RSU deployment optimization method based on MF-GAPSO algorithm

By adopting the improved GAPSO algorithm of the Memetic framework in the Internet of Vehicles environment, combining OpenStreetMap and SUMO technologies, RSU deployment is optimized, and the coverage and cost problems in RSU deployment are solved, efficient and stable RSU deployment is achieved, and the adaptability is significantly improved. and convergence speed.

CN118984460APending Publication Date: 2024-11-19HEILONGJIANG UNIVERSITY OF SCIENCE AND TECHNOLOGY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411100667.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-12
Publication Date
2024-11-19

AI Technical Summary

Technical Problem

The deployment of RSUs in the Internet of Vehicles environment is difficult to maximize coverage and minimize coverage overlap within cost-effectiveness, and traditional group intelligence algorithms are prone to fall into local optimal solutions, have low adaptability, and are difficult to implement in real-world maps. Efficient deployment.

Method used

Using the GAPSO algorithm based on the Memetic framework, the traditional GAPSO algorithm is improved by introducing the local search process, and the MF-GAPSO algorithm is designed, combined with OpenStreetMap and SUMO technology, the deployment location of RSU in the road network is optimized, coverage is maximized and coverage overlap is minimized. .

Benefits of technology

It improves the fitness, convergence speed and stability of RSU deployment, significantly improves the quality of combination optimization, and realizes efficient RSU deployment in real-world maps, which is better than other group intelligent algorithms in comparison.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118984460A_ABST
    Figure CN118984460A_ABST
Patent Text Reader

Abstract

The invention discloses an RSU deployment optimization method based on an MF-GAPSO algorithm. The method comprises the following steps: step 1, constructing a system model; step 2, based on a Memetic framework, designing an MF-GAPSO algorithm; and step 3, an MF-GAPSO algorithm is used in an Internet of Vehicles environment, and an OpenStreetMap and SUMO combined conversion map file technology is adopted to determine the optimal position of the RSU on the map. According to the method, the swarm intelligence algorithm is combined and optimized, a more scientific, reasonable and efficient optimization decision model is established, and the problems that the RSU is difficult to deploy in a real world map and a traditional algorithm is low in fitness, prone to falling into a local optimal solution and poor in deployment effect in the RSU deployment process are effectively solved. The MF-GAPSO algorithm provided by the invention is superior to other comparative swarm intelligence algorithms in the aspects of final fitness, convergence speed, stability, exploration and utilization balance.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to a roadside unit (RSU) deployment method, and in particular to an RSU deployment optimization method based on a Memetic framework-based GAPSO (MF-GAPSO) algorithm. Background Art

[0002] Vehicular Ad Hoc Networks (VANETs) are self-organizing networks that use wireless modes to propagate information between driving cars, transmitting data in a multi-hop manner between vehicles. However, the fundamental nature of traffic systems is highly uncertain, and the dynamic changes in the relative positions and time of fast-moving vehicles will cause corresponding changes in the communication topology. When the effective communication range between vehicles is exceeded, the multi-hop strategy used for data communication fails, making VANETs unsuitable for vehicular communication and unstable.

[0003] In recent years, the Internet of Vehicles (IoV) has attracted great attention from the academic community and has become a very promising application of the Internet of Things (IoT). IoV refers to the mutual communication between intelligent systems (such as RSU) running on advanced network infrastructure, other embedded processing units in vehicles, and various cyber-physical systems (such as sensors, radars, mobile devices, etc.), which can realize information exchange and promote vehicle-to-anything (V2X) communication, thus contributing to a safer and more efficient transportation system.

[0004] Various devices such as sensors and cameras equipped on vehicles collect route information as well as the vehicle's position and acceleration, which can be exchanged with neighboring vehicles or RSUs. This distributed network of smart vehicles shares data among themselves through vehicle-to-vehicle (V2V) communication or with RSUs through vehicle-to-infrastructure (V2I) communication.

[0005] RSUs are high-end processing and communication devices installed near roads, with stronger communication capabilities than vehicles. They act as communication gateways with Internet backhaul access rights and play a key role in V2I communications. They are responsible for collecting traffic flow information and road anomaly information, that is, during vehicle driving, through information interaction with RSUs, real-time road traffic conditions are obtained, vehicle information is released, and vehicles communicate with external networks as intermediaries. The optimized deployment of RSUs enables high-performance communication between vehicles and between vehicles and RSUs. However, the installation and maintenance costs of RSUs in the Internet of Vehicles are high, so it is necessary to balance the optimal network coverage and deployment costs in order to reasonably deploy RSUs. This situation is particularly serious in urban areas, because the cost of deploying many RSUs to cover the entire urban area is too high. In addition, the location and number of RSUs also affect quality of service (QoS) parameters such as network latency.

[0006] The optimized deployment of RSU can improve network efficiency, facilitate the rational use of resources, and expand the coverage area, which is an important task in the Internet of Vehicles environment. Therefore, efficient Internet of Vehicles requires the optimization of RSU deployment in a given road layout. At the same time, how to deploy RSU in the optimal position in the real-world roadmap has become a difficult problem faced by the Internet of Vehicles field.

[0007] Swarm intelligence algorithms are the mainstream solutions for RSU deployment optimization, such as genetic algorithms (GA), particle swarm optimization (PSO), and genetic algorithms with random restart hill climbing (GARHC). In order to improve search capabilities and optimization quality, researchers have proposed a series of improved methods based on the above algorithms, which have achieved good results in the combinatorial optimization solution process. However, existing studies still have room for further improvement in the search quality and fitness of the combined process when constructing a combinatorial optimization solution model.

[0008] Among the numerous swarm intelligence algorithms, the GA algorithm is favored in various optimization problems due to its few parameters, fast convergence, and simple calculation. However, some studies have not improved the traditional GA algorithm, and may fall into the local optimal solution during the search process; some studies have improved the GA algorithm but the effect is not very ideal; some studies have not applied RSU deployment in real-world maps. In order to improve the performance of the GA algorithm, Agoston E Eiben et al. used a simple hill climbing algorithm as a local search process based on the GA algorithm, but there is still room for further improvement in terms of its fitness and other aspects. Summary of the invention

[0009] In order to deploy RSUs at intersections or roadsides in an efficient manner, maximize the coverage of RSUs within cost-effectiveness, and minimize the coverage overlap of different RSUs, the present invention combines the characteristics of the Internet of Vehicles communication environment and real-world maps to provide an RSU deployment optimization method based on the MF-GAPSO algorithm. This method improves network efficiency and optimizes network coverage in the Internet of Vehicles by placing RSUs at the optimal location of a given road network.

[0010] The objective of the present invention is achieved through the following technical solutions:

[0011] An RSU deployment optimization method based on the MF-GAPSO algorithm includes the following steps:

[0012] Step 1: Build a system model:

[0013] Step 1-1, Problem Description: The goal of RSU deployment is to create an RSU network in an environment with a given roadmap layout, thereby maximizing the coverage of RSUs and minimizing the coverage overlap of different RSUs;

[0014] Step 1-2: Build a system model based on the problem description;

[0015] Step 2: Based on the Memetic framework, the MF-GAPSO algorithm is designed by improving the traditional GAPSO algorithm. The specific steps of the MF-GAPSO algorithm are:

[0016] Step 2-1, initialize the population, global optimal position, optimal fitness, fitness list, and the initial solution space is composed of possible positions, probabilities, corresponding indicators, mutations and crossover random numbers;

[0017] Step 2-2, create a variable to store fitness and calculate the fitness of each individual in the current population P;

[0018] Step 2-3: Evaluate the individuals in the population according to the fitness function, then sort the population in a non-increasing manner and create a new population λ n ;

[0019] Step 2-4: Using the elite strategy, according to the selection ratio P s Select the individual with the highest fitness from the current population and copy it to the new population λ n Among these best individuals, two parents are randomly selected to pass their genes to the next generation;

[0020] Step 2-5: Select two parent individuals to perform single-point crossover to generate offspring individuals, where the chromosomes of both parents are randomly selected to ensure that the offspring contains genes from both parents;

[0021] Step 2-6: If the random number R m Less than or equal to the mutation probability P m , perform mutation operations on the offspring individuals to increase the diversity of the population, and add these individuals to the new population λ n middle;

[0022] Step 2-7: After the new individual is generated, according to the local search probability P l Choose whether to perform a local search for the individual. If the random number R l Less than or equal to P l , then perform local search for each individual to try to find a better solution. If R l Greater than P l , then no local search is performed on the current individual, but the solution of the current individual is kept unchanged;

[0023] Step 2-8: Update population P to new population λ n ;

[0024] Step 2-9, check and update the global optimal position and optimal fitness. If a better individual is found, record the optimal fitness of each generation in the optimal fitness list;

[0025] Step 2-10: Use the generated solution operator to initialize the particle position, individual optimal position and individual optimal fitness in PSO, and perform the same number of iterations as the GA algorithm. In each iteration, update the speed and position of each particle. If the new position fitness of the particle is higher, update the global optimal position and optimal fitness to evaluate the fitness of the generated new population again.

[0026] Step 2-11, if the maximum number of iterations G is reached, the set of the best R positions and the best fitness are output as the final solution, otherwise, repeat steps 2-1 to 2-10;

[0027] Step 2-12, draw a graph of the fitness changing with the number of iterations to visualize the optimization process and results of the algorithm;

[0028] Step 3: Use the MF-GAPSO algorithm in the Internet of Vehicles environment and adopt the OpenStreetMap and SUMO joint conversion map file technology to determine the optimal location of the RSU on the map. The specific steps are as follows:

[0029] Step 3-1: Based on the system model constructed in step 1, define the connected vehicle environment where the RSU is deployed;

[0030] Step 3-2: Use the MF-GAPSO algorithm designed in step 2 to output the number of nodes in the target area, the initial population, the final population, the set of optimal R positions, the optimal fitness value, and an image of the fitness changing with the number of iterations;

[0031] Step 3-3: First, obtain the map of the target area and the OSM file map.osm through OpenStreetMap; then use SUMO's netconvert tool to convert the map.osm file into a road network information file map.net.xml. After the conversion is completed, open it in SUMO's NETEDIT tool to obtain the road network of the target area for further viewing and processing, and use the processed map.net.xml in the algorithm code of step 3-2; finally, convert the map.net.xml file into a node information file map.nod.xml, open the map.nod.xml file, and display the id, coordinates and type of each node; the data of the optimal R position set generated by the MF-GAPSO algorithm represents the node number of the area, search for "node" in the map.nod.xml file, find the coordinates of the corresponding numbered node, find the node in NETEDIT according to the coordinates, check whether its id and type are consistent with those in the map.nod.xml file, and mark it at the same position in the corresponding deployment area map, and repeat the process until the final deployment location of all RSUs is determined.

[0032] Compared with the prior art, the present invention has the following advantages:

[0033] 1. The present invention combines and optimizes swarm intelligence algorithms to establish a more scientific, reasonable and efficient optimization decision model, which effectively solves the difficulties in deploying RSUs in real-world maps and the problems of low adaptability of traditional algorithms during RSU deployment, easy to fall into local optimal solutions and poor deployment effects.

[0034] 2. Since the PSO calculation process can be parallelized to speed up the search, the present invention proposes an optimal RSU deployment algorithm based on the Memetic framework, namely the MF-GAPSO algorithm. The traditional GAPSO algorithm uses the mutated solution operator of GA as the initial population of PSO, but the effect is not ideal. The present invention improves it based on the Memetic framework and introduces a local search process into the traditional GAPSO algorithm to improve fitness.

[0035] 3. The MF-GAPSO algorithm proposed in the present invention is superior to other swarm intelligence algorithms in terms of final fitness, convergence speed, stability, and balance between exploration and utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is the system model;

[0037] Figure 2 It is the flow chart of MF-GAPSO algorithm;

[0038] Figure 3 It is a simulation flow chart;

[0039] Figure 4 It is the SUMO road network in the Matsuura Avenue area;

[0040] Figure 5 It is the SUMO road network in the Central Street area;

[0041] Figure 6 The relationship between fitness and the number of RSUs deployed in the Songpu Avenue area;

[0042] Figure 7 It is the relationship between fitness and the number of RSUs deployed in the Central Street area;

[0043] Figure 8 Comparison of the fitness of each algorithm when iterating 20 times for the Songpu Avenue area;

[0044] Fig. 9 Comparison of the fitness of each algorithm when iterating 20 times for the Central Avenue area;

[0045] Fig.10 Iterated 20 RSU deployments for the Songpu Avenue area;

[0046] Fig.11 Iterated 20 RSU deployments for the Central Street area;

[0047] Fig.12 Comparison of the fitness of each algorithm when iterating 40 times for the Songpu Avenue area;

[0048] Fig.13 Comparison of the fitness of each algorithm when iterating 40 times for the Central Avenue area;

[0049] Fig.14 Comparison of the fitness of each algorithm when iterating 60 times for the Songpu Avenue area;

[0050] Fig.15 Comparison of the fitness of each algorithm when iterating 60 times for the Central Avenue area;

[0051] Fig.16 Iterated 40 RSU deployments for the Songpu Avenue area;

[0052] Fig.17 40 iterations of RSU deployment for the Central Avenue area;

[0053] Fig.18 Iterated 60 RSU deployments in the Songpu Avenue area;

[0054] Fig.19 RSU deployment was iterated 60 times for the Central Street area. DETAILED DESCRIPTION

[0055] The technical solution of the present invention is further described below in conjunction with the accompanying drawings, but is not limited thereto. Any modification or equivalent replacement of the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention should be included in the protection scope of the present invention.

[0056] The present invention provides an RSU deployment optimization method based on the MF-GAPSO algorithm. First, according to the problem description, a system model is constructed. Then, based on the Memetic framework, by improving the traditional GAPSO algorithm, the rationality of the algorithm combination optimization is improved, and an MF-GAPSO algorithm is designed, and the optimization problem is established and solved. Finally, a simulation experiment is carried out based on the road map of the real world - Songpu Avenue area in Songbei District and Central Street area in Daoli District, and the optimal RSU position with different iterations is displayed to obtain real results. The map, road network information and node information of the target area are generated by OpenStreetMap and SUMO, and an efficient solution for deploying RSU in the selected area is designed. The technical solution of the present invention is compared with the existing advanced swarm intelligence solution to prove the effectiveness of the proposed RSU deployment solution. The simulation results prove the superiority of the MF-GAPSO algorithm in RSU deployment optimization. In terms of improving fitness, the algorithm is better than the other four swarm intelligence algorithms, which effectively improves the quality of the combination process.

[0057] 1. Establishment and solution of optimization problem

[0058] 1. Problem description and system model

[0059] The goal of RSU deployment is to create an RSU network in an environment with a given roadmap layout, thereby maximizing the coverage of RSUs and minimizing the coverage overlap of different RSUs; in addition, the high installation and maintenance costs of RSUs limit the number of their use. According to the actual needs of the intersection, each intersection may require 1-2 RSU devices associated with traffic lights, cameras, etc. In view of the fact that the existing RSU deployment mechanism cannot achieve high efficiency and is not suitable for diverse real road network environments, the MF-GAPSO algorithm is proposed, which is based on the evolutionary computing part of artificial intelligence. For example, if a 4×4 road network consisting of 16 intersections is considered, then each intersection is a potential location for deploying RSUs. The present invention considers a road network in two areas in the real world, which is divided into different sections and consists of multiple lanes.

[0060] According to the deployment goal proposed in this invention, the MF-GAPSO algorithm is used in the Internet of Vehicles environment, and the OpenStreetMap and SUMO joint conversion map file technology are adopted to determine the optimal location of RSU on a given map. Different from the traditional GAPSO algorithm, the Memetic framework involves local search technology, which starts from a solution and explores the solutions in its neighborhood through small adjustments to find a better solution, making it efficient in achieving RSU deployment optimization. The system model is as follows Figure 1 As shown in the figure, the connected vehicle environment with RSU deployed is displayed, where the black circle represents the coverage of RSU.

[0061] 2. Algorithm description and optimization problem solving

[0062] Since the convergence accuracy of the traditional GAPSO algorithm mixed with the early GA algorithm in finding suboptimal solutions is not high, and the later PSO algorithm is prone to fall into the local optimal solution, the effect of the algorithm is not ideal. Therefore, the present invention improves it based on the Memetic framework, introduces local search into the GAPSO algorithm, uses the hill climbing algorithm to fine-tune the search process, and forms the MF-GAPSO mechanism, which is an extension of the traditional algorithm to improve fitness.

[0063] For the RSU deployment problem, the phenotype is the possible location where the RSU can be placed, the corresponding genotype representing the RSU location is a set of integers, and it is necessary to avoid assigning multiple RSUs to one location.

[0064] The input parameters of the MF-GAPSO algorithm include a set of k possible locations, the number of RSUs to be deployed R, the population size n, and the number of iterations G. It is also necessary to input a map of the target area, including the layout of different roads and intersections, and generate a map of the target area, road network information, and node information through OpenStreetMap and SUMO.

[0065] The concept behind the MF-GAPSO algorithm of the present invention is as follows:

[0066] The algorithm focuses on an objective function - the RSU coverage represented by θ, which aims to find the number of intersections covered by each RSU, because these intersections are possible deployment locations of RSUs. The RSU coverage in a given map layout is expressed as:

[0067]

[0068] Among them, R i and R j represents two different RSUs, numbered i and j, η i and η j Respectively represent R i and R jThe number of intersections covered. Once η is calculated for an RSU, it is not recalculated in the derivation of θ. i +η j -(η i ∩η j ) represents R i and R j The optimal number of intersections covered, N represents the total number of RSUs.

[0069] The fitness function is unique to each problem and represents how close the generated solution is to meeting the problem requirements. The output of the function is a number that should be maximized or minimized depending on the problem. The algorithm is to maximize the coverage of the RSU, so the fitness function is expressed as:

[0070]

[0071] After each iteration, the particle's velocity and position are updated, expressed as:

[0072]

[0073] This formula is used to calculate the velocity V of particle i at the m+1th iteration. i,m+1 Among them, V i,m is the velocity of the particle at the mth iteration; ω is the inertia weight, which is used to balance the previous velocity of the particle; c1 and c2 are acceleration coefficients, which are usually set to constants and are used to adjust the particle to the individual optimal position P best and the global optimal position G best speed; r1 and r2 are random numbers in the range [0,1] used to introduce randomness; h i,m is the position of the particle at the mth iteration.

[0074] h i,m+1 =h i,m +V i,m+1 (4)

[0075] This formula is used to calculate the new position h of particle i at the m+1th iteration. i,m+1 It is done by taking the particle's current position h i,m The new speed V calculated in formula (1) i,m+1 Obtained by addition.

[0076]

[0077] This formula is used to calculate the inertia weight ω, where ω max ,ω min They represent the initial weight and final weight respectively, iter represents the current number of iterations, and iter maxRepresents the maximum number of iterations, and the weight ω gradually decreases as the number of iterations increases.

[0078] After each iteration, the particle’s velocity and position are updated using equations (3) and (4), respectively, and the weight factor is updated according to equation (5).

[0079] The pseudo code of the MF-GAPSO algorithm that outputs R sets of positions is shown below.

[0080]

[0081] Step 1 represents the initialization phase, which initializes the population, global optimal position, optimal fitness, and fitness list. The initial solution space is composed of random arrays of possible positions, probabilities, corresponding indicators, mutations, and crossovers.

[0082] Steps 2 to 23 constitute the main iteration process of the GA algorithm. In step 3, a variable is created to store the fitness, and the fitness of each individual in the current population P is calculated. Step 4 evaluates the individuals in the population according to the fitness function, then sorts the population in a non-increasing manner and creates a new population λ n .

[0083] Steps 5 to 7 use the elite strategy, according to the selection ratio P s Select the individual with the highest fitness from the current population and copy it to the new population λ n In the process, two parents are randomly selected from the best individuals and their genes are passed on to the next generation, which ensures that the excellent individuals can be retained.

[0084] Steps 8 to 14 describe the crossover and mutation process of the population. First, two parent individuals are selected for single-point crossover to generate offspring individuals, where the chromosomes of both parents are randomly selected to ensure that the offspring contains genes from both parents. Then in step 11, if the random number R m Less than or equal to the mutation probability P m , perform mutation operations on the offspring individuals to increase the diversity of the population, and add these individuals to the new population λ n middle.

[0085] Steps 15 to 19 involve a local search process, where local search is applied to a portion of the new generation of individuals. Specifically, after the new individuals are generated, the local search probability P l Choose whether to perform a local search for the individual. If the random number R l Less than or equal to P l, then perform local search for each individual to try to find a better solution, because local search helps to find the local optimal value faster, and based on the local search probability, it can avoid the algorithm from falling into the local optimal solution and increase the chance of finding the global optimal solution; if R l Greater than P l , then no local search is performed on the current individual, but the solution of the current individual is kept unchanged.

[0086] The bitclimber function is used to perform local optimization on individuals. The specific local search process is as follows: select a current individual current as the starting point of the local search and iterate; in each iteration, apply the mutate function to the current individual to generate a candidate individual candidate. The mutate function simulates gene mutation by randomly selecting a gene in the individual and replacing it with a random position, trying to find a solution with higher fitness; if the fitness of the candidate individual is not lower than that of the current individual, the candidate individual is used as the new current individual, otherwise, the current individual is kept unchanged; after the iteration, the best individual found in the local search process is returned.

[0087] Step 20: Update the population P to the new population λ n , step 21 checks and updates the global optimal position and optimal fitness. If a better individual is found, step 22 records the optimal fitness of each generation into the optimal fitness list.

[0088] Steps 24 to 32 are the iterative process of the PSO algorithm, using the generated solution operator to initialize the particle position, individual optimal position, and individual optimal fitness in the PSO, and perform the same number of iterations as the GA algorithm. In each iteration, the speed and position of each particle are updated. In steps 28 and 29, if the fitness of the new position of the particle is higher, the global optimal position and optimal fitness are updated to evaluate the fitness of the generated new population again.

[0089] Steps 33 to 37 are the termination condition checks of the algorithm. If the maximum number of iterations G is reached, the set of the best R positions and the best fitness are output as the final solution. Otherwise, steps 2 to 33 are repeated.

[0090] Finally, step 38 is to draw a graph showing the fitness changing with the number of iterations to visualize the optimization process and results of the algorithm.

[0091] Through the above steps, the MF-GAPSO algorithm combines the advantages of the GAPSO algorithm and local search technology. Its randomness helps to find new potential deployment locations. Through operations such as elite strategy, selection, crossover, mutation, local search, and updating the speed position of particles, it finally returns a set of optimal R locations, aiming to find the optimal RSU deployment location to achieve maximum RSU coverage and network performance.

[0092] Figure 2 The process followed by the algorithm is shown. Initially, the RSUs are randomly deployed at various intersections and the fitness of the current population is determined using the fitness function. This is followed by the application of genetic operators: parent selection, crossover and mutation processes, followed by the replacement of a part of the current population with the generated offspring and the application of a local search process on the generated offspring. The iterative process of the PSO algorithm is then performed and a check is made as to whether the termination condition is reached, which is equal to the maximum number of iterations reached. If the termination condition is not met, the process is repeated. Finally, the optimal position of the deployment and an image of the fitness changing with the number of iterations Generation are output.

[0093] 2. Simulation Experiment

[0094] This section introduces the simulation environment used to check the performance of the algorithm. In terms of the site and cost involved, it is not feasible to deploy the vehicle network in real time, so a simulation experiment was chosen and run multiple times to determine the final result. The program simulation of different regional maps was performed using Python, and the processed data was visualized using Matplotlib, NumPy, and SciPy. The experiment selected five algorithms from the swarm intelligence algorithm for comparison, including three algorithms based on the Memetic framework and two traditional algorithms.

[0095] In order to conduct a detailed evaluation and diversity of results, and to test the performance of the algorithm in different road networks, real-world locations were selected: Songpu Avenue area in Songbei District and Central Street area in Daoli District as target areas for RSU deployment. These areas were selected because of their diversity and different structural characteristics: the Songpu Avenue area represents a situation with a small number of road intersections and a sparse road network; the Central Street area represents a dense road network layout consisting of a large number of intersections, which provides more possibilities for deploying RSUs. These locations help to verify the potential of the algorithm proposed in this invention to select the optimal location of RSUs under different conditions, and finally reasonably deploy RSUs in these two areas.

[0096] The proposed modules are as follows:

[0097] Map conversion module: This module uses OpenStreetMap to generate OSM files, and uses SUMO's netconvert command to convert OSM files into road network files map.net.xml and node files mapnodes.nod.xml, thereby generating a map of the target area to enable the deployment of RSU on the real-world road layout. OpenStreetMap, referred to as OSM, is an open source, editable world map project that aims to create a detailed and updated map, including but not limited to information such as roads, water bodies, buildings, and public facilities; the open source software SUMO is a realistic traffic generator used for urban traffic simulation, and NETEDIT is a tool in SUMO for creating and editing traffic networks, which helps to visualize routes and road networks.

[0098] RSU deployment module based on Memetic framework: This module uses MF-GAPSO algorithm to deploy RSU in the optimal position, thereby increasing RSU coverage and improving network efficiency. The simulation process is as follows: Figure 3 shown.

[0099] The SUMO road networks of the two regions are as follows: Figure 4 and Figure 5 shown.

[0100] Various simulation parameters are defined below to guide the Memetic framework search process, assuming that the coverage area of ​​each RSU is equidistant. These parameters are summarized in Table 1.

[0101] Table 1 Simulation parameters

[0102]

[0103] Table 2 shows the road information of the selected locations obtained from SUMO. The use of different road maps helps to verify the potential of the algorithm to deploy the optimal locations of RSUs in different situations.

[0104] Table 2 Detailed information of regional layout

[0105]

[0106] The nodes represent the intersections and the starting and ending points of roads in the transportation network, and the edges represent the road segments connecting the nodes.

[0107] The algorithm performance is improved by using local search to find the best possible location of RSU among all available locations. Several swarm intelligence algorithms already exist to suggest the best possible location of RSU in a given area map. The performance of MF-GAPSO algorithm is compared with MFRD, GAPSO, GA and GARHC algorithms for two-area layout.

[0108] The overall RSU deployment stage of the road area is related to the density of the road network. When deploying, not only should the sections with higher traffic density be considered, but also the intersections with large traffic volume cannot be completely installed. Since RSU deployment and maintenance are expensive, the average cost of an RSU is 20,000 yuan, so the number of RSUs also needs to be considered. First, consider the coverage area of ​​each RSU (see Table 1), the number of regional nodes (see Table 2), the selected regional scope (see Table 3) and cost-effectiveness, and determine how many RSUs are most appropriate to deploy in the two areas, such as Figure 6 and Figure 7 shown.

[0109] Figure 6 It means that the adaptability is highest when 14 RSUs are deployed in the Songpu Avenue area. Figure 7 It indicates that the fitness is highest when 23 RSUs are deployed in the Central Street area. As can be seen from the figure, the increase in the number of RSUs has increased the fitness of both areas. This result is more obvious in the Central Street area because it has a denser road network and contains more intersections, which provides the possibility of using RSUs to cover more intersections.

[0110] The GA algorithm is a traditional genetic algorithm; the GAPSO algorithm is a traditional hybrid genetic particle swarm algorithm; the MFRD algorithm and the GARHC algorithm integrate the hill climbing algorithm and the random restart technology into the local search process of the GA algorithm respectively; the MF-GAPSO algorithm integrates the hill climbing algorithm into the local search process of the GAPSO algorithm. The three algorithms based on the Memetic framework also have the following differences: MF-GAPSO and MFRD algorithms both use the bitclimber function to implement local search, and simulate individual mutations based on local mutation probabilities; the GARHC algorithm uses the hillClimbing function to implement local search, randomly generates neighbor solutions of the current solution, and explores the solution space by exchanging positions, which does not depend on the mutation probability. These algorithms all follow an evolutionary process, which is represented in the form of optimal and average chromosome fitness in two different road layouts. The graphs of fitness changing with the number of iterations are shown below. Figure 8 and Fig. 9 shown.

[0111] from Figure 8 and Fig. 9It can be seen that for the above five algorithms, initially, it can be observed that the fitness has improved. As the number of iterations increases, the fitness gradually converges to the global optimal value. For these two areas, the performance of the algorithm is analyzed from the aspects of final fitness, convergence speed, stability, and exploration and utilization balance. The final fitness refers to the highest fitness value reached by the algorithm during iteration; the convergence speed reflects how fast the fitness value of the algorithm improves during the iteration process; stability can be evaluated by the smoothness of the average fitness curve of the algorithm; exploration refers to the ability of the algorithm to find new solutions in the solution space, and utilization refers to the ability of the algorithm to use known information to improve the current solution.

[0112] (1) Final fitness: The MF-GAPSO algorithm performs best by providing the highest fitness value, thereby finding a better deployment location for the RSU. It tries to find the optimal solution during the iteration process, and its fitness value is much higher than that of other algorithms, indicating that the algorithm's optimization ability is significantly higher than other algorithms in the experiment.

[0113] (2) Convergence speed: The MF-GAPSO algorithm fine-tunes the search process. As the number of iterations increases, the fitness curve rises rapidly, and a higher fitness value is obtained within a fewer number of iterations, and tends to be flat in the later stage, indicating that the algorithm has a fast convergence speed, high efficiency, and can quickly approach the optimal solution.

[0114] (3) Stability: The average fitness curves of the five algorithms all show a certain degree of smoothness, but the average fitness curves of MF-GAPSO, GAPSO and GA algorithms are smoother, indicating that their performance is more stable during the iteration process.

[0115] (4) Exploration and exploitation balance: The fitness initially increases continuously and then gradually approaches the global optimum. In some cases, a sudden increase in fitness can be seen, which can be attributed to the mutation operator and the local search process that allows searching for new maxima in the solution space.

[0116] The optimal fitness values ​​of the RSUs in the two regions after 20 iterations are recorded in Table 3.

[0117] Table 3 The optimal fitness values ​​of RSU in two regions after 20 iterations

[0118]

[0119] It can be seen from Table 3 that the coverage of the Central Avenue area is larger, and as mentioned above, its road network is denser. At the same time, the optimal fitness value obtained is also increased compared with the Matsuura Avenue area.

[0120] In general, the algorithms based on the Memetic framework generally outperform the traditional algorithms. The traditional algorithms cannot produce excellent results in complex combinatorial search environments. In addition, the MF-GAPSO algorithm proposed in this invention performs better than the MFRD and GARHC algorithms in terms of fitness in the two regions. The reason why it performs best is that it combines the global search capability of the GAPSO algorithm with the fine-tuning capability of the local search of the Memetic framework, and further strengthens the exploration and utilization of the optimal solution through the hill climbing algorithm.

[0121] After 20 iterations, the final RSU deployment locations in the two areas are as follows: Fig.10 and Fig.11 As shown. Fig.10 and Fig.11 It can be seen that after 20 iterations, some RSUs have a large overlapping coverage area, and the overall coverage is not comprehensive, and the number of iterations is insufficient. In order to better illustrate the evolution of the algorithm, 40 and 60 iterations are also selected to demonstrate the optimal position of the RSU deployment by the MF-GAPSO algorithm.

[0122] Figure 12 to Figure 15 The fitness comparison of each algorithm for RSU in two regions at different numbers of iterations is shown. Figure 16 to Figure 19 The optimal location determination of the RSUs in the two regions is shown for different numbers of iterations. It can be observed that when considering 23 RSUs compared to 14 RSUs, the number of potential deployment locations also increases, so more iterations are required to converge. At the same time, as the number of RSUs increases, the search space also increases. Due to the randomness of the algorithm, it is possible to find the optimal solution in the later generations. Figure 16 to Figure 19 This is illustrated because the MF-GAPSO algorithm tries to find the optimal solution until the last generation, making the optimal position of the deployment constantly changing.

[0123] Finally, it can be seen from the simulation experiment that the application of the MF-GAPSO algorithm based on the Memetic framework can obtain the optimal results.

Claims

1. A RSU deployment optimization method based on MF-GAPSO algorithm, characterized in that The method comprises the following steps: Step 1: Build a system model; Step 2: Based on the Memetic framework, the MF-GAPSO algorithm is designed by improving the traditional GAPSO algorithm. The specific steps of the MF-GAPSO algorithm are: Step 2-1, initialize the population, global optimal position, optimal fitness, fitness list, and the initial solution space is composed of possible positions, probabilities, corresponding indicators, mutations and crossover random numbers; Step 2-2, create a variable to store fitness and calculate the fitness of each individual in the current population P; Step 2-3: Evaluate the individuals in the population according to the fitness function, then sort the population in a non-increasing manner and create a new population λ n ; Step 2-4: Using the elite strategy, according to the selection ratio P s Select the individual with the highest fitness from the current population and copy it to the new population λ n Among these best individuals, two parents are randomly selected to pass their genes to the next generation; Step 2-5: Select two parent individuals to perform single-point crossover to generate offspring individuals, where the chromosomes of both parents are randomly selected to ensure that the offspring contains genes from both parents; Step 2-6: If the random number R m Less than or equal to the mutation probability P m , perform mutation operations on the offspring individuals to increase the diversity of the population, and add these individuals to the new population λ n middle; Step 2-7: After the new individual is generated, according to the local search probability P l Choose whether to perform a local search for the individual. If the random number R l Less than or equal to P l , then perform local search for each individual to try to find a better solution. If R l Greater than P l , then no local search is performed on the current individual, but the solution of the current individual is kept unchanged; Step 2-8: Update population P to new population λ n ; Step 2-9, check and update the global optimal position and optimal fitness. If a better individual is found, record the optimal fitness of each generation in the optimal fitness list; Step 2-10: Use the generated solution operator to initialize the particle position, individual optimal position and individual optimal fitness in PSO, and perform the same number of iterations as the GA algorithm. In each iteration, update the speed and position of each particle. If the new position fitness of the particle is higher, update the global optimal position and optimal fitness to evaluate the fitness of the generated new population again. Step 2-11, if the maximum number of iterations G is reached, the set of the best R positions and the best fitness are output as the final solution, otherwise, repeat steps 2-1 to 2-10; Step 2-12, draw a graph of the fitness changing with the number of iterations to visualize the optimization process and results of the algorithm; Step 3: Use the MF-GAPSO algorithm in the Internet of Vehicles environment and adopt the OpenStreetMap and SUMO joint conversion map file technology to determine the optimal location of the RSU on the map.

2. The RSU deployment optimization method based on the MF-GAPSO algorithm according to claim 1 is characterized in that The specific steps of step one are as follows: Step 1-1, Problem Description: The goal of RSU deployment is to create an RSU network in an environment with a given roadmap layout, thereby maximizing the coverage of RSUs and minimizing the coverage overlap of different RSUs; Step 1-2: Build a system model based on the problem description.

3. The RSU deployment optimization method based on the MF-GAPSO algorithm according to claim 1 is characterized in that The specific steps of step 3 are as follows: Step 3-1: Based on the system model constructed in step 1, define the connected vehicle environment where the RSU is deployed; Step 3-2: Use the MF-GAPSO algorithm designed in step 2 to output the number of nodes in the target area, the initial population, the final population, the set of optimal R positions, the optimal fitness value, and an image of the fitness changing with the number of iterations; Step 3-3: First, obtain the map of the target area and the OSM file map.osm through OpenStreetMap; then use SUMO's netconvert tool to convert the map.osm file into a road network information file map.net.xml. After the conversion is completed, open it in SUMO's NETEDIT tool to obtain the road network of the target area, conduct further viewing and processing, and use the processed map.net.xml in the MF-GAPSO algorithm; finally, convert the map.net.xml file into a node information file map.nod.xml, open the map.nod.xml file, and display the id, coordinates and type of each node; the data of the optimal R position set generated by the MF-GAPSO algorithm represents the node number of the area, search for "node" in the map.nod.xml file, find the coordinates of the corresponding numbered node, find the node in NETEDIT according to the coordinates, check whether its id and type are consistent with those in the map.nod.xml file, and mark it at the same position in the corresponding deployment area map, and repeat the process until the final deployment location of all RSUs is determined.