CN110414803B

CN110414803B - Method and device for evaluating intelligent level of automatic driving system under different internet connection degrees

Info

Publication number: CN110414803B
Application number: CN201910611202.2A
Authority: CN
Inventors: 王建强; 刘艺璁; 郑讯佳; 许庆; 黄荷叶; 李克强; 崔明阳; 林学武
Original assignee: Tsinghua University
Current assignee: Tsinghua University
Priority date: 2019-07-08
Filing date: 2019-07-08
Publication date: 2022-01-07
Anticipated expiration: 2039-07-08
Also published as: CN110414803A

Abstract

The invention discloses an evaluation method for the intelligent level of an automatic driving system under different internet connection degrees, which comprises the following steps: s1, selecting an evaluation index of the intelligent level of the automatic driving system; s2, obtaining the quantitative evaluation basis of the intelligent level of the automatic driving system according to the numerical difference between the actual action amount and the theoretical minimum action amount in the traffic test process; s3, determining the variation range of the quantitative evaluation basis, and dividing at least two evaluation intervals for evaluating the intelligent level of the tested automatic driving system, wherein each evaluation interval corresponds to one intelligent level; s4, acquiring multiple groups of quantitative evaluation basis data of the tested automatic driving system under different internet connection degrees; and S5, carrying out statistical analysis on the quantitative evaluation basis data, and evaluating the intelligence level of the tested automatic driving system according to each statistical analysis result. The invention can three-dimensionally and truly evaluate the intelligent level of the automatic driving system under different internet connection degrees.

Description

Method and device for evaluating intelligent level of automatic driving system under different internet connection degrees

Technical Field

The invention relates to the technical field of intelligent networked automobile and automatic driving automobile evaluation, in particular to an evaluation method and device for the intelligent level of an automatic driving system under different networking degrees.

Background

Intellectualization and networking are inevitable trends of automatic automobile driving, and in the long-term networking development process, networked vehicles and non-networked vehicles simultaneously exist in traffic. Different internet connection degrees in mixed traffic are important factors influencing the intelligent level of the automatic driving automobile, and the evaluation of the intelligent level of the automatic driving system under different internet connection degrees can reveal the back influence law. In recent years, the technology of the automatic driving intelligent test field is continuously developed, and compared with the traditional test method (such as software simulation test, hardware-in-loop test, test scene test and real vehicle road test), the automatic driving intelligent test field has stronger capabilities of automation, repeatability, trueness and the like, and provides an effective platform for the evaluation of an automatic driving system under different internet connection degrees.

In the prior art, an important factor of different internet connection degrees existing in mixed traffic is not considered, most of the prior art can be regarded as an evaluation technology for fixing the internet connection degree to be 0, and the prior art cannot be used for evaluating the intelligent level of an automatic driving system under different internet connection degrees. Even from the perspective of reference, the existing evaluation technology still has many problems:

1. the prior art is not reasonable enough to select the evaluation index of the intelligent level of the automatic driving system. The three unmanned challenge games organized by DARPA (united states defense advanced planning research institute) all take the time consumption for completing all specified projects as evaluation indexes, but the single index cannot reflect the intelligence level of the automatic driving system. The intelligent levels are layered and indexes are selected according to task classification at each level for comprehensive evaluation, but the comprehensive indexes do not directly reflect the intelligent levels (such as safety and high efficiency).

2. The prior art is not objective enough due to the fact that subjective factors are mixed in evaluation. The complexity is subjectively determined according to task categories and corresponds to intelligent level grades, but the complexity difference of each type of task under different scenes is huge. Subjective factors also exist in selecting the weight corresponding to each evaluation index.

3. In the prior art, the running track can be quantitatively evaluated, but the selected reference track is based on the static ideal geometric track and the running track of a driver, and the intelligent level is corresponded by quantifying the difference between the track to be evaluated and the reference track. However, the reference trajectory selected by the method is not suitable for a dynamic traffic environment, and cannot embody objective safety and high efficiency.

4. The comprehensive evaluation index and the method obtained by the prior art are not enough to process the randomness of the networking degree in the test.

Therefore, in order to solve the above problems, it is necessary to develop an evaluation method and apparatus for the intelligence level of the automatic driving system under different internet connection levels.

Disclosure of Invention

The invention aims to provide an evaluation method and an evaluation device for the intelligent level of an automatic driving system under different internet connection degrees, which can three-dimensionally and truly evaluate the intelligent level of the automatic driving system under different internet connection degrees.

In order to achieve the purpose, the invention provides an evaluation method for the intelligent level of an automatic driving system under different internet connection degrees, which comprises the following steps:

s1, selecting the action quantity as an evaluation index of the intelligent level of the automatic driving system based on the driving safety field theory;

s2, obtaining the quantitative evaluation basis of the intelligent level of the automatic driving system according to the numerical difference between the actual acting quantity and the theoretical minimum acting quantity in the traffic testing process, wherein the quantitative evaluation basis can monotonously change along with the change of the numerical difference between the actual acting quantity and the theoretical minimum acting quantity;

s3, determining the variation range of the quantitative evaluation basis according to the acquisition mode of the quantitative evaluation basis in S2, and dividing at least two evaluation intervals for evaluating the intelligent level of the tested automatic driving system, wherein each evaluation interval corresponds to one intelligent level;

s4, acquiring multiple groups of quantitative evaluation basis data of the tested automatic driving system under different internet connection degrees in a set automatic driving intelligent test scene;

and S5, performing statistical analysis on each group of quantitative evaluation obtained in S4 according to data, and evaluating the intelligence level of the tested automatic driving system according to the evaluation interval in S3 to which each statistical analysis result belongs.

Further, the quantitative evaluation in S4 specifically includes, according to the data acquisition method:

acquiring an actual acting quantity S of an actual measurement path traveled by the vehicle to be tested according to the driving parameters of the vehicle to be tested, the driving parameters of other road users and road environment test data acquired in the automatic driving intelligent test scene;

acquiring theoretical minimum action quantity S in the automatic driving intelligent test scene according to reference paths planned in advance for other road users in the automatic driving intelligent test scene and road environment test data;

and obtaining the quantitative evaluation basis data according to the numerical difference between the actual acting amount and the theoretical minimum acting amount under the automatic driving intelligent test scene and by combining the acquisition mode of the quantitative evaluation basis in the S2.

Further, the actual acting amount S in S41 is obtained from the following formulas (1) to (7):

G_i＝m_ig (4)

F_ai＝E_ai·M_i·P_i (5)

in equations (1) to (7), A is the start position of the travel path of the vehicle i under test, B is the end position of the travel path of the vehicle i under test, t_AIs the time corresponding to A, t_BAt the time corresponding to B, L is the measured timeLagrange quantity, x, in the travel path of vehicle i_iIs the measured path longitudinal displacement, y, of the measured vehicle i_iIs the lateral displacement of the measured path of the measured vehicle i,

for the longitudinal running speed of the vehicle i under test along the measured path,

the longitudinal acceleration of the vehicle i under test along the measured path,

for the transverse running speed, R, of the vehicle i under test along the measured path_iAs a field of resistance, G_iIs a constant force field, F_aiFor the risk forces of the lane or road boundary a on the vehicle i under test, E_aiTo be located at (x)_a,y_a) The potential energy field formed by the lane line or the road boundary a is in (x)_i,y_i) The vector field strength of (d); v_jiPotential energy generated by other road users j to the tested vehicle i, F_jiFor the risk forces, E, produced by the other road users j on the vehicle i under test_jiThe kinetic energy field formed for the rest of the road users j is (x)_i,y_i) The vector field strength of (d); m is_iThe quality of the tested vehicle i is measured; g is gravity acceleration, f is rolling resistance coefficient, i_αIs a gradient, C_DiIs the wind resistance coefficient, W, of the vehicle i under test_iIs the windward area, lambda, of the vehicle i to be tested_iIs a rotating mass conversion coefficient, P, of the vehicle i to be measured_aAs road influencing factor, P, at the lane line a or at the road boundary_iIs an influence factor, P, at the i position of the vehicle to be tested_jIs the road influence factor of the other road users j, D is the lane width, r_aiTo point from lane line a or road boundary to the centroid (x) of the vehicle i under test_i,y_i) Distance vector of, M_iIs the equivalent mass of the vehicle i to be tested, M_jIs the equivalent mass of the other road users j, K is the adjustment coefficient, r_jiIs the mass center (x) of the user j on the other roads_j,y_j) I mass center (x) of the vehicle to be detected_i,y_i) Distance vector between, v_jIs the velocity vector, θ, of the remaining road users j_jIs r_jiAnd v_jThe angle of (a) is a lane line or a road boundary, b is the number of the lane line or the road boundary, and n is the number of other road users.

Further, the theoretical minimum amount of action S is obtained by the following formula (16):

in the formula (I), the compound is shown in the specification,

is the longitudinal displacement of the reference path of the tested vehicle i;

is the transverse displacement of the reference path of the vehicle i under test,

for the longitudinal travel speed of the vehicle i under test along the reference path,

is the transverse running speed of the tested vehicle i along the reference path.

Further, the manner of obtaining the basis for the quantitative evaluation in S2 includes:

the first case: the quantitative evaluation criterion can be monotonically increased as the difference between the actual acting amount and the theoretical minimum acting amount increases, and is expressed by the following formula (8):

or

The second case: the quantitative evaluation criterion is expressed as the following formula (9) which can be monotonically decreased as the difference between the actual amount of action and the theoretical minimum amount of action increases:

or

Further, the sets of quantitative evaluation in S4 are (y) in terms of data_k1,y_k2 ...y_km) Wherein, k is a serial number corresponding to one of the network connection degrees, and m is the number of different network connection vehicle distribution forms under the network connection degree; the method of "statistically analyzing each of the quantitative evaluations by data obtained in S4" in S5 includes: and calculating the fitting of the average value, the standard deviation, the extreme value, the frequency and the frequency distribution characteristic or specific distribution of each group of quantitative evaluation basis data, and presenting the statistical analysis result in a distribution diagram and/or a form.

Further, the method of "performing statistical analysis on each set of the quantitative assessments obtained in S4 according to data" in S5 specifically includes:

respectively calculating the average value and the extreme value of each group of quantitative evaluation basis data to obtain the intelligent level grade which can be reached and at least can be reached by the measured automatic driving system under different networking degrees or the average value and the lower limit of the intelligent level grade of the measured automatic driving system;

wherein the quantitative evaluation is expressed by an average value of data as formula (17):

in the first case, the quantitative evaluation is based on the maximum y of the extremum of the data expressed by equation (18)_kmax：

y_kmax＝max{y_k1,y_k2,...,y_km} (18)

In the second situation, theThe extreme value of the quantitative evaluation criterion data is a minimum value y expressed by formula (19)_kmin：

y_kmin＝min{y_k1,y_k2,...,y_km} (19)。

Further, S4 is followed by:

s6, storing the actual action amount S obtained through calculation in the S4, the internet connection degree C in the automatic driving intelligent test scene and the distribution form F of internet connection vehicles into a (C, F, S) form;

and S7, under the condition that the intelligent automatic driving test scenes are the same, executing a step S41 by changing the internet connection degree C and the distribution form F of the internet connection vehicles, testing the automatic driving system to be tested and recording the actual action S corresponding to the test process.

The invention also provides an evaluation device for the intelligent level of the automatic driving system under different internet connection degrees, which comprises the following components:

the information acquisition module is used for acquiring the running parameters of the tested vehicle, the running parameters of other road users and road environment test data in the automatic driving intelligent test scene;

the action quantity calculation module is used for acquiring the actual action quantity of the actual measurement path traveled by the tested vehicle according to the traveling parameters of the tested vehicle, the traveling parameters of other road users and the road environment test data acquired by the information acquisition module, and acquiring the theoretical minimum action quantity in the automatic driving intelligent test scene according to the reference path planned in advance for the other road users in the automatic driving intelligent test scene and the road environment test data; and

and the statistical evaluation module is used for storing an evaluation interval for evaluating the intelligent level grade of the tested automatic driving system, acquiring multiple groups of quantitative evaluation basis data of the tested automatic driving system under different internet connection degrees according to the actual action quantity and the theoretical minimum action quantity obtained by the action quantity operation module, carrying out statistical analysis on each group of quantitative evaluation basis data, and evaluating the intelligent level of the tested automatic driving system according to the evaluation interval to which each statistical analysis result belongs.

Further, the action amount calculation module includes:

and the actual acting quantity calculating unit of the tested vehicle calculates the actual acting quantity S in the actual measuring path of the tested vehicle according to the running parameters of the tested vehicle and the road environment test data₀；

A road constraint actual acting amount calculating unit which establishes a static risk field of a lane line, a road boundary or a static obstacle according to the road environment test data and the driving data of the tested vehicle based on a driving safety field theory and calculates the road constraint actual acting amount S of the lane line, the road boundary or the static obstacle in the tested vehicle actual measurement path₁；

The actual acting amount calculating unit of the other road users calculates the actual acting amount S in the actual measuring paths of the other road users according to the driving parameters of the other road users and the road environment test data₂；

A collecting means for collecting the results of the calculation by the actual acting amount calculating means of the vehicle under test, the road constraint actual acting amount calculating means, and the actual acting amount calculating means of the other road users, and calculating the actual acting amount S of the actual path on which the vehicle under test travels according to the following formula:

S＝S₀-S₁-S₂。

the invention has the beneficial effects that: the invention provides an evaluation method and device for the intelligent level of an automatic driving system under different internet connection degrees based on an automatic driving intelligent test field platform aiming at evaluation under a long-term stage that the popularization process of the automatic driving system of mixed traffic is bound. In the specific evaluation process, the intelligent level of the automatic driving system is quantified by establishing an action quantity index in a more objective and interpretable mode, and two important factors of reflecting the intelligent level, namely safety and high efficiency, can be directly reflected. In the grading evaluation of the intelligent level, the mode of artificially assuming a scene and task difficulty degree in the prior art is abandoned, the intelligent level grade of the automatic driving system in the test scene is objectively and reasonably evaluated from multiple angles by constructing the numerical difference mapping between the actual action quantity and the theoretical minimum action quantity, and the evaluation method is not limited to a specific test scene and has wide application range. In addition, the evaluation method provided by the invention is combined with statistical analysis, and can well cope with randomness caused by internet connection vehicle distribution under different internet connection degrees, so that the evaluation result of the intelligent level of the automatic driving system under different internet connection degrees can be more three-dimensional and real from the whole angle.

Drawings

FIG. 1 is a schematic diagram of a control logic structure of an automatic driving intelligent test field based on the invention;

FIG. 2 is a flow chart of the evaluation of the intelligent level of the automatic driving system under different internet connection degrees in the present invention;

FIG. 3 is a schematic view of a scene for an automatic driving system evaluation at a signal-free intersection by an automatic driving intelligent test field according to the present invention;

FIG. 4 is a schematic diagram of test scenario generation and reproduction under different networking degrees in the present invention;

fig. 5 is a schematic diagram of an internal module of the vehicle-mounted intelligent level evaluation device provided by the invention.

Detailed Description

In the drawings, the same or similar reference numerals are used for the same or similar elements or elements having the same or similar functions. Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

The method for evaluating the intelligence level of the automatic driving system under different internet connection degrees provided by the embodiment is implemented based on the automatic driving intelligent test field, fig. 1 shows a schematic control logic structure diagram of the automatic driving intelligent test field, and fig. 3 shows a schematic scene diagram of the automatic driving intelligent test field provided by the embodiment for evaluating the signal-lamp-free intersection automatic driving system. As shown in fig. 1 and 3, the autonomous driving intelligent test site can perform information interaction with the control center 4. The automatic driving intelligent test field is a completely networked test environment and is provided with equipment such as a vehicle networking communication facility, a roadside communication facility 5, a test road 6, a differential GPS high-precision positioning system 7 and the like, and various data acquired by the equipment are fed back to the control center 4. The control center 4 can plan, simulate, monitor and remotely control the tested autonomous driving vehicle 1 (hereinafter, referred to as the "tested vehicle i or the tested vehicle 1" for short).

Specifically, as shown in fig. 2, the method for evaluating the intelligence level of the automatic driving system under different internet connection degrees provided by this embodiment includes the following steps:

and S1, selecting the action quantity as an evaluation index of the intelligent level of the automatic driving system based on the driving safety field theory. The acting quantity S includes both a traffic risk variable directly reflecting safety and an integral at a time level directly reflecting high efficiency, and may be formula (1):

in the formula (1), S is the acting quantity of the tested vehicle in the test traffic process, and the test traffic process refers to the whole process that the tested vehicle passes through the intersection without the signal lamp from the initial position A, turns left, and passes over the stop line of the target lane to reach the final position B. t is t_AThe starting time of the traffic process (the time corresponding to the arrival of the tested vehicle at the starting position A) can be tested according to the starting condition of the test task. t is t_BThe terminal time of the test traffic process (the time corresponding to the test traffic process vehicle reaching the terminal position B) can correspond to the terminal condition of the test task. And L is the Lagrange quantity of the tested vehicle in the traffic test process.

By the selected evaluation index, S1 can quantify the intelligence level of the automatic driving system in an objective and interpretable manner.

Of course, in a complex test traffic process, the traffic risk of the vehicle under test includes road constraints and the risk impact on it by the remaining road users. According to the driving safety field theory, various traffic risk influences can be quantified in the form of lagrangian quantity, and then the lagrangian quantity L can be expressed as formula (2):

G_i＝m_ig (4)

F_ai＝E_ai·M_i·P_i (5)

in equations (1) to (7), A is the start position of the travel path of the vehicle i under test, B is the end position of the travel path of the vehicle i under test, t_AIs the time corresponding to A, t_BThe moment corresponding to B; l is the Lagrange quantity in the driving path of the tested vehicle i; x is the number of_iIs the measured path P1 longitudinal displacement, y of the measured vehicle i_iThe measured path P1 of the measured vehicle i is laterally displaced,

for the longitudinal travel speed of the vehicle i under test along the measured path P1,

the longitudinal acceleration of the vehicle i under test along the measured path P1,

the kinematic information can be acquired by a GPS (global positioning system) for the transverse running speed of the tested vehicle i along the tested path P1; r_iAs a field of resistance, G_iIs a constant force field; f_aiFor the risk forces of the lane or road boundary a on the vehicle i under test, E_aiTo be located at (x)_a,y_a) The potential energy field formed by the lane line or the road boundary a is in (x)_i,y_i) The vector field strength of (d); v_jiPotential energy generated by other road users j to the tested vehicle i, F_jiFor the risk forces, E, produced by the other road users j on the vehicle i under test_jiThe kinetic energy field formed for the rest of the road users j is (x)_i,y_i) The vector field strength of (d); m is_iThe quality of the tested vehicle i is measured; g is the acceleration of gravity, generally taken to be 9.81m/s²(ii) a f is a rolling resistance coefficient which is determined according to the conditions of the tire and the road surface and is generally selected from 0.015 to 0.02; i.e. i_αThe grade is determined according to the road geometric conditions of the test field; c_DiThe wind resistance coefficient of the tested vehicle i is determined by the appearance shape of the tested vehicle and is generally selected from 0.25 to 0.5; w_iThe frontal area of the tested vehicle i can be obtained by calculation according to the geometric shape of the tested vehicle; lambda [ alpha ]_iThe conversion coefficient of the rotating mass of the tested vehicle i can be generally 1.05 according to the relevant knowledge of the vehicle theory; l is_T,aSelecting a corresponding value for the type of the lane line a or the road boundary according to the definition of the driving safety field theory; p_aAs road influencing factor, P, at the lane line a or at the road boundary_iIs an influence factor, P, at the i position of the vehicle to be tested_jReferring to the related papers which are already published, the influence factors of the roads at the j positions of other road users are generally set to be 1; d is lane width which can be obtained by measuring the lane width of a test field; r is_aiTo point from lane line a or road boundary to the centroid (x) of the vehicle i under test_i,y_i) A distance vector of (d); m_iIs the equivalent mass of the vehicle i to be tested, M_jThe equivalent mass of the other road users j is a function of the vehicle type, the real mass and the running speed, and can be calculated by referring to a related paper; k is an adjustment factor, which is generally set to 0.5 in reference to related papers; k is a radical of₁，k₂And k₃For constant coefficients, the reference related papers are generally set to 1,1.2 and 45, respectively; r is_jiIs the mass center (x) of the user j on the other roads_j,y_j) I mass center (x) of the vehicle to be detected_i,y_i) A distance vector therebetween; v. of_jFor the speed of the rest of the road user jA vector; theta_jIs r_jiAnd v_j(iv) an included angle (counterclockwise is positive); a is a lane line or a road boundary; b is the number of lane or road boundaries; n is the number of remaining road users.

S2, theoretically, by solving the extreme value analysis of the acting quantity S, there must be a reference path P2 corresponding to the theoretical minimum acting quantity S, i.e., the vehicle travels according to the reference path P2, and at this time, the safety and efficiency of the corresponding vehicle reach the optimal state. In view of this, from the numerical difference between the actual quantity of action S and the theoretical minimum quantity of action S during the test traffic, a quantitative evaluation basis y of the intelligence level of the autopilot system is obtained, which can vary monotonically with the variation of the numerical difference between said actual quantity of action S and said theoretical minimum quantity of action S. Such as: the quantitative evaluation criterion y may become monotonically larger as the numerical difference between the actual acting amount S and the theoretical minimum acting amount S becomes larger, or the quantitative evaluation criterion y may become monotonically smaller as the numerical difference between the actual acting amount S and the theoretical minimum acting amount S becomes larger. The terms "become larger and monotonously larger" and "become larger and monotonously smaller" here mainly depend on the form of acquisition of the selected quantitative evaluation criterion y according to the quantitative evaluation criterion y.

In one embodiment, the quantitative evaluation is represented in the present embodiment in accordance with the form of acquisition of y as a mapping f (S, S) in the formulae (8) and (9), by means of which the numerical difference between the actual acting quantity S and the theoretical minimum acting quantity S can be quantified, i.e. the quantitative evaluation varies monotonically in accordance with the change in the difference between the actual acting quantity S and the theoretical minimum acting quantity S, the scale of which is related to the mapping f (S, S). The mapping relationship f (S, S) includes two cases:

or

or

Of course, the mapping relation f (S, S) may be in other forms than the above-described four forms as long as a requirement that the quantitative evaluation monotonously changes according to the change of the difference between the actual amount of action S and the theoretical minimum amount of action S in terms of y is satisfied.

Preferably, S2 specifically includes the following S21 and S22:

s21, calculating a theoretical minimum acting quantity S and its corresponding reference path P2 based on the actual acting quantity S during the test traffic. According to the expression of the actual acting quantity S (functional) provided by the formula (1), the variation is calculated and made 0 according to the formula (10), a specific reference path P2 (which is a function of time) corresponding to the theoretical minimum acting quantity S can be obtained, and then the reference path P2 is substituted into the formula (1), so that a specific numerical value of the theoretical minimum acting quantity S can be obtained:

and S22, establishing a quantitative evaluation basis y for dividing the intelligent level grade of the automatic driving system by constructing an acquisition form of the quantitative evaluation basis y of the intelligent level of the automatic driving system according to the actual action quantity S and the theoretical minimum action quantity S obtained in the step S21. In equations (8) and (9), it is obvious that the numerical difference between the actual amount of action S and the theoretical minimum amount of action S is represented by the difference therebetween. By definition of the amount of action S, it is necessary that the actual amount of action S is greater than the theoretical minimum amount of action S.

S3, according to the acquisition mode of the quantitative evaluation basis in S2, determining the variation range of the quantitative evaluation basis y, and dividing at least two evaluation intervals (shown as the following intervals) for evaluating the intelligent level of the tested automatic driving system, wherein each evaluation interval corresponds to one intelligent level:

[y₀,y₁)[y₁,y₂)...[y_n-1,y_n]

wherein, y₀,y₁,…,y_nThe endpoint values of each interval. In the test process, each endpoint value is determined according to the following rules:

1. randomly recruiting a plurality of drivers, completely controlling the tested vehicle by the drivers under the same test scene, calculating the acting amount and the quantitative evaluation basis y under the control of different drivers, and calculating the mean value of the quantitative evaluation basis

Standard deviation sigma_hm；

2. When n is an even number, let

When n is an odd number, let

Preferably, n-5;

3. length equalization is performed to construct each interval, and the standard deviation sigma of the quantitative evaluation is set_hmOf the ratio, i.e. y_k-y_k-1＝k_p·σ_hm,

Coefficient of interval length k_pThe selection can be combined with the requirement of the refinement degree; preferably, k is_p＝2；

4. Adjusting the coverage of the far left and right regions, mainly y₀And y_nThe coverage of the mapping relation is just consistent with the theoretical coverage of the quantitative evaluation according to y.

For example: in this embodiment, the first case is adopted and the mapping relationship corresponding to the left side is adopted, that is, y ═ S/S. At this time, the smaller the quantitative evaluation criterion y, the higher the intelligence level. In this example, assume that

σ_hm0.25, the range [0, + ∞ ] according to y is quantized and evaluated, and is divided into 5 intervals, the interval length coefficient k_pEach interval corresponds to an intelligent level, and is sequentially "very high", "higher", "medium", "low", and "very low", as shown in the following intelligent level table 1:

TABLE 1

y interval	[0,0.5)	[0.5,1.0)	[1.0,1.5)	[1.5,2.0)	[2.0,+∞)
						Intelligent level	Is very high	Is higher than	Medium and high grade	Is lower than	Is very low

And S4, acquiring multiple groups of quantitative evaluation basis data of the tested automatic driving system under different internet connection degrees in the set automatic driving intelligent test scene shown in the figures 2 to 4.

Preferably, the quantitative evaluation in S4 specifically includes, according to the data acquisition method:

and S41, before each test, generating and reproducing an automatic driving intelligent test scene and changing necessary test conditions through the control center 4 to obtain the running parameters of the tested vehicle, the running parameters of other road users and road environment test data under each test condition.

The control center 4 can monitor and control the automatic driving intelligent test field, and data acquisition is carried out through feedback information of the automatic driving intelligent test field, so that a data center in the test process is formed. The "control center 4" can also form a high-precision map, and has important functions of communicating with and planning paths of various vehicles in a test scene.

Specifically, the software layer of the control center 4 has different software modules docked with it according to different categories of objects of the received data, and the modules operate independently from each other and are processed in a fusion optimization manner. The modules are suitable to adopt a distributed communication mode, the software scheme can select 'zeroMQ' as a communication library, and selects a 'publish-subscribe' working mode of the zeroMQ. All modules can be mounted on a zeroMQ type information bus, and after each module acquires the latest test information, the latest test information is published on the information bus so as to be subscribed by the required modules. It should be noted that, different marks are respectively made on the networked vehicle and the non-networked vehicle in the test scene before the test starts, and these marks are reflected on the information bus along with the transmitted parameters, so that the module for assisting "subscription" can be used for identifying information sources. For example, when the intelligent level of the process that the tested vehicle turns left and passes through the intersection without the signal lamp is tested to evaluate the intelligent level of the automatic driving system in the test scene: when the tested vehicle turns left and passes through the intersection without the signal lamp, the control center 4 can accurately capture the information through a high-precision map and a differential GPS positioning function of the control center, and the test process of the networking degree under the condition of the networking vehicle distribution form is terminated, the information interaction among all modules of a software layer of the control center 4 is terminated along with the termination of the information interaction, but all previous test information is stored by the control center 4, and a data center of the test process is formed.

The "remaining road users" mainly include the networked vehicles 2 and the non-networked vehicles 3.

The 'necessary test condition' is a preset series of limited networking degrees represented by the occupation ratio of the networked vehicles, the distribution form of the networked vehicles and whether the independent sensing system of the tested vehicle is activated under the full networking degree. That is, in the intelligent test field, the network connection degree and the network connection vehicle distribution can be changed, thereby establishing different test conditions.

The traffic elements in the "test scenario" include: as shown in fig. 2 and 3, the vehicle to be tested 1, the networked vehicle 2, the non-networked vehicle 3, the test road 6, the test facility and the test background object. In this embodiment, the test scenario is set as: the vehicle 1 to be tested directly approaches the stop line of the traffic intersection without the signal lamp from south to north, safely passes through the traffic intersection, turns left to turn to the west lane, crosses the stop line and leaves the traffic intersection.

In the 'test scene', the automatic driving system to be evaluated is installed on the tested vehicle 1, and the automatic driving system is accessed into the centralized internet of vehicles in the automatic driving intelligent test scene, so that the automatic driving system is allowed to acquire required external information through the data center of the control center 4 according to the algorithm and the test condition of the automatic driving system. And the networked vehicle and the non-networked vehicle are not provided with an automatic driving system to be evaluated. The networked vehicle 2 and the non-networked vehicle 3 are set as vehicles simulating manual driving by the control center 4, and both the networked vehicle 2 and the non-networked vehicle 3 can communicate with the control center 4 and plan the path of the vehicles. The networked vehicle 2 and the non-networked vehicle 3 are different in whether they can communicate with the vehicle 1 under test through the control center 4 and exchange necessary information (such as position, velocity, acceleration, etc.). The test road 6 is a part of road structure suitable for the test scene selected in the automatic driving intelligent test field, in the embodiment, the test road 6 is a crossroad, and software and hardware equipment such as traffic lights and other traffic indication signals are removed, so that the signal-lamp-free intersection test road is formed. The test facilities mainly comprise a differential GPS base station 7, a traffic camera, roadside communication equipment 5, communication equipment DSRC/LTE-V) and the like. The test backdrop mainly comprises background objects near the signalless intersection for simulating buildings, blocks, etc.

In each test, the running parameters of the vehicle to be tested, the running parameters of the networked vehicle 2 and the non-networked vehicle 3 and the road environment test data are extracted from the test data fed back to the control center 4 from the automatic driving intelligent test field. The "running parameters" include longitudinal and lateral displacements, velocities, accelerations, and the like. The "road environment test data" includes path information of the road users other than the vehicle 1 to be tested, the networked vehicle 2 and the non-networked vehicle 3, road surface roughness, gradient, wind power of the road environment, lane line or road boundary position, static obstacle position and size, and the like.

The test data all adopt a centralized communication topological structure to realize the interaction of information and data. The tested vehicle, the networked vehicle and the non-networked vehicle respectively pack driving parameters which CAN be extracted from the CAN, including parameters such as longitudinal and transverse displacement, speed, acceleration and the like, transmit the driving parameters to the roadside communication facility in real time through a DSRC or LTE-V communication technology, and feed back the driving parameters to the control center 4 through a special communication mode of the roadside communication facility and the control center 4. In this embodiment, the other road users (such as simulated moving pedestrians, model moving cyclists, bicycles, etc.) are not considered, but the effectiveness of the evaluation method is not affected, and if the other road users exist, the path information and parameters thereof can be fed back to the control center 4 by the same communication technology and manner. As for the test data of the road environment, including the road roughness, the road gradient, the wind power, the direction and the speed of the road environment, the position of the lane line or the road boundary, the position and the size of the static obstacle, etc., it can be basically considered as static data, so at the beginning of the test process, the test data of all the road environments are packaged and transmitted to the control center 4 at one time by the aforementioned communication technology and manner.

In the test process, the tested vehicle needs to acquire information of other vehicles and road environments through sensing to assist intelligent decision and control of an internal automatic driving system. On one hand, the tested vehicle is provided with a plurality of sensors, including a camera, a laser radar, a millimeter wave radar, an ultrasonic radar and the like, so that the requirement of the automatic driving system on independent sensing hardware in the test process is met, and the connection and disconnection of a software layer with the automatic driving system to be tested in the vehicle can be realized; on the other hand, the automatic driving system can also acquire information in a sensing mode of internet connection, the automatic driving system performs information interaction with a road side communication facility and a control center 4 software layer tested vehicle module, the control center 4 software layer tested vehicle module performs information interaction with other modules through a zeroMQ type information bus, data with a non-internet vehicle specific mark is not allowed to be transmitted to the control center 4 software layer tested vehicle module, and therefore transmission of the information through a communication link of the control center 4-road side communication equipment-tested vehicle is prevented from the source, and the fact that the tested vehicle cannot acquire any information of the non-internet vehicle in the test process is guaranteed.

And S42, after each test, acquiring the actual acting quantity S of the measured path P1 traveled by the measured vehicle according to the driving parameters of the measured vehicle, the driving parameters of other road users and the road environment test data acquired in the automatic driving intelligent test scene. Wherein, by using the above equations (1) and (2), the acting quantity S is a time integral of the lagrangian quantity L, which includes three parts: kinetic energy and resistance potential energy L of tested vehicle₀Potential energy V generated by lane line, road boundary or static barrier to the tested vehicle₁The total potential energy V generated by other road users to the tested vehicle₂。

Wherein, the kinetic energy and the resistance potential energy L of the tested vehicle₀"the calculation method is as follows:

the obtained running parameters of the tested vehicle and part of road environment test data are combined with the static characteristic parameters of the tested vehicle, including the mass, the rolling resistance coefficient,The wind resistance coefficient, the windward area, the conversion coefficient of the rotating mass and the like can be used for calculating the kinetic energy and the resistance potential energy L of the tested vehicle₀The formula is calculated as the following formula (11):

wherein, the potential energy V generated by the lane line, the road boundary or the static barrier to the tested vehicle₁"the calculation method is as follows:

based on the driving safety field theory, a static risk field of a lane line, a road boundary or a static obstacle can be established. According to the obtained road environment test data and the driving data of part of the tested vehicles, the potential energy V generated by the lane line, the road boundary or the static barrier to the tested vehicles can be calculated₁The calculation formula is shown as the following formula (12):

wherein, the total potential energy V generated by other road users to the tested vehicle₂"the calculation method is as follows:

based on the driving safety field theory, the dynamic risk fields of other road users can be established. The other road users comprise networked vehicles, non-networked vehicles, vulnerable road users and the like. According to the obtained other vehicles and the road environment test data, the total potential energy V generated by other road users on the tested vehicle can be calculated₂The formula is as follows (13):

the lagrangian quantity L is calculated as shown in the following formula (14):

L＝L₀-V₁-V₂ (14)

according to the time that the tested vehicle completes the task in the test process, the Lagrange quantity L in the test process is integrated, the action quantity S corresponding to the tested vehicle actual measurement path P1 under the network connection degree and the distribution condition of the test can be calculated, and the calculation formula of the actual action quantity S is the following formula (15):

after determining the test scenario of left-turn traffic at the intersection without signal lamps selected in this embodiment, the control center 4 may reproduce the test scenario at the beginning of each test, and the method for reproducing the test scenario specifically includes:

when some traffic elements (such as a tested vehicle, an internet vehicle and a non-internet vehicle) are not in the preset initial test condition, firstly, the control center 4 carries out high-precision path planning on the traffic elements which do not meet the test condition by virtue of the automatic high-efficiency monitoring control capability of the control center on the elements in the test field, so that the traffic elements can arrive at the position meeting the initial test condition as soon as possible and accurately. Then, the control center 4 performs high-precision attitude control on all the traffic elements in position, so that the traffic elements meet the initial test attitude requirement. Finally, the control center 4 performs initial motion control on all the traffic elements in the preset position and attitude setting, so that each traffic element simultaneously reaches the motion parameters (such as speed, acceleration and the like) required by the initial test at the preset initial position and in the preset attitude.

Before the test, as shown in fig. 4, the control center 4 needs to set and mark the internet connection features of all vehicles except the vehicle to be tested. The settings and flags include two dimensions. One of the dimensions is the setting of the networking degree: and selecting a certain internet connection degree (such as 50%), namely the proportion of the internet connection vehicles in all vehicles except the vehicle to be detected. And the other dimension is the setting of the distribution form of the networked vehicles: since only a portion of the vehicles are networked vehicles and need to be distributed among all the vehicles, the selected distribution form of the embodiment is random distribution, which conforms to the unpredictability of the distribution of networked vehicles in a mixed traffic environment, and one form of marking is performed on the distributed networked vehicles, while another form of marking is performed on the rest of the non-networked vehicles. It should be noted that, this embodiment does not describe a distribution form of the internet vehicles in a network connection degree repeatedly, and in a series of cyclic tests under a real situation, the distribution form of the network connection degree and the internet vehicles may be variable rather than fixed.

And S43, acquiring theoretical minimum action quantity S in the automatic driving intelligent test scene according to the reference path P2 planned in advance for the other road users in the automatic driving intelligent test scene and the road environment test data. Wherein the theoretical minimum amount of action S is obtained from the following formula (16):

in the formula (I), the compound is shown in the specification,

is the longitudinal displacement of the reference path P2 of the vehicle i under test;

for the lateral displacement of the reference path P2 of the vehicle i under test,

for the longitudinal running speed of the vehicle i under test along the reference path P2,

is the lateral running speed of the vehicle i under test along the reference path P2.

It should be noted that, no matter how the networking degree and the distribution condition thereof in the test scenario change, the vehicle under test should have a uniform theoretical minimum acting amount S, which is determined by the continuous reproduction of the test scenario and the calculation manner of the theoretical minimum acting amount.

S44, obtaining multiple groups according to the numerical difference between the actual action amount in S42 and the theoretical minimum action amount in S43 and the acquisition mode of the quantitative evaluation basis in S2The quantitative evaluation is based on data. Specifically, a set of action amount data (S) corresponding to all the Internet connection vehicle distribution patterns in each Internet connection degree_k1,S_k2 ...S_km) And calculating the difference degree by combining the theoretical minimum action quantity S, and outputting a corresponding group of quantitative evaluation basis data. Each set of quantitative assessments was passed on the basis of data (y)_k1,y_k2 ...y_km) And presenting, wherein k is a serial number corresponding to one of the network connection degrees, and m is the number of different network connection vehicle distribution forms under the network connection degree.

And S5, performing statistical analysis on each group of quantitative evaluation obtained in S4 according to data, evaluating the intelligent level of the tested automatic driving system according to the evaluation interval in S3 to which each statistical analysis result belongs, and finally printing and outputting the intelligent level grading evaluation results of the tested automatic driving system under different internet connection degrees in a visual mode.

In one embodiment, the method of "statistically analyzing by data each of the sets of quantitative assessments obtained at S4" in S5 includes:

and calculating the fitting of the average value, the standard deviation, the extreme value, the frequency and the frequency distribution characteristic or specific distribution of each group of quantitative evaluation basis data, and presenting the statistical analysis result in a distribution diagram and/or a form. Specifically, the average value and the extreme value of each group of quantitative evaluation are respectively obtained according to data, so as to obtain the intelligent level which can be reached and at least can be reached by the tested automatic driving system under different networking degrees, or the average value and the lower limit of the intelligent level.

in the case where the quantitative evaluation criterion is monotonously increased with an increase in the difference between the actual amount of action and the theoretical minimum amount of action, the maximum value of the quantitative evaluation criterion determines the intelligenceLower limit of energy level at which the quantitative evaluation is based on the maximum value y represented by the data extreme value of equation (18)_kmax：

y_kmax＝max{y_k1,y_k2,...,y_km} (18)

In the case where the quantitative evaluation criterion is a criterion which monotonically decreases with an increase in the difference between the actual amount of action and the theoretical minimum amount of action, the minimum value of the quantitative evaluation criterion determines the lower limit of the intelligent level, and the extreme value of the quantitative evaluation criterion data at this time is the minimum value y expressed by the expression (19)_kmin：

y_kmin＝min{y_k1,y_k2,...,y_km} (19)

In one embodiment, the step S5 of "evaluating the intelligence level of the measured automatic driving system" is to perform multi-angle rating evaluation on the intelligence level of the measured automatic driving system under different internet access degrees according to the statistical analysis result of the quantitative evaluation basis and the intelligence level rating interval. Wherein the "multi-angle" points of interest include: generally, from the average meaning, the intelligent level levels that the tested automatic driving system can reach under different internet connection degrees, the intelligent level levels that the tested automatic driving system can at least reach under different internet connection degrees, the distribution characteristics of the intelligent level levels of the tested automatic driving system under different internet connection degrees, and the like. In particular, the intelligent level grade which can be reached by the tested automatic driving system is only sensed based on the tested vehicle independently under the zero networking degree, and the intelligent level grade which can be reached by the tested automatic driving system is only sensed based on the multi-vehicle networking degree under the full networking degree.

In one embodiment, S4 is followed by:

s6, storing the actual action S of the tested vehicle in the no-signal intersection testing process, the internet connection degree C and the distribution form F of the internet connection vehicles in the automatic driving intelligent testing scene, which are obtained through calculation in the step S4, into a data storage unit in an intelligent level evaluation device statistical analysis module in the form of a group of data, wherein the storage form is as follows:

(C,F，S) (20)

in the formula (20), C is the networking degree, and can be in a decimal form or a percentage form; and F, marking the data in the distribution form of the internet connection vehicles, and establishing a marking association method by marking different distribution forms of the internet connection vehicles through the data. In this embodiment, the serial numbers are associated with the internet connection distribution form in advance, and the serial numbers can mark the internet connection distribution form; s is the actual acting amount of the test process.

As shown in fig. 5, the present invention further provides an evaluation device for the intelligent level of the automatic driving system under different internet connection degrees, the evaluation device for the intelligent level is disposed on the vehicle to be tested, after a test is terminated, the evaluation device for the intelligent level mounted on the vehicle to be tested will make information and data requests to the control center 4 through roadside equipment, and the data center of the control center 4 will open all data that can be acquired by the control center 4 in the test process to the evaluation device for the intelligent level, no matter whether the driving parameters of the internet connection vehicle or the driving parameters of the non-internet connection vehicle, at this time, the specific marks of the two devices will not work.

The intelligent level evaluation device comprises an information acquisition module, an action quantity operation module, a statistic evaluation module and a result display module, wherein:

the information acquisition module is used for acquiring the running parameters of the tested vehicle, the running parameters of other road users and road environment test data in the automatic driving intelligent test scene.

The action quantity calculation module obtains the actual action quantity of the measured path P1 of the vehicle to be tested according to the running parameters of the vehicle to be tested, the running parameters of the other road users and the road environment test data which are obtained by the information acquisition module, and obtains the theoretical minimum action quantity in the automatic driving intelligent test scene according to the reference path P2 planned in advance for the other road users in the automatic driving intelligent test scene and the road environment test data.

The statistical evaluation module is used for storing an evaluation interval for evaluating the intelligent level grade of the tested automatic driving system, acquiring multiple groups of quantitative evaluation basis data of the tested automatic driving system under different internet connection degrees according to the actual action quantity and the theoretical minimum action quantity obtained by the action quantity operation module, carrying out statistical analysis on each group of quantitative evaluation basis data, and evaluating the intelligent level of the tested automatic driving system according to the evaluation interval to which each statistical analysis result belongs.

And the result display module is used for printing and outputting the intelligent level grading evaluation results of the tested automatic driving system under different internet connection degrees in a visual mode.

In one embodiment, the action calculation module includes an actual action calculation unit of the vehicle to be tested, a road constraint actual action calculation unit, an actual action calculation unit of the remaining road users, and a summary unit, wherein:

the unit for calculating the actual acting quantity of the vehicle to be tested requests and obtains the running parameters and part of road environment test data of the vehicle to be tested, which are stored in the data center of the control center 4 through communication technology and mode, and then calls the static characteristic parameters of the vehicle to be tested, which comprise mass, rolling resistance coefficient, wind resistance coefficient, windward area, rotating mass conversion coefficient and the like, and by utilizing the following formulas (21) to (24), the kinetic energy and resistance potential energy L of each time step length in the measured path P1 of the vehicle to be tested can be calculated₀Integrating the measured values and calculating the actual action S of the vehicle to be tested in the test process₀：

G_i＝m_ig (24)

In formulae (21) to (24), R_iIs a resistance field; g_iIs a constant force field; m is_iThe quality of the tested vehicle i is measured; x is the number of_iThe longitudinal displacement of the tested vehicle i is obtained; y is_iThe transverse displacement of the tested vehicle i is obtained; g is the acceleration of gravity; f is a rolling resistance coefficient; i.e. i_αIs a slope; c_DiThe wind resistance coefficient of the tested vehicle i is obtained; w_iThe frontal area of the tested vehicle i is shown; lambda [ alpha ]_iAnd the conversion coefficient is the rotating mass of the tested vehicle i.

The road constraint actual acting quantity calculating unit requests and acquires the road environment test data stored in the data center of the control center 4 and the driving data of part of the tested vehicle through a communication technology and a communication mode, and based on the driving safety field theory, a static risk field of a lane line, a road boundary or a static obstacle can be established, so that the potential energy V generated by the lane line, the road boundary or the static obstacle at each time step in the tested vehicle measured path P1 can be calculated₁And integrating the data to obtain the actual road constraint action S₁The calculation method is as follows:

F_ai＝E_ai·M_i·P_i (27)

in the formula, E_aiTo be located at (x)_a,y_a) The potential energy field formed by the lane line a or the road boundary is in (x)_i,y_i) Of the vector field strength, L_T,aOf the type of lane line a or road boundary, P_aIs the road influence factor at the lane line a or the road boundary, D is the lane width, r_aiTo point from lane line a or road boundary to the centroid (x) of the vehicle i under test_i,y_i) K is the adjustment coefficient, M_iIs the equivalent mass of the vehicle i to be tested, P_iIs the road influence factor at the detected vehicle i, a is the lane line or the road boundary, b is the number of the lane line or the road boundary, n is the number of other road users, k₂Is a constant coefficient.

The rest of the road users only comprise networked vehicles and non-networked vehicles. The actual action amount calculation units of other road users request and acquire the running parameters of the networked vehicles and the non-networked vehicles and the road environment test data stored in the data center of the control center 4 through communication technology and mode, and based on the driving safety field theory, the dynamic risk fields of the networked vehicles and the non-networked vehicles can be established, so that the total potential energy V generated by the networked vehicles and the non-networked vehicles at each time step in the measured vehicle path P1 can be calculated₂And integrating the data to obtain the actual action S of the users on other roads₂The calculation method is as follows:

in the formula, V_jiPotential energy generated by other road users j to the tested vehicle i, F_jiFor the risk forces, E, produced by the other road users j on the vehicle i under test_jiThe kinetic energy field formed for the rest of the road users j is (x)_i,y_i) At a vector field strength, grad E_jiThe field intensity gradient generated by the other road users j to the tested vehicle i, K is an adjusting coefficient, P_iIs an influence factor, P, at the i position of the vehicle to be tested_jIs a road influence factor, M, at the location of the other road users j_iIs the equivalent mass of the vehicle i to be tested, M_jIs the equivalent mass, r, of the remaining road users j_jiIs the mass center (x) of the user j on the other roads_j,y_j) I mass center (x) of the vehicle to be detected_i,y_i) Distance vector between, V_ijPotential energy v generated by the tested vehicle i to other road users j_jIs the velocity vector, θ, of the remaining road users j_jIs r_jiAnd v_jThe angle of,

is v is_jAngle with the x-axis, k₁And k₃Is a constant coefficient.

The summarizing unit summarizes the calculated results of the actual action amount calculating unit of the tested vehicle, the road constraint actual action amount calculating unit and the actual action amount calculating units of other road users, and can calculate the action amount S corresponding to the measured path P1 of the tested vehicle under the network connection degree and the distribution condition of the network connection degree, and the calculating mode is as follows:

finally, it should be pointed out that: the above examples are only for illustrating the technical solutions of the present invention, and are not limited thereto. Those of ordinary skill in the art will understand that: modifications can be made to the technical solutions described in the foregoing embodiments, or some technical features may be equivalently replaced; such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.

Claims

1. An evaluation method for the intelligent level of an automatic driving system under different internet connection degrees is characterized by comprising the following steps:

s5, carrying out statistical analysis on each group of quantitative evaluation obtained in S4 according to data, and evaluating the intelligence level of the tested automatic driving system according to the evaluation interval in S3 to which each statistical analysis result belongs;

the method for acquiring the quantitative evaluation reference data in S4 specifically includes:

acquiring the actual acting quantity S of the actual measurement path of the vehicle to be tested according to the driving parameters of the vehicle to be tested, the driving parameters of other road users and the road environment test data acquired in the automatic driving intelligent test scene;

2. The method for evaluating the intelligence level of an automatic driving system according to claim 1, wherein the actual acting amount S in S4 is obtained from the following equations (1) to (7):

G_i＝m_ig (4)

F_ai＝E_ai·M_i·P_i (5)

in equations (1) to (7), A is the start position of the travel path of the vehicle i under test, B is the end position of the travel path of the vehicle i under test, t_AIs the time corresponding to A, t_BAt the time corresponding to B, L is the running of the vehicle iAmount of lagrange in path, x_iIs the measured path longitudinal displacement, y, of the measured vehicle i_iIs the lateral displacement of the measured path of the measured vehicle i,

for the transverse running speed, R, of the vehicle i under test along the measured path_iAs a field of resistance, G_iIs a constant force field, F_aiFor the risk forces of the lane or road boundary a on the vehicle i under test, E_aiTo be located at (x)_a,y_a) The potential energy field formed by the lane line or the road boundary a is in (x)_i,y_i) The vector field strength of (d); v_jiPotential energy generated by other road users j to the tested vehicle i, F_jiFor the risk forces, E, produced by the other road users j on the vehicle i under test_jiThe kinetic energy field formed for the rest of the road users j is (x)_i,y_i) The vector field strength of (d); m is_iThe quality of the tested vehicle i is measured; g is gravity acceleration, f is rolling resistance coefficient, i_αIs a gradient, C_DiIs the wind resistance coefficient, W, of the vehicle i under test_iIs the windward area, lambda, of the vehicle i to be tested_iIs a rotating mass conversion coefficient, P, of the vehicle i to be measured_aAs road influencing factor, P, at the lane line a or at the road boundary_iIs an influence factor, P, at the i position of the vehicle to be tested_jIs the road influence factor of the other road users j, D is the lane width, r_aiTo point from lane line a or road boundary to the centroid (x) of the vehicle i under test_i,y_i) Distance vector of, M_iIs the equivalent mass of the vehicle i to be tested, M_jIs the equivalent mass of the other road users j, K is the adjustment coefficient, r_jiIs the mass center (x) of the user j on the other roads_j,y_j) Direction to be measuredVehicle i centroid (x)_i,y_i) Distance vector between, v_jIs the velocity vector, θ, of the remaining road users j_jIs r_jiAnd v_jA is a lane line or a road boundary, b is the number of lane lines or road boundaries, n is the number of other road users, k₁，k₂And k₃Is a constant coefficient, L_T,aIs the type of lane line a or road boundary.

3. The method for evaluating the intelligence level of an automatic driving system according to claim 2, wherein the theoretical minimum amount of action S is obtained by the following formula (16):

in the formula (I), the compound is shown in the specification,

is the longitudinal displacement of the reference path of the tested vehicle i;

4. The method for evaluating the intelligence level of an automatic driving system according to claim 1, wherein the quantitative evaluation in S2 is obtained according to a method comprising:

or

or

5. The method for assessing the intelligence level of an autopilot system according to claim 4 wherein the step of "statistically analyzing each of said quantitative assessments obtained at step S4 based on data" at step S5 specifically comprises the steps of:

y_kmax＝max{y_k1,y_k2,...,y_km} (18)

In the second case, the quantitative evaluation is a minimum y expressed by equation (19) depending on the extreme value of the data_kmin：

y_kmin＝min{y_k1,y_k2,...,y_km} (19)。

6. The method for assessing the intelligence level of an autopilot system according to any one of claims 1 to 5 in which the sets of quantitative assessments at S4 are based on data of (y)_k1,y_k2 ... y_km) Wherein, k is a serial number corresponding to one of the network connection degrees, and m is the number of different network connection vehicle distribution forms under the network connection degree; the method of "statistically analyzing each of the quantitative evaluations by data obtained in S4" in S5 includes: and calculating the fitting of the average value, the standard deviation, the extreme value, the frequency and the frequency distribution characteristic or specific distribution of each group of quantitative evaluation basis data, and presenting the statistical analysis result in a distribution diagram and/or a form.

7. The method for assessing the intelligence level of an autopilot system at different levels of internet connectivity as claimed in claim 5, further comprising, after S4:

8. The utility model provides an evaluation device of autopilot system intelligence level under different networking degree, its characterized in that, evaluation device of intelligence level establishes on being surveyed the car, includes:

the statistical evaluation module is used for storing an evaluation interval for evaluating the intelligent level grade of the tested automatic driving system, acquiring a plurality of groups of quantitative evaluation basis data of the tested automatic driving system under different internet access degrees according to the actual action quantity and the theoretical minimum action quantity obtained by the action quantity operation module, carrying out statistical analysis on each group of quantitative evaluation basis data, and evaluating the intelligent level of the tested automatic driving system according to the evaluation interval to which each statistical analysis result belongs;

the method for acquiring the quantitative evaluation basis data specifically comprises the following steps:

9. The apparatus of claim 8, wherein the action calculation module comprises:

S＝S₀-S₁-S₂。