An interpretable prediction method for autonomous driving emissions considering the influence of road geometric conditions

Abstract

The present invention proposes an interpretable prediction method for autonomous driving emissions considering the influence of road geometric conditions, including the following steps: Step S1, through on-site vehicle tests, record the data and control strategy-related information of autonomous vehicles, sort out their autonomous driving trajectories and emission data, and perform preprocessing to construct a dataset; Step S2, use the constructed dataset to infer the autonomous driving operating conditions and road geometry-related information; Step S3, implement feature engineering using machine learning methods; Step S4, select machine learning models with good interpretability from the classical machine learning model library for training; Step S5, evaluate based on the model prediction performance and interpretability to determine the optimal interpretable prediction model for autonomous driving emissions. The present invention can consider the influence of different road geometries and control strategies on autonomous driving emissions, and effectively improve the emission prediction accuracy of autonomous driving in real road scenarios.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic carbon emission prediction, in particular to an interpretable prediction method for autonomous driving emissions considering the influence of road geometric conditions. Background Art

[0002] With the rapid increase in the number of global vehicles, the problems of energy consumption and greenhouse gas emissions caused by road traffic have become increasingly severe. Reducing road traffic carbon emissions has become an urgent global issue. As an emerging technology that has developed rapidly in the contemporary traffic field, autonomous driving technology carries the expectation of improving energy utilization efficiency and reducing tail gas emissions. As the main carrier for vehicle travel, the geometric characteristics of roads have a significant impact on the operating states of both manually driven vehicles and autonomous driving vehicles. Different driving modes will inevitably lead to different operating states of vehicles when adapting to changes in road geometric conditions, resulting in differences in tail gas emissions. However, the emission prediction models constructed based on manually driven vehicle data are difficult to accurately predict and interpret autonomous driving emissions. Therefore, how to accurately predict and interpret the tail gas emissions of autonomous driving vehicles with different control strategies under the influence of road geometry is particularly important.

[0003] In response to this, existing research in the field has used methods such as real vehicle tests, virtual simulation tests, and theoretical analysis to mainly explore the influence mechanism of different road geometric conditions on tail gas emissions based on manually driven vehicle data and construct emission prediction models considering the influence of road geometric conditions. There is little research that takes autonomous driving vehicles as the research object and simultaneously considers the road curvature characteristics and road slope characteristics to construct an interpretable emission prediction model. Therefore, accurately predicting and interpreting the tail gas emissions of autonomous driving vehicles with different control strategies under the influence of road geometry has become a key problem that needs to be solved urgently and has important practical significance. Summary of the Invention

[0004] The present invention proposes an interpretable prediction method for autonomous driving emissions considering the influence of road geometric conditions, which can consider the influence of different road geometries and control strategies on autonomous driving emissions and effectively improve the emission prediction accuracy of autonomous driving in real road scenarios.

[0005] The present invention adopts the following technical solutions.

[0006] An interpretable prediction method for autonomous driving emissions considering the influence of road geometric conditions includes the following steps;

[0007] Step S1: Through on-site real vehicle test experiments, record the data of autonomous driving vehicles and information related to control strategies, sort out the autonomous driving trajectories and emission data, and perform preprocessing to construct a data set;

[0008] Step S2: Use the constructed data set to calculate the information related to autonomous driving operating conditions and road geometry;

[0009] Step S3: Implement feature engineering using machine learning methods;

[0010] Step S4: Screen machine learning models with good interpretability from the classical machine learning model library for training;

[0011] Step S5: Evaluate based on model prediction performance and interpretability to determine the optimal interpretable prediction model for autonomous driving emissions.

[0012] In Step S1, the vehicles selected for the experiment are: commercial autonomous driving vehicles of different brands and specifications on the existing market;

[0013] The experimental site is selected as: a protected closed test site where the terrain slope and curvature of the road have large fluctuations;

[0014] The experimental design is as follows: autonomous driving vehicles of different brands and specifications drive on the test site using different control strategies and set speeds, collect emission data using a Portable Emission Measurement System (PEMS), and record vehicle trajectory data using a Global Positioning System (GPS) and an On-Board Diagnostics Ⅱ (OBDⅡ) simultaneously;

[0015] In the experiment, the relevant information recorded for autonomous driving vehicles includes: fuel type, total load mass, maximum power, engine displacement, emission standard, fuel injection type, vehicle type classification, number of passenger seats, and maximum speed;

[0016] The relevant information recorded for the autonomous driving control strategy includes: Adaptive Cruise Control (ACC), time headway, ACC set speed (30 - 120 km / h), acceleration control strategy, and equipped auxiliary driving system;

[0017] The collected emission data includes: instantaneous emission rates of pollutants such as carbon dioxide, carbon monoxide, hydrocarbons, nitric oxide, and nitrogen dioxide;

[0018] The converted trajectory data includes: the abscissa of the vehicle x , the ordinate y , the vertical coordinate z , the speed V and the driving distance S of the vehicle per second;

[0019] The road geometric data includes: road slope G , road curvatureC , road slope change rate GCR , road curvature change rate CCR ;

[0020] The operating condition data includes: acceleration a , acceleration change rate, vehicle specific power VSP ;

[0021] In step S1, when sorting out the data, the trajectory data is converted into a three-dimensional rectangular coordinate system.

[0022] Step S1 includes the following steps;

[0023] Step S11, preprocess the obtained trajectory and emission data, including missing value processing, duplicate value processing, outlier detection and processing;

[0024] The method for processing missing values is to fill in the missing values using the cubic spline interpolation method;

[0025] The method for processing duplicate values is to find and delete exactly the same duplicate data;

[0026] The method for outlier detection and processing is to detect and eliminate outliers using the interquartile range method IQR;

[0027] Step S12, deduce the road geometric data, including road slope G , road curvature C , road slope change rate GCR , road curvature change rate CCR ;

[0028] The road slope G The calculation formula is:

[0029] ;

[0030] The road curvature C The calculation formula is:

[0031] ;

[0032] The road slope change rate GCR The calculation formula is:

[0033] ;

[0034] The road curvature change rate CCR The calculation formula is:

[0035] ;

[0036] In the formula, z t ist Vertical coordinate of the moment ΔS t is t- from 1 to t Change in driving distance at the moment Δx t is t- from 1 to t Change in abscissa at the moment Δy t is t- from 1 to t Change in ordinate at the moment G t is t Gradient at the moment C t is t Curvature at the moment

[0037] Step S13, Deduce the operating condition data of autonomous driving, including acceleration a , acceleration change rate, vehicle specific power VSP etc.;

[0038] The vehicle specific power VSP The calculation formula is:

[0039] ;

[0040] In the formula, V t is t Vehicle speed at the moment a t is t Vehicle acceleration at the moment G t is t Road gradient at the moment

[0041] Step S2 includes the following steps;

[0042] Step S21, Encode the category features related to autonomous vehicles and control strategies;

[0043] The features to be encoded include ACC headway, acceleration control strategy, sensors equipped on the vehicle, auxiliary driving systems equipped on the vehicle, fuel type, emission standard, vehicle type classification, fuel injection type, which together with the data obtained from the above records and deductions form the original training data set;

[0044] Step S22: Using the geometric path of the road, information related to the trajectory and control strategy of the autonomous vehicle, and the operating conditions of autonomous driving as input features, form model input data, and use autonomous driving emissions as the output result to construct and train multiple machine learning models with embedded feature importance. The machine learning models include XGBoost, LightGBM, Random Forest, Decision Tree, and Extra Trees;

[0045] Step S23: Use the above machine learning models to output the basic feature importance, and select the machine learning model with the largest proportion of feature importance related to road geometry and autonomous driving control strategy as the base model sensitive to road geometry conditions;

[0046] Step S24: Based on the model selected in S23, use the recursive feature elimination method with the root mean square error as the cross-validation evaluation index to screen the best feature combination.

[0047] The rules for encoding category features related to autonomous vehicles and control strategies in Step S21 are as follows:

[0048] If the feature name is ACC headway, the digital encoding rule is: take 1 for short time intervals; take 2 for medium time intervals; take 3 for long time intervals;

[0049] If the feature name is acceleration control strategy, the digital encoding rule is: take 1 for the constant headway strategy; take 2 for the safety distance model strategy; take 3 for the comfort optimization strategy; take 4 for the energy consumption-based optimization strategy;

[0050] If the feature name is lane keeping assist, the digital encoding rule is: take 1 for not installed; take 2 for installed;

[0051] If the feature name is lane centering control, the digital encoding rule is: take 1 for not installed; take 2 for installed;

[0052] If the feature name is automatic emergency braking, the digital encoding rule is: take 1 for not installed; take 2 for installed;

[0053] If the feature name is traffic congestion assist, the digital encoding rule is: take 1 for not installed; take 2 for installed;

[0054] If the feature name is blind spot monitoring and assist, the digital encoding rule is: take 1 for not installed; take 2 for installed;

[0055] If the feature name is highway assist driving, the digital encoding rule is: take 1 for not installed; take 2 for installed;

[0056] If the feature name is lidar, the digital encoding rule is: take 1 for not installed; take 2 for installed;

[0057] If the feature name is millimeter wave radar, the digital encoding rule is: take 1 for not installed; take 2 for installed;

[0058] If the feature name is camera, the digital coding rule is: 1 for not assembled; 2 for assembled;

[0059] If the feature name is ultrasonic sensor, the digital coding rule is: 1 for not assembled; 2 for assembled;

[0060] If the feature name is inertial measurement unit, the digital coding rule is: 1 for not assembled; 2 for assembled;

[0061] If the feature name is global navigation satellite system, the digital coding rule is: 1 for not assembled; 2 for assembled;

[0062] If the feature name is fuel type, the digital coding rule is: 1 for diesel; 2 for gasoline; 3 for CNG; 4 for LNG (liquefied natural gas); 5 for LPG (liquefied petroleum gas); 6 for hybrid electric.

[0063] If the feature name is emission standard, the digital coding rule is: 1 for National I; 2 for National II; 3 for National III;

[0064] 4 for National IV; 5 for National V; 6 for National VI;

[0065] If the feature name is vehicle type classification, the digital coding rule is: 1 for light gasoline vehicle; 2 for light diesel vehicle;

[0066] 1 for heavy gasoline vehicle; 2 for heavy diesel vehicle;

[0067] If the feature name is fuel injection type, the digital coding rule is: 1 for carburetor; 2 for electronic fuel injection; 3 for compression ignition; 4 for non-direct injection; 5 for direct injection.

[0068] Step S3 includes the following steps;

[0069] Step S31: Using the feature combinations screened in step S2, classify the data based on vehicle specific power VSP , road slope G and road curvature C ;

[0070] Step S32: Using road geometry, information related to autonomous vehicles and control strategies, and autonomous driving operating conditions as input variables, and autonomous driving emissions as the output variable, normalize the samples composed of input variables and output variables, and randomly divide the samples according to a preset ratio to form a model training set and a test set;

[0071] Step S33: Screen multiple machine learning models with good interpretability from the classical machine learning model library, complete parameter optimization through Bayesian optimization respectively, and then substitute them into training and testing;

[0072] The machine learning models include XGBoost, LightGBM, Random Forest, Decision Tree, Extra Trees, Support Vector Machine, Deep Forest, and Gradient Boosting Tree.

[0073] In step S32, the samples are randomly divided in a ratio of 7:3 as the model training set and the test set respectively.

[0074] Step S4 includes the following steps;

[0075] Step S41: Construct a hierarchical model, including an objective layer and an index layer;

[0076] The objective layer is: the interpretability of the model;

[0077] The index layer includes evaluation indexes of the consistency of feature importance distribution, the stability of feature interpretation, the interpretability of feature interaction, the clarity of global interpretation, and the stability of local interpretation;

[0078] The consistency of feature importance distribution is: the degree of conformity between the feature importance distribution output by the model and the expert expectation;

[0079] The stability of feature interpretation is: the stability degree of the SHAP (Shapley Additive Explanations) value distribution of different samples;

[0080] The interpretability of feature interaction is: the clarity degree of the model reflecting the interaction between features;

[0081] The clarity of global interpretation is: the clarity degree of the global SHAP interpretation of the model;

[0082] The accuracy of local interpretation is: the reasonable degree of the SHAP value of a single sample explaining the prediction result;

[0083] Step S42: Construct an importance judgment matrix based on expert scoring and perform normalization processing;

[0084] The importance judgment matrix is: formed by pairwise comparison and scoring of each evaluation index to form an importance judgment matrix;

[0085] The matrix form is:

[0086] ;

[0087] The normalization processing formula of the importance judgment matrix is:

[0088] ;

[0089] In the formula, the matrix element d ij is the iThe importance of the j th evaluation index relative to the d norm,ij th evaluation index, i is the normalization result of the importance of the j th evaluation index relative to the

[0090] Step S43: Calculate the weights using the eigenvector method and conduct a consistency test. If the consistency test passes, output the weights of each evaluation index. If the consistency test fails, return to Step S42 to adjust the importance judgment matrix;

[0091] The calculation of weights using the eigenvector method is as follows: Solve the maximum eigenvalue λ max and the corresponding eigenvector of the importance judgment matrix, and normalize the eigenvector to obtain the weight vector W , where the i th element W i is the weight of the i th evaluation index;

[0092] The consistency test is as follows: Calculate the consistency index CI and the consistency ratio CR . If CR <0.1, the importance judgment matrix has acceptable consistency. Otherwise, the importance judgment matrix needs to be adjusted;

[0093] The consistency index CI is calculated as follows:

[0094] ;

[0095] The consistency ratio CR is calculated as follows:

[0096] ;

[0097] In the formula, λ max is the maximum eigenvalue of the importance judgment matrix; W is the weight vector; W i is the weight of the i th evaluation index; CI is the consistency index; n is the dimension of the judgment matrix; CR is the consistency ratio; RI is the random consistency index depending on the dimension of the importance judgment matrix;

[0098] Step S44: Conduct interpretability analysis on each machine learning model based on the SHAP method. Invite experts to score respectively around the performance of the model on the above five evaluation indicators based on the interpretability analysis results of each machine learning model, and then calculate the weighted average as the comprehensive score according to the weights of each evaluation indicator, and sort the interpretability of the models in descending order based on the comprehensive score;

[0099] The interpretability analysis includes: feature importance analysis, feature interaction analysis, swarm plot analysis, force diagram analysis, decision path analysis, waterfall plot analysis;

[0100] Step S45: Sort the models in ascending order respectively based on the single machine learning model prediction performance evaluation indicators, and comprehensively sort the prediction performance of the models according to the average value of the sorting;

[0101] The model prediction performance evaluation indicators include root mean square error, mean absolute error, and mean absolute percentage error;

[0102] Step S46: Select the model with the smallest average value of prediction performance and interpretability sorting as the optimal emission prediction model.

[0103] In step S42, the method for obtaining the importance judgment matrix is: adopt the 1-9 scale method, invite experts to compare and score each pair of evaluation indicators to form an importance judgment matrix.

[0104] The present invention records the information related to the self-driving vehicle and control strategy of the automatic driving through on-site real vehicle tests, obtains the automatic driving trajectory and emission data, and calculates the information related to the automatic driving operation condition and road geometry; uses the machine learning method to implement feature engineering; screens a variety of machine learning models with good interpretability from the classic machine learning model library for training; finally determines the optimal automatic driving emission interpretable prediction model based on the model prediction performance and interpretability. Compared with the prior art, the present invention has the following beneficial effects:

[0105] (1) Using the self-driving vehicle equipped with PEMS to collect trajectory and emission data, and calculating the road geometry condition data according to the trajectory data, it can collect data and construct models on the actual road section, and more truly reflect the influence of road geometry conditions on automatic driving emissions;

[0106] (2) It can consider the influence of different road geometries and control strategies on automatic driving emissions, and effectively improve the emission prediction accuracy of automatic driving in the real road scenario;

[0107] (3) It can take into account interpretability while improving the prediction accuracy, can explain the contribution of road geometric condition-related features to autonomous driving emissions, provides a reference for future improvement of autonomous driving control and optimization of road geometric design, and makes up for the deficiency of existing technical solutions that only target straight sections and are difficult to reduce autonomous driving emissions from the perspective of adapting to road geometric conditions.

[0108] In summary, the present invention designs a method for predicting autonomous driving emissions considering the influence of road geometry and control strategies. The method for predicting autonomous driving emissions considering the influence of road geometry and control strategies of the present invention uses information related to the autonomous vehicle and control strategies of autonomous driving, the operating conditions of autonomous driving, and road geometry information as model inputs, adopts a variety of machine learning models to predict autonomous driving emissions, and determines the optimal model by comprehensively considering the model prediction performance and interpretability, providing an effective technical means for improving the prediction accuracy of autonomous driving emissions; through the design method of the present invention, data collection and model construction can be carried out on any actual section, more truly reflecting the influence of road geometric conditions on autonomous driving emissions; it can consider the influence of different road geometries and control strategies on autonomous driving emissions, effectively improving the prediction accuracy of autonomous driving emissions in real road scenarios; it takes into account interpretability while improving the prediction accuracy, can explain the contribution of road geometric condition-related features to the emissions of autonomous driving vehicles, provides a reference for future improvement of autonomous driving control and optimization of road geometric design, and makes up for the deficiency of existing technical solutions that only target straight sections and are difficult to reduce autonomous driving emissions from the perspective of adapting to road geometric conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0109] The following further details the present invention in conjunction with the drawings and specific embodiments:

[0110] Attached Figure 1 is a schematic flow chart of an interpretable prediction method for autonomous driving emissions considering the influence of road geometric conditions provided by an embodiment of the present invention;

[0111] Attached Figure 2 is a schematic flow chart of an embodiment of the present invention for obtaining emission and trajectory data through on-road vehicle tests and calculating information related to road geometry and autonomous driving operating conditions;

[0112] Attached Figure 3 is a schematic flow chart of an embodiment of the present invention for implementing feature engineering using machine learning methods;

[0113] Attached Figure 4 is a schematic flow chart of an embodiment of the present invention for screening and training a variety of machine learning models with good interpretability;

[0114] Attached Figure 5 is a schematic flow chart of the present invention for determining the optimal interpretable prediction model for autonomous driving emissions based on model prediction performance and interpretability. Detailed implementation manners

[0115] The technical solutions of the present invention will be specifically described below in conjunction with the accompanying drawings.

[0116] To make the features and advantages of this patent application more obvious and understandable, specific embodiments are given below for detailed description as follows:

[0117] It should be noted that the following detailed descriptions are all illustrative and are intended to provide further explanations for this application. Unless otherwise specified, all technical and scientific terms used in this specification have the same meaning as commonly understood by those of ordinary skill in the technical field to which this application belongs.

[0118] It should be noted that the terms used here are only for describing specific implementation manners and are not intended to limit the exemplary implementation manners according to this application. As used here, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should also be understood that when the terms "include" and / or "comprise" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0119] As Figure 1 shown, an interpretable prediction method for autonomous driving emissions considering the influence of road geometric conditions includes the following steps;

[0120] Step S1: Through on-site real vehicle test experiments, record the data and control strategy-related information of autonomous driving vehicles, sort out their autonomous driving trajectories and emission data, and perform preprocessing to construct a data set; that is, record the information related to the autonomous driving vehicle itself (the vehicle itself is the autonomous driving vehicle) and the control strategy through on-site real vehicle tests, obtain the autonomous driving trajectory and emission data, and calculate the information related to the autonomous driving operating conditions and road geometry; the flow chart of this step is as Figure 2 shown;

[0121] Step S2: Use the constructed data set to calculate the information related to the autonomous driving operating conditions and road geometry;

[0122] Step S3: Implement feature engineering using machine learning methods; that is, based on the acquired data, implement feature engineering using machine learning methods; the flow chart of this step is as Figure 3 shown;

[0123] Step S4: Screen machine learning models with good interpretability from the classical machine learning model library for training; the flow chart of this step is as Figure 4 shown;

[0124] Step S5: Evaluate based on the model prediction performance and interpretability to determine the optimal interpretable prediction model for autonomous driving emissions; the flow chart of this step is as Figure 5 shown.

[0125] In step S1, the vehicles selected for the experiment are: commercial autonomous vehicles of different brands and specifications on the existing market;

[0126] The experimental site is selected as: a protected closed test site where the terrain slope and curvature of the road fluctuate widely;

[0127] The experiment is designed as follows: autonomous vehicles of different brands and specifications drive on the test site using different control strategies and set speeds, and a Portable Emission Measurement System (PEMS) is used to collect emission data. At the same time, a Global Positioning System (GPS) and an On-Board Diagnostics Ⅱ (OBDⅡ) are used to record the vehicle trajectory data;

[0128] During the experiment, the relevant information recorded for autonomous vehicles includes: fuel type, total load mass, maximum power, engine displacement, emission standard, fuel injection type, vehicle type classification, number of passenger seats, and maximum speed;

[0129] The relevant information recorded for the autonomous driving control strategy includes: Adaptive Cruise Control (ACC), time headway, ACC set speed (30 - 120 km / h), acceleration control strategy, and equipped assisted driving system;

[0130] The collected emission data includes: instantaneous emission rates of pollutants such as carbon dioxide, carbon monoxide, hydrocarbons, nitric oxide, and nitrogen dioxide;

[0131] The converted trajectory data includes: the abscissa of the vehicle x , the ordinate y , the vertical coordinate z , the speed V and the driving distance S of the vehicle per second;

[0132] The road geometric data includes: road slope G , road curvature C , road slope change rate GCR , road curvature change rate CCR ;

[0133] The operating condition data includes: acceleration a , acceleration change rate, vehicle specific power VSP ;

[0134] In step S1, when sorting out data, the trajectory data is converted into a three-dimensional rectangular coordinate system.

[0135] Step S1 includes the following steps;

[0136] Step S11: Preprocess the obtained trajectory and emission data, including missing value processing, duplicate value processing, outlier detection and processing;

[0137] The method for processing missing values is to fill in the missing values using the cubic spline interpolation method;

[0138] The method for processing duplicate values is to find and delete exactly the same duplicate data;

[0139] The method for outlier detection and processing is to detect and eliminate outliers using the interquartile range (IQR) method;

[0140] Step S12: Deduce the road geometric data, including road slope G , road curvature C , road slope change rate GCR , road curvature change rate CCR ;

[0141] The road slope G The calculation formula is:

[0142] ;

[0143] The road curvature C The calculation formula is:

[0144] ;

[0145] The road slope change rate GCR The calculation formula is:

[0146] ;

[0147] The road curvature change rate CCR The calculation formula is:

[0148] ;

[0149] Where z t is t the vertical coordinate at a certain moment; ΔS t is t- the change in the driving distance from 1 to t at a certain moment; Δx t is t- the change in the horizontal coordinate from 1 to t at a certain moment; Δyt is t- from 1 to t the change in the vertical coordinate at a moment; G t is t the slope at a moment; C t is t the curvature at a moment;

[0150] Step S13, deduce the operating condition data of autonomous driving, including acceleration a , acceleration change rate, vehicle specific power VSP etc.;

[0151] The vehicle specific power VSP The calculation formula is:

[0152] ;

[0153] In the formula, V t is t the speed at a moment; a t is t the acceleration at a moment; G t is t the road slope at a moment.

[0154] Step S2 includes the following steps;

[0155] Step S21, encode the category features related to autonomous driving vehicles and control strategies;

[0156] The features to be encoded include the ACC headway, acceleration control strategy, sensors equipped on the vehicle, assisted driving systems equipped on the vehicle, fuel type, emission standard, vehicle type classification, fuel injection type, which together with the data obtained from the above records and deductions form the original training data set; see Table 1 below for details.

[0157] Table 1:

[0158]

[0159] Step S22, use the geometric path of the road, the trajectory and control strategy related information of the autonomous driving vehicle, and the operating condition of autonomous driving as input features to form model input data, and use the autonomous driving emissions as the output result to construct and train multiple machine learning models with embedded feature importance. The machine learning models include XGBoost, LightGBM, Random Forest, Decision Tree, and Extra Trees;

[0160] Step S23: Use the above machine learning model to output the importance of basic features, and select the machine learning model with the largest proportion of feature importance related to road geometry and autonomous driving control strategy as the base model sensitive to road geometry conditions;

[0161] Step S24: Based on the model selected in S23, use the recursive feature elimination method with the root mean square error as the cross-validation evaluation index to screen the best feature combination.

[0162] The rules for encoding category features related to autonomous vehicles and control strategies in Step S21 are as follows:

[0163] If the feature name is ACC headway, the digital encoding rule is: take 1 for short time intervals; take 2 for medium time intervals; take 3 for long time intervals;

[0164] If the feature name is acceleration control strategy, the digital encoding rule is: take 1 for the constant headway strategy; take 2 for the safety distance model strategy; take 3 for the comfort optimization strategy; take 4 for the energy consumption-based optimization strategy;

[0165] If the feature name is lane keeping assist, the digital encoding rule is: take 1 if not equipped; take 2 if equipped;

[0166] If the feature name is lane centering control, the digital encoding rule is: take 1 if not equipped; take 2 if equipped;

[0167] If the feature name is automatic emergency braking, the digital encoding rule is: take 1 if not equipped; take 2 if equipped;

[0168] If the feature name is traffic congestion assist, the digital encoding rule is: take 1 if not equipped; take 2 if equipped;

[0169] If the feature name is blind spot monitoring and assist, the digital encoding rule is: take 1 if not equipped; take 2 if equipped;

[0170] If the feature name is highway assist driving, the digital encoding rule is: take 1 if not equipped; take 2 if equipped;

[0171] If the feature name is lidar, the digital encoding rule is: take 1 if not equipped; take 2 if equipped;

[0172] If the feature name is millimeter wave radar, the digital encoding rule is: take 1 if not equipped; take 2 if equipped;

[0173] If the feature name is camera, the digital encoding rule is: take 1 if not equipped; take 2 if equipped;

[0174] If the feature name is ultrasonic sensor, the digital encoding rule is: take 1 if not equipped; take 2 if equipped;

[0175] If the feature name is inertial measurement unit, the digital encoding rule is: take 1 if not equipped; take 2 if equipped;

[0176] If the feature name is Global Navigation Satellite System, the digital coding rule is: 1 for unassembled; 2 for assembled;

[0177] If the feature name is fuel type, the digital coding rule is: 1 for diesel; 2 for gasoline; 3 for CNG; 4 for LNG (liquefied natural gas); 5 for LPG (liquefied petroleum gas); 6 for hybrid electric;

[0178] If the feature name is emission standard, the digital coding rule is: 1 for National I; 2 for National II; 3 for National III;

[0179] 4 for National IV; 5 for National V; 6 for National VI;

[0180] If the feature name is vehicle type classification, the digital coding rule is: 1 for light gasoline vehicle; 2 for light diesel vehicle;

[0181] 1 for heavy gasoline vehicle; 2 for heavy diesel vehicle;

[0182] If the feature name is fuel injection type, the digital coding rule is: 1 for carburetor; 2 for electronic fuel injection; 3 for compression ignition; 4 for non-direct injection; 5 for direct injection.

[0183] Step S3 includes the following steps;

[0184] Step S31: Using the feature combinations screened in step S2, classify the data based on vehicle specific power VSP , road slope G and road curvature C ;

[0185] Step S32: Using road geometry, information related to autonomous vehicles and control strategies, and autonomous driving operating conditions as input variables, and autonomous driving emissions as output variables, normalize the samples composed of input variables and output variables, and randomly divide the samples according to a preset ratio to form a model training set and a test set;

[0186] Step S33: Screen multiple machine learning models with good interpretability from the classical machine learning model library, complete parameter optimization through Bayesian optimization respectively, and then substitute them into training and testing;

[0187] The machine learning models include XGBoost, LightGBM, Random Forest, Decision Tree, Extra Trees, Support Vector Machine, Deep Forest, and Gradient Boosting Tree.

[0188] In step S32, the samples are randomly divided according to a ratio of 7:3 as the model training set and the test set respectively.

[0189] Step S4 includes the following steps;

[0190] Step S41: Construct a hierarchical structure model, including an objective layer and an index layer;

[0191] The objective layer is: the interpretability of the model;

[0192] The index layer includes evaluation indexes of the consistency of feature importance distribution, the stability of feature interpretation, the interpretability of feature interaction, the clarity of global interpretation, and the stability of local interpretation;

[0193] The consistency of feature importance distribution is: the degree of conformity between the feature importance distribution output by the model and the expert expectation;

[0194] The stability of feature interpretation is: the stability degree of the SHAP (Shapley Additive Explanations) value distributions of different samples;

[0195] The interpretability of feature interaction is: the clarity degree of the model reflecting the interaction between features;

[0196] The clarity of global interpretation is: the clarity degree of the SHAP global interpretation of the model;

[0197] The accuracy of local interpretation is: the reasonable degree of the SHAP value of a single sample explaining the prediction result;

[0198] Step S42: Construct an importance judgment matrix based on expert scoring and perform normalization processing;

[0199] The importance judgment matrix is: by comparing and scoring each pair of evaluation indexes, an importance judgment matrix is formed;

[0200] The matrix form is:

[0201] ;

[0202] The normalization processing formula of the importance judgment matrix is:

[0203] ;

[0204] In the formula, the matrix element d ij is the importance of the i th evaluation index relative to the j th evaluation index, d norm,ij is the normalization processing result of the importance of the i th evaluation index relative to the j th evaluation index;

[0205] Step S43: Calculate the weights using the eigenvector method and conduct a consistency test. If the consistency test passes, output the weights of each evaluation index. If the consistency test fails, return to Step S42 to adjust the importance judgment matrix;

[0206] The calculation of weights using the eigenvector method is as follows: Solve the maximum eigenvalue of the importance judgment matrix λ max and the corresponding eigenvector, and normalize the eigenvector to obtain the weight vector W , where the i th element W i is the weight of the i th evaluation index;

[0207] The consistency test is as follows: Calculate the consistency index CI and the consistency ratio CR . If CR < 0.1, the importance judgment matrix has acceptable consistency. Otherwise, the importance judgment matrix needs to be adjusted;

[0208] The consistency index CI is calculated as follows:

[0209] ;

[0210] The consistency ratio CR is calculated as follows:

[0211] ;

[0212] In the formula, λ max is the maximum eigenvalue of the importance judgment matrix; W is the weight vector; W i is the i th evaluation index weight; CI is the consistency index; n is the dimension of the judgment matrix; CR is the consistency ratio; RI is the random consistency index depending on the dimension of the importance judgment matrix;

[0213] Step S44: Conduct an interpretability analysis of each machine learning model based on the SHAP method. Invite experts to score respectively around the performance of the model on the above five evaluation indexes based on the interpretability analysis results of each machine learning model, and then calculate the weighted average as the comprehensive score according to the weights of each evaluation index. Sort the interpretability of the models in descending order based on the comprehensive score;

[0214] The interpretability analysis includes: feature importance analysis, feature interaction analysis, swarm plot analysis, force plot analysis, decision path analysis, and waterfall plot analysis;

[0215] Step S45: Sort the models in ascending order based on the prediction performance evaluation metrics of a single machine learning model, and comprehensively sort the prediction performance of the models according to the average of the rankings.

[0216] The model prediction performance evaluation metrics include root mean square error, mean absolute error, and mean absolute percentage error;

[0217] Step S46: Select the model with the smallest average of prediction performance and interpretability rankings as the optimal emission prediction model.

[0218] In step S42, the method for obtaining the importance judgment matrix is as follows: Using the 1-9 scale method, invite experts to compare and score each pair of evaluation indicators to form an importance judgment matrix.

[0219] The autonomous driving in this example is L2-level autonomous driving. For currently in-use low-level autonomous driving vehicles, including electric and fuel vehicle types of autonomous driving vehicles, through the data collected by the on-vehicle exhaust detection device PMES, predict the emission problems of low-level autonomous driving.

[0220] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0221] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0222] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to work in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction device that implements the functions specified in one process or a plurality of processes and / or blocks Figure 1 one process or a plurality of processes and / or blocks Figure 1 specified in one block or a plurality of blocks.

[0223] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, such that a series of operation steps are executed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing steps for implementing the functions specified in one process or a plurality of processes and / or blocks Figure 1 one process or a plurality of processes and / or blocks Figure 1 specified in one block or a plurality of blocks.

[0224] As described above, it is only the preferred embodiment of the present invention, and is not a limitation to the present invention in other forms. Any person skilled in the art may use the technical content disclosed above to make changes or modifications into equivalent embodiments with equivalent changes. However, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the technical solution content of the present invention still fall within the protection scope of the technical solution of the present invention.

[0225] This patent is not limited to the above best mode. Anyone inspired by this patent can obtain other various forms of evaluation methods for energy conservation and emission reduction of pure electric vehicles under complex road alignment conditions. All equal changes and modifications made according to the scope of the patent application of the present invention shall fall within the coverage of this patent.

Claims

1. An explainable prediction method for autonomous driving emissions considering the influence of road geometry conditions, characterized by: The steps include: Step S1: Record the data of the autonomous driving vehicle and the relevant information of the control strategy through on-site real vehicle test experiments, sort out its autonomous driving trajectory and emission data and perform preprocessing to construct a data set; Step S2, using the constructed data set to infer information related to the autonomous driving operating conditions and road geometry; Step S3: Implement feature engineering using machine learning methods; Step S4: Select a machine learning model with good interpretability from a classic machine learning model library for training; Step S5: Evaluate the model prediction performance and interpretability to determine the optimal interpretable prediction model for autonomous driving emissions; Step S3 comprises the following steps: Step S31: Using the feature combination selected in step S2, based on the vehicle specific power VSP , road slope G and road curvature C Categorize the data; Step S32: using road geometry, information related to the autonomous driving vehicle and control strategy, and autonomous driving operating conditions as input variables, and autonomous driving emissions as output variables, normalizing the samples consisting of the input variables and the output variables, and randomly dividing the samples according to a preset ratio to form a model training set and a test set; Step S33: select a variety of machine learning models with good interpretability from the classic machine learning model library, and perform parameter optimization through Bayesian optimization and then perform training and testing; The machine learning models include XGBoost, lightweight gradient boosting machine, random forest, decision tree, extreme random tree, support vector machine, deep forest, ladder boosting tree; Step S4 comprises the following steps: Step S41, constructing a hierarchical structure model, including a target layer and an indicator layer; Step S42: construct an importance judgment matrix based on expert scores and perform normalization processing; Step S43, using the eigenvector method to calculate the weights and perform a consistency check. If the consistency check passes, the weights of each evaluation index are output; if the consistency check fails, the process returns to step S42 to adjust the importance judgment matrix; Step S44: performing interpretability analysis on each machine learning model based on the SHAP method; Step S45: sort the models in ascending order based on the prediction performance evaluation index of the single machine learning model, and comprehensively sort the prediction performance of the models according to the average of the sorting; Step S46: Select the model with the smallest average values ​​of prediction performance and explainability as the optimal emission prediction model.

2. The explainable prediction method for autonomous driving emissions considering the influence of road geometry conditions according to claim 1, characterized in that: In step S1, the vehicles selected for the on-site real vehicle test experiment are: commercial autonomous driving vehicles of different brands and specifications on the existing market; The experimental site for on-site real vehicle testing is selected as: a protected closed test site with a wide range of fluctuations in road terrain slope and curvature; The on-site real-vehicle test experiment is designed as follows: autonomous driving vehicles of different brands and specifications use different control strategies and set speeds to drive on the test site, and use the portable emission measurement system PEMS to collect emission data. At the same time, the global positioning system GPS and the on-board automatic diagnostic system OBDⅡ are used to record vehicle trajectory data. In the experiment, the relevant information recorded about the autonomous vehicle included: fuel type, gross load, maximum power, engine displacement, emission standard, injection type, vehicle type classification, passenger seats and maximum speed; The recorded information related to the control strategy of the autonomous driving vehicle includes: adaptive cruise control ACC, headway, ACC set speed, acceleration control strategy and equipped auxiliary driving system; The collected emission data include: instantaneous emission rates of pollutants such as carbon dioxide, carbon monoxide, hydrocarbons, nitric oxide, and nitrogen dioxide; The trajectory data includes: the horizontal coordinate of the vehicle x , vertical coordinate y , vertical coordinate z ,speed V and driving distance S Second-by-second data; The road geometry data of the on-site real vehicle test experiment includes: road slope G , road curvature C , Road slope change rate GCR , Road curvature change rate CCR ; In the on-site real vehicle test experiment, the vehicle's operating condition data includes: acceleration a , acceleration rate, vehicle specific power VSP .

3. The explainable prediction method for autonomous driving emissions considering the influence of road geometry conditions according to claim 2, characterized in that: In step S1, during data combing, the trajectory data is converted into a three-dimensional rectangular coordinate system.

4. The explainable prediction method for autonomous driving emissions considering the influence of road geometry conditions according to claim 2, characterized in that: Step S1 includes the following steps: Step S11, preprocessing the acquired trajectory and emission data, including missing value processing, duplicate value processing, and outlier detection and processing; The missing value processing method is to use the third-order spline interpolation method to fill the missing values; The duplicate value processing method is to find and delete identical duplicate data; The outlier detection and processing method is to use the quartile method IQR to detect outliers and remove them; Step S12: Calculate road geometry data, including road slope G , road curvature C , Road slope change rate GCR , Road curvature change rate CCR ; The road slope G The calculation formula is: ； The road curvature C The calculation formula is: ； The road slope change rate GCR The calculation formula is: ； The road curvature change rate CCR The calculation formula is: ； In the formula, z t for t The vertical coordinate of the moment; ΔS t for t- 1 to t Change in driving distance at any moment; Δx t for t- 1 to t The change of the horizontal axis at a certain moment; Δy t for t- 1 to t The change of vertical coordinate at a certain moment; G t for t The slope of the moment; C t for t The curvature of the moment; Step S13: Calculate the automatic driving operation data, including acceleration a , acceleration rate, vehicle specific power VSP ; The vehicle specific power VSP The calculation formula is: ； In the formula, V t for t Time speed; a t for t Time acceleration; G t for t The slope of the road at any time.

5. The explainable prediction method for autonomous driving emissions considering the influence of road geometry conditions according to claim 2, characterized in that: Step S2 comprises the following steps: Step S21, encoding the category features related to the autonomous driving vehicle and the control strategy; The features to be encoded include ACC headway, acceleration control strategy, vehicle sensors, vehicle auxiliary driving systems, fuel type, emission standards, vehicle model classification, and injection type, which together with the above recorded and inferred data constitute the original training data set; Step S22: using the geometric path of the road, the trajectory of the autonomous driving vehicle and related information of the control strategy, and the autonomous driving operating conditions as input features to form model input data, using the autonomous driving emissions as output results, and constructing and training multiple machine learning models with embedded feature importance, the machine learning models including XGBoost, lightweight gradient boosting machine, random forest, decision tree, and extreme random tree; Step S23: Use the above machine learning model to output the basic feature importance, and select the machine learning model with the largest proportion of road geometry and autonomous driving control strategy related feature importance as the base model sensitive to the influence of road geometry conditions; Step S24: Based on the model selected in S23, a recursive feature elimination method is used to select the best feature combination with the root mean square error as the cross-validation evaluation index.

6. The explainable prediction method for autonomous driving emissions considering the influence of road geometry conditions according to claim 5, characterized in that: Step S21 encodes the rules for the category features related to the autonomous driving vehicle and the control strategy as follows: If the feature name is ACC headway, the digital coding rule is 1 for short time interval, 2 for medium time interval, and 3 for long time interval. If the feature name is acceleration control strategy, the digital coding rule is 1 for constant headway strategy; 2 for safety distance model strategy; 3 for comfort optimization strategy; and 4 for energy consumption optimization strategy. If the feature name is Lane Keeping Assist, the digital coding rule is 1 if not installed and 2 if installed. If the feature name is Lane Centering Control, the digital coding rule is 1 if not equipped and 2 if equipped. If the feature name is automatic emergency braking, the digital coding rule is 1 if not installed and 2 if installed. If the feature name is Traffic Congestion Assist, the digital coding rule is 1 if not equipped and 2 if equipped. If the feature name is blind spot monitoring and assistance, the digital coding rule is 1 if not equipped and 2 if equipped. If the feature name is high-speed assisted driving, the digital coding rule is 1 if not equipped and 2 if equipped. If the feature name is LiDAR, the digital coding rule is 1 if not equipped and 2 if equipped. If the feature name is millimeter wave radar, the digital coding rule is 1 if not equipped and 2 if equipped. If the feature name is camera, the digital coding rule is 1 if not installed and 2 if installed. If the feature name is ultrasonic sensor, the digital coding rule is 1 for unassembled and 2 for assembled. If the feature name is inertial measurement unit, the digital coding rule is 1 if not assembled and 2 if assembled. If the feature name is Global Navigation Satellite System, the digital coding rule is 1 if not equipped and 2 if equipped. If the feature name is fuel type, the digital coding rule is: diesel takes 1; gasoline takes 2; CNG takes 3; LNG takes 4; LPG takes 5; hybrid takes 6; If the characteristic name is emission standard, the digital coding rule is 1 for National I, 2 for National II, and 3 for National III. Country IV takes 4; Country V takes 5; Country VI takes 6; If the feature name is vehicle type classification, the digital coding rule is 1 for light gasoline vehicles and 2 for light diesel vehicles; For heavy-duty gasoline vehicles, take 1; for heavy-duty diesel vehicles, take 2; If the feature name is injection type, the digital coding rule is 1 for carburetor, 2 for electronic injection, 3 for compression ignition, 4 for indirect injection, and 5 for direct injection.

7. The method for interpretable prediction of autonomous driving emissions considering the influence of road geometry conditions according to claim 1, characterized in that: In step S32, the samples are randomly divided into a ratio of 7:3 as a model training set and a test set respectively.

8. The method for interpretable prediction of autonomous driving emissions considering the influence of road geometry conditions according to claim 1, characterized in that: Step S4 comprises the following steps: Step S41, constructing a hierarchical structure model, including a target layer and an indicator layer; The target layer is: model interpretability; The indicator layer includes evaluation indicators of feature importance distribution consistency, feature interpretation stability, feature interaction interpretability, global interpretation clarity, and local interpretation stability; The feature importance distribution consistency is: the degree of conformity between the feature importance distribution output by the model and the expert's expectations; The feature interpretation stability is: the stability of the distribution of SHAP values ​​of different samples; The feature interaction interpretability is: the clarity of the model in reflecting the interaction between features; The global explanation clarity is: the clarity of the global explanation of the model SHAP; The local explanation accuracy is: the reasonableness of the SHAP value of a single sample to explain the prediction result; Step S42: construct an importance judgment matrix based on expert scores and perform normalization processing; The importance judgment matrix is ​​formed by comparing and scoring each evaluation index in pairs; The matrix form is: ； The normalization processing formula of the importance judgment matrix is: ； In the formula, the matrix elements d ij For the i The evaluation index is relative to j The importance of the evaluation indicators d norm,ij For the i The evaluation index is relative to j The importance normalization results of the evaluation indicators; Step S43, using the eigenvector method to calculate the weights and perform a consistency check. If the consistency check passes, the weights of each evaluation index are output; if the consistency check fails, the process returns to step S42 to adjust the importance judgment matrix; Step S44: perform interpretability analysis on each machine learning model based on the SHAP method, invite experts to score the performance of each machine learning model on the above five evaluation indicators based on the interpretability analysis results of each machine learning model, and then calculate the weighted average value as the comprehensive score according to the weight of each evaluation indicator, and sort the interpretability of the model in descending order based on the comprehensive score; Step S45: sort the models in ascending order based on the prediction performance evaluation index of the single machine learning model, and comprehensively sort the prediction performance of the models according to the average of the sorting; Step S46: Select the model with the smallest average values ​​of prediction performance and explainability as the optimal emission prediction model.

9. The explainable prediction method for autonomous driving emissions considering the influence of road geometry conditions according to claim 8, characterized in that: In step S42, the importance judgment matrix is ​​obtained by using a 1-9 scale method and inviting experts to compare and score each evaluation index in pairs.

Images