A method and device for evaluating driving behavior of an autonomous driving vehicle

By calculating the driving behavior characteristic parameters of the autonomous driving vehicle and calculating the distance from the clustering center point of the human driver, the driving behavior score is obtained, which solves the problem that quantitative evaluation cannot be performed in the existing technology, and realizes the driving of autonomous driving vehicles Accurate and objective evaluation of behavior.

CN114692713BActive Publication Date: 2024-11-22ZHENGZHOU YUTONG BUS CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202011592294.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-12-29
Publication Date
2024-11-22
Estimated Expiration
2040-12-29

AI Technical Summary

Technical Problem

The existing methods for evaluating driving behavior of autonomous vehicles cannot conduct objective and accurate quantitative evaluations, cannot absorb the correct driving methods of human drivers and abandon bad driving habits.

Method used

By obtaining the driving behavior information of the autonomous driving vehicle, calculating the driving behavior characteristic parameters, and using it as a sample point to calculate the distance from the clustering center point of the human driver, the driving behavior score is obtained, and quantitative evaluation is performed based on the reference distance and score.

Benefits of technology

The objective and accurate quantitative evaluation of the driving behavior of autonomous vehicles is achieved, and the impact of the most standardized driving behavior categories and other categories is comprehensively considered, which improves the accuracy of the evaluation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114692713B_ABST
    Figure CN114692713B_ABST
Patent Text Reader

Abstract

The present invention provides a method and device for evaluating the driving behavior of an autonomous driving vehicle, and belongs to the technical field of autonomous driving vehicle evaluation. The method includes: obtaining the driving behavior information of the autonomous driving vehicle on the route to be evaluated, calculating the driving behavior characteristic parameters of the autonomous driving vehicle according to the driving behavior information of the autonomous driving vehicle, taking the driving behavior characteristic parameters of the autonomous driving vehicle as sample points, and calculating the distance from the sample point to the reference point; calculating the driving behavior score of the autonomous driving vehicle according to the distance from the sample point to the reference point, the reference distance and the score corresponding to the reference distance; wherein, the smaller the distance from the sample point to the reference point, the higher the driving behavior score of the autonomous driving vehicle; the reference point is the cluster center point of the reference class, and the reference distance is a certain value in the distance from the cluster center point of other driving behavior categories to the reference point. The present invention can realize the quantitative evaluation of the driving behavior of the autonomous driving vehicle.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to a method and device for evaluating driving behavior of an autonomous driving vehicle, and belongs to the technical field of autonomous driving vehicle evaluation. Background Art

[0002] Driving behavior is an important factor that must be considered for both human-driven vehicles and self-driving vehicles. Most traffic accidents are caused by improper driving operations, and bad driving habits are the main cause of traffic accidents. Self-driving vehicles help improve the intelligence of the "human-vehicle-environment" system, thereby improving traffic efficiency and traffic safety. With the development of self-driving technology, self-driving vehicles are gradually being promoted in closed scenes, freeing human drivers from driving operations. However, there is still no reliable and effective evaluation method for how to objectively and accurately evaluate the driving behavior of self-driving vehicles, so that self-driving vehicles can absorb the correct driving methods of human drivers, abandon the bad driving habits of human drivers, and make the driving behavior of self-driving vehicles as safe and effective as excellent drivers.

[0003] Although a driving behavior evaluation method based on FCM clustering and BP neural network is disclosed in the Chinese invention patent application document with application publication number CN107153916A, the method clusters the driving behavior characteristic parameters extracted from the driving behavior information, uses the clustering results as training samples to train a BP neural network model, and uses the trained BP neural network model to evaluate the driving behavior to be evaluated, but the method can only make a qualitative evaluation of the driving behavior, and cannot make a quantitative evaluation. Summary of the invention

[0004] The purpose of the present invention is to provide a driving behavior evaluation method and device for an autonomous driving vehicle, so as to make an objective and accurate quantitative evaluation of the driving behavior of the autonomous driving vehicle.

[0005] To achieve the above object, the present invention provides a driving behavior evaluation method for an autonomous driving vehicle, the method comprising the following steps:

[0006] (1) Obtaining driving behavior information of the autonomous driving vehicle on the route to be evaluated, the driving behavior information including: instantaneous vehicle speed, acceleration / deceleration, steering wheel angle, total vehicle operation time, total vehicle operation mileage, vehicle cumulative energy consumption, vehicle surrounding obstacle information, and vehicle surrounding road data information; the vehicle surrounding road data information includes ramps, curves, tunnels, accident-prone areas, road section speed limits, and traffic light information;

[0007] (2) calculating driving behavior characteristic parameters of the autonomous driving vehicle based on the driving behavior information of the autonomous driving vehicle, wherein the driving behavior characteristic parameters include: the proportion of speeding time, the average number of sudden acceleration / deceleration times, the average number of lane changes, the average energy consumption, the number of times the vehicle failed to avoid obstacles, and the number of times the vehicle correctly responded to road signs and markings;

[0008] Among them, the speeding time ratio is: the ratio of the cumulative value of the duration of the vehicle speeding on the route to be evaluated to the total running time of the vehicle. When the instantaneous vehicle speed is greater than the set speed upper limit or greater than the road section speed limit, the vehicle is determined to be speeding; the average number of sudden acceleration / deceleration times is: the ratio of the cumulative value of the number of sudden acceleration / deceleration times of the vehicle on the route to be evaluated to the total running mileage of the vehicle. When the vehicle acceleration / deceleration is greater than the set acceleration / deceleration upper limit, the vehicle is determined to have sudden acceleration / deceleration; the average number of lane changes is: the ratio of the cumulative value of the number of lane changes on the route to be evaluated to the total running mileage of the vehicle. When the steering wheel angle is greater than the set angle upper limit, the vehicle is determined to have changed lanes; the average energy consumption is the ratio of the cumulative energy consumption of the vehicle on the route to be evaluated to the total running mileage of the vehicle; combined with the obstacle information around the vehicle, it is judged whether the vehicle successfully avoids the obstacle when encountering it, and then the number of unsuccessful obstacle avoidance times of the vehicle on the route to be evaluated is obtained; combined with the road data information around the vehicle, it is judged whether the vehicle correctly responds to the road marking line, and then the number of correct responses to the road marking line by the vehicle on the route to be evaluated is obtained;

[0009] (3) Taking the driving behavior characteristic parameters of the autonomous driving vehicle as sample points, the distance from the sample points to the reference points is calculated;

[0010] (4) Calculating the driving behavior score of the autonomous driving vehicle based on the distance from the sample point to the reference point, the benchmark distance, and the score corresponding to the benchmark distance; wherein the smaller the distance from the sample point to the reference point, the higher the driving behavior score of the autonomous driving vehicle;

[0011] The reference point and the reference distance are obtained by the following steps:

[0012] Acquire driving behavior information of human drivers of different time periods, different drivers, and different vehicles on the route to be evaluated as a sample data set, and correspondingly obtain a driving behavior characteristic parameter data set of human drivers; perform cluster analysis on the driving behavior characteristic parameters of human drivers in the driving behavior characteristic parameter data set to obtain multiple different driving behavior categories and cluster center points of each driving behavior category; select from the multiple different driving behavior categories a category with the smallest proportion of speeding time, the smallest average number of sudden acceleration / deceleration times, the smallest average number of lane changes, the smallest average energy consumption, the smallest number of unsuccessful obstacle avoidances, and the largest number of correct responses to road signs and markings by vehicles as a reference category, the reference point is the cluster center point of the reference category, and the reference distance is a value of the distance from the cluster center point of other driving behavior categories to the reference point.

[0013] The present invention also provides a driving behavior evaluation device for an autonomous driving vehicle, which includes a processor and a memory, and the processor executes a computer program stored in the memory to implement the above-mentioned driving behavior evaluation method for the autonomous driving vehicle.

[0014] The beneficial effect of the present invention is that the present invention uses the driving behavior characteristic parameters of the autonomous driving vehicle as sample points, and calculates the driving behavior score of the autonomous driving vehicle according to the distance from the sample point to the reference point, the reference distance and the score corresponding to the reference distance, so as to realize the quantitative evaluation of the driving behavior of the autonomous driving vehicle. Among them, the reference point and the reference distance are obtained by clustering analysis of the driving behavior characteristic parameters of human drivers, the reference point is the cluster center point of the most standardized driving behavior category among all driving behavior categories obtained by cluster analysis (i.e., the driving behavior category with the smallest proportion of speeding time, the smallest average number of rapid acceleration / deceleration, the smallest average number of lane changes, the smallest average energy consumption, the smallest number of unsuccessful obstacle avoidances, and the largest number of correct responses to road signs and markings by the vehicle), and the reference distance is a value in the distance from the cluster center point of other driving behavior categories to the reference point. Therefore, when the present invention quantitatively evaluates the driving behavior of the autonomous driving vehicle, it comprehensively considers the influence of the most standardized driving behavior category and other driving behavior categories on the driving behavior evaluation, which is more objective and the evaluation result is more accurate.

[0015] Furthermore, in the above method and device, the calculation formula for the driving behavior score of the autonomous driving vehicle is:

[0016]

[0017] In the formula, S represents the driving behavior score of the autonomous vehicle, M represents the score corresponding to the benchmark distance, D represents the benchmark distance, and d represents the distance from the sample point to the reference point.

[0018] Furthermore, in the above method and device, the step of performing cluster analysis includes: using principal component analysis to reduce the dimension of the driving behavior characteristic parameters of the human driver; and performing cluster analysis on the driving behavior characteristic parameters of the human driver after dimension reduction in the driving behavior characteristic parameter data set.

[0019] Furthermore, in the above method and device, the method also includes a step of qualitatively evaluating the driving behavior of the autonomous driving vehicle, which step includes: inputting the driving behavior characteristic parameters of the autonomous driving vehicle into a trained driving behavior classification model to obtain a driving behavior category corresponding to the driving behavior of the autonomous driving vehicle; the trained driving behavior classification model is obtained by the following steps: using the driving behavior characteristic parameters of each human driver in the driving behavior characteristic parameter data set and their corresponding driving behavior categories as training samples, and using the training samples to train the classification model to obtain a trained driving behavior classification model.

[0020] Furthermore, in the above method and device, the classification model is a collaborative filtering algorithm classification model based on user behavior, a k-nearest neighbor algorithm classification model, a decision tree classification model or a SVM classification model. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a flow chart of a driving behavior evaluation method for an autonomous driving vehicle in an embodiment of the method of the present invention;

[0022] Figure 2 It is a schematic diagram of the structure of a driving behavior evaluation device for an autonomous driving vehicle in an embodiment of the device of the present invention. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0024] Method Example:

[0025] like Figure 1 As shown, the driving behavior evaluation method of the automatic driving vehicle of this embodiment (hereinafter referred to as the driving behavior evaluation method) includes the following steps:

[0026] (1) Obtain driving behavior information of the autonomous driving vehicle on the route to be evaluated;

[0027] Among them, driving behavior information includes: instantaneous vehicle speed, acceleration / deceleration, steering wheel angle, total vehicle operation time, total vehicle operation mileage, vehicle cumulative energy consumption, vehicle surrounding obstacle information, vehicle surrounding road data information; vehicle surrounding road data information includes ramps, curves, tunnels, accident-prone areas, road speed limits, and traffic light information.

[0028] The driving behavior information is obtained as follows:

[0029] A monitoring terminal is installed on the autonomous driving vehicle, which is used to collect the vehicle's instantaneous speed, acceleration, deceleration, steering wheel angle, total vehicle operating time, total vehicle operating mileage, vehicle energy consumption (including instantaneous energy consumption, cumulative energy consumption, average energy consumption), information on obstacles around the vehicle, whether the vehicle successfully avoids obstacles when encountering them, and road data information around the vehicle (including information on ramps, curves, tunnels, accident-prone areas, road speed limits, the location of traffic lights, etc.). The vehicle's position information (including longitude, latitude, elevation, direction, GPS speed) can also be obtained through GPS positioning technology. As other implementation methods, enhanced positioning can also be achieved through vehicle speed signals, inertial navigation, differential positioning and other technologies.

[0030] The steering wheel angle can be output by the steering wheel angle sensor. The steering wheel angle is a key indicator to measure whether the driver frequently turns, changes lanes, makes sudden turns, and drives smoothly. It is essential for evaluating the driver's driving style. In actual applications, the left / right turn sign can also be used to determine whether the vehicle changes lanes, but it is more accurate to use the steering wheel to determine whether the vehicle changes lanes.

[0031] Whether the vehicle successfully avoids obstacles is reflected in whether the vehicle collides, which can measure the safety of the autonomous vehicle. Specifically, it can be judged based on the data output by the acceleration sensor or the collision sensor whether the vehicle collides, or based on the obstacle information around the vehicle, it can be judged whether the distance between the vehicle and other surrounding traffic participants or obstacles is less than a preset threshold to determine whether the vehicle collides.

[0032] The vehicle mainly collects information about key components in the vehicle and real-time information detected by sensors installed on the vehicle (longitude, latitude, elevation, direction, GPS speed, vehicle speed obtained by the speed sensor, information and status of other traffic participants in the surrounding area, surrounding road data information) through the CAN bus and Ethernet.

[0033] Vehicle surrounding road data information: Based on map information or sensors, the forward signal light data information and real-time status information are obtained, including local area intersection information, road section information (curves, slopes, tunnels, etc.), the connection between roads, traffic signs and markings (such as accident-prone areas, road speed limit signs, stop and yield, lane lines, crosswalks), etc. Vehicle surrounding road data information is mainly used to determine the responsiveness of autonomous driving vehicles to different roads and road signs, markings, etc. Among them, the real-time status of the forward signal light: contains the current status information of one or more intersection signal lights. Combined with map information and perception information, provide the vehicle with real-time forward signal light phase information.

[0034] Information and status of other surrounding traffic participants: Based on its own corresponding perception system, the vehicle obtains real-time status information of surrounding traffic participants (including surrounding vehicles, non-motor vehicles, pedestrians, obstacles, etc.), providing support for human driver operation or autonomous driving vehicle decision-making.

[0035] The vehicle's position information can be used as auxiliary information and combined with surrounding road data information to calculate the number or probability of a certain bad driving behavior occurring in a certain scene or location in subsequent calculations.

[0036] (2) calculating driving behavior characteristic parameters of the autonomous driving vehicle based on the driving behavior information of the autonomous driving vehicle;

[0037] Among them, the driving behavior characteristic parameters include: the proportion of speeding time, the average number of sudden acceleration / deceleration, the average number of lane changes, the average energy consumption, the number of times the vehicle failed to avoid obstacles, and the number of times the vehicle correctly responded to road signs and markings.

[0038] The specific calculation method of driving behavior characteristic parameters is:

[0039] When the instantaneous vehicle speed is greater than the set upper speed limit or the speed limit of the road section, the vehicle is judged to be speeding, and the ratio of the cumulative duration of the vehicle speeding on the evaluated route to the total vehicle operation time is taken as the speeding time ratio, in percentage;

[0040] When the vehicle acceleration / deceleration is greater than the set acceleration / deceleration upper limit, it is determined that the vehicle has experienced sudden acceleration / deceleration. The ratio of the cumulative number of sudden acceleration / deceleration of the vehicle on the route to be evaluated to the total mileage of the vehicle is taken as the average number of sudden acceleration / deceleration times, in units of times / 100km.

[0041] When the steering wheel angle is greater than the set upper limit of the steering angle, the vehicle is judged to have changed lanes. The ratio of the cumulative number of lane changes on the route to be evaluated to the total mileage of the vehicle is taken as the average number of lane changes, in units of times / 100km.

[0042] The ratio of the cumulative energy consumption of vehicles on the evaluated route to the total mileage of vehicles is taken as the average energy consumption, in units of L / 100km;

[0043] Combined with the obstacle information around the vehicle, it is determined whether the vehicle successfully avoids the obstacle when encountering it, and then the number of times the vehicle on the route to be evaluated fails to avoid the obstacle is obtained; when the distance between the vehicle and the obstacle is less than the preset threshold, it is determined that the vehicle fails to avoid the obstacle;

[0044] Combined with the road data information around the vehicle, it is determined whether the vehicle correctly responds to the road markings, and then the number of times the vehicle on the evaluated route correctly responds to the road markings is obtained; for example, the vehicle stops before the stop line, and the vehicle travels below the speed limit in the speed limit area, which are all considered correct responses to the signs and markings.

[0045] (3) Taking the driving behavior characteristic parameters of the autonomous driving vehicle as sample points, calculating the distance from the sample point to the reference point; calculating the driving behavior score of the autonomous driving vehicle based on the distance from the sample point to the reference point, the benchmark distance, and the score corresponding to the benchmark distance (the smaller the distance from the sample point to the reference point, the higher the driving behavior score of the autonomous driving vehicle), thereby achieving a quantitative evaluation of the driving behavior of the autonomous driving vehicle;

[0046] Among them, the reference point and the reference distance are obtained by the following steps:

[0047] The driving behavior information of human drivers of different time periods, different drivers, and different vehicles on the route to be evaluated is obtained as a sample data set, and a driving behavior characteristic parameter data set of human drivers is obtained accordingly; cluster analysis is performed on the driving behavior characteristic parameters of human drivers in the driving behavior characteristic parameter data set to obtain multiple different driving behavior categories and cluster center points of each driving behavior category; from the multiple different driving behavior categories obtained, a category with the smallest proportion of speeding time, the smallest average number of sudden acceleration / deceleration, the smallest average number of lane changes, the smallest average energy consumption, the smallest number of unsuccessful obstacle avoidances, and the largest number of correct responses to road signs and markings by vehicles is selected as a reference category; the reference point is the cluster center point of the reference class, and the benchmark distance is the maximum value of the distances from the cluster center points of other driving behavior categories to the reference point.

[0048] Research has found that excellent driving behavior can directly affect the reduction of energy consumption. In most cases, we can only say that most drivers can achieve relatively good results, but cannot achieve the best of all indicators. This is normal. In this case, we can first find the optimal union that can be achieved among different drivers, and then exclude some conflicting data sets to obtain the optimal values ​​of each parameter as a reference class (if we use an indicator as the time required to pass a certain distance, and the driver with the shortest time in the data set speeds during the journey, then we exclude this optimal value and choose the data set with no illegal driving behavior and the shortest time as the reference class). This reference class can be virtual, serving as the goal of excellent human drivers and autonomous driving vehicles.

[0049] In this embodiment, the calculation formula for the driving behavior score of the autonomous driving vehicle is:

[0050]

[0051] In the formula, S represents the driving behavior score of the autonomous vehicle, M represents the score corresponding to the benchmark distance, D represents the benchmark distance, and d represents the distance from the sample point to the reference point.

[0052] Since the reference distance D in this embodiment is the maximum value of the distances from the cluster center points of other driving behavior categories to the reference point (i.e., the reference distance D is the distance from the cluster center point farthest from the reference point to the reference point), and since the farther the distance from the reference point is, the smaller the driving behavior score is, the value of M needs to be set to a lower score value (which can be set according to actual needs), for example, M=60; as another embodiment, the reference distance D can also be the minimum value of the distances from the cluster center points of other driving behavior categories to the reference point (i.e., the reference distance D is the distance from the cluster center point closest to the reference point to the reference point), and the value of M needs to be set to a higher score value (which can be set according to actual needs), for example, M=90; of course, the reference distance D can also be any value of the distances from the cluster center points of other driving behavior categories to the reference point (i.e., the reference distance D is the distance from the cluster center point of any driving behavior category other than the reference category to the reference point), and at this time, the value of M only needs to be determined according to the distance of the cluster center point from the reference point and the principle that the farther the distance from the reference point is, the smaller the driving behavior score is.

[0053] Among them, by setting the number of clusters during cluster analysis, the number of driving behavior categories can be determined according to actual needs. For example, when the number of clusters is set to 3, the corresponding driving behavior categories are 3, and the cluster analysis algorithm can select an existing cluster analysis algorithm, such as K-means clustering or FCM clustering; by analyzing the driving behavior characteristic parameters under each driving behavior category, the driving style corresponding to each driving behavior category can be obtained, such as mild, moderate, and aggressive.

[0054] (4) Use the trained driving behavior classification model to qualitatively evaluate the driving behavior of the autonomous vehicle.

[0055] It is easy to know that by performing cluster analysis on the driving behavior characteristic parameters of human drivers in the driving behavior characteristic parameter data set, multiple different driving behavior categories are obtained. That is to say, through cluster analysis, the driving behavior characteristic parameters of each human driver have their own corresponding driving behavior categories. Therefore, the driving behavior characteristic parameters of each human driver in the driving behavior characteristic parameter data set of human drivers and their corresponding driving behavior categories are used as training samples. The training samples are used to train the classification model (collaborative filtering algorithm classification model based on user behavior, k-nearest neighbor algorithm classification model, decision tree classification model or SVM classification model), and a trained driving behavior classification model can be obtained. Therefore, the driving behavior characteristic parameters of the autonomous driving vehicle are input into the trained driving behavior classification model, and the driving behavior category corresponding to the driving behavior characteristic parameters of the autonomous driving vehicle (that is, the driving behavior category corresponding to the driving behavior of the autonomous driving vehicle) can be obtained, and the driving style corresponding to the driving behavior of the autonomous driving vehicle can be obtained, thereby achieving a qualitative evaluation of the driving behavior of the autonomous driving vehicle.

[0056] As another implementation method, when performing cluster analysis, the driving behavior characteristic parameters of the human driver can be first reduced in dimension using the principal component analysis method, and only several variables in the driving behavior characteristic parameters of the human driver that have a greater impact on the driving behavior type are retained to obtain the driving behavior characteristic parameters of the human driver after dimensionality reduction, and then cluster analysis is performed on the driving behavior characteristic parameters of the human driver after dimensionality reduction in the driving behavior characteristic parameter data set; for example, in this embodiment, the driving behavior characteristic parameters are 5-dimensional, and if the driving behavior characteristic parameters after dimensionality reduction obtained by the principal component analysis method are 3-dimensional, then all the driving behavior characteristic parameters in the driving behavior characteristic parameter data set are replaced with 3-dimensional driving behavior characteristic parameters, and cluster analysis is performed using the 3-dimensional driving behavior characteristic parameters. Accordingly, when calculating the driving behavior score of the autonomous driving vehicle, the 3-dimensional driving behavior characteristic parameters of the autonomous driving vehicle are also used for calculation.

[0057] In addition, after the cluster analysis is completed, the outliers of each driving behavior category can be deleted (outliers refer to points that are far away from the cluster center of each driving behavior category. The number of outliers is small. Deleting outliers can avoid interfering with the training of subsequent driving behavior classification models), and the driving behavior characteristic parameter data set of human drivers is updated to obtain an updated driving behavior characteristic parameter data set of human drivers (referred to as updated characteristic parameter data set), so that the driving behavior characteristic parameters of each human driver in the updated characteristic parameter data set and their corresponding driving behavior categories are used as training samples to train a driving behavior classification model.

[0058] The driving behavior evaluation method in this embodiment can realize both quantitative evaluation of the driving behavior of the autonomous driving vehicle and qualitative evaluation of the driving behavior of the autonomous driving vehicle. As another implementation method, when only a quantitative evaluation of the driving behavior of the autonomous driving vehicle is required, step (4) can be omitted.

[0059] To summarize, the driving behavior evaluation method of this embodiment is mainly based on the real-time data of human drivers driving vehicles on the evaluated route collected by the Internet of Vehicles platform, to construct a sample data set of human drivers' driving behavior, and to derive driving behavior categories of different driving styles by clustering analysis of the human drivers' driving behavior characteristic parameters (such as the average number of lane changes within a fixed distance or fixed time interval, the proportion of speeding time, the average number of sudden acceleration / deceleration, the number of unsuccessful obstacle avoidance, the average energy consumption, the number of times the vehicle correctly responds to road signs and markings, etc.); then, combined with the real-time data of the autonomous driving vehicle, the intelligent driving behavior of the autonomous driving vehicle is evaluated.

[0060] Device Example:

[0061] The driving behavior evaluation device of the autonomous driving vehicle of this embodiment is as follows: Figure 2 As shown, the device includes a processor and a memory, wherein a computer program that can be run on the processor is stored in the memory, and the processor implements the method in the above method embodiment when executing the computer program.

[0062] That is, the method in the above method embodiment should be understood as a process of the driving behavior evaluation method of the autonomous driving vehicle that can be implemented by computer program instructions. These computer program instructions can be provided to the processor so that the processor executes these instructions to generate functions specified in the above method flow.

[0063] The processor referred to in this embodiment refers to a processing device such as a microprocessor MCU or a programmable logic device FPGA.

[0064] The memory referred to in this embodiment includes a physical device for storing information, which is usually to digitize the information and then store it in a medium using electrical, magnetic or optical methods. For example: various memories that use electrical energy to store information, such as RAM, ROM, etc.; various memories that use magnetic energy to store information, such as hard disks, floppy disks, tapes, magnetic core memories, bubble memories, and USB flash drives; various memories that use optical methods to store information, such as CDs or DVDs. Of course, there are other types of memories, such as quantum memories, graphene memories, and so on.

[0065] The device is composed of the above-mentioned memory, processor and computer program, and is implemented by the processor executing corresponding program instructions in the computer. The processor can be equipped with various operating systems, such as Windows operating system, Linux system, Android, iOS, ROS system, etc.

Claims

1. A method for evaluating driving behavior of an autonomous driving vehicle, characterized in that: The method comprises the following steps: (1) Obtaining driving behavior information of the autonomous driving vehicle on the route to be evaluated, the driving behavior information including: instantaneous vehicle speed, acceleration / deceleration, steering wheel angle, total vehicle operation time, total vehicle operation mileage, vehicle cumulative energy consumption, vehicle surrounding obstacle information, and vehicle surrounding road data information; the vehicle surrounding road data information includes ramps, curves, tunnels, accident-prone areas, road section speed limits, and traffic light information; (2) calculating driving behavior characteristic parameters of the autonomous driving vehicle based on the driving behavior information of the autonomous driving vehicle, wherein the driving behavior characteristic parameters include: the proportion of speeding time, the average number of sudden acceleration / deceleration times, the average number of lane changes, the average energy consumption, the number of times the vehicle failed to avoid obstacles, and the number of times the vehicle correctly responded to road signs and markings; Among them, the speeding time ratio is: the ratio of the cumulative value of the duration of the vehicle speeding on the route to be evaluated to the total vehicle operation time. When the instantaneous vehicle speed is greater than the set speed upper limit or greater than the road section speed limit, the vehicle is determined to be speeding; the average number of sudden acceleration / deceleration times is: the ratio of the cumulative value of the number of sudden acceleration / deceleration times of the vehicle on the route to be evaluated to the total mileage of the vehicle operation. When the vehicle acceleration / deceleration is greater than the set acceleration / deceleration upper limit, the vehicle is determined to have sudden acceleration / deceleration; the average number of lane changes is: the ratio of the cumulative value of the number of lane changes on the route to be evaluated to the total mileage of the vehicle operation. When the steering wheel angle is greater than the set angle upper limit, the vehicle is determined to have changed lanes; the average energy consumption is the ratio of the cumulative energy consumption of the vehicle on the route to be evaluated to the total mileage of the vehicle operation; combined with the obstacle information around the vehicle, it is judged whether the vehicle successfully avoids the obstacle when encountering it, and then the number of unsuccessful obstacle avoidance times of the vehicle on the route to be evaluated is obtained; combined with the road data information around the vehicle, it is judged whether the vehicle correctly responds to the road marking line, and then the number of correct responses to the road marking line by the vehicle on the route to be evaluated is obtained; (3) Taking the driving behavior characteristic parameters of the autonomous driving vehicle as sample points, the distance from the sample points to the reference points is calculated; (4) Calculating the driving behavior score of the autonomous driving vehicle based on the distance from the sample point to the reference point, the benchmark distance, and the score corresponding to the benchmark distance; wherein the smaller the distance from the sample point to the reference point, the higher the driving behavior score of the autonomous driving vehicle; The reference point and the reference distance are obtained by the following steps: Acquire driving behavior information of human drivers of different time periods, different drivers, and different vehicles on the route to be evaluated as a sample data set, and correspondingly obtain a driving behavior characteristic parameter data set of human drivers; perform cluster analysis on the driving behavior characteristic parameters of human drivers in the driving behavior characteristic parameter data set to obtain multiple different driving behavior categories and cluster center points of each driving behavior category; select from the multiple different driving behavior categories a category with the smallest proportion of speeding time, the smallest average number of sudden acceleration / deceleration times, the smallest average number of lane changes, the smallest average energy consumption, the smallest number of unsuccessful obstacle avoidances, and the largest number of correct responses to road signs and markings by vehicles as a reference category, the reference point is the cluster center point of the reference category, and the reference distance is a value of the distance from the cluster center point of other driving behavior categories to the reference point.

2. The driving behavior evaluation method of an autonomous driving vehicle according to claim 1, characterized in that: The calculation formula of the driving behavior score of the autonomous driving vehicle is: In the formula, S represents the driving behavior score of the autonomous vehicle, M represents the score corresponding to the benchmark distance, D represents the benchmark distance, and d represents the distance from the sample point to the reference point.

3. The driving behavior evaluation method of an automatic driving vehicle according to claim 1 or 2, characterized in that: The step of performing cluster analysis includes: using principal component analysis to reduce the dimension of the driving behavior characteristic parameters of human drivers; and performing cluster analysis on the driving behavior characteristic parameters of human drivers after the dimension reduction in the driving behavior characteristic parameter data set.

4. The driving behavior evaluation method of an automatic driving vehicle according to claim 1 or 2, characterized in that: The method also includes a step of qualitatively evaluating the driving behavior of the autonomous driving vehicle, which step includes: inputting the driving behavior characteristic parameters of the autonomous driving vehicle into a trained driving behavior classification model to obtain a driving behavior category corresponding to the driving behavior of the autonomous driving vehicle; the trained driving behavior classification model is obtained by the following steps: using the driving behavior characteristic parameters of each human driver in the driving behavior characteristic parameter data set and their corresponding driving behavior categories as training samples, and using the training samples to train the classification model to obtain a trained driving behavior classification model.

5. The driving behavior evaluation method of an automatic driving vehicle according to claim 4, characterized in that: The classification model is a collaborative filtering algorithm classification model based on user behavior, a k-nearest neighbor algorithm classification model, a decision tree classification model or a SVM classification model.

6. A driving behavior evaluation device for an autonomous driving vehicle, characterized in that: The device includes a processor and a memory, and the processor executes a computer program stored in the memory to implement the driving behavior evaluation method of an autonomous driving vehicle as described in any one of claims 1-5.