Automatic driving expressway linear design method based on control simulation and multi-objective optimization

Abstract

The invention discloses an automatic driving expressway linear multi-target automatic line selection method based on control and simulation optimization. The method is based on a road design platform and a vehicle dynamics simulation platform, is used for verifying the influence of road geometric parameters on automatic driving vehicle dynamics, and is combined with a plurality of automatic driving path tracking control algorithms to dynamically select the most suitable control mode for different road curvatures, vehicle speeds and environmental conditions. Road design optimization is carried out through a multi-target optimization algorithm and an evaluation algorithm, the targets of safety, comfort, environmental protection and the like are balanced, a comprehensive optimal design line shape scheme can be automatically selected and designed, and the defects of subjectivity and single-target consideration existing in traditional manual line selection and alignment are overcome. According to the method, the adaptability of different path tracking algorithms to a design scheme is verified by combining sensitivity analysis. Organic combination of scientific decision and engineering practice is realized, and the intelligent level and engineering feasibility of the expressway line shape are remarkably improved while the design efficiency is ensured.

Description

Technical Field

[0001] The present invention belongs to the technical field of autonomous driving and intelligent transportation, and particularly relates to an autonomous driving highway alignment design method based on multi-modal control algorithms, high-precision simulation verification, and multi-objective optimization, aiming to optimize the autonomous driving road design method and improve the safety, passenger comfort, and ecological adaptability of autonomous driving vehicles traveling thereon. Background Art

[0002] As the core driving force of intelligent transportation systems, autonomous driving technology has made breakthroughs in recent years in areas such as environmental perception, decision-making and planning, and vehicle control. Low-level autonomous driving vehicles have gradually achieved commercial applications, and high-level autonomous driving systems have also entered the large-scale testing stage, marking the transition of intelligent connected vehicle technology from the laboratory to actual roads. However, traditional highway alignment design methods have long been constructed based on the behavioral characteristics of human driving, and their geometric parameter selection, safety evaluation systems, and design specifications are all based on the physiological and psychological responses of human drivers. With the acceleration of the large-scale deployment process of autonomous driving vehicles, the structural contradiction between existing road infrastructure and vehicle intelligentization needs has become increasingly prominent: the path tracking control algorithms relied on by autonomous driving vehicles (such as model predictive control and sliding mode control) are highly sensitive to geometric parameters such as road curvature, slope, and superelevation, while traditional design methods do not fully consider such dynamic coupling relationships, resulting in an increase in vehicle trajectory tracking errors and a decrease in lateral stability. Especially in complex alignment sections, it is easy to cause control command oscillations, severely restricting the full performance of autonomous driving systems.

[0003] Current highway alignment design methods have significant limitations both theoretically and methodologically. From a theoretical perspective, existing design specifications are dominated by engineering economy and construction feasibility, lacking a quantitative collaborative optimization mechanism for multi-dimensional objectives such as safety, comfort, and environmental protection, and it is difficult to meet the advanced requirements of intelligent transportation systems for energy efficiency management, carbon emission control, and passenger experience. Methodologically, the design process overly relies on static mechanical models and empirical formulas, and a high-precision vehicle-road joint simulation platform has not been constructed, unable to quantitatively analyze the influence laws of road geometric parameters on vehicle dynamic responses (such as yaw rate and lateral acceleration), resulting in a lack of data support for safety and comfort evaluation indicators. In addition, the traditional design process still centers around trial-and-error iteration dominated by manual experience, without introducing intelligent algorithms to achieve automated optimization of alignment parameters and comprehensive evaluation of multiple schemes, making it difficult to adapt to the need to improve the design efficiency of large-scale road networks and unable to respond to the industry trend of deep integration of artificial intelligence technology.

[0004] To address the above bottlenecks, it is urgent to break through the theoretical framework of traditional design methods and construct a highway alignment design method that integrates dynamic adaptation of multi-modal control strategies, high-precision joint simulation verification, and multi-objective intelligent decision-making. This method needs to reduce the dynamic response deviation of vehicles through the closed-loop interaction between the path tracking control algorithm and road geometric parameters, integrate safety, comfort, and environmental protection indicators, establish a data-driven comprehensive evaluation system, use high-fidelity simulation platforms such as CarSim / Simulink to analyze the mapping relationship between road parameters and vehicle dynamics characteristics, and introduce the multi-objective genetic algorithm (NSGA-II) and the CRITIC objective weighting method to realize the automatic generation and optimization of alignment schemes. Through the deep integration of the dynamic switching mechanism of multi-modal control strategies, the vehicle-road joint simulation verification system, and the multi-objective collaborative optimization framework, this method aims to significantly improve the adaptability of road alignments and autonomous driving systems, providing key technical support for the standardization and intelligent upgrade of intelligent transportation infrastructure. Summary of the Invention

[0005] In view of the above problems, the present invention proposes an autonomous driving highway alignment design method based on control simulation and multi-objective optimization for highway alignment design. It can comprehensively consider safety, comfort, and ecological comprehensive objective functions, and automatically select and determine the route. It not only combines various autonomous driving path tracking algorithm technologies to closely fit the driving characteristics of autonomous driving vehicles, but also fully simulates and analyzes the relationship between the engineering characteristics, shape characteristics of different road design schemes and the optimal solution, ensuring that the obtained optimal alignment is suitable for the engineering characteristics. The alignment design method provided by the present invention has strong robustness and generalization, and can be directly combined with existing road design software as an integrated automatic design solution method.

[0006] The autonomous driving highway alignment design method designed by the present invention mainly relies on the following devices and systems: a road design platform, an autonomous driving simulation platform, an algorithm integration and simulation platform, and a platform that supports the execution of multi-objective optimization and evaluation algorithms. Through the coordinated action of these devices, the present invention realizes the autonomous driving highway alignment design method based on control simulation and multi-objective optimization. The specific implementation steps are as follows:

[0007] An intelligent highway alignment design method for autonomous driving, including the steps of:

[0008] S1: Based on road design software, conduct preliminary route determination to obtain a series of candidate alignment sets;

[0009] S2: Perform parameterization processing on the alignment;

[0010] S3: Import the parameterized alignment into the vehicle simulation platform and conduct simulation using the autonomous driving path tracking algorithm;

[0011] S4: Analyze the simulation results using a multi-objective optimization algorithm, construct the objective function and solve it;

[0012] S5: Establish a comprehensive route evaluation system based on the multi-objective optimization algorithm and the evaluation algorithm to screen the optimal alignment scheme.

[0013] Furthermore, in S1, use Autodesk Civil3D for preliminary road alignment design. First, determine the starting and ending points of the highway, and select a preliminary alignment scheme in combination with the road functional requirements. Then, optimize using the straight-spiral-circular design mode, set curve segments with different radii to ensure smooth transition of each curve segment, and adjust its parameters to meet the design specification requirements. Finally, obtain candidate alignments that meet the actual engineering requirements.

[0014] Furthermore, in S2, perform parametric modeling on the candidate alignments, extract the three-dimensional coordinate sequences of each alignment, convert the data into a standardized format through the Civil3D secondary development interface, integrate additional parameters, and output it as an input file compatible with the CarSim / Simulink co-simulation platform.

[0015] Furthermore, in S3, in the CarSim / Simulink co-simulation platform, import the parametric alignment data, and select pure pursuit control, model predictive control, and sliding mode control algorithms for simulation using the autonomous driving path tracking algorithm. Among them, the pure pursuit control algorithm uses a simplified two-wheeled vehicle model, does not consider the slip between the tire and the ground, and selects the look-ahead distance L = 5m to obtain the optimal control effect; the model predictive control algorithm simplifies the vehicle into a three-degree-of-freedom single-track model, adopts the small-angle assumption, and establishes a controller model in MATLAB / Simulink; the sliding mode control algorithm designs the switching function as Determine the system parameters as: c1 = 2, k1 = 0.1, k2 = 5 to reduce the chattering of the controller; so that by adopting different path tracking algorithms, it is ensured that the path control can be optimized in complex traffic environments and at different vehicle speeds, enhancing the driving stability and responsiveness of autonomous vehicles.

[0016] Furthermore, in S4, for multi-objective optimization modeling, select safety, comfort, and ecological environmental protection as the objective functions, and use the non-dominated genetic algorithm NSGA-II for solution, with a population size of 10,000 and a maximum number of iterations of 1,000; the total number of optimization variables is 4, which are the vehicle speed, acceleration, yaw angle, and yaw rate along a certain route; so that multiple objectives can be comprehensively considered during the optimization process.

[0017] Further, in S5, the CRITIC evaluation algorithm is used to calculate the weight coefficients of each objective; the Pareto solution set is normalized by Min-Max to standardize the data; the conflict degree between objectives is quantified based on the Pearson correlation coefficient matrix, and based on the comprehensive objective variation coefficient and the conflict index, a weight vector W = [w_safety, w_comfort, w_ecology] is generated. The Pareto front solution set is sorted according to the weighted comprehensive score, and the comprehensively optimal linear scheme is selected.

[0018] Further, after S5, it also includes: S6: Combining sensitivity analysis to verify the adaptability of different path tracking algorithms to the design scheme. By simulating the performance of different path tracking algorithms under different road conditions, the influence of vehicle dynamic parameters on path selection is studied. Sensitivity analysis is carried out by changing key parameters for multiple simulations, observing the influence of different parameter changes on the objective function, and ensuring that the design scheme has sufficient robustness and adaptability to cope with the uncertainties in the actual road environment.

[0019] Beneficial effects

[0020] The present invention provides a highway alignment design method for autonomous driving with high intelligence, strong multi-objective collaboration, excellent computational efficiency, and outstanding engineering adaptability. Through this method, the automatic generation and dynamic optimization of the road alignment scheme can be realized, significantly improving the path tracking accuracy, driving safety, and passenger comfort of autonomous vehicles. This method achieves technological breakthroughs in three dimensions: geometric parameter optimization, multi-objective collaborative decision-making, and algorithm adaptability verification, based on a multi-modal control strategy and high-precision joint simulation technology. Through the deep coupling of the NSGA-II multi-objective genetic algorithm and the CRITIC objective weighting method, a scientific balance between the safety, comfort, and ecology objectives of the design scheme is ensured; combined with the vehicle-road joint simulation platform, the influence law of road geometric parameters on vehicle dynamic characteristics is quantitatively analyzed, providing a data-driven basis for design decisions. The application of this technology can greatly shorten the traditional design cycle, reduce the cost of manual trial and error, and provide key technical support for the standardized upgrade of intelligent transportation infrastructure.

[0021] The method proposed by the present invention exhibits excellent performance in various scenarios such as straight-line segments, gradually changing curvature segments, and composite alignment segments. Through simulation verification, the optimized alignment scheme can increase the average safety score by 17%, the average comfort score by 7%, and the average ecological benefit score by 7%. At the same time, based on the design scheme optimized by the Pareto front, its comprehensive score is improved by more than 10% compared with the traditional method, and the time-consuming for optimization and solution is controlled within 5 minutes, meeting the timeliness requirements of large-scale road network design. In addition, sensitivity analysis shows that the adaptability error of the design scheme to different path tracking algorithms such as SMC, PP, and MPC is less than 0.1 meter, and the oscillation frequency of the control command does not exceed 2 Hz, fully verifying the robustness and engineering feasibility of the scheme. This method provides an innovative solution for the collaborative optimization of autonomous driving vehicles and road infrastructure, and has important practical value for promoting the efficient deployment of intelligent transportation systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The drawings are only for the purpose of showing specific embodiments and are not considered to be a limitation of the present invention.

[0023] Figure 1 It is a flowchart of a method for designing the alignment of an autonomous driving highway based on control simulation and multi-objective optimization in an embodiment of the present invention.

[0024] Figure 2 It is a framework of the CarSim / Simulink co-simulation platform in an embodiment of the present invention, showing the interaction process of the road model, vehicle dynamics, and control algorithm.

[0025] Figure 3 It is the structural flow of the multi-objective optimization algorithm and evaluation algorithm used in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the drawings of the embodiments of the present invention. Many specific details are set forth in the following description to facilitate a thorough understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar extensions without departing from the connotation of the present invention, so the present invention is not limited by the specific embodiments disclosed below.

[0027] As Figures 1-3 , the present invention provides an intelligent alignment design method for highways for autonomous driving, including the steps:

[0028] S1. Use Autodesk Civil3D for preliminary road alignment design. First, determine the starting and ending points of the highway, and select a preliminary alignment scheme in combination with road functional requirements (such as the number of lanes, road type, design speed, etc.). Then, optimize it using the straight-spiral-circular design mode, set curve segments with different radii to ensure smooth transitions of each curve segment, and adjust its parameters such as design speed and minimum curvature radius to meet the requirements of design specifications, and finally obtain candidate alignments that meet the actual engineering needs.

[0029] Use the Autodesk Civil3D road design software to construct an initial alignment layout based on the straight-spiral-circular geometric design mode, and generate a basic alignment by inputting the starting and ending point coordinates. By setting the horizontal curve radius threshold (≥650 m) and offsetting the alignment laterally in 5-m steps, a set of candidate alignments including curvature continuity verification is generated to ensure that the geometric parameters meet the requirements of the "Highway Alignment Design Specification".

[0030] S2. Parametrize the preliminarily selected alignment. The specific steps are as follows: Use the parametrization tool provided in Civil3D to convert the preliminarily selected alignment into road centerline data presented in the form of a three-dimensional coordinate point set. Set a step size of 1 m to accurately extract the three-dimensional coordinate points along the road. And save these data in a standard file format (such as CSV or Excel) to ensure that the data can be seamlessly imported into the subsequent simulation platform for further processing.

[0031] Perform parametric modeling on the candidate alignments, extract the three-dimensional coordinate sequences of each alignment (including planar coordinates X / Y, elevation Z, and curvature κ), convert the data into a standardized format (such as CSV or Excel) through the Civil3D secondary development interface, and integrate additional parameters such as road slope and superelevation rate, and output it as an input file compatible with the CarSim / Simulink co-simulation platform.

[0032] S3. Import the parametrized alignment into the autonomous driving simulation platform and select a suitable path tracking algorithm for simulation. The selected path tracking algorithms include Pure Pursuit (PP), Model Predictive Control (MPC), and Sliding Mode Control (SMC). Among them, the PP algorithm uses a simplified two-wheeled vehicle model, sets the look-ahead distance L to 5 m, and ensures good path tracking stability at different vehicle speeds; the MPC algorithm simplifies the vehicle model to a three-degree-of-freedom single-track model, considers the small-angle assumption, combines the vehicle dynamics model, and uses MATLAB / Simulink for real-time simulation; the SMC algorithm designs a switching function, and the control parameters are adjusted by To ensure control accuracy, it is applicable to situations with strong response capabilities to emergencies.

[0033] In the CarSim / Simulink co-simulation platform, import parameterized linear data and embed three path tracking algorithm modules of sliding mode control (SMC), pure pursuit (PP), and model predictive control (MPC) respectively. Set the simulation scenario parameters: select the dynamic parameters of a B-class passenger car for the vehicle model, the control period is 0.1 second, and the lateral deviation threshold ≤ 0.2 meter. Run the simulation and collect the dynamic response data of the vehicle driving along each line, including longitudinal speed, lateral acceleration, yaw rate, and tire slip ratio, and set the sampling frequency to 10 Hz.

[0034] S4. Use a multi-objective optimization algorithm to analyze the simulation results, construct the objective function and solve it. In this stage, the selected optimization objectives include: 1) Safety (such as the lateral stability during vehicle driving); 2) Comfort (such as the acceleration change and deflection angle of the vehicle); 3) Ecological and environmental protection (such as fuel consumption and emissions during vehicle driving). In order to obtain a reasonable optimization solution, use the non-dominated sorting genetic algorithm (NSGA-II) for solution, set the population size to 10,000, and the maximum number of iterations to 1,000. The optimization variables include: the driving speed, acceleration, deflection angle, and deflection angular velocity of the vehicle along the route, etc., to ensure that the obtained alternative routes can achieve the optimal balance under multiple objectives.

[0035] Construct a multi-objective optimization model based on the simulation data. The objective function includes a safety objective function, a comfort objective function, and an ecological objective function. Use the non-dominated sorting genetic algorithm (NSGA-II) for optimization and solution, and set the algorithm parameters: population size 10,000, maximum number of iterations 1,000, crossover probability 0.9, mutation probability 0.1. The optimization variables are the vehicle driving state parameter set (speed v, acceleration a, deflection angle θ, deflection angular velocity ω), and the constraint condition is v ∈ [60, 120] km / h, |a| ≤ 3.5 m / s 2 。

[0036] S5. Use the CRITIC evaluation algorithm to evaluate all alternative routes and comprehensively score the routes according to the weights of each objective function. The CRITIC algorithm reasonably determines the weights of each objective by normalizing each objective function and combining its sensitivity to the final result. During the evaluation process, first normalize all objective function values to ensure that they are within the same order of magnitude range and eliminate the influence caused by the differences in different objective units. Finally, select the route with the highest comprehensive score as the optimal linear scheme.

[0037] Using the CRITIC evaluation algorithm, calculate the weight coefficients of each objective. Perform Min-Max normalization on the Pareto solution set to standardize the data; quantify the conflict degree between objectives based on the Pearson correlation coefficient matrix, and generate the weight vector W = [w_safety, w_comfort, w_ecology] based on the comprehensive objective coefficient of variation (standard deviation / mean) and the conflict index. Sort the Pareto front solution set according to the weighted comprehensive score, and screen out the comprehensively optimal linear scheme.

[0038] S6. Combine sensitivity analysis to verify the adaptability of different path tracking algorithms to the design scheme. In this step, by simulating the performance of different path tracking algorithms under different road conditions, study the influence of vehicle dynamic parameters (such as vehicle speed, deflection angle) on path selection. Sensitivity analysis is carried out by changing key parameters (such as vehicle speed, curve radius, error of the starting point, etc.) for multiple simulations, and observe the influence of different parameter changes on the objective function to ensure that the design scheme has sufficient robustness and adaptability to cope with the uncertainties in the actual road environment.

[0039] For the optimized linear scheme, switch different path tracking algorithms (SMC / PP / MPC) in the simulation platform, and set extreme scenarios (curvature mutation section, slope transition area) for sensitivity testing. Record the mean value of the lateral position deviation (μerr) and the oscillation frequency of the control command (fctrl). If μerr ≤ 0.3 m and fctrl ≤ 2 Hz are satisfied, it is determined that the scheme has algorithm adaptability. Finally, output the linear design parameters that meet the multi-objective constraints and the adaptation suggestions for the control algorithm.

[0040] Example 1:

[0041] S1. Use Autodesk Civil3D road design software to generate the terrain according to the project terrain data, determine the starting and ending coordinates, and generate the initial alignment based on the straight-spiral-circular geometric pattern. Change the radius lateral offset in steps of 5 m, set the radius threshold, and generate a series of candidate alignments.

[0042] S2. Export the three-dimensional coordinate sequence (X / Y / Z), curvature κ, slope i, and superelevation rate parameters of the candidate alignment, and save them as a CSV format file. Import it into the CarSim platform to generate the same road alignment

[0043] S3. Import the road file through the CarSim vehicle dynamics module, select the B-class passenger car dynamics model (wheelbase 2.8 m, curb weight 1500 kg), and set the initial vehicle speed to 80 km / h and the maximum speed to 120 km / h.

[0044] S4. Use Simulink to build a path tracking algorithm, and set the control period and deviation threshold. Build a pure pursuit (PP) algorithm module, simplify the vehicle into a two-wheeled vehicle model to simplify the vehicle force model, and do not consider the slip between the tire and the ground. Calculate the required front wheel steering angle according to the error between the current position of the vehicle and the preview point. The forward view distance L = 5m is the optimal value obtained by simulation and is applicable to the case where the vehicle speed is 120 km / h.

[0045] S5. Run the co-simulation, collect the dynamic response data of the vehicle driving along each alignment, including longitudinal speed v (m / s), lateral acceleration a (m / s 2 ), yaw rate ω (rad / s), vertical forces of the four tires, etc., and the sampling frequency is 10 Hz. Import the generated dynamic response parameters into the matlab workspace.

[0046] S6. Build a multi-objective optimization model based on the dynamic parameters, and set the safety, comfort, and ecological objectives.

[0047] S7. Use the NSGA-II algorithm for optimization, set the algorithm parameters and optimize the vehicle driving parameters.

[0048] S8. Output 50 sets of solution sets, perform normalization processing, calculate the objective weights, and select the alignment scheme with the highest comprehensive score.

[0049] S9. Switch the path tracking algorithm in the preferred alignment to conduct extreme scenario tests, record the lateral deviation and control oscillation frequency, and verify the robustness of the scheme.

[0050] Example 2:

[0051] S1. Use Autodesk Civil3D to generate the terrain and determine the starting and ending coordinates, and generate the initial alignment based on the straight-spiral-circular geometric pattern. Adjust the radius at a 5-meter step and generate candidate alignments, export the three-dimensional coordinates, curvature, slope, and superelevation rate parameters as a CSV file, import them into the CarSim platform, and set the C-class passenger car model, with an initial speed of 80 km / h and a maximum speed of 120 km / h.

[0052] S2. Use Simulink to build a path tracking algorithm, and set the control period and deviation threshold. Build a model predictive control (MPC) algorithm module. By predicting the future state of the system and combining control constraints, optimize the current control quantity. Use the linearized and discretized three-degree-of-freedom vehicle model for path tracking, and simplify the tire model through small-angle assumptions to improve the calculation efficiency. The constraint conditions include the sideslip angle constraint of the center of mass, the vehicle adhesion condition constraint, and the tire sideslip angle constraint to ensure the stable driving of the vehicle during the control process. Set the objective function, and minimize the path tracking error and the control cost of the system by optimizing the control increment. The control parameters are set as follows: the prediction horizon Np = 8, the control horizon Nc = 1, optimize the control increment, and obtain the optimal front wheel steering angle.

[0053] S3. Run the co-simulation, and collect the dynamic response data of the vehicle driving along each alignment, including the longitudinal speed v (m / s), the lateral acceleration a (m / s 2 ), the yaw rate ω (rad / s), the vertical forces of the four tires, etc. The sampling frequency is 10 Hz. Import the generated dynamic response parameters into the matlab workspace.

[0054] S4. Build a multi-objective optimization model based on the dynamic parameters, and set the safety, comfort, and ecological objectives.

[0055] S5. Use the NSGA-II algorithm for optimization, set the algorithm parameters, and optimize the vehicle driving parameters.

[0056] S6. Output 50 groups of solution sets, perform normalization processing, calculate the objective weights, and select the alignment scheme with the highest comprehensive score.

[0057] S7. Switch the path tracking algorithm in the optimized alignment to conduct extreme scenario tests, record the lateral deviation and the control oscillation frequency, and verify the robustness of the scheme.

[0058] Example 3

[0059] S1. Use Autodesk Civil3D road design software to generate the terrain based on the project terrain data, determine the start and end coordinates, and generate the initial alignment based on the straight-spiral-circular geometric pattern. Change the radius lateral offset in 5-meter steps, set the radius threshold, and generate a series of candidate alignments.

[0060] S2. Export the three-dimensional coordinate sequence (X / Y / Z), curvature κ, slope i, and superelevation rate parameters of the candidate alignments, and save them as a CSV format file. Import them into the CarSim platform to generate the same road alignment

[0061] S3. Import the road file using the CarSim vehicle dynamics module. Select a Class B passenger car dynamics model (wheelbase 2.8 meters, curb weight 1500 kg). Set the vehicle's initial speed to 80 km / h and its maximum speed to 120 km / h.

[0062] S4. Use Simulink to build a path tracking algorithm and set the control period and deviation threshold. Build a sliding mode control (SMC) algorithm module. By designing the sliding surface and switching function, the system state is guided to the predetermined sliding surface. Control inputs are used to force the system to move along the sliding surface, eliminating tracking errors. The Lyapunov stability criterion is used to ensure system stability. The control parameters are c1 = 2, k1 = 0.1, and k2 = 5. Multiple simulations are performed to determine the optimal controller parameters to reduce control chattering.

[0063] S5. Run the joint simulation to collect the dynamic response data of the vehicle along each line, including the longitudinal speed v (m / s), lateral acceleration a (m / s 2 ), yaw rate ω (rad / s), and the vertical forces acting on the four tires, with a sampling frequency of 10 Hz. The generated dynamic parameters are imported into the MATLAB workspace.

[0064] S6. Build a multi-objective optimization model based on dynamic parameters and set safety, comfort, and ecological goals.

[0065] S7. Use the NSGA-II algorithm for optimization, set algorithm parameters, and optimize vehicle driving parameters.

[0066] S8. Output 50 sets of solutions, perform standardization, calculate the target weights, and select the linear solution with the highest comprehensive score.

[0067] S9. Switch the path tracking algorithm in the preferred alignment to perform extreme scenario testing, record the lateral deviation and control oscillation frequency, and verify the robustness of the solution.

[0068] Example 4

[0069] S1. Use Autodesk Civil 3D to generate the terrain, determine the start and end coordinates, and generate an initial alignment based on a straight-slope-circular geometry. Change the radius lateral offset in 5-meter increments to generate candidate alignments. Export the 3D coordinates, curvature, slope, and superelevation parameters, save them as a CSV file, and import them into the CarSim platform. Import the road file into CarSim, select the vehicle model, and set the initial speed. Use Simulink to build a path tracking control algorithm and eliminate tracking errors through control inputs. Run a co-simulation, collect dynamic response data, and import it into Matlab for analysis.

[0070] S2. Build a multi-objective optimization model based on dynamic parameters, set safety, comfort, and ecological objectives. The dynamic parameters exported from Carsim include longitudinal velocity v, lateral acceleration a, angular velocity ωz, yaw angle θ, total vehicle vertical force Fz,total, and road curvature κ.

[0071] Optimization of safety benefits:

[0072]

[0073] Among them, LTR represents the rollover index; m represents the mass of the vehicle, h is the height of the vehicle's center of mass, v represents the vehicle's speed, R is the turning radius of the vehicle's travel, g represents the acceleration due to gravity, and Fz,total is the total vertical ground force of the vehicle. The value of LTR reflects the rollover trend of the vehicle. When its value is close to 1, it indicates that the vehicle is in an extremely stable state and there is a high risk of rollover.

[0074]

[0075] Among them, e y represents the lateral deviation, θ is the heading angle error, δ is the steering angle, ωz is the yaw angular velocity of the vehicle, and ωdesired represents the desired yaw angular velocity. Q1, Q2, R, S are the corresponding weight coefficients used to balance the influence of different control variables on the optimization objective.

[0076] Optimization of ecological benefits:

[0077]

[0078] E total = E1 + E2

[0079] The ecological benefit optimization part includes two energy consumption terms E1 and E2. Among them, f(v) and g(a) respectively represent the influence of speed v and acceleration a on energy consumption, and the coefficients α, β, γ, λ, μ are used to adjust their influence degrees. Vcut and Vfill respectively represent the volumes of excavation and filling, Eequipment is the equipment usage cost, Mmaterial is the material cost, κ(x) represents the curvature, g(x) represents the terrain change, Crepair represents the repair cost, Eequipment,maintenance is the equipment maintenance cost, and Mreplacement represents the equipment replacement cost.

[0080] Optimization of passenger comfort:

[0081]

[0082] The passenger comfort optimization part involves the fluctuations of acceleration and speed. Among them, σ ais the standard deviation of acceleration, which measures the degree of acceleration fluctuation. σv is the standard deviation of velocity, which measures the degree of velocity change. w1, w2, w3, and w4 are weight coefficients used to balance the influence of different comfort factors. The integral term reflects the cumulative curvature of the path, while quantifies the influence of road surface unevenness. The overall goal is to optimize the comfort of passengers by minimizing these indicators.

[0083] S3. Use the NSGA-II algorithm for optimization. The objective functions are selected as safety, comfort, and ecological environmental protection. The non-dominated genetic algorithm NSGA-II is used for solution. The population size is 10,000, and the maximum number of iterations is 1,000. The total number of optimization variables is 4, namely the speed, acceleration, yaw angle, and yaw rate of the vehicle along a certain route, so that multiple objectives can be comprehensively considered during the optimization process, ensuring that the designed road alignment reaches a good balance among safety, comfort, and environmental friendliness, and improving the comprehensive benefits of road design.

[0084] S4. Output the Pareto front, perform Min-Max normalization on the Pareto solution set, calculate the objective weights using the CRITIC evaluation method, and screen out the alignment scheme with the highest comprehensive score. Screen out the alignment scheme with the highest comprehensive score. For example, for the alignment scheme with the highest score (ID = 23), its parameters are R = 850 m and e = 4%. Replicate this scheme in CarSim, and the measured standard deviation of lateral acceleration is 0.38 m / s 2 , which is better than the optimization objective (≤0.5 m / s 2 ).

[0085] Table 1

[0086]

[0087] Table 1 above shows the performance of four road design methods (Sliding Mode Control (SMC), Pure Pursuit Control (PP), Model Predictive Control (MPC), and traditional manual design method) in terms of score improvement, safety, comfort, ecological benefits, etc. Compared with the traditional manual design method, SMC, PP, and MPC have all improved in terms of score, especially in terms of comfort and ecological benefits scores, and the MPC and PP algorithms show more significant improvements. In addition, the safety score has also increased compared with the traditional method, reflecting the advantages of modern optimization algorithms in road design. Although the score differences among these three algorithms are not large, especially in terms of ecological benefits and comfort, the gap in their overall performance is relatively small, showing the consistency of these methods in optimization goals. It is worth noting that although there are some minor differences in parameters among the three, in practical applications, they can all provide relatively stable and reliable performance, showing strong robustness. We found that the design results using the multi-objective optimization algorithm show better comprehensive performance in multiple objectives such as safety, comfort, and ecological protection. Traditional methods usually rely on a single objective or local optimization, while our method incorporates multiple performance indicators into the optimization model, thus ensuring that the design of the highway not only meets the technical requirements but also provides higher safety guarantees and comfortable experiences in all aspects of vehicle operation, with significant advantages. This method can effectively overcome the limitations of traditional design, improve the reliability of the overall system and the optimization effect of multiple objectives while ensuring the design efficiency.

[0088] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all belong to the protection scope of the present invention.

Claims

1. A method for the alignment design of an autonomous driving highway based on control simulation and multi-objective optimization, characterized in that, Including the steps: S1: Based on road design software, conduct preliminary route alignment to obtain a series of candidate alignments; S2: Parametrize the alignments; S3: Import the parameterized alignments into a vehicle simulation platform and perform simulations using an autonomous driving path tracking algorithm; S4: Use a multi-objective optimization algorithm to analyze the simulation results, construct an objective function, and solve it; S5: Establish a comprehensive route evaluation system based on the multi-objective optimization algorithm and evaluation algorithm to screen the optimal alignment scheme.

2. The method for designing the alignment of an autonomous driving highway based on control simulation and multi-objective optimization according to claim 1, wherein In S1, use Autodesk Civil3D for preliminary road alignment design. First, determine the starting and ending points of the highway, and select a preliminary alignment scheme in combination with the road functional requirements. Then, optimize using the straight-spiral-circular design mode, set curve segments with different radii to ensure smooth transitions of each curve segment, and adjust its parameters to meet the design specification requirements. Finally, obtain candidate alignments that meet the actual engineering requirements.

3. The method for designing the alignment of an autonomous driving highway based on control simulation and multi-objective optimization according to claim 2, wherein In S2, perform parametric modeling on the candidate alignments, extract the three-dimensional coordinate sequences of each alignment, convert the data into a standardized format through the Civil3D secondary development interface, integrate additional parameters, and output an input file compatible with the CarSim / Simulink co-simulation platform.

4. The method for designing the alignment of an autonomous driving highway based on control simulation and multi-objective optimization according to claim 3, wherein In S3, in the CarSim / Simulink co-simulation platform, parametric linear data is imported, and the pure pursuit control, model predictive control, and sliding mode control algorithms are selected for the simulation of the autonomous driving path tracking algorithm. Among them, the pure pursuit control algorithm uses a simplified two-wheeled vehicle model without considering the slip between the tire and the ground, and the forward look-ahead distance L = 5m is selected to obtain the optimal control effect; the model predictive control algorithm simplifies the vehicle into a three-degree-of-freedom single-track model, adopts the small-angle assumption, and establishes a controller model in MATLAB / Simulink; the sliding mode control algorithm designs the switching function as where s is the switching function, which is used to construct the core variable of the sliding mode controller; e is the tracking error, that is, the deviation between the target trajectory and the actual trajectory; is the change rate of the tracking error, which represents the change trend of the error over time; c1 is a constant weight coefficient, which is used to adjust the weight ratio of the error and the error change rate in the switching function. The system parameters are determined as: c1 = 2 to reduce the chattering of the controller; by adopting different path tracking algorithms, it is ensured that the path control can be optimized under complex traffic environments and different vehicle speeds, and the driving stability and responsiveness of autonomous driving vehicles are enhanced.

5. The method for designing the alignment of an autonomous driving highway based on control simulation and multi-objective optimization according to claim 4, wherein In S4, the selected optimization objectives include: safety; comfort; ecological and environmental protection; Use the non-dominated sorting genetic algorithm to solve, set the population size to 10,000, and the maximum number of iterations to 1,000; The optimization variables include: the driving speed, acceleration, yaw angle, and yaw rate of the vehicle along the route, ensuring that the obtained alternative routes can achieve the optimal balance under multiple objectives.

6. The method for designing the alignment of an autonomous driving highway based on control simulation and multi-objective optimization according to claim 1, characterized in that In S5, use the CRITIC evaluation algorithm to evaluate all alternative routes, and comprehensively score the routes according to the weights of each objective function; The CRITIC algorithm determines the weights of each objective by standardizing each objective function and combining its sensitivity to the final result; During the evaluation process, first normalize all objective function values to ensure that they are within the same order of magnitude range, eliminating the influence caused by differences in units of different objectives; Finally, select the route with the highest comprehensive score as the optimal alignment scheme.

7. The method for designing the alignment of an autonomous driving highway based on control simulation and multi-objective optimization according to claim 1, wherein After S5, it also includes; S6: Combine sensitivity analysis to verify the adaptability of different path tracking algorithms to the design scheme. By simulating the performance of different path tracking algorithms under different road conditions, study the influence of vehicle dynamics parameters on path selection. Sensitivity analysis conducts multiple simulations by changing key parameters, observes the influence of different parameter changes on the objective function, and ensures that the design scheme has sufficient robustness and adaptability to cope with the uncertainties in the actual road environment.

Images