Automatic driving vehicle chassis domain evaluation method based on comprehensive intelligent measurement

Abstract

The invention belongs to the technical field of automatic driving test, and discloses an automatic driving vehicle chassis domain evaluation method with comprehensive intelligent measurement, which comprises the following steps: step 1, a vehicle chassis adopts a nonlinear system integrated with multivariable control, the nonlinear system is decoupled through a decoupling pre-compensator, and the decoupled nonlinear system is connected with a decoupled pre-compensator; therefore, evaluation element data in the urban high traffic conflict density scene are extracted; 2, establishing an evaluation system according to related evaluation elements and traffic flow characteristics, and selecting evaluation indexes; and step 3, synthesizing a CRITIC method, an entropy evaluation method and an order relation analysis method to obtain a weight set, and obtaining an evaluation matrix according to a fuzzy comprehensive evaluation method so as to construct an evaluation model. According to the method, extraction of chassis dynamic performance elements is realized, and objective evaluation of the vehicle chassis performance in an urban high traffic conflict density scene is also realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of autonomous driving testing, and in particular relates to a chassis domain evaluation method for autonomous driving vehicles with comprehensive intelligent metrics. Background Art

[0002] In the field of autonomous driving evaluation, a set of standard evaluation methods has not yet been formed. Among them, subjective evaluation methods such as hierarchical analysis method and G1 are weighted according to expert experience, which is difficult to avoid the influence of subjectivity. Objective evaluation methods such as entropy method and gray theory method have a narrow scope of application. Moreover, when the vehicle is running, longitudinal movement and lateral movement are often coupled. Under steering conditions, the change of tire lateral force simultaneously forms the lateral movement and longitudinal movement of the whole vehicle. When steering and braking, the longitudinal force generated by steering simultaneously causes the longitudinal force of the front and rear tires to change. The influence of the coupling of longitudinal and lateral effects in the above steering process is specifically manifested in that when the vehicle turns, the direction of the tire force changes with the change of the front wheel angle, the longitudinal component force decreases, and the lateral component force increases, resulting in a decrease in longitudinal acceleration, thereby affecting the acceleration of the vehicle. Obviously, the current vehicle evaluation system is difficult to accurately analyze and evaluate the above coupling situation, and cannot provide a more objective and accurate evaluation of various aspects such as the vehicle's operating reliability. Summary of the invention

[0003] The purpose of the present invention is to provide a comprehensive intelligent metric evaluation method for the chassis domain of an autonomous driving vehicle, which is used to solve the technical problem that the prior art is difficult to provide objective and accurate evaluation results when the longitudinal and lateral effects of the vehicle are coupled under steering conditions.

[0004] The method for evaluating the chassis domain of an autonomous driving vehicle based on a comprehensive intelligent metric comprises the following steps.

[0005] Step 1: The vehicle chassis adopts a nonlinear system with integrated multivariable control, and the nonlinear system is decoupled through a decoupling precompensator to extract evaluation factor data in a high traffic conflict density scenario in the city.

[0006] Step 2: Establish an evaluation system and select evaluation indicators based on relevant evaluation factors and traffic flow characteristics.

[0007] A criterion layer is established based on the evaluation factors obtained from the decoupling of the chassis model and the analysis of traffic flow and other aspects. The criterion layer includes safety, instability, and acceleration. By analyzing the main influencing factors of chassis safety, instability, and acceleration under high traffic conflict density scenarios in cities, six evaluation indicators are finally determined, namely, comprehensive steering collision index, acceleration, TLC, yaw moment distribution, number of emergency braking times under high traffic conflict flow, and number of successful merging times under high traffic conflict density.

[0008] The weight set is obtained by combining the CRITIC method, entropy method and order relationship analysis method, and then the evaluation matrix is ​​obtained according to the fuzzy comprehensive evaluation method to construct the evaluation model.

[0009] Preferably, the step 2 comprises.

[0010] 1) By integrating TTC and TLC, a comprehensive index of steering collision is obtained to reflect the combined influence of longitudinal collision and lateral deviation in steering collision. The calculation formula is as follows:

[0011]

[0012] Among them, CISC stands for the comprehensive index of steering collision. Used to measure the speed of collision with the vehicle in front, which is positively correlated with the longitudinal acceleration. It is used to measure the speed of lane deviation and is negatively correlated with the longitudinal acceleration. A is the weight constant of the same lane and B is the weight constant of the adjacent lane. The two constants A and B are determined according to the changes in the heading angle and yaw rate in the experimental data during the collision time experiments of the same lane and adjacent lane. This step measures the acceleration of the vehicle by combining the steering collision comprehensive index and the acceleration index.

[0013] Preferably, the step 2 also includes.

[0014] 2) When the vehicle turns, lateral motion and yaw motion occur simultaneously and affect the vehicle stability. The indicator layer applied in this step measures the vehicle instability through TLC and yaw moment distribution. The calculation formula of TLC is:

[0015]

[0016] Where v is the driving speed, y is rr is the vertical distance between the right front wheel and the lane line, and θ is the relative heading angle. The vehicle generates additional yaw moment through the distribution of yaw moment to suppress the "head-pushing" or "tail-swinging" phenomenon and maintain a stable vehicle driving state. The yaw moment distribution is used to measure the stability of the vehicle under harsh working conditions.

[0017] 3) At the traffic flow level, this step extracts the number of emergency braking times under high traffic conflict flow and the number of successful merging times under high traffic conflict density, and measures vehicle safety in an urban traffic environment by the number of emergency braking times and the number of successful merging times.

[0018] Preferably, the step three uses the CRITIC method to analyze the influence of the correlation between indicators to obtain the CRITIC weights; uses the entropy method to analyze the difference in the amount of information of each indicator to obtain the entropy method weights; uses the ordinal relationship analysis method to analyze the importance of each indicator to obtain the ordinal relationship weights; and combines the various weights to obtain a comprehensive weight set.

[0019] Preferably, the CRITIC method is used for low-dimensional matrices [m i ' j ] to perform standard deviation calculation:

[0020]

[0021] Among them, m′ ij Represents the importance index of each evaluation index, that is, the importance index of the i-th evaluation index of the j-th level, the maximum value of level j is n, and the maximum value of label i is m; δ y is the standard deviation of the importance index of the yth group, which represents the size of the difference in the importance index values ​​of each evaluation indicator; next, the correlation coefficient σ between the importance index data of the ith group and the jth group is calculated ij :

[0022]

[0023] The amount of information contained in each weight vector is expressed by the following formula:

[0024]

[0025] The jth weight vector K j The corresponding CRITIC weights are:

[0026]

[0027] Where y represents the weight vector K j The total number, T j That is the desired CRITIC weight.

[0028] Preferably, the entropy value method includes: the entropy value calculation formula of the i-th evaluation index of the j-th level related to the test evaluation of the autonomous driving vehicle is:

[0029]

[0030] In the formula, p ij represents the probability of indicator i at the jth level, and n represents the number of states of a single indicator. The entropy value calculation formula of the i-th indicator weight is:

[0031]

[0032] According to the entropy value H i The properties of the obtained: 0≤w i ≤1, The entropy method is used to finally obtain the set of entropy weights of the evaluation index: w = (w1, w2, ..., w m ) T .

[0033] Preferably, the ordinal relation analysis method is performed by in the index set {x1,x1,…,x m}, the evaluation indicators are organized into a sequence relationship, namely:

[0034]

[0035] in, It represents the single factor index after forming the order relationship, and gives the evaluation index x k-1 With x k The importance ratio r k , the corresponding calculation formula is: Indicates the order relationship weight corresponding to each ranked evaluation index; order relationship weight It is calculated by the following two formulas:

[0036]

[0037] Among them, m represents the total number of evaluation indicators in the indicator set, r k The value range is [1,2).

[0038] Preferably, the comprehensive weight is obtained by the entropy weight w i , CRITIC weight T j and order relation weight The comprehensive calculation formula of the comprehensive weight in this step is:

[0039]

[0040] Among them, w it , T jt , represents the entropy weight, CRITIC weight and order relation weight of the t-th level evaluation index, ξ t represents the t-th level evaluation index, and finally the comprehensive weight of each level of evaluation index of the vehicle chassis control system is calculated as ξ=(ξ1,ξ2,…,ξ n ) T Based on the analysis in step 2, there are six evaluation indicators at each level involved, ξ1, ξ2, ξ3, ξ4, ξ5, and ξ6, which correspond to the comprehensive steering collision index, acceleration, TLC, yaw moment distribution, the number of emergency braking times under high traffic conflict flow, and the number of successful merging times under high traffic conflict density.

[0041] Preferably, the specific steps of establishing the autonomous driving vehicle chassis domain test evaluation model in step 3 in combination with the fuzzy comprehensive evaluation matrix are as follows:

[0042] (1) Based on the determination of the indicator system, for the single factor U iPerform single factor evaluation, i = 1, 2, 3, ..., m, m is the number of factors; get the evaluation level V corresponding to the i-th factor j The membership degree c ij , where j = 1, 2, 3, ..., n, where n is the number of levels; thus, the membership of the evaluation results of the m factors constitutes an evaluation matrix C, which is used to represent the fuzzy relationship from the factor set U to the evaluation level set V, and the expression is: C = (c ij ) m×n ,This step also needs to be normalized by row or column to obtain the final evaluation matrix C, and based on the six evaluation factors determined in step 2, n is 6;

[0043] (2) Determine the weight and single factor evaluation model. The comprehensive weight ξ obtained previously is combined with the evaluation matrix C to obtain the chassis domain test evaluation model of the autonomous driving vehicle, which is expressed as follows: E = C·ξ = (e1, e2, …, e n ), where the fuzzy comprehensive evaluation model E is the obtained model result, e i It represents the membership degree of the evaluation index corresponding to the evaluation level after weighting in the fuzzy comprehensive evaluation model E;

[0044] (2) Calculate the autonomous driving vehicle test evaluation score. In order to intuitively see the evaluation results, the membership evaluation level of each evaluation indicator is quantized into a score μ, and the comprehensive score S is generated by the calculation formula:

[0045] S=100Eμ=(e1,e2,...,e n )·μ T ×100

[0046] Among them, E is the fuzzy comprehensive evaluation model.

[0047] Preferably, in the step one, the dynamic and static information of the vehicle is first converted into the input of the vehicle chassis control system through the decision layer; then the reversibility of the system is determined by the Interactor algorithm, and the corresponding inverse system is obtained; for the multivariable control system, by giving a single input excitation, its steady-state response is measured, and the pairing relationship of the input and output is determined by establishing a variable pairing relationship; the inverse system model is used to compensate the original system into a pseudo-linear system with a linear transfer relationship, so as to realize a one-to-one mapping relationship between input and output; after performing the decoupling design, the output variables longitudinal velocity, center of mass sideslip angle and yaw angular velocity are extracted as the element features of the evaluation.

[0048] The advantages of the present invention are: based on the analysis of the requirements of safety, instability and acceleration in the criterion layer of vehicle operation, the present invention determines six evaluation indicators. These evaluation indicators can not only achieve the decoupling of longitudinal and lateral effects under the condition that the longitudinal and lateral effects of the vehicle are coupled under the steering condition, but also comprehensively consider the two factors of longitudinal collision and lateral deviation, and objectively and accurately evaluate the vehicle operation conditions under the steering condition and the vehicle chassis performance in the high traffic conflict density scene in the city. In the step of constructing the model to establish the evaluation index system, the present method combines the subjective and objective evaluation methods such as the entropy method, the order relationship analysis method (i.e., the G1 method) and the CRITIC method to determine the index weights, which reduces the subjectivity of the evaluation and expands the adaptability of the evaluation method. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 The present invention is a flowchart of a comprehensive intelligent metric chassis domain evaluation method for an autonomous driving vehicle.

[0050] Figure 2 Diagram of the vehicle chassis domain evaluation system used in the present invention. DETAILED DESCRIPTION

[0051] The specific implementation modes of the present invention are further explained in detail below by describing the embodiments with reference to the accompanying drawings, so as to help those skilled in the art to have a more complete, accurate and in-depth understanding of the inventive concept and technical solution of the present invention.

[0052] like Figure 1 to Figure 2 As shown, the present invention provides a chassis domain evaluation method for an autonomous driving vehicle based on a comprehensive intelligent metric, comprising the following steps.

[0053] Step 1: The vehicle chassis adopts a nonlinear system with integrated multivariable control, and the nonlinear system is decoupled through a decoupling precompensator to extract evaluation factor data in a high traffic conflict density scenario in the city.

[0054] In this step, the dynamic and static information of the vehicle is first converted into the input of the vehicle chassis control system through the decision layer. The control law of the chassis control system is formed according to the vehicle dynamics model. The model of the three-degree-of-freedom vehicle nonlinear system includes longitudinal motion, lateral motion and yaw motion.

[0055] The calculation formula for longitudinal movement is:

[0056]

[0057] The calculation formula for lateral movement is:

[0058]

[0059] The calculation formula for yaw motion is:

[0060]

[0061] In the above three calculation formulas, m is the vehicle mass, v is x is the longitudinal velocity, β is the sideslip angle of the center of mass, a is the distance from the front axle to the center of mass, b is the distance from the rear axle to the center of mass, J is the moment of inertia of the vehicle around the z axis, c f 、c r are the equivalent tire cornering stiffness of the front and rear wheels, δ f is the front wheel turning angle, F xtotal The total longitudinal force generated by the four tires of the car is M z It is the active yaw moment generated by the difference in longitudinal force between the left and right wheels of the vehicle.

[0062] In the model, u=[F xtotal δ f M z ] T is the system input, x=[v x βγ] T is the state variable, y=[v x βγ] T is the output variable.

[0063] Next, the reversibility of the system is determined by the Interactor algorithm. The calculation formula is:

[0064]

[0065] Let Y1 = y1, then the rank of the Jacobian matrix for the system input u is:

[0066]

[0067] From the calculation formula of lateral movement:

[0068]

[0069] Re-order The rank of the Jacobian matrix for the system input u is:

[0070]

[0071] From the calculation formula of yaw motion:

[0072]

[0073] Re-order The rank of the Jacobian matrix for the system input u is:

[0074]

[0075] From this we can get the relative vector order of the system: α = [α1 α2 α3] T =[1 1 1] T , the sum of the relative orders is 3, which is equal to the order of the system. According to the existence theorem of implicit functions, this system has an inverse system; the inverse system is expressed as: Connecting the inverse system in series in front of the vehicle model can achieve system decoupling.

[0076] Next, for the multivariable control system, by giving a single input excitation and measuring its steady-state response, determine the relative gain matrix of the system's front wheel angle and yaw moment to the center of mass slip angle. When the relative gain is close to 1, the system can determine the input-output pairing relationship through this variable pairing relationship, that is, select the front wheel angle to control the center of mass slip angle, the arm moment to control the yaw rate, and the longitudinal force generated by the four tires to control the longitudinal speed. Use the inverse system model to compensate the original system into a pseudo-linear system with a linear transfer relationship to weaken the coupling effect between the system loops and realize a one-to-one mapping relationship between input and output. After the decoupling design, extract the output variables longitudinal speed, center of mass slip angle and yaw rate as the element features of the evaluation.

[0077] For the element features obtained above, unify each of the element features into a preset range to obtain the normalized value of each evaluation element; then determine the data spectrum according to the normalized value, and each data point in the data spectrum is the element feature corresponding to the normalized value; for each data point, find K adjacent points for the data point; connect the data point with the surrounding adjacent points to form a local point cluster structure; obtain the weights between the data points; obtain the eigenvector and eigenvalue after dimensionality reduction according to the weight calculation; obtain a low-dimensional matrix according to the eigenvalue and eigenvector, complete the clustering of element features, and thus extract the evaluation element data under the urban high traffic conflict density scenario.

[0078] Step 2: Establish an evaluation system and select evaluation indicators based on relevant evaluation factors and traffic flow characteristics.

[0079] Based on the evaluation factors obtained by decoupling the chassis model and the analysis of traffic flow and other aspects, the criterion layer is established, which includes safety, instability, and acceleration. In the steering condition, the front wheel angle controls the sideslip angle of the center of mass, and the cross arm torque controls the yaw rate.

[0080] 1) The change in longitudinal force generated by the four tires will affect the vehicle acceleration, and as time accumulates, it will affect the longitudinal speed. The longitudinal speed reflects the standard-layer acceleration of the vehicle. In the steering process, there is coupling between longitudinal movement and lateral movement. Specifically, when the vehicle turns, the direction of the tire force changes with the change of the front wheel angle, the longitudinal component of force decreases, and the lateral component of force increases, resulting in a decrease in longitudinal acceleration, which affects the acceleration of the vehicle. The commonly used acceleration indicator TTC (collision time) takes into account the collision time of obstacles in the front lane. In actual situations, collisions often have certain angles and lane deviations, which can be regarded as a combination of longitudinal collision and lateral deviation. Therefore, this step integrates TTC and TLC to obtain a comprehensive steering collision indicator to reflect the impact of this coupling. The calculation formula is as follows:

[0081]

[0082] Among them, CISC stands for the comprehensive index of steering collision. Used to measure the speed of collision with the vehicle in front, which is positively correlated with the longitudinal acceleration. It is used to measure the speed of lane deviation and is negatively correlated with longitudinal acceleration. A is the weight constant of the same lane and B is the weight constant of the adjacent lane. The two constants A and B are determined by the changes in the heading angle and yaw rate in the experimental data during the collision time experiments of the same lane and adjacent lane. This step measures the acceleration of the vehicle by combining the steering collision comprehensive index and the acceleration index.

[0083] 2) When the vehicle is turning, if the sideslip angle of the center of mass is too large or too small, it will affect the stability of the vehicle. An appropriate yaw rate can make the vehicle easier to control. Specifically, the sideslip angle of the center of mass plus the yaw angle is equal to the heading angle of the vehicle. The heading angle affects the lateral movement of the vehicle. In the DYC (direct yaw moment control) control, the additional yaw angle is controlled by allocating the yaw torque, thereby affecting the yaw motion of the vehicle. The lateral motion and yaw motion occur simultaneously and jointly affect the stability of the vehicle. Therefore, the indicator layer applied in this step measures the vehicle instability through TLC (trigger-activated lane change assist) and yaw moment distribution. The calculation formula of TLC is:

[0084]

[0085] Where v is the driving speed, y is rr is the vertical distance between the right front wheel and the lane line, and θ is the relative heading angle. Under high speed or bad road conditions, the vehicle is very easy to lose stability and easily produce "head-pushing" or "tail-swinging" phenomena. The vehicle can effectively suppress these phenomena by generating additional yaw moment through yaw moment distribution to maintain a stable vehicle driving state. Therefore, yaw moment distribution is used to measure the stability of the vehicle under bad conditions.

[0086] 3) At the traffic flow level, traffic conflict is one of the main reasons affecting vehicle safety. Considering the traffic flow characteristics of cars in typical scenarios such as merging areas and intersections with high traffic conflicts, this step extracts the number of emergency braking times under high traffic conflict flow and the number of successful merging times under high traffic conflict density. The number of emergency braking times and the number of successful merging times are used to measure vehicle safety in urban traffic environments.

[0087] This step is based on the analysis of safety, instability, and acceleration in the criterion layer, and ultimately determines six evaluation indicators, namely, comprehensive steering collision index, acceleration, TLC, yaw moment distribution, number of emergency braking under high traffic conflict flow, and number of successful merging under high traffic conflict density.

[0088] Step 3: Integrate the CRITIC method, entropy method and order relationship analysis method (i.e., G1 method) to obtain the weight set, and then obtain the evaluation matrix based on the fuzzy comprehensive evaluation method to construct the evaluation model.

[0089] From step 2, we can see that the indicators of the evaluation system are interrelated. For example, TTC and TLC are both related to speed. Therefore, this step uses the CRITIC method to analyze the impact of the mutual correlation between indicators and obtain the CRITIC weight; the entropy method is used to analyze the difference in the amount of information of each indicator and obtain the entropy method weight; the order relationship analysis method is used to analyze the importance of each indicator and obtain the order relationship weight; and the various weights are combined to obtain a comprehensive weight set.

[0090] 1) CRITIC (determining the importance of indicators through correlation between indicators) method: for low-dimensional matrices:

[0091]

[0092] Perform standard deviation calculation:

[0093]

[0094] Among them, m′ ij Represents the importance index of each evaluation index, that is, the importance index of the i-th evaluation index of the j-th level, the maximum value of level j is n, and the maximum value of label i is m; δ y It is the standard deviation of the importance index of the yth group, which represents the size of the difference in the importance index values ​​of each evaluation indicator.

[0095] Next, calculate the correlation coefficient σ between the importance index data of the i-th group and the j-th group ij :

[0096]

[0097] The amount of information contained in each weight vector is expressed by the following formula:

[0098]

[0099] The jth weight vector K j The corresponding CRITIC weights are:

[0100]

[0101] Where y represents the weight vector K j The total number, T j That is the desired CRITIC weight.

[0102] 2) The entropy method uses the actual value of the required evaluation index to measure the amount of effective information, thereby determining the corresponding weight of the evaluation index. The larger the entropy value, the smaller the amount of system information and the smaller the indicator weight; conversely, the more certain the system is, the greater the amount of information and the greater the indicator weight. According to the definition of entropy, the entropy value calculation formula for the i-th evaluation index of the j-th level related to the test evaluation of autonomous driving vehicles is:

[0103]

[0104] In the formula, p ij represents the probability of indicator i at the jth level, and n represents the number of states of a single indicator. The entropy value calculation formula of the i-th indicator weight is:

[0105]

[0106] According to the entropy value H i The properties of the obtained: 0≤w i ≤1, The entropy method is used to finally obtain the set of entropy weights of the evaluation index: w = (w1, w2, ..., w m ) T .

[0107] 3) The order relationship analysis method is to use the index set {x1,x1,…,x m}, the evaluation indicators are organized into a sequence relationship, namely:

[0108]

[0109] in, It represents the single factor index after forming the order relationship, and gives the evaluation index x k-1 With x k The importance ratio r k , the corresponding calculation formula is: Indicates the order relationship weight corresponding to each ranked evaluation index. Order relationship weight It is calculated by the following two formulas:

[0110]

[0111] Among them, m represents the total number of evaluation indicators in the indicator set, r k The value range is [1,2), as shown in Table 1.

[0112] Table 1: Reference table for order relation assignment.

[0113]

[0114] 4) The comprehensive weight is calculated by the entropy weight w i , CRITIC weight T j and order relation weight The comprehensive calculation formula of the comprehensive weight in this step is:

[0115]

[0116] Among them, w it , T jt , represents the entropy weight, CRITIC weight and order relation weight of the t-th level evaluation index, ξ t represents the t-th level evaluation index, and finally the comprehensive weight of each level of evaluation index of the vehicle chassis control system is calculated as ξ=(ξ1,ξ2,…,ξ n ) T In this method, based on the analysis in step 2, there are six evaluation indicators at each level involved, ξ1, ξ2, ξ3, ξ4, ξ5, and ξ6, which correspond to the steering collision comprehensive index, acceleration, TLC, yaw moment distribution, the number of emergency braking times under high traffic conflict flow, and the number of successful merging times under high traffic conflict density.

[0117] Based on the comprehensive weight, the specific steps of establishing the autonomous driving vehicle test evaluation model through the fuzzy comprehensive evaluation method are as follows.

[0118] (1) Based on the determination of the indicator system, for the single factor U i Perform single factor evaluation, i = 1, 2, 3, ..., m, m is the number of factors; get the evaluation level V corresponding to the i-th factor j The membership degree c ij , where j = 1, 2, 3, ..., n, n is the number of levels. Therefore, the membership of the evaluation results of m factors constitutes an evaluation matrix C, which is used to express the fuzzy relationship from the factor set U to the evaluation level set V, and the expression is:

[0119]

[0120] This step also requires normalization by row or column to obtain the final evaluation matrix C, and based on the six evaluation factors determined in step two, n is 6.

[0121] (2) Determine the weight and single factor evaluation model. The comprehensive weight ξ obtained previously is combined with the evaluation matrix C to obtain the chassis domain test evaluation model of the autonomous driving vehicle, which is expressed as follows:

[0122]

[0123] Among them, the fuzzy comprehensive evaluation model E is the obtained model result, e i It represents the membership degree of the evaluation index of the i-th level corresponding to the evaluation grade after weighting in the fuzzy comprehensive evaluation model E.

[0124] For multi-level evaluation models, a multi-level fuzzy evaluation model is used. The calculation starts from the bottom layer, and the results are used to form the evaluation matrix of the evaluation factors of the previous level. Similarly, the autonomous driving vehicle test evaluation model E is obtained.

[0125] (3) Calculate the autonomous driving vehicle test evaluation score. In order to intuitively see the evaluation results, the membership evaluation level set of each evaluation indicator (good, relatively good, average, relatively poor, poor) is quantified into a score μ, and the score μ is integrated to generate a comprehensive score S in a percentage system. The calculation formula is:

[0126] S=100Eμ=(e1,e2,…,e n )·μ T ×100

[0127] Wherein, E is a fuzzy comprehensive evaluation model. In this embodiment, μ=[1.0 0.8 0.6 0.4 0.2].

[0128] The present invention is described above by way of example in conjunction with the accompanying drawings. It is obvious that the specific implementation of the present invention is not limited to the above-mentioned method. As long as various non-substantial improvements are made using the inventive concept and technical solution of the present invention, or the inventive concept and technical solution are directly applied to other occasions without improvement, they are all within the protection scope of the present invention.

Claims

1. A chassis domain evaluation method for autonomous driving vehicles based on comprehensive intelligent metrics, characterized by: The following steps are involved: Step 1: The vehicle chassis adopts a nonlinear system with integrated multivariable control, and the nonlinear system is decoupled by a decoupling precompensator to extract evaluation factor data in a high traffic conflict density scenario in the city; Step 2: Establish an evaluation system and select evaluation indicators based on relevant evaluation factors and traffic flow characteristics; Based on the evaluation factors obtained by decoupling the chassis model and the analysis of traffic flow and other aspects, a criterion layer is established, which includes safety, instability, and acceleration. By analyzing the main influencing factors of chassis safety, instability, and acceleration in urban high traffic conflict density scenarios, six evaluation indicators are finally determined, namely, steering collision comprehensive index, acceleration, TLC, yaw moment distribution, number of emergency braking under high traffic conflict flow, and number of successful merging under high traffic conflict density. Step 3: The weight set is obtained by integrating the CRITIC method, entropy method and order relationship analysis method, and then the evaluation matrix is ​​obtained according to the fuzzy comprehensive evaluation method to construct the evaluation model.

2. The comprehensive intelligent metric chassis domain evaluation method for autonomous driving vehicles according to claim 1 is characterized by: The second step comprises: 1) By integrating TTC and TLC, a comprehensive index of steering collision is obtained to reflect the combined influence of longitudinal collision and lateral deviation in steering collision. The calculation formula is as follows: Among them, CISC stands for the comprehensive index of steering collision. Used to measure the speed of collision with the vehicle in front, which is positively correlated with the longitudinal acceleration. It is used to measure the speed of lane deviation and is negatively correlated with the longitudinal acceleration. A is the weight constant of the same lane and B is the weight constant of the adjacent lane. The two constants A and B are determined according to the changes in the heading angle and yaw rate in the experimental data during the collision time experiments of the same lane and adjacent lane. This step measures the acceleration of the vehicle by combining the steering collision comprehensive index and the acceleration index.

3. The comprehensive intelligent metric chassis domain evaluation method for autonomous driving vehicles according to claim 2 is characterized by: The step 2 also includes: 2) When the vehicle turns, lateral motion and yaw motion occur simultaneously and affect the vehicle stability. The indicator layer applied in this step measures the vehicle instability through TLC and yaw moment distribution. The calculation formula of TLC is: Where v is the driving speed, y is rr is the vertical distance between the right front wheel and the lane line, and θ is the relative heading angle. The vehicle generates additional yaw moment through the distribution of yaw moment to suppress the "head-pushing" or "tail-swinging" phenomenon and maintain a stable vehicle driving state. The yaw moment distribution is used to measure the stability of the vehicle under adverse working conditions. 3) At the traffic flow level, this step extracts the number of emergency braking times under high traffic conflict flow and the number of successful merging times under high traffic conflict density, and measures vehicle safety in an urban traffic environment by the number of emergency braking times and the number of successful merging times.

4. The comprehensive intelligent metric chassis domain evaluation method for autonomous driving vehicles according to claim 1 is characterized by: The step three uses the CRITIC method to analyze the influence of the correlation between indicators and obtains the CRITIC weight; uses the entropy method to analyze the difference in the amount of information of each indicator and obtains the entropy method weight; uses the order relationship analysis method to analyze the importance of each indicator and obtains the order relationship weight; and combines each weight to obtain a comprehensive weight set.

5. The comprehensive intelligent metric chassis domain evaluation method for autonomous driving vehicles according to claim 4 is characterized by: The CRITIC method is used for low-dimensional matrices [m′ ij ] to perform standard deviation calculation: Among them, m′ ij Represents the importance index of each evaluation index, that is, the importance index of the i-th evaluation index of the j-th level, the maximum value of level j is n, and the maximum value of label i is m; δ y is the standard deviation of the importance index of the yth group, which represents the size of the difference in the importance index values ​​of each evaluation indicator; next, the correlation coefficient σ between the importance index data of the ith group and the jth group is calculated ij : The amount of information contained in each weight vector is expressed by the following formula: The jth weight vector K j The corresponding CRITIC weights are: Where y represents the weight vector K j The total number, T j That is the desired CRITIC weight.

6. The chassis domain evaluation method for an autonomous driving vehicle based on a comprehensive intelligent metric according to claim 4 is characterized by: The entropy value method includes: the entropy value calculation formula of the i-th evaluation index of the j-th level related to the test evaluation of the autonomous driving vehicle is: In the formula, p ij represents the probability of indicator i at the jth level, and n represents the number of states of a single indicator; the entropy value calculation formula of the i-th indicator weight is: According to the entropy value H i The properties of the obtained: 0≤w i ≤1, The entropy method is used to finally obtain the set of entropy weights of the evaluation index: w = (w1, w2, ..., w m ) T .

7. The comprehensive intelligent metric chassis domain evaluation method for autonomous driving vehicles according to claim 4 is characterized by: The order relation analysis method is to use the index set {x1,x1,…,x m }, the evaluation indicators are organized into a sequence relationship, namely: in, It represents the single factor index after forming the order relationship, and gives the evaluation index x k-1 With x k The importance ratio r k , the corresponding calculation formula is: Indicates the order relationship weight corresponding to each ranked evaluation index; order relationship weight It is calculated by the following two formulas: Among them, m represents the total number of evaluation indicators in the indicator set, r k The value range is [1,2).

8. The comprehensive intelligent metric chassis domain evaluation method for autonomous driving vehicles according to claim 4 is characterized by: The comprehensive weight is calculated by the entropy weight w i , CRITIC weight T j and order relation weight The comprehensive calculation formula of the comprehensive weight in this step is: Among them, w it 、T jt , represents the entropy weight, CRITIC weight and order relation weight of the t-th level evaluation index, ξ t represents the t-th level evaluation index, and finally the comprehensive weight of each level of evaluation index of the vehicle chassis control system is calculated as ξ=(ξ1,ξ2,…,ξ n ) T Based on the analysis in step 2, there are six evaluation indicators at each level, ξ1, ξ2, ξ3, ξ4, ξ5, and ξ6, which correspond to the comprehensive steering collision index, acceleration, TLC, yaw moment distribution, the number of emergency braking times under high traffic conflict flow, and the number of successful merging times under high traffic conflict density.

9. The comprehensive intelligent metric chassis domain evaluation method for autonomous driving vehicles according to claim 1 is characterized by: The specific steps of establishing the chassis domain test evaluation model of the autonomous driving vehicle in step 3 by combining the fuzzy comprehensive evaluation matrix are as follows: (1) Based on the determination of the indicator system, for the single factor U i Perform single factor evaluation, i = 1, 2, 3, ..., m, m is the number of factors; get the evaluation level V corresponding to the i-th factor j The membership degree c ij , where j = 1, 2, 3, ..., n, where n is the number of levels; thus, the membership of the evaluation results of the m factors constitutes an evaluation matrix C, which is used to represent the fuzzy relationship from the factor set U to the evaluation level set V, and the expression is: C = (c ij ) m×n ,This step also needs to be normalized by row or column to obtain the final evaluation matrix C, and based on the six evaluation factors determined in step 2, n is 6; (2) Determine the weight and single factor evaluation model. The comprehensive weight ξ obtained previously is combined with the evaluation matrix C to obtain the chassis domain test evaluation model of the autonomous driving vehicle, which is expressed as follows: E = C ξ = (e1, e2, ..., e n ), where the fuzzy comprehensive evaluation model E is the obtained model result, e i It represents the membership degree of the evaluation index corresponding to the evaluation level after weighting in the fuzzy comprehensive evaluation model E; (3) Calculate the test evaluation score of the autonomous driving vehicle. In order to intuitively see the evaluation results, the membership evaluation level of each evaluation indicator is quantized into a score μ, and the comprehensive score S is generated by the calculation formula: S=100Eμ=(e1,e2,...,e n )·m T ×100 Among them, E is the fuzzy comprehensive evaluation model.

10. The comprehensive intelligent metric chassis domain evaluation method for autonomous driving vehicles according to claim 1, characterized in that: In the step 1, the dynamic and static information of the vehicle is first converted into the input of the vehicle chassis control system through the decision layer; then the reversibility of the system is determined by the Interactor algorithm, and the corresponding inverse system is obtained; for the multivariable control system, a single input excitation is given, its steady-state response is measured, and the pairing relationship of input and output is determined by establishing a variable pairing relationship; The inverse system model is used to compensate the original system into a pseudo-linear system with a linear transfer relationship, realizing a one-to-one mapping relationship between input and output; After the decoupling design, the output variables longitudinal velocity, center of mass sideslip angle and yaw rate are extracted as the element features of the evaluation.

Images