A multi-objective optimization method for vehicle safety based on NSGA-II

Abstract

The present invention discloses a multi-objective optimization method for vehicle safety based on NSGA‑II, comprising the following steps: (1) determining a vehicle simulation simulator and setting a vehicle operation scenario; (2) determining the range and default value of vehicle physical parameters; (3) establishing a multi-objective optimization model, including: establishing a vehicle physical parameter selection model; performing averaging processing according to the value ranges of different parameters, and establishing a vehicle physical parameter configuration difference model; establishing a vehicle obstacle avoidance effect model according to the obstacle avoidance related data returned by configuring the vehicle physical parameters in the automobile simulation simulator; (4) solving the multi-objective optimization model using a multi-objective search algorithm. The present invention uses an automobile simulation simulator to solve the problem that vehicle physical parameters are difficult to combine and configure in real scenarios; by setting a multi-objective optimization model, the search space is reduced, and a Pareto optimal solution set is obtained in a relatively short time, which is convenient for analyzing parameters that have a greater impact on vehicle safety performance.

Description

Technical Field

[0001] The present invention relates to a vehicle safety target optimization method, and in particular to a vehicle safety multi-objective optimization method based on NSGA-II. Background Art

[0002] As the number of cars continues to grow, the frequency of car-related traffic accidents is also increasing. The issue of car driving safety has attracted more and more attention from people, who are eager for cars to have better handling stability. According to relevant statistical information, traffic accidents caused by car loss of stability have always accounted for a high proportion of all traffic accidents. At the same time, the speed of cars is getting faster and faster with the development of technology and the improvement of roads, which has led to higher requirements for the handling stability and active safety of cars. How to improve the handling stability of cars has become a research hotspot for many scholars and has important research value and significance.

[0003] This becomes more difficult due to the high variability of automotive products and the rapid iteration of individual model versions. In this context, the concept of automotive product lines is proposed as a unified framework to accommodate different sources of variability in automotive products. The concept of product lines is a way to capture the variability in a product line with the same characteristics but differences. By considering the commonality and variability of product lines, product line engineering includes the development, maintenance and analysis of product lines. In order to address the complexity of heterogeneous systems with multiple product lines interacting with each other, the concept of product lines is proposed. In the context of multiple automotive production lines, different configurations of vehicle physical parameters are simulated using automotive simulation simulators. For safety and stability analysis, it is necessary to solve high-dimensional multi-objective optimization problems. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to provide a multi-objective optimization method for vehicle safety based on NSGA-II for analyzing the selection of vehicle physical parameters that affect the vehicle obstacle avoidance effect.

[0005] Technical solution: The vehicle safety multi-objective optimization method of the present invention comprises the following steps:

[0006] (1) Determine the vehicle simulation simulator and set the vehicle operation scenario;

[0007] (2) Determine the range and default values ​​of vehicle physical parameters;

[0008] (3) Establishing a multi-objective optimization model, including: establishing a vehicle physical parameter selection model; performing averaging processing according to the value ranges of different parameters to establish a vehicle physical parameter configuration difference model; establishing a vehicle obstacle avoidance effect model based on the obstacle avoidance related data returned by configuring the vehicle physical parameters in the automobile simulation simulator;

[0009] (4) Use multi-objective search algorithm to solve the multi-objective optimization model.

[0010] Furthermore, in step (2), the vehicle physical parameter range is determined based on the specific parameters of the vehicle production line; the default value is the parameter value configured by the vehicle simulation simulator under specific conditions so that the simulated vehicle can take timely braking measures in front of an obstacle to ensure that the vehicle does not collide with the obstacle.

[0011] Furthermore, the multi-objective optimization model in step (3) is:

[0012]

[0013] Where, num_paras_selected is the number of selected physical parameters; selected parameters is the selected physical parameters, distance is the distance in front of the obstacle, velocity is the collision speed of the vehicle when it collides; f(2) is the average value of the selected physical parameters and the average difference, v i is the default value of parameter i, v' i is the changed value of parameter i, is the difference in the value range of parameter i; f(3) is the calculation formula for the vehicle's obstacle avoidance effect. If the vehicle does not collide with the obstacle, the collision speed is 0; if the vehicle collides with the obstacle, the distance the vehicle stops in front of the obstacle is 0.

[0014] Furthermore, three optimization objectives are set for the multi-objective optimization model. The first objective O1 is to minimize the number of parameter selections, the second objective O2 is to minimize the average value of the difference between the selected parameters and their default values, and the third objective O3 is to maximize the vehicle obstacle avoidance effect. The optimization objectives of the multi-objective optimization model are:

[0015]

[0016] Furthermore, in step (4), the process of solving the multi-objective optimization model based on the NSGA-II method includes:

[0017] (41) Initialize the population;

[0018] (42) inputting the established multi-objective optimization model into the evaluation function;

[0019] (43) Draw a graph of the Pareto front and obtain the optimal solution set.

[0020] Furthermore, the initialization of the population includes setting the population size, the number of evolution generations, the crossover operator and its crossover rate, the mutation operator and its mutation rate, the number of independent variables, the range of independent variables, and the number of objective functions.

[0021] Compared with the prior art, the present invention has the following significant effects: 1. The use of an automobile simulation simulator can effectively solve the problem that the physical parameters of a vehicle are difficult to combine and configure in a real scenario. At the same time, due to the operational safety and repeatability of the simulator, the test risk and cost are greatly reduced; 2. By setting a multi-objective optimization problem, the search space can be reduced, and the Pareto optimal solution set can be obtained in a shorter time, which is convenient for studying and analyzing the parameters that have a greater impact on the vehicle safety performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a schematic diagram of the process of the present invention;

[0023] Figure 2 Flowchart for implementing automatic obstacle avoidance for a car simulator;

[0024] Figure 3 Run the flow chart for the vehicle simulation simulator. DETAILED DESCRIPTION

[0025] The present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0026] Figure 1 The overall flow chart of the vehicle safety multi-objective optimization method of the present invention includes the following steps:

[0027] Step 1: Determine the vehicle simulation simulator and set the vehicle operation scenario. This includes the following steps:

[0028] Step 11, analyzing various open source unmanned driving simulation simulators, such as AirSim, CARLA, LGSVL, etc., CARLA is selected as the unmanned driving simulation simulator of the present invention due to its good openness.

[0029] Step 12: According to the virtual vehicle provided by CARLA, an Audi vehicle is selected as the experimental vehicle, and a straight road with a length of 100 meters and an obstacle at the end is selected in the Town3 map as the experimental scene.

[0030] Step 2: Determine the vehicle physical parameters, default values ​​and their ranges. The specific implementation steps are as follows:

[0031] Step 21, by collecting the product configuration parameters of Audi A3-A8 series and the changeable vehicle physical parameters officially provided by CARLA, and determining the parameter range of the vehicle physical parameters, 12 physical parameters are selected, specifically max_rpm (maximum engine speed), mass (vehicle mass), radius (tire radius), damping_rate_full_throttle (damping ratio when the throttle is maximum), damping_rate_zero_throttle_clutch_engaged (damping ratio when the throttle is zero and the clutch is engaged), damping_rate_zero_throttle_clutch_disengaged (damping ratio when the throttle is zero and the clutch is disengaged), gear_switch_time (switching time between gears), clutch_strength (clutch vehicle strength), drag_coefficient (vehicle chassis drag coefficient), tire_friction (wheel friction), damping_rate (wheel damping rate), max_brake_torque (maximum braking torque).

[0032] Step 22, in the vehicle operation scenario designed in step 12, manually drive the vehicle until it stops in front of an obstacle, collect pictures and corresponding vehicle operation instructions obtained by the front camera of the car during driving every 0.1 seconds, use a neural network to train the data set, and then obtain images through the front camera of the vehicle. The model trained by the neural network predicts the throttle and brake values ​​of the vehicle at each moment, and the flow chart is as follows: Figure 2 As shown. According to the vehicle physical parameter range determined in step 21, a set of parameter values ​​within the parameter range is set as default values. This set of default values ​​enables the vehicle to stop in front of obstacles in a straight line scene through the implemented unmanned driving algorithm prediction model, and is set as v1, v2, ... v12. Table 1 shows the vehicle parameters, their ranges and default values.

[0033] Table 1 Vehicle parameters

[0034]

[0035] Step 3: Establish a multi-objective optimization mathematical model, including establishing a vehicle physical parameter selection model, a vehicle physical parameter configuration difference model, and a vehicle obstacle avoidance effect model. The multi-objective optimization model is:

[0036]

[0037] In formula (1), num_paras_selected is the number of selected physical parameters; selected parameters is the selected physical parameters, distance is the distance in front of the obstacle, velocity is the collision speed of the vehicle when it collides; f(2) is the average value of the selected physical parameters and the difference, v i is the default value of parameter i, v' i is the changed value of parameter i, is the difference in the value range of parameter i; f(3) is the calculation formula for the vehicle's obstacle avoidance effect. If the vehicle does not collide with the obstacle, the collision speed is 0; if the vehicle collides with the obstacle, the distance the vehicle stops in front of the obstacle is 0.

[0038] The specific implementation steps are as follows:

[0039] Step 31, establish a vehicle physical parameter selection model: use the NSGA-II algorithm (non-dominated sorting genetic algorithm with elite strategy) to average the parameter values ​​given by the search. The formula for average processing is:

[0040]

[0041] By setting a threshold, it is determined whether the NSGA-II search algorithm selects the physical parameter. Furthermore, the threshold is set according to the value range of the physical parameter:

[0042] The parameter value range is greater than 1000, and the threshold is 0.005;

[0043] The parameter value range is 100-1000, and the threshold is 0.01;

[0044] The parameter value range is 1-100, and the threshold is 0.02;

[0045] The parameter value range is 0-1, and the threshold is 0.04.

[0046] When P is greater than the threshold, it is considered that the search algorithm has selected the physical parameter, that is, the physical parameter has changed; if it is less than or equal to the threshold, it is considered that the search algorithm has not selected the parameter, that is, the physical parameter is the default value, so that the search algorithm can select parameters more reasonably, and finally count the number of all changed parameters.

[0047] Step 32, establish a vehicle physical parameter configuration difference model. For the vehicle physical parameters selected by the NSGA-II search algorithm in step 31, use the averaged sum of all changed parameters divided by the number of changed parameters to calculate the average change rate of the changed value and its corresponding default value, that is, the calculation formula is f(2).

[0048] Step 33, establish the vehicle obstacle avoidance effect model, according to the vehicle physical parameters selected by the search algorithm in step 31, use the parameter values ​​provided by the NSGA-II search algorithm, and the unselected parameters are the default values. All parameter values ​​are passed into the CARLA vehicle simulation simulator, such as Figure 3 The operation flow chart shown in the figure runs in a 100-meter straight line scene, starting from the starting point. Every 0.1s, the vehicle records the picture through the front camera, and uses the training model in step 22 to predict the vehicle's behavior until the driving is completed. When the vehicle moves forward, if the vehicle collides with the obstacle at the end point, the collision speed velocity is recorded, and the distance between the vehicle and the obstacle is 0; if the vehicle does not collide with the obstacle at the end point, after the vehicle stops in front of the obstacle, the distance between the vehicle and the obstacle is recorded as distance, and the collision speed is 0, and f(3) = -distance + velocity is calculated.

[0049] The specific steps are as follows:

[0050]

[0051] Step 34, according to formula (1), three optimization objectives are set. The first objective O1 is to minimize the number of parameter selections, the second objective O2 is to minimize the average value of the difference between the selected parameters and their default values, and the third objective O3 is to maximize the vehicle's obstacle avoidance effect, that is, it is more desirable for the vehicle to collide with obstacles under smaller parameter changes.

[0052] The optimization objectives of the multi-objective optimization model are:

[0053]

[0054] Step 4: Solve the multi-objective optimization mathematical model based on the NSGA-II method, obtain the Pareto solution set, and draw the Pareto frontier diagram. The implementation process is:

[0055] Step 41, initialize the population.

[0056] Initializing the population includes setting the population size, evolutionary generations, crossover operator and its crossover rate, mutation operator and its mutation rate, number of independent variables, range of independent variables, and number of objective functions.

[0057] Step 42, input the established multi-objective optimization mathematical model into the evaluation function. The multi-objective optimization framework used is jMetalPy, and jMetalPy provides an algorithm template for NSGA-II. According to step 21, the parameter value range is set to set the population upper limit and population lower limit, and the established multi-objective optimization model is input into the evaluation function, the population size is set to 50, the evolutionary generation is set to 100, the crossover operator is simulated binary crossover, the crossover rate is 0.9, the mutation operator is polynomial mutation, the mutation rate is equal to the inverse of the number of variables, that is, 1 / 12, the number of independent variables is 12, the independent variable range is given in step 21, and the number of objective functions is 3.

[0058] The specific parameter codes are as follows:

[0059]

[0060] Step 43, draw the image of the Pareto front to obtain the optimal solution set. Through the continuous iterative calculation of the search algorithm, the image of the Pareto front can be automatically drawn to obtain the optimal solution set. By analyzing the optimal solution set, the parameters that are more likely to have a greater impact on the obstacle avoidance effect of the vehicle under smaller changes are selected.

[0061] The vehicle safety multi-objective optimization method of the present invention comprehensively considers the impact of different physical parameters of the vehicle on the safety performance of the vehicle, uses a vehicle simulator to conduct experiments to establish a mathematical model, and uses the NSGA-II algorithm to solve it to obtain the optimal vehicle parameter selection solution set, which helps to improve the scientific nature of vehicle parameter selection and improve the safety and stability analysis of the vehicle.

[0062] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

Claims

1. A multi-objective optimization method for vehicle safety based on NSGA-II, characterized in that: The steps include: (1) Determine the vehicle simulation simulator and set the vehicle operation scenario; (2) Determine the range and default values ​​of vehicle physical parameters; (3) Establishing a multi-objective optimization model, including: establishing a vehicle physical parameter selection model; performing averaging processing according to the value ranges of different parameters to establish a vehicle physical parameter configuration difference model; establishing a vehicle obstacle avoidance effect model based on the obstacle avoidance related data returned by configuring the vehicle physical parameters in the automobile simulation simulator; (4) Use a multi-objective search algorithm to solve the multi-objective optimization model; The multi-objective optimization model in step (3) is: Where, num_paras_selected is the number of selected physical parameters; selected parameters is the selected physical parameters, distance is the distance in front of the obstacle, velocity is the collision speed of the vehicle when it collides; f(2) is the average value of the selected physical parameters and the average difference, v i is the default value of parameter i, v' i is the changed value of parameter i, is the difference between the value ranges of parameter i; f(3) is the calculation formula for the vehicle's obstacle avoidance effect. If the vehicle does not collide with the obstacle, the collision speed is 0; if the vehicle collides with the obstacle, the distance the vehicle stops in front of the obstacle is 0; Three optimization objectives are set for the multi-objective optimization model. The first objective O1 is to minimize the number of parameter selections, the second objective O2 is to minimize the average value of the difference between the selected parameters and their default values, and the third objective O3 is to maximize the vehicle obstacle avoidance effect. The optimization objectives of the multi-objective optimization model are: The vehicle physical parameter range is determined according to the specific parameters of the vehicle production line; the default value is the parameter value configured by the vehicle simulation simulator under specific conditions to enable the simulated vehicle to take timely braking measures in front of obstacles to ensure that the vehicle does not collide with the obstacle.

2. The vehicle safety multi-objective optimization method based on NSGA-II according to claim 1 is characterized in that: In the step (4), the process of solving the multi-objective optimization model based on the NSGA-II method includes: (41) Initialize the population; (42) inputting the established multi-objective optimization model into the evaluation function; (43) Draw a graph of the Pareto front and obtain the optimal solution set.

3. The vehicle safety multi-objective optimization method based on NSGA-II according to claim 2 is characterized in that: The initialization of the population includes setting the population size, the number of evolution generations, the crossover operator and its crossover rate, the mutation operator and its mutation rate, the number of independent variables, the range of independent variables, and the number of objective functions.

Images