An automatic driving function-oriented perception system sensor automatic optimization configuration method

Abstract

The application belongs to the technical field of automatic driving cars, and in particular relates to an automatic driving function-oriented perception system sensor automatic optimal configuration method, which comprises the following steps: step 1: defining task input and output, performing perception system sensor selection configuration design in a function-oriented manner, giving a single or multiple automatic driving functions, and meeting the perception requirements of the functions; step 2: analyzing and modeling the sensor system; step 3: completing the optimization model construction of the sensor system design by defining the decision variables, the objective function and the constraint conditions; step 4: based on the optimization model construction of the sensor system design in step 3, solving the optimization model based on an intelligent genetic algorithm, and outputting a final sensor selection configuration scheme; and step 5: simulating and verifying the effectiveness of the design model and the solving method. Through the application, the given automatic driving function can be automatically selected and configured by means of quantitative optimal design and solving the optimization model based on an intelligent genetic algorithm.

Description

Technical Field

[0001] This invention belongs to the field of autonomous vehicle technology, and more specifically, relates to an automatic optimal configuration method for sensors in an autonomous driving function-oriented perception system. Background Technology

[0002] In recent decades, the rapid development of automotive technology has brought immense convenience to human society, but it has also exacerbated problems such as road congestion, traffic accidents, energy shortages, and environmental pollution. Autonomous driving technology offers new ideas and technical means to alleviate these problems. Its technical architecture can be broadly divided into three main aspects: perception, decision-making and planning, and motion control. Among these, the perception system's sensors are responsible for measuring and collecting real-time information to perceive and understand the environment. Commonly used perception sensors include cameras, millimeter-wave radar, lidar, and ultrasonic radar. Different types of sensors have different detection performance and costs. Furthermore, different autonomous driving functions or levels have varying requirements for the configuration of perception system sensors. Therefore, the selection and placement of sensors have a significant impact on the performance and cost of autonomous driving systems.

[0003] Currently, the selection and configuration of environmental perception sensors for autonomous driving are mostly based on qualitative selection and configuration using engineering experience. However, with the continuous enrichment of autonomous driving functions, research and attention to higher levels of autonomous driving are also increasing. Furthermore, the rapid development of sensor technology has led to a rapid increase in the number of available sensors. Traditional experience-based qualitative configuration methods are insufficient to meet the new system design requirements. Therefore, this patent proposes an automatic optimization configuration method for perception system sensors guided by autonomous driving functions. Through quantitative optimization design, it achieves automatic selection and configuration of perception system sensors for a given autonomous driving function. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes an automatic optimal configuration method for sensors in a perception system oriented towards autonomous driving functions. Through quantitative optimization design and solving the optimization model based on an intelligent genetic algorithm, the method achieves automatic selection and configuration of sensors for a given autonomous driving function.

[0005] To achieve the above objectives, the present invention provides a method for automatic optimal configuration of sensors in a perception system guided by autonomous driving functions, comprising:

[0006] Step 1: Define the task input and output, and design the sensor selection and configuration of the perception system based on the function, given one or more autonomous driving functions to meet the perception requirements of the functions;

[0007] Step 2: Analysis and modeling of the sensor system, which includes function-related modeling and design-related modeling;

[0008] Step 3: Based on the functional-related modeling and design-related modeling in Step 2, the optimization model for the sensor system design is constructed by defining decision variables, objective functions, and constraints.

[0009] Step 4: Based on the sensor system design in Step 3, construct an optimization model, solve the optimization model using an intelligent genetic algorithm, and output the final sensor selection and configuration scheme.

[0010] Step 5: Based on the results of Step 3 and Step 4, perform simulation to verify the effectiveness of the design model and solution method.

[0011] Furthermore, the analysis and modeling of the sensor system in step 2 specifically includes:

[0012] Step 2.1: Record the x and y coordinates and orientation angles of the candidate sensor installation locations in the vehicle's top view using a matrix;

[0013] Step 2.2: Determine the sensor parameters;

[0014] Step 2.3: Determine the detection range of the sensor;

[0015] Step 2.4: Define a target detection area coupled with the sensor function. By modeling the target detection range coupled with the function, determine the detection area that the function focuses on.

[0016] Furthermore, in step 2.2, the parameters of the sensor are determined by defining an 8-dimensional vector Para. i ∈R 8 Let i ∈ N to describe the parameters of the i-th sensor, as follows:

[0017]

[0018] In the formula, type i The sensor types include four types: C, R, L, and S, where C is a camera, R is millimeter-wave radar, L is lidar, and S is ultrasonic radar; d i For the maximum detection distance, φ i Field of view (FoV) For distance detection accuracy, For higher velocity detection accuracy, pr i For price, pw i To utilize power, T i For service life;

[0019] Step 2.3, determining the sensor's detection range, specifically involves simplifying the detection ranges of the four sensors—camera, millimeter-wave radar, lidar, and ultrasonic radar—to d. i and φ i Zhang Cheng's fan shape or circle.

[0020] Furthermore, the construction of the optimization model for the sensor system design in step 3 specifically includes:

[0021] Step 3.1: Determine the sensor configuration design scheme under n candidate sensors and m candidate installation locations;

[0022] Step 3.2: Determine the objective function J(X), which is used to evaluate the functionality, reliability, and economy of design scheme X;

[0023] Step 3.3: Determine the constraints for the functionality, reliability, and economy of design scheme X in step 3.2.

[0024] Furthermore, step 3.1, determining the sensor configuration design scheme, specifically includes:

[0025] Define the decision variable vector X∈R n The vector X has a dimension corresponding to the number of n candidate sensors. The i-th element of the vector X represents the installation location of the i-th candidate sensor, i.e., X[i]∈M;

[0026] Where X represents the sensor configuration design scheme given n alternative sensors and m alternative installation locations;

[0027] When X[i] = 0, it means that the i-th candidate sensor is not selected and therefore not installed. When X[i] is any other value, it means that the sensor is selected and installed at the installation position of that sequence number.

[0028] Furthermore, the formula for calculating the objective function J(X) in step 3.2 is as follows:

[0029] J(X)=αJ1(X)+βJ2(X)+γJ3(X)

[0030]

[0031] J2(X)=w 21 C pr (X)+w 22 C pw (X)

[0032] J3(X) = -T system (X)

[0033] In the formula, α, β, γ, w 11 w12 w 21 w 22 Let J1(X) be a positive constant, J2(X) be functionality, J3(X) be reliability, and r(X) be economy, and r(X) be detection coverage. For the key detection point p i The distance detection error under design scheme X, For the key detection point p i The velocity detection error under design scheme X, C pr (X) represents the system price cost, C pw (X) represents the cost of power usage, T system For the system's lifespan;

[0034] Step 3.2 specifically includes:

[0035] Step 3.1.1: Functionality is evaluated by detection coverage and detection accuracy at key detection points;

[0036] Step 3.1.2: Economic efficiency through system price cost C pr (X) and power usage cost C pw (X) is evaluated;

[0037] Step 3.1.3: Reliability is measured by the system's lifespan.

[0038] Furthermore, the formula for calculating the detection coverage r(X) in step 3.1.1 is as follows:

[0039]

[0040] In the formula, S(X) is the area of ​​the intersection between the actual detection area and the target detection area of ​​scheme X, S target The area of ​​the target detection region;

[0041] Distance detection error The calculation formula is:

[0042]

[0043] Speed ​​detection error The calculation formula is:

[0044]

[0045] In the formula, In design scheme X, the detection range should be able to cover the key detection point p. i The number of sensors;

[0046] The system price cost C in step 3.1.2 prThe formula for calculating (X) is:

[0047]

[0048] Power cost C pw The formula for calculating (X) is:

[0049]

[0050] In the formula, S(X[i]) is a function to determine whether the i-th candidate sensor is selected, where 1 represents that it is selected and 0 represents that it is not selected;

[0051] The system lifespan T in step 3.1.3 system The calculation formula is:

[0052]

[0053] In the formula, n s This represents the number of sensors selected in design scheme X.

[0054] Furthermore, in step 3.3, constraints are determined regarding the functionality, reliability, and economy of design scheme X in step 3.2, specifically including:

[0055] Step 3.3.1: Boundary constraints for decision variable X, define C1 constraint;

[0056] Step 3.3.2: Define constraints C2-C9 for functional constraints;

[0057] Step 3.3.3: Define constraints C10-C13 for economic constraints;

[0058] Step 3.3.4: Define the C14 constraint for the reliability constraint;

[0059] Step 3.3.5: Based on engineering experience and patterns, use vectors Let C15 represent a set of installation locations where the same sensor is selected;

[0060] Step 3.3.6: Based on the C1-C15 constraints defined in Steps 3.3.1 to 3.3.4, establish the overall optimization model.

[0061] Furthermore, the constraint C1 in step 3.3.1 is:

[0062] C1:

[0063] in, This represents the alternative installation locations for the camera. These represent alternative installation locations for millimeter-wave radar. These represent alternative installation locations for the lidar. This represents the alternative installation locations for the ultrasonic radar;

[0064] The constraints C2-C5 in step 3.3.2 are the minimum constraints that the farthest detection distance around the vehicle achievable by design scheme X should satisfy:

[0065] C2:d front (X)≥d front,min

[0066] C3:d back (X)≥d back,min

[0067] C4:d left (X)≥d left,min

[0068] C5:d right (X)≥d right,min

[0069] Where, d front,min d back,min d left,min and d right,min This represents the minimum allowable detection distance in the four directions: front, back, left, and right.

[0070] The constraints C6 and C7 in step 3.3.2 are constraints that the key detection points should be covered by the sensor for detection, or, when visual information is required, should be covered by the camera for detection.

[0071] C6:NumCover(p i X)≥1, i=1,2,..,n C

[0072] C7:

[0073] Among them, the function NumCover(p i The detection area in the design scheme X can cover the key detection point p. i The number of sensors, function The detection area in design scheme X can cover the key detection point p. i The number of cameras, These represent key detection points that require visual inspection.

[0074] The constraints C8 and C9 in step 3.3.2 are the constraints that the distance detection error and velocity detection error at the key detection points should satisfy:

[0075] C8:

[0076] C9:

[0077] in, To the maximum permissible distance detection error, The maximum permissible speed detection error;

[0078] The constraints C10-C13 in step 3.3.3 have maximum allowable values ​​for system price cost, system function cost, single sensor price, and single sensor power consumption, respectively.

[0079] C10:C pr (X)≤C pr,max

[0080] C11:C pw (X)≤C pw,max

[0081] C12:

[0082] C13:

[0083] Among them, C pr,max C pw,max pr max ,pw max This is the maximum allowed value set.

[0084] The constraint C14 in step 3.3.4 is that the system service life should be greater than the minimum required value.

[0085] C14:T system (X)≥T system,min

[0086] Among them, T system,min This is the minimum requirement value set.

[0087] Furthermore, the final sensor selection and configuration scheme output in step 4 is obtained based on the design solution of a genetic algorithm, specifically including:

[0088] Step 4.1: Initialize n in the population ind Individual Y i , where n ind Let i be the population size, i = 1, 2, ..., n ind ;

[0089] Step 4.2: Set the algebraic counter k = 0 and calculate the objective function J(X) for each individual;

[0090] Step 4.3: Run the calculation process, k <n iter n iter This represents the maximum number of iterations.

[0091] Based on the objective function J(X) of an individual, select pairs of individuals from the previous generation, perform chromosome crossover on the selected pairs of individuals with a certain crossover probability, perform mutation on the chromosomes after crossover with a certain mutation probability, generate a new population, and increment the generation counter k by 1.

[0092] Step 4.4: End the calculation and obtain the optimal individual Y. opt Y opt This is for the final sensor selection and configuration scheme.

[0093] The beneficial effects of this invention are:

[0094] This application first satisfies the perception requirements of a given autonomous driving function and the selection and configuration of sensors in the perception system. Then, it analyzes the specific characteristics and perception requirements of the given function to help establish a function-oriented critical point set (CPS) and related constraints. Simultaneously, it determines the set of candidate sensors and the set of candidate installation locations on the vehicle. Multi-dimensional modeling is performed on individual sensors and the overall sensor system, including at least the dimensions of detection functionality, reliability, and economy. Then, by defining decision variables, objective functions, and constraints, an optimization model for the sensor system design is constructed. Based on this optimization model, an intelligent genetic algorithm is used to solve the optimization model, outputting the final sensor selection and configuration scheme. Finally, simulation is used to verify the effectiveness of the design model and solution method, enabling closed-loop control of the entire method. Through quantitative optimization design, automatic selection and configuration of sensors for a given autonomous driving function is achieved. Attached Figure Description

[0095] Figure 1 This is a general flowchart of an embodiment of the present invention;

[0096] Figure 2 This is a schematic diagram of the sensor detection range modeling according to an embodiment of the present invention;

[0097] Figure 3 This is a schematic diagram of the target detection range modeling of sensor functional coupling according to an embodiment of the present invention;

[0098] Figure 4 This is a schematic diagram of all the alternative installation positions on the vehicle and the vehicle coordinate system according to an embodiment of the present invention;

[0099] Figure 5 These are the alternative sensor parameters for embodiments of the present invention;

[0100] Figure 6 These are the simulation parameters for the ACC adaptive cruise control in this embodiment of the invention;

[0101] Figure 7 This is a diagram showing the actual deployment and effect of the sensors for ACC adaptive cruise control according to an embodiment of the present invention;

[0102] Figure 8 This is a graph showing the change of the objective function of ACC adaptive cruise control in an embodiment of the present invention with genetic evolution;

[0103] Figure 9 This is a diagram showing the actual deployment and effect of the sensors in the APA automatic parking assist according to an embodiment of the present invention;

[0104] Figure 10 These are the simulation parameters of the APA automatic parking assist in this embodiment of the invention;

[0105] Figure 11 This is a graph showing the change of the objective function of the APA automatic parking assist according to an embodiment of the present invention with genetic evolution. Detailed Implementation

[0106] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0107] The purpose of this application is to achieve automatic selection and configuration of sensors for a given autonomous driving function by designing a function-oriented optimization model and solving it based on a genetic algorithm, and to verify the effectiveness of the automatic selection and configuration method through the design of sensors with multiple functions.

[0108] like Figure 1 As shown, this embodiment provides an automatic optimal configuration method for sensors in an autonomous driving function-oriented perception system. This configuration method includes:

[0109] Step 1: Define the task inputs and outputs, and design the sensor selection and configuration of the perception system based on the function, given one or more autonomous driving functions to meet the perception requirements of the functions.

[0110] Step 2: Analysis and modeling of the sensor system, which includes function-related modeling and design-related modeling.

[0111] Step 3: Based on the functional and design-related modeling in Step 2, the optimization model for the sensor system design is constructed by defining decision variables, objective functions, and constraints.

[0112] Step 4: Based on the sensor system design in Step 3, construct an optimization model, solve the optimization model using an intelligent genetic algorithm, and output the final sensor selection and configuration scheme.

[0113] Step 5: Based on the results of Steps 3 and 4, perform simulation to verify the effectiveness of the design model and solution method.

[0114] Specifically, in this embodiment, the autonomous driving functions include at least Adaptive Cruise Control (ACC), Automated Parking Assistance (APA), Lane Keeping Assistance (LKA), and Automated Emergency Braking (AEB).

[0115] The sensor selection and configuration design for a perception system includes selection and configuration. Selection refers to choosing the appropriate product type and model from the candidate sensors, while configuration refers to deploying the selected sensors to specific locations on the vehicle.

[0116] It should be noted that different sensor types and models, as well as different deployment locations, have a significant impact on perception performance and cost. In order to find the best solution, this paper provides an optimized configuration method for sensor solutions in perception systems, which is oriented towards autonomous driving functions.

[0117] Specifically, in this embodiment, the function-related modeling mainly analyzes the specific characteristics and specific perception requirements of a given function to help establish a function-oriented critical point set (CPS) and CPS-related constraints, which will serve as some of the constraints for subsequent optimization modeling; the design-related modeling mainly determines the set of candidate sensors and the set of candidate installation locations on the vehicle, while performing multi-dimensional modeling of individual sensors and the overall sensor system, including at least the dimensions of detection functionality, reliability, and economy.

[0118] Furthermore, the analysis and modeling of the sensor system in step 2 specifically includes:

[0119] Step 2.1: Record the x and y coordinates and orientation angles of the candidate sensor installation locations in the vehicle top view in a matrix format.

[0120] First, define the vehicle length as L. veh Width is W vehThe system has n alternative sensor products and m alternative installation locations.

[0121] Then, define N = {1, 2, ..., n} as the set of candidate sensors, and define M = {0, 1, 2, ..., m} as the set of candidate installation locations. Here, 0 represents no installation, i.e., an empty location.

[0122] By defining matrix P deploy ∈R m×3 Record the x and y coordinates and orientation angles of m alternative installation locations in the top view of the vehicle.

[0123] Step 2.2: Determine the parameters of the sensor.

[0124] Define an 8-dimensional vector Para i ∈R 8 Let i ∈ N to describe the parameters of the i-th sensor, as follows:

[0125]

[0126] In the formula, type i The sensor types include four types: C, R, L, and S, where C is a camera, R is millimeter-wave radar, L is lidar, and S is ultrasonic radar; d i For the maximum detection distance, φ i Field of view (FoV) For distance detection accuracy, For higher velocity detection accuracy, pr i , for price, pw i To utilize power, T i For service life.

[0127] Step 2.3: Determine the detection range of the sensor.

[0128] like Figure 2 As shown, the detection range of the four sensors—camera, millimeter-wave radar, lidar, and ultrasonic radar—is simplified to d. i and φ i Zhang Cheng's fan shape or circle.

[0129] The sensor's detection range is a fan-shaped area φ i <360°, the sensor's detection range is a circular φ i =360°.

[0130] Step 2.4: Determine the key detection area based on the sensor's function.

[0131] like Figure 3As shown, a target detection area coupled with sensor function is defined. By modeling the target detection range coupled with function, the detection area emphasized by this function is determined.

[0132] Define the length of the target detection region as L field Width is W field The vehicle's position in the XOY coordinate system is (x veh ,y veh The vehicle's orientation angle in the XOY coordinate system is θ. veh .

[0133] Where, θ veh =0 indicates parallel to the X-axis, and counterclockwise rotation increases the size of the angle.

[0134] It should be noted that different autonomous driving functions have different perception requirements and areas of focus. For ACC adaptive cruise control, the focus is mainly on the area in front of the vehicle, while for APA automatic parking assist, the focus is mainly on the surrounding area around the vehicle.

[0135] Furthermore, the optimization model construction for the sensor system design in step 3 specifically includes:

[0136] Step 3.1: Determine the sensor configuration design scheme under n candidate sensors and m candidate installation locations.

[0137] First, define the decision variable vector X∈R. n Having a system dimension with n candidate sensors, the i-th element of vector X represents the installation location of the i-th candidate sensor, i.e., X[i]∈M.

[0138] Where X represents the sensor configuration design scheme given n alternative sensors and m alternative installation locations.

[0139] When X[i] = 0, it means that the i-th candidate sensor is not selected and therefore not installed. When X[i] is any other value, it means that the sensor is selected and installed at the installation position of that sequence number.

[0140] Step 3.2: Determine the objective function J(X). The objective function J(X) is used to evaluate the functionality, reliability, and economy of design scheme X. The formula for calculating the objective function J(X) is:

[0141] J(X)=αJ1(X)+βJ2(X)+γJ3(X)

[0142]

[0143] J2(X)=w 2i C pr (X)+w 22 Cpw (X)

[0144] J3(X) = -T system (X)

[0145] In the formula, α,β,γ,w 11 ,w 12 ,w 21 ,w 22 Let J1(X) be a positive constant, J2(X) be functionality, J3(X) be reliability, and r(X) be economy, and r(X) be detection coverage. For the key detection point p i The distance detection error under design scheme X, For the key detection point p i The velocity detection error under design scheme X, C pr (X) represents the system price cost, C pw (X) represents the cost of power usage, T system This refers to the system's lifespan.

[0146] The objective function J(X) is used to evaluate the functionality, reliability, and economy of design scheme X, specifically including:

[0147] Step 3.1.1: Functionality is evaluated through detection coverage and detection accuracy at key detection points. The formula for calculating detection coverage r(X) is:

[0148]

[0149] In the formula, S(X) is the area of ​​the intersection between the actual detection area and the target detection area of ​​scheme X, S target The area of ​​the target detection region.

[0150] The detection accuracy at critical detection points consists of distance detection error and speed detection error. Critical detection points are points that require focused detection for a given function. These points are located within the target detection area, indicating that the detection of the area near these points has a significant impact on the current function. For example, ACC (Adaptive Cruise Control) needs to define critical detection points at medium and long distances in front of the vehicle, while APA (Adaptive Assistive Assist) has critical detection points distributed throughout the surrounding area of ​​the vehicle. These critical detection points collectively form the Critical Detection Point Set (CPS).

[0151]

[0152] in, This represents the position of the i-th key detection point in the vehicle coordinate system (the vehicle coordinate system has its origin at the vehicle's geometric center, the front of the vehicle as the positive x-axis, and the left side of the vehicle as the positive y-axis). To distinguish the total n... C The importance of each key detection point is defined by an importance vector. The i-th element of vector I represents the importance of the i-th key detection point, and its value ranges from [0,1].

[0153] definition For the key detection point p i The distance detection error under design scheme X, For the key detection point p i Velocity detection error under design scheme X.

[0154] It should be noted that, and Based on the detection accuracy of the sensors, it can be calculated that when an object is detected by multiple sensors simultaneously, the detection error can be reduced through multi-sensor fusion.

[0155] Distance detection error The calculation formula is:

[0156]

[0157] Speed ​​detection error The calculation formula is:

[0158]

[0159] In the formula, In design scheme X, the detection range should be able to cover the key detection point p. i The number of sensors.

[0160] Step 3.1.2: Economic efficiency through system price cost C pr (X) and power usage cost C pw (X) is evaluated.

[0161]

[0162]

[0163]

[0164] In the formula, S(X[i]) is a function to determine whether the i-th candidate sensor is selected, where 1 represents that it is selected and 0 represents that it is not selected.

[0165] Step 3.1.3: Reliability is measured by system lifespan. System lifespan T system The calculation formula is:

[0166]

[0167] In the formula, n s This represents the number of sensors selected in design scheme X.

[0168] It should be noted that the designed sensor system is a series system, requiring all sensors to work together to complete the task, and the system lifespan T system It is measured by the minimum lifespan of a single sensor.

[0169] Step 3.3: Determine the constraints regarding the functionality, reliability, and economy of design scheme X from Step 3.2. Specifically, this includes:

[0170] Step 3.3.1: Boundary constraints on decision variable X.

[0171] First, the decision variable X must satisfy the boundary constraints, and the constraint condition C1 is defined as follows:

[0172] C1:

[0173] in, This represents the alternative installation locations for the camera. These represent alternative installation locations for millimeter-wave radar. These represent alternative installation locations for the lidar. This represents the alternative installation locations for the ultrasonic radar.

[0174] Step 3.3.2: Define the following constraints C2-C9 for functional constraints.

[0175] The maximum detection distance achievable by design scheme X for the area around the vehicle should satisfy a minimum constraint, namely:

[0176] C2:d front (X)≥d front,min

[0177] C3:d back (X)≥d back,min

[0178] C4:d left (X)≥d left,min

[0179] C5:d right (X)≥d right,min

[0180] Where, d front,min d back,min d left,min and d right,min This represents the minimum allowable detection distance in the four directions: front, back, left, and right.

[0181] Key detection points should be covered and detected by sensors, or, when visual information is required, by cameras.

[0182] C6:NumCover(p i X)≥1, i=1,2,..,n C

[0183] C7:

[0184] Among them, the function NumCover(p i The detection area in the design scheme X can cover the key detection point p. i The number of sensors, function The detection area in design scheme X can cover the key detection point p. i The number of cameras, This represents the key detection points that require visual inspection.

[0185] Furthermore, the distance detection error and velocity detection error at key detection points should meet certain constraints, namely:

[0186] C8:

[0187] C9:

[0188] in, To the maximum permissible distance detection error, This represents the maximum permissible speed detection error.

[0189] Step 3.3.3: Define constraints C10-C13 for economic constraints.

[0190] The system price cost, system function cost, individual sensor price, and individual sensor power consumption have maximum allowable values, namely:

[0191] C10:C pr (X)≤C pr,max

[0192] C11:C pw (X)≤C pw,max

[0193] C12:

[0194] C13:

[0195] Among them, C pr,max C pw,max pr max ,pw max This is the maximum allowed value set.

[0196] Step 3.3.4: Define the C14 constraint for reliability constraints.

[0197] The system's lifespan should exceed the minimum required value.

[0198] C14:T system (X)≥T system,min

[0199] Among them, T system,min This is the minimum requirement value set.

[0200] Step 3.3.5: Based on engineering experience and patterns, use vectors Let C15 represent a set of installation locations where the same sensor is selected.

[0201] C15:s i =s i′ if X[i] and X[i′]∈D j for anyj and i,i′∈N; and if type i

[0202] =type i′ .

[0203] It should be noted that in the sensor configuration of the autonomous driving perception system, the same sensor can be selected in some different installation locations, which is beneficial for subsequent system maintenance and facilitates replacement and repair.

[0204] Step 3.3.5: Based on the C1-C15 constraints defined in Steps 3.3.1 to 3.3.4, establish the overall optimization model.

[0205] In another embodiment, for a specific autonomous driving function, this optimization model It can be simplified according to actual functional characteristics and requirements.

[0206] Furthermore, the final sensor selection and configuration scheme output in step 4 is obtained based on the design solution of a genetic algorithm.

[0207] The main operators of genetic algorithms include encoding operators, selection operators, crossover operators, and mutation operators. In this embodiment, 0-1 binary encoding, random traversal selection operators, two-point crossover operators, and binary mutation operators are applied.

[0208] The computational principle of genetic algorithms is to establish the fitness of individuals as the objective function J(X), and their phenotype as the decision variable X. Constraints are applied to the individual's objective function J(X) in the form of a penalty function. That is, when the design scheme represented by an individual's phenotype violates the defined constraints, the objective function J(X) of that individual increases, making it exceed the objective functions J(X) of all other individuals that do not violate the constraints. This greatly reduces the probability of the individual that violates the constraints being selected for the next generation, thereby achieving the purpose of survival of the fittest. The specific calculation method is as follows:

[0209] Step 4.1: Initialize n in the population ind Individual Y i .

[0210] Where, n ind Let i be the population size, i = 1, 2, ..., n ind .

[0211] Step 4.2: Set the algebraic counter k = 0 and calculate the objective function J(X) for each individual.

[0212] Step 4.3: Run the calculation process, k <n iter n iter This represents the maximum number of iterations.

[0213] Based on the objective function J(X) of an individual, select pairs of individuals from the previous generation, perform chromosome crossover on the selected pairs of individuals with a certain crossover probability, perform mutation on the chromosomes after crossover with a certain mutation probability, generate a new population, and increment the generation counter k by 1.

[0214] Step 4.4: End the calculation and obtain the optimal individual Y. opt .

[0215] Y opt This is for the final sensor selection and configuration scheme.

[0216] Example 1:

[0217] This application provides an automatic optimal configuration method for sensors in a perception system guided by autonomous driving functions, which sets the basic parameters of alternative installation locations and alternative sensors on the vehicle in a simulation.

[0218] like Figure 4 As shown, first set the vehicle length L veh = 4.5m, width W veh =2m, there are m=14 alternative installation positions on the vehicle, i.e., M={0,1,2,…,14}. The specific positions and orientation angles of all alternative installation positions on the vehicle in the vehicle coordinate system are represented by the matrix P. deploy for:

[0219]

[0220] It should be noted that, in order to speed up the solution, not all of the 14 alternative installation locations on the vehicle need to be included in every autonomous driving function. For a specific function, a subset of the 14 locations can be established as its alternative installation location set according to its own characteristics and needs.

[0221] like Figure 5 As shown, the simulation section provides 56 alternative sensors.

[0222] This completes the basic parameter settings for the alternative installation locations and alternative sensors on the vehicle in the simulation.

[0223] Example 2:

[0224] This application provides an automatic optimal configuration method for sensors in a perception system guided by autonomous driving functions. Based on Embodiment 1, the proposed design model and solution method are verified by taking ACC adaptive cruise control as an example for sensor selection and configuration.

[0225] The ACC function can control the longitudinal movement of the vehicle, so that the vehicle can maintain a stable and safe distance from the vehicle in front or travel at a set longitudinal speed. Therefore, ACC mainly focuses on the detection capability of medium and long distances in front of the vehicle.

[0226] like Figure 5 As shown, based on the functional characteristics of ACC, 26 sensors (including cameras, LiDAR, and millimeter-wave radar) with serial numbers 1-26 are selected from the 56 sensors in Example 1 as candidate sensors for ACC. Therefore, the decision variable X is a 26-dimensional vector.

[0227] like Figure 4 As shown, based on engineering design experience, locations 1, 2, 3, 5, 8, 11, and 14 are selected as seven alternative installation locations. The set in constraint C1 is defined as: M C ={0,2}, M R ={0,1,5,8,11,14}, M L ={0,3}.

[0228] It should be noted that different types of sensors are suitable for different installation locations. For example, installation location 2 refers to the area behind the rearview mirror, which is suitable for installing camera-type sensors. Additionally, since ultrasonic radar was not selected as a candidate sensor, it is not necessary to define M in step 3.3.1. S .

[0229] The critical detection point set for ACC includes two critical detection points: CPS = {p1(150,0), p2(50,0)}, that is, p1 and p2 are located 150 meters and 50 meters directly in front of the vehicle, respectively. The importance vector of the critical detection point is I = [1,1], so both points need to be visually detected.

[0230] Since ACC focuses primarily on the detection area directly in front of the vehicle, constraints C3-C5 can be omitted; furthermore, ACC typically involves a small number of sensing sensors, so constraint C15 can also be omitted. Therefore, the objective function of the optimization model for ACC is J(X), and the constraints are C1, C2, and C6-C14.

[0231] like Figure 6 , Figure 7 As shown, the number of individuals n in the population in the genetic algorithm ind Set to 20, maximum number of iterations n iter Set it to 500, and set other simulation parameters accordingly. Figure 6 After setting the simulation parameters, the solution results are as follows:

[0232] X=[0,0,2,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,1,0,0,0]

[0233] The results indicate that the third sensor (camera) was selected and installed at location 2, the 20th and 23rd sensors (millimeter-wave radar) were selected and installed at location 1, and none of the lidar sensors were selected.

[0234] like Figure 8 As shown, fitness (objective function) changes with genetic evolution. The horizontal line represents the optimal fitness of the population, and the vertical line represents the average fitness of the population.

[0235] Example 3:

[0236] This application provides an automatic optimal configuration method for sensors in a perception system guided by autonomous driving functions. Based on Embodiment 1, the proposed design model and solution method are verified by taking APA (Automatic Parking Assist) as an example for sensor selection and configuration.

[0237] APA primarily focuses on detecting targets within a relatively short range around the vehicle. Considering its functional requirements, 40 candidate sensors, numbered 5-14 and 27-56, were selected, mostly short-range cameras, fisheye cameras, and ultrasonic radars.

[0238] like Figure 4 As shown, there are 12 alternative installation locations, including locations 1, 4-14.

[0239] Define the set in constraint C1 as: M C ={0,1,4,9,10}, M S ={0,1,4,5,6,7,8,9,10,11,12,13,14}. Define the sets in constraint C15 as: D1 = {1,4,9,10}, D2 = {1,4,6,7,12,13}, D3 = {5,8,11,14}.

[0240] like Figure 9 As shown, the key detection point set is defined as 12 points around the vehicle.

[0241] like Figure 10 As shown, the genetic population of individual n ind Set to 10, maximum number of iterations n iter Set to 100, and set other simulation parameters accordingly. Figure 10 After setting the simulation parameters, the solution results are as follows:

[0242] X=[0,0,0,0,0,0,1,4,9,10,0,0,0,0,0,0,0,0,0,0,2,2,2,2,0,0,0,0,0,0,1,1,1,1,1,1,0,0,0,0]

[0243] like Figure 5 As shown, sensors 11-14 (cameras) were selected to be installed at positions 1, 4, 9, and 10, respectively; sensors 37-40 (ultrasonic radars) were selected to be installed at positions 5, 8, 11, and 14, respectively; and sensors 47-52 (ultrasonic radars) were selected to be installed at positions 1, 4, 6, 7, 12, and 13.

[0244] like Figure 11 As shown, fitness (objective function) changes with genetic evolution. The horizontal line represents the optimal fitness of the population, and the vertical line represents the average fitness of the population.

[0245] The above descriptions are merely embodiments of the present invention, and common knowledge regarding specific structures and characteristics of the solutions is not described in detail here. It will be apparent to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the present invention is defined by the appended claims rather than the foregoing description. Therefore, it is intended that all variations falling within the meaning and scope of equivalents of the claims be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for automatic optimal configuration of sensors in a perception system oriented towards autonomous driving functions, characterized in that, The automatic optimal configuration method for sensors in the sensing system includes: Step 1: Define the task input and output, and design the sensor selection and configuration of the perception system based on the function, given one or more autonomous driving functions to meet the perception requirements of the functions; Step 2: Analysis and modeling of the sensor system, which includes function-related modeling and design-related modeling; Step 2.1: Record the x and y coordinates and orientation angles of the candidate sensor installation locations in the vehicle's top view using a matrix; Step 2.2: Determine the sensor parameters; Step 2.3: Determine the detection range of the sensor; Step 2.4: Define a target detection area coupled with the sensor function. By modeling the target detection range coupled with the function, determine the detection area that the function focuses on. Step 3: Based on the functional-related modeling and design-related modeling in Step 2, the optimization model for the sensor system design is constructed by defining decision variables, objective functions, and constraints. Step 3.1: Determine the sensor configuration design scheme under n candidate sensors and m candidate installation locations; Step 3.1.1: Functionality is evaluated by detection coverage and detection accuracy at key detection points; Detection coverage The calculation formula is: ； In the formula, For the plan The area of ​​the intersection between the actual detection area and the target detection area. The area of ​​the target detection region; Distance detection error The calculation formula is: ； Speed ​​detection error The calculation formula is: ； In the formula, In the design scheme In the middle, the detection range can cover key detection points. The number of sensors; For distance detection accuracy; For speed detection accuracy; Step 3.1.2: Economic efficiency through system price and cost and power consumption cost ; Step 3.1.3: Reliability is measured by system lifespan; Step 3.2: Determine the objective function objective function Used to evaluate design schemes Functionality, reliability, and economy; Step 3.3: Regarding the design scheme in Step 3.2 Functionality, reliability, and economy; determine the constraints. The constraint that key detection points should be covered by sensors, or, when visual information is required, by cameras: ； ； Among them, the function Return to design scheme The central detection area can cover key detection points. The number of sensors, function Return to design scheme The central detection area can cover key detection points. The number of cameras, These represent key detection points that require visual inspection. The importance of key detection points; The constraints that the distance detection error and velocity detection error at key detection points should meet are as follows: ； ； in, To the maximum permissible distance detection error, The maximum permissible speed detection error; Step 4: Based on the sensor system design in Step 3, construct an optimization model, solve the optimization model using an intelligent genetic algorithm, and output the final sensor selection and configuration scheme; Step 5: Based on the results of Step 3 and Step 4, perform simulation to verify the effectiveness of the design model and solution method.

2. The automatic optimal configuration method for sensors in an autonomous driving function-oriented perception system according to claim 1, characterized in that, In step 2.2, the sensor parameters are determined by defining an 8-dimensional vector. To describe the first The parameters of each sensor are as follows: ； In the formula, The sensor types include four types: C, R, L, and S. C is a camera, R is millimeter-wave radar, L is lidar, and S is ultrasonic radar. For the longest detection distance, For the field of view, For distance detection accuracy, For higher speed detection accuracy, For price, To use power, For service life; Step 2.3, determining the sensor's detection range, specifically involves simplifying the detection ranges of the four sensors—camera, millimeter-wave radar, lidar, and ultrasonic radar—to... and Zhang Cheng's fan shape or circle.

3. The automatic optimal configuration method for sensors in an autonomous driving function-oriented perception system according to claim 1, characterized in that, Step 3.1, which determines the sensor configuration design scheme, specifically includes: Define decision variable vector A vector system having dimensions corresponding to the number of n candidate sensors. The i-th element represents the installation location of the i-th candidate sensor, i.e. ; in, Design a sensor configuration scheme given n candidate sensors and m candidate installation locations; A set of alternative installation locations; The time indicates that the i-th alternative sensor was not selected and therefore will not be installed. Other values ​​indicate that the item is selected and installed at the installation location of that sequence number.

4. The automatic optimal configuration method for sensors in an autonomous driving function-oriented perception system according to claim 3, characterized in that, The objective function in step 3.2 The calculation formula is: ； ； ； ； In the formula, A positive constant. , For reliability, For economic reasons, To detect coverage Key detection points In the design scheme The distance detection error is as follows. Key detection points In the design scheme The speed detection error is below. For system price cost, For power cost, For the system's lifespan, The importance of the i-th key detection point is denoted by , and its value ranges from [0,1].

5. The automatic optimal configuration method for sensors in an autonomous driving function-oriented perception system according to claim 1, characterized in that, The system price cost in step 3.1.2 The calculation formula is: ； Power cost The calculation formula is: ； In the formula, , This is a function to determine whether the i-th candidate sensor is selected, where 1 represents selection and 0 represents non-selection. For price, For power usage; The system lifespan in step 3.1.3 The calculation formula is: ； In the formula, For design scheme The number of sensors selected.

6. The automatic optimal configuration method for sensors in an autonomous driving function-oriented perception system according to claim 4, characterized in that, Step 3.3 refers to the design scheme in step 3.

2. Functionality, reliability, and economy are considered, and constraints are determined, specifically including: Step 3.3.1: Decision Variables Define the boundary constraints, specifically the C1 constraint; Step 3.3.2: Define constraints C2-C9 for functional constraints; Step 3.3.3: Define constraints C10-C13 for economic constraints; Step 3.3.4: Define the C14 constraint for the reliability constraint; Step 3.3.5: Based on engineering experience and patterns, use vectors Let C15 represent a set of installation locations where the same sensor is selected. To select the number of sets of installation locations for the same sensor; Step 3.3.6: As defined in steps 3.3.1 to 3.3.4 Constraints, establish an overall optimization model .

7. The automatic optimal configuration method for sensors in an autonomous driving function-oriented perception system according to claim 6, characterized in that, The constraint C1 in step 3.3.1 is: ； in, These represent alternative installation locations for the camera. These represent alternative installation locations for millimeter-wave radar. This represents the alternative installation locations for the lidar. This represents the alternative installation locations for the ultrasonic radar; The constraints C2-C5 in step 3.3.2 are the design scheme. The maximum achievable detection distance around the vehicle should meet a minimum constraint: ； ； ； ； in, , , and This represents the minimum allowable detection distance in the four directions: front, back, left, and right. The constraints C10-C13 in step 3.3.3 have maximum allowable values ​​for system price cost, system function cost, single sensor price, and single sensor power consumption, respectively. ； ； ； ； in, , , , This is the maximum allowed value set. The constraint C14 in step 3.3.4 is that the system service life should be greater than the minimum required value. ； in, This is the minimum requirement value set.

8. The automatic optimal configuration method for sensors in an autonomous driving function-oriented perception system according to claim 1, characterized in that, The final sensor selection and configuration scheme output in step 4 is obtained based on the design solution of a genetic algorithm, specifically including: Step 4.1: Initialize the population individual ,in, For population size, ; Step 4.2: Set up the algebraic counter Calculate the objective function for each individual. ; Step 4.3: Run the calculation process. , This represents the maximum number of iterations. Individual-based objective function Select pairs of individuals from the previous generation, perform chromosome crossover on the selected pairs with a certain crossover probability, and perform mutation on the crossover chromosomes with a certain mutation probability to generate a new population. (Generation counter) Increase by 1; Step 4.4: End the calculation and obtain the optimal individual. , This is for the final sensor selection and configuration scheme.

Images