and the uncalibrated version of AM-NSGA-III. This fully demonstrates the overwhelming technical advantages and industrial application value of this invention in real-world urban vehicular networks.

[0092] In high-density scenarios (such as Tianfu Square): As shown in Table 1, the algorithm outperforms the comparison algorithms (NSGA-III, MOEA / D) in terms of latency, cost and service quality, and the IGD index is reduced by 15-25%.

[0093] Table 1: Algorithm Performance Comparison for High-Density Scenes (e.g., Tianfu Square):

[0094] Performance indicators The algorithm of this invention (AM-NSGA-III-c) Comparison with Algorithm 1 (NSGA-III) Comparison Algorithm 2 (MOEA / D) Number of vehicles providing services in each scenario (in vehicles) 56704 56704 56704 Scene grid division (number) 50×50 (2500 decision variables) 50×50 (2500 decision variables) 50×50 (2500 decision variables) Optimal total delay (seconds) 202880.4 (IBRSG strategy) 268975.4 No feasible solution (NFS=0) Minimum number of RSUs to deploy (units) 18 24 No solution Maximum delay (seconds) in sensitive area 0.735~0.857 0.747~0.843 No solution Number of Pareto optimal solutions 22.16 (mean of 25 experiments) 10.56 (mean of 25 experiments) 0 IGD (Inverse Generation Distance) 0.2408 (95% confidence interval [0.2190, 0.2625]) 0.301 \

[0095] In low-density scenarios (such as Du Fu Thatched Cottage): As shown in Table 2, the algorithm remains efficient, with better solution set diversity (HV improved by 10-20%), and adapts to sparse traffic.

[0096] Table 2: Algorithm Performance Comparison for Low-Density Scenarios (e.g., Du Fu Thatched Cottage)

[0097] Performance indicators The algorithm of this invention (AM-NSGA-III-c) Comparison with Algorithm 1 (NSGA-III) Comparison Algorithm 2 (MOEA / D) Number of vehicles providing services in each scenario (in vehicles) 36405 36405 36405 Percentage of sensitive areas (%) 5 (corresponding to 125 grids) 5 (corresponding to 125 grids) 5 (corresponding to 125 grids) HV (Hypervolute, Solution Variety) 0.8312 0.755 0.692 Number of feasible solutions 19.88 (mean of 25 experiments) 13.28 (mean of 25 experiments) 8.96 (mean of 25 experiments) Minimum number of RSUs to deploy (units) 18 (uniform distribution) 24 (Locally overly dense) No effective deployment Data offloading load variance reduction 25% (relative to traditional strategies) No optimization No convergent solution

[0098] Sensitivity analysis: Adjust the proportion of sensitive areas (5-20%), as shown in Table 3.

[0099] Table 3: Sensitive Area Proportion Adjustment Value stability table:

[0100] Percentage of sensitive areas (%) Number of grid cells in sensitive areas <![CDATA[Objective function f2 (maximum delay in the sensitive area)]]> Delay constraint threshold (seconds) Does the constraint satisfy? 5 125 0.735~0.812 2 yes 10 250 0.747~0.835 2 yes 15 375 0.781~0.853 2 yes 20 500 0.805~0.857 2 yes

[0101] The robustness of the algorithm is verified using Table 3. The value remains stable within the threshold.

[0102] The above are merely preferred embodiments of the present invention and do not constitute any limitation on the present invention. Any equivalent substitutions or modifications made by those skilled in the art to the technical solutions and content disclosed in the present invention without departing from the scope of the present invention shall be deemed to have remained within the protection scope of the present invention.

Claims

1. A method for deploying roadside units of an urban vehicle-mounted network, characterized in that, Comprising the following steps: S1: grid processing of the target city area, taking each grid as a potential RSU deployment location, and establishing a multi-objective optimization model based on identified obstacles and delay-sensitive areas; S2: configuring various parameters required for algorithm running, and randomly generating an initial population, which is evenly divided into multiple sub-populations; S3: multiple sub-populations run NSGA-III algorithm synchronously and independently and perform adaptive parameter adjustment, multiple sub-populations iterate multiple times, and collaborative optimization between sub-populations is completed through migration operation until the maximum iteration number is reached or the convergence condition is met; S4: merging all sub-populations, screening all Pareto optimal solutions from the final population through non-dominated sorting, and outputting a complete and decision-making Pareto optimal deployment scheme set based on the optimal solutions; S5: selecting a deployment scheme and running a data offloading strategy; S6: multi-dimensional verification of the running IBRSG strategy to provide guidance for parameter setting in actual application; In S3, the iterative optimization of multiple sub-populations specifically comprises the following steps: S31: Calculate total delay with IBRSG strategy , maximum delay of sensitive area and number of RSUs ; S32: running NSGA-III algorithm to generate offspring through selection, crossover and mutation operations; S33: adopting a calibration mechanism to process density constraints: S34: dynamically adjusting the crossover rate and the mutation rate wherein the crossover rate is increased if the current migration operation did not improve the sub-population and the mutation rate is decreased if the current migration operation improved the sub-population ; and the mutation rate is decreased if the current migration operation improved the sub-population S35: before 10-20% individuals of each sub-population are migrated to other sub-populations every 5-10 generations, replacing the worst 20-30% individuals in the sub-population to promote population diversity.

2. The urban vehicular network road side unit deployment method of claim 1, wherein, In the S1, the gridding processing specifically includes: dividing a target urban area into K grids, each grid k corresponding to a variable When is 1, it indicates that the RSU has been deployed; when is 0, it indicates that the RSU has not been deployed; The multi-objective optimization model includes a minimum target function of total delay, a minimum target function of maximum delay in sensitive areas, a minimum target function of RSU number, and obstacle constraints and minimum spacing constraints.

3. The urban vehicular network road side unit deployment method of claim 2, wherein, Total delay The minimization objective function includes transmission delay, processing delay, queue delay, and propagation delay, expressed as: ； wherein, for a vehicle at a time The latency, delay calculation is based on vehicle trajectory and RSU coverage, using Shannon formula to estimate transmission rate; Minimized sensitive zone maximum delay For ensuring communication reliability for critical areas, expressed as: ； The target function guarantees the QoS of high-priority areas; The number of RSUs A minimization objective function for controlling construction and maintenance costs is represented as: ； The multi-objective optimization model indirectly minimizes the budget through the target function while balancing the conflict between delay and delay targets; The obstacle constraints are expressed by binary variables representing obstacle positions, the constraints are expressed as ; The minimum spacing constraint is used to prevent RSUs from being overly dense, avoid signal interference, by setting a minimum RSU spacing is 50 meters, the constraint is expressed as: where, is the Euclidean distance.

4. The urban vehicular network road side unit deployment method of claim 1, wherein, The parameters to be set in S2 include: total delay , maximum delay of sensitive area , number of RSUs , obstacle position , distance constraint , population size N is between 360 and 500, subpopulation number m=3-5, and maximum iteration number G=50-100; Randomly generate a population P, which is divided into m sub-populations, and evaluate the feasibility and target value of the initial solution.

5. The method of claim 1, wherein, In S33, the calibration mechanism specifically comprises the following steps: S331: Calculate all RSU pair distances, identify RSU pairs that violate minimum RSU pair distance S331: Calculate all RSU pair distances, identify RSU pairs that violate minimum RSU pair distance S332: according to the traffic volume ranking in the coverage range, retain the RSU with relatively large traffic volume, and remove another RSU to avoid over-dense deployment; S333: re-evaluate the target function after calibration to ensure the feasibility of the solution.

6. The urban vehicular network road side unit deployment method of claim 1, wherein, The S3 further comprises a constraint handling mechanism, which is specifically: The comparative method handles the constraint violation solution, allowing the quality of the solution to be compared under constraint relaxation, denoted as: ； wherein, is a dynamic tolerance threshold, which is a non-negative real number and is adjusted adaptively in the algorithm iteration; is a constraint violation degree, which is used to quantify the distance from the solution to the feasible region and is calculated based on the summation of the violation degrees of inequality constraints and equality constraints; denotes that solution A dominates solution B, i.e. all objective function values of A are not inferior to B and at least one objective is strictly better.

7. The method of claim 1, wherein, S4 specifically is: after algorithm termination iteration, merging all sub-populations, screening all Pareto optimal solutions from the final population through non-dominated sorting, decoding these optimal solutions into a specific RSU position coordinate list, calculating the performance index corresponding to each deployment scheme, and finally outputting a complete and decision-making Pareto optimal scheme set.

8. The method of claim 1, wherein, In S5, the data offloading strategy comprises the following steps: S51 : initialize random policy and null policy ; S52: sequentially update the connection strategy of each vehicle, and select the RSU that minimizes the total delay; S53: iterate until the strategy change is less than the error threshold, i.e. error=0.01, to ensure convergence to a stable Nash equilibrium.

9. The method of claim 1, wherein, The S6 is specifically: before and after deployment, by testing in different urban scenes, and compared with traditional algorithms, the comprehensive performance of the scheme is evaluated, at the same time, the influence of key parameters on the results is analyzed to verify the robustness of the scheme and provide guidance for parameter setting in actual application.

Images