A data-driven method for generating vehicle economy simulation test scenarios

Abstract

The present invention discloses a data-driven method for generating vehicle economy simulation test scenarios, comprising: dividing the vehicle's driving speed into speed segments based on the vehicle's historical driving data, dividing the speed segments into speed states and defining a state space, and calculating the state transition probability between speed states to obtain a Markov state transition probability matrix for the vehicle's driving speed; simultaneously, semantically describing the speed segments with keywords, selecting scene elements and rationally combining scene elements, and reproducing sub-scenarios; outputting a speed state chain based on the Markov state transition probability matrix, randomly selecting sub-scenarios corresponding to the speed states, converting the speed state chain into a sub-scenario chain, and outputting a complete simulation test scenario. The present invention uses the dynamic transition characteristics of vehicle driving data as constraints to randomly generate simulation test scenarios, thereby improving the test level of vehicle economy simulation tests.

Description

Technical Field

[0001] The present invention relates to the field of vehicle engineering technology, and more particularly, to a method for generating a vehicle economy simulation test scenario based on data-driven. Background Art

[0002] Vehicle economy testing methods typically include laboratory chassis dynamometer testing, road testing, and simulation testing. Compared to road testing, laboratory chassis dynamometer testing is conducted under controlled conditions and has relatively high accuracy and repeatability. Simulation methods significantly save research time and costs because they eliminate the need for on-site experiments. In vehicle economy simulation testing, vehicle energy consumption or emission results are typically derived based on driving conditions. The vehicle's driving conditions reflect the most representative driving characteristics of a particular vehicle model, a certain road section, or a certain region. However, the vehicle energy consumption or emission results derived from driving conditions often differ significantly from the actual situation.

[0003] Virtual simulation testing, as one of the foundational testing technologies for autonomous vehicle testing, has garnered widespread attention in the automotive industry in recent years. The construction of simulation scenarios is a crucial component of simulation testing. In the driving field, scenarios are considered a comprehensive reflection of the driving environment and vehicle behavior within a specific time and space. They describe external road conditions, weather conditions, and traffic conditions, as well as the vehicle's own driving tasks and status information. In autonomous driving test scenario research, some scholars categorize scenario generation into two main approaches: mechanistic modeling and data-driven modeling. Mechanistic modeling focuses on creating scenarios based on theory, while data-driven approaches focus on reproducing scenarios and deriving key scenarios based on data. Compared to driving conditions, simulation test scenarios are an abstraction and generalization of the real world. In particular, data-driven scenario generation can achieve a certain degree of scenario reproducibility. Through appropriate scenario construction and vehicle dynamics modeling, the closeness of vehicle economy simulation testing to real-world conditions can be improved.

[0004] How to extract the real-world vehicle driving patterns and reproduce and synthesize the scenarios is a problem that must be solved in data-driven vehicle economy simulation test scenario generation. Summary of the Invention

[0005] The purpose of this invention is to design and develop a data-driven vehicle economy simulation test scenario generation method, which uses the state transition probability matrix to extract the dynamic transfer characteristics of the vehicle's driving speed and reproduce the sub-scene of the speed segment to achieve scenario synthesis.

[0006] The technical solution provided by the present invention is:

[0007] A data-driven vehicle economy simulation test scenario generation method includes the following steps:

[0008] Step 1: Collect the vehicle's driving speed and divide it into speed segments:

[0009] When v≤1.5km / h, it is the idling segment;

[0010] When v>1.5km / h and a>0.15m / s 2 For accelerated segments;

[0011] When v>1.5km / h and a≤0.15m / s 2 For the deceleration segment;

[0012] When v>1.5km / h and -0.15m / s 2 <a≤0.15m / s 2 When is a uniform speed segment;

[0013] Among them, v is the vehicle speed and a is the vehicle acceleration;

[0014] Step 2: Calculate the state transition probability matrix of the speed segment and perform sub-scene reproduction on the speed segment;

[0015] Among them, the state transition probability matrix is:

[0016]

[0017] Where P is the state transition probability matrix, p ij is the transition probability from state i to state j, where state i and state j refer to any state in n speed states, and i = 1, 2, …, n, j = 1, 2, …, n;

[0018] Step 3: Output the speed state chain based on the state transition probability matrix, randomly select a sub-scenario corresponding to the speed state, and convert the speed state chain into a sub-scenario chain to obtain a complete simulation test scenario.

[0019] Preferably, the step 1 further comprises:

[0020] When the speed segment is less than 5 seconds and the speed segment types before and after the speed segment are the same, the type of the speed segment is kept consistent with both sides.

[0021] Preferably, the step 2 of calculating the state transition probability matrix of the speed segment specifically includes:

[0022] Step 1: Calculate the average speed of each speed segment, define a state based on the average speed of each speed segment at intervals of 5 km / h, and divide the speed states into state 1, state 2, state 3, ..., state n;

[0023] Step 2: All velocity states constitute the state space;

[0024] Step 3: Calculate the transition probability between different speed states:

[0025]

[0026] Where p ij is the transition probability from state i to state j, N ij is the frequency of the segment from state i to state j, where state i and state j refer to any state in n speed states, ∑ j N ij is the frequency of the fragments of all states that state i can transfer to;

[0027] Step 4. Calculate the state transition probability matrix:

[0028]

[0029] Preferably, the sub-scene reproduction of the speed segment includes:

[0030] Semantic description of keywords and selection of scene elements.

[0031] Preferably, the semantic description of the keyword includes sub-scene distance, sub-scene termination condition and scene elements;

[0032] Wherein, the sub-scene distance is the travel distance of the speed segment;

[0033] The sub-scenario termination condition is to terminate according to the sub-scenario termination time.

[0034] Preferably, the sub-scenario termination time satisfies:

[0035]

[0036] Where, L senario is the sub-scene distance, is the average speed of the speed segment, i is the speed state to which the speed segment belongs, and k is the number of the speed segment in the speed segment set under the speed state i.

[0037] Preferably, the scene elements include static scene elements, dynamic scene elements, meteorological environment elements and measured vehicle elements.

[0038] Preferably, the static scene elements include:

[0039] Road types, traffic facilities, geographic information and static obstacles;

[0040] The dynamic scene elements include:

[0041] Dynamic traffic signs and traffic participants;

[0042] The meteorological environment factors include:

[0043] precipitation, light, temperature, humidity, and climate;

[0044] The tested vehicle elements include:

[0045] The initial state of the vehicle, the vehicle's driving goals and behavioral elements.

[0046] Preferably, the outputting of the speed state chain based on the state transition probability matrix specifically includes:

[0047] The initial speed state is selected, and a random number uniformly distributed within (0,1) is generated using the Monte Carlo method to determine the next speed state, thereby obtaining a speed state chain.

[0048] Preferably, the sub-scenario corresponding to the randomly selected speed state specifically includes:

[0049] A speed segment is selected from state 1 with a speed range of [0,5) as the beginning of the speed state chain, and the speed difference between adjacent speed segments does not exceed 0.5 km / h.

[0050] The beneficial effects of the present invention are:

[0051] (1) The present invention designs and develops a data-driven vehicle economy simulation test scenario generation method, which uses the Markov state transition probability matrix to extract the dynamic transition characteristics of the actual vehicle speed, thereby improving the similarity between the simulation test scenario and the real world;

[0052] (2) The data-driven vehicle economy simulation test scenario generation method designed and developed by the present invention has stronger randomness and selectivity compared with the vehicle economy simulation test based on working conditions, making the simulation test results closer to the actual results;

[0053] (3) The present invention designs and develops a data-driven vehicle economy simulation test scenario generation method, which greatly reduces the combination of scenario elements and improves the efficiency of scenario generation by reproducing sub-scenarios of speed segments;

[0054] (4) The present invention designs and develops a data-driven vehicle economy simulation test scenario generation method, which takes the scenario as the information basis and the vehicle model as the material basis, provides a more complete information environment for hardware-in-the-loop and driver-in-the-loop tests, and improves the level of vehicle economy testing. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1This is a flow chart of the data-driven vehicle economy simulation test scenario generation method described in the present invention.

[0056] Figure 2 This is a schematic diagram of vehicle speed correlation when the time scale is 1s according to the present invention.

[0057] Figure 3 This is a schematic diagram of vehicle speed correlation when the time scale is 2s according to the present invention.

[0058] Figure 4 This is a schematic diagram of vehicle speed correlation when the time scale is 5s as described in the present invention.

[0059] Figure 5 This is a schematic diagram of vehicle speed correlation when the time scale is 10s according to the present invention.

[0060] Figure 6 This is a schematic diagram of the speed segment division described in the present invention.

[0061] Figure 7 Schematic diagram of the Markov state transition probability distribution of the present invention.

[0062] Figure 8 Schematic diagram of accelerating sub-scene reproduction in the embodiment of the present invention.

[0063] Figure 9 Schematic diagram of reproducing the deceleration sub-scene in the embodiment of the present invention.

[0064] Figure 10 Schematic diagram of uniform speed sub-scene reproduction in the embodiment of the present invention.

[0065] Figure 11 Schematic diagram of the complete scene synthesis process described in the present invention. DETAILED DESCRIPTION

[0066] The present invention is described in further detail below so that those skilled in the art can implement the invention with reference to the description.

[0067] like Figure 1 As shown, the present invention provides a data-driven vehicle economy simulation test scenario generation method, which divides the vehicle's driving speed into speed segments based on the vehicle's historical driving data, divides the speed segments into speed states and defines the state space, and calculates the state transition probability to obtain the Markov state transition probability matrix of the vehicle's driving speed; semantic description of keywords is performed based on the speed segments, scene elements are selected, and scene elements are reasonably combined to reproduce the sub-scenes; speed state chains are output based on the Markov state transition probability matrix, sub-scenes corresponding to the speed states are randomly selected, the speed state chains are converted into sub-scene chains, and a complete simulation test scenario is output.

[0068] Before describing the working process of the present invention in detail, the Markov property of the vehicle speed is first verified:

[0069] Suppose X(t) is a random process that changes with time t(t∈T). If the state X(t0) at time t0 is known, then the state X(t1) of the process at time t1 has nothing to do with the state before time t0, and is only related to the state X(t0) at time t0. This characteristic is called no aftereffect or Markov property. A random process with no aftereffect or Markov property is called a Markov process.

[0070] Use probability distribution function to describe Markov property, let S be the state space of random process X(t), in X(t i )=x i ,x i ∈S, i=1,2,…,n, for any time t i (t i ∈T,t1<t2<…t n ), there is X(t n ) is equal to the conditional probability distribution function of X(t n-1 )=x n-1 When X(t n ) is as follows:

[0071] P{X(t n )≤x n |X(t1)≤x1,…,X(t n-1 )≤x n-1}=P{X(t n )≤x n |X(t n-1 )≤x n-1},x n ∈S;

[0072] During the vehicle's driving process, the vehicle's current speed is only related to the speed of the previous state, and has nothing to do with the previous historical state. n , we have the following formula:

[0073] P{V(t n )≤v n |V(t1)≤v1,…,V(t n-1 )≤v n-1}=P{V(t n )≤v n |V(t n-1 )≤v n-1},v n ∈S;

[0074] That is, the vehicle driving process can be regarded as a Markov process.

[0075] Use the Pearson correlation coefficient to calculate the correlation between two adjacent states within a certain time interval:

[0076]

[0077] Among them, X is the velocity variable, and Y is the velocity variable after a certain time scale adjacent to X.

[0078] Table 1 Correlation coefficients of vehicle speed at different time scales

[0079]

[0080] like Figure 2-5 As shown in Table 1, it can be proved that in a small time scale (i.e. within 10s), the vehicle driving process can be regarded as a Markov process.

[0081] The present invention specifically comprises the following steps:

[0082] Step 1: Figure 6 As shown, the vehicle's driving speed is collected and divided into speed segments:

[0083] The acceleration is zero and the idle segment is divided into the uniform speed segment and the acceleration is greater than zero and the deceleration segment is divided into the acceleration segment and the acceleration is less than zero. However, since the actual collected vehicle driving condition data will be affected by interference and equipment noise, the acceleration of zero almost does not exist and the vehicle speed will also fluctuate. Therefore, the acceleration threshold is set. When the acceleration is at -0.15m / s 2 Up to 0.15m / s 2 When the acceleration at this point is marked as zero, it is classified as a zero acceleration segment. In order to distinguish between idle segments and non-idle segments, the speed threshold is set to 1.5 km / h, and the speed segments are divided as follows:

[0084] (a) When v≤1.5 km / h, it is the idling segment;

[0085] (b) When v>1.5 km / h and a>0.15 m / s 2 For accelerated segments;

[0086] (c) When v>1.5 km / h and a≤0.15 m / s 2 For the deceleration segment;

[0087] (d) When v>1.5km / h and -0.15m / s 2 <a≤0.15m / s 2 When is a uniform speed segment;

[0088] Where v is the vehicle speed and a is the vehicle acceleration.

[0089] In order to maintain the continuity of the vehicle state, the division results are corrected for speed segments with too short a time length: if the speed segment is less than 5s and the types of the speed segments before and after are the same, the type of the speed segment is kept consistent with both sides;

[0090] Step 2: Figure 7 As shown, the state transition probability matrix of the speed segment is calculated, and the speed segment is reproduced in a sub-scene;

[0091] Calculating the state transition probability matrix of the speed segment specifically includes:

[0092] (1) Divide the speed state:

[0093] Calculate the average speed of each speed segment, define a state based on the average speed of each speed segment at intervals of 5 km / h, and divide the speed states into state 1, state 2, state 3, ..., state n;

[0094] (2) Define the state space:

[0095] All speed states (state 1, state 2, state 3, ..., state n) constitute the state space;

[0096] (3) Calculate the state transition probability:

[0097] The state transition frequency is the maximum likelihood estimate of the state transition probability. Based on the n speed states divided above, the number of speed segments within the speed state and the state relationship between two adjacent segments are counted. The probability is estimated by the frequency, and the transition probability between different states is calculated. The calculation formula is as follows:

[0098]

[0099] Where p ij is the transition probability from state i to state j, N ij is the frequency of the segment from state i to state j, where state i and state j refer to any state in n speed states, ∑ j N ij is the frequency of the fragments of all states that state i can transfer to;

[0100] (4) The one-step transition probability between all speed states constitutes the state transition probability matrix P, and the state transition probability matrix is ​​obtained as follows:

[0101]

[0102] The sub-scene refers to a scene segment reproduced for a certain speed segment, and is the smallest unit for scene synthesis.

[0103] The sub-scene reproduction of the speed segment specifically includes:

[0104] Based on the speed segments, the semantic description of keywords and the selection of scene elements are performed to reproduce the sub-scenes.

[0105] The semantic description of the keyword includes:

[0106] 1. Sub-scene distance:

[0107] Set the travel distance of the speed segment to the sub-scene distance, where the sub-scene distance refers to the length of the road in the sub-scene;

[0108] 2. Sub-scenario termination conditions:

[0109] Set the sub-scenario termination condition to time-based termination, where the sub-scenario termination time is calculated according to the following formula:

[0110]

[0111] Where, L senario is the sub-scene distance, is the average speed of the speed segment, subscript i is the speed state to which the speed segment belongs, and subscript j is the number of the speed segment in the speed segment set under the speed state i.

[0112] 3. Scene elements:

[0113] The scene elements include static scene elements within a certain spatial range, dynamic scene elements within a certain time and space range, traffic participant elements, meteorological environment elements and measured vehicle elements.

[0114] Among them, static scene elements include road types, traffic facilities, geographic information and static obstacles. Static scene elements are selected and their attributes are set. The attribute settings include the size, position, quantity, etc. of the elements.

[0115] Dynamic scene elements include dynamic traffic signs and traffic participant elements. Taking traffic lights as an example, the attributes of dynamic traffic signs are set to the size, position and change frequency of traffic lights; traffic participant elements include motor vehicles, non-motor vehicles, pedestrians and animals; traffic participant elements are selected and their attributes are set, and their attributes include size, position, distance / direction from the vehicle, speed (size / direction), acceleration, motion trajectory, etc.

[0116] Meteorological environmental factors include light, temperature, humidity, climate, etc.

[0117] The elements of the vehicle under test include the initial state of the vehicle under test, the vehicle's driving target and behavior elements.

[0118] The details are shown in Table 2:

[0119] Table 2 Scene elements

[0120]

[0121]

[0122] The selected scene elements are static scene elements, dynamic scene elements, traffic participant elements, meteorological environment elements and measured vehicle elements, and the above scene elements are combined.

[0123] Step 3: Output the speed state chain based on the Markov state transition probability matrix, randomly select the sub-scenario corresponding to the speed state, convert the speed state chain into a sub-scenario chain, and output the complete simulation test scenario:

[0124] (1) Output speed status chain:

[0125] Select the initial speed state, use Monte Carlo simulation to generate uniformly distributed random numbers that conform to the state transition probability matrix, and determine the next speed state. From the calculated state transition probability matrix, it can be seen that the state transition probability of state i to state 1, 2, 3, ..., n is (p i1 p i2 … p in ), the cumulative probability value is 1, and the interval (0,1) is divided into n intervals according to the probability value. The Monte Carlo method is used to generate a random number uniformly distributed in (0,1). The interval into which the random number falls corresponds to the speed state at the next moment, and all subsequent speed states are generated in this way, and a speed state chain of the required length is output.

[0126] (2) Randomly extract the speed segments in each speed state:

[0127] Select a speed segment from the lower speed state as the beginning of the speed state chain, and the speed difference between the adjacent segments should not exceed 0.5 km / h;

[0128] Preferably, a speed segment is selected from state 1 with a speed range of [0,5) as the beginning of the speed state chain.

[0129] (3) Randomly extract sub-scenes under each speed segment.

[0130] (4) The speed state chain is converted into a sub-scene chain, while meeting the continuity requirements between sub-scenes, and outputting the synthesized complete scene.

[0131] Example

[0132] The vehicle's driving speed is collected and divided into speed segments. The average speed of each speed segment is defined as a state at intervals of 5 km / h. The speed states are divided into state 1, state 2, state 3, ..., state 12. The range of the average speed of the speed segment in state 1 is [0,5), the range of the average speed of the speed segment in state 2 is [5,10), and so on. The range of the average speed of the speed segment in state 12 is [55,60).

[0133] All speed states (state 1, state 2, state 3, ..., state 12) constitute a state space.

[0134] Based on the 12 speed states divided above, the number of speed segments within the speed state and the state relationship between two adjacent segments are counted, the probability is estimated by frequency, and the transition probability between different states is calculated. The one-step transition probability between all speed states constitutes the state transition probability matrix P, and the state transition probability matrix is ​​obtained:

[0135]

[0136] The semantic description of keywords based on speed segments can be as follows: "The sub-scenario distance is 80 meters, and the scenario end time is 10 seconds. On a two-lane straight road, there is a 50 km / h speed limit sign on the right side of the lane. A small passenger car is 20 meters in front of the vehicle under test and is traveling at a constant speed of 30 km / h. The vehicle under test is in the left lane, with an initial speed of 30 km / h, and accelerates to the right to change lanes, ensuring the speed is below the 50 km / h speed limit."

[0137] like Figure 8 As shown in the figure, ① static scene elements are selected based on the semantic description of the keyword: a three-lane straight road (80m); a speed limit sign (50km / h). ② dynamic scene elements are selected based on the semantic description of the keyword: a passenger car (<3.5m). ③ auxiliary scene elements are selected: a green belt; a curb; and a building. ④ scene elements are combined based on the initial position of the test vehicle: the speed limit sign is placed on the right side of the lane, 30m away from the test vehicle; the passenger car is placed 20 meters in front of the test vehicle, traveling at a constant speed of 30km / h.

[0138] The keyword semantic description of a deceleration sub-scenario might be: "The sub-scenario distance is 100 meters, and the scenario end time is 15 seconds. On a two-lane straight road, there is a crosswalk 100 meters away from the vehicle under test. A 1.75-meter-tall adult male crosses the crosswalk at a speed of 1.5 meters per second when the vehicle under test is 20 meters away from the crosswalk. The vehicle under test is located in the right lane and travels at a constant speed of 36 kilometers per hour from its initial position. When it is 10 meters away from the crosswalk, it decelerates to 18 kilometers per hour until it reaches zero speed before the crosswalk."

[0139] like Figure 9 As shown in the figure, ① static scene elements are selected based on the semantic description of the keyword: a two-lane straight road (100m); a pedestrian crossing traffic marking. ② Dynamic scene elements are selected based on the semantic description of the keyword: pedestrians (≥1.2m). ③ Auxiliary scene elements are selected: green belts; curbs; buildings. ④ Scene elements are combined based on the initial position of the tested vehicle: the pedestrian crossing traffic marking is placed 100m away from the tested vehicle; the pedestrian is placed to the right of the tested vehicle's direction of travel. The behavior is triggered by an event, and the event is when the tested vehicle moves to 10m from the crosswalk.

[0140] The keyword semantic description of a uniform speed sub-scenario might be "The sub-scenario distance is 100 meters, and the scenario end time is 5 seconds. The vehicle under test travels at a uniform speed of 45 km / h from its initial position on a single-lane straight road, without interfering with traffic participants in any direction."

[0141] like Figure 10 As shown in the figure, ① static scene elements are selected based on the semantic description of the keyword: a single lane straight road (100m). ② dynamic scene elements are selected based on the semantic description of the keyword: none. ③ auxiliary scene elements are selected: green belt, curb, building. ④ scene elements are combined based on the initial position of the vehicle under test as the reference position.

[0142] Based on the Markov state transition probability matrix, the speed state chain is output, and the sub-scenario corresponding to the speed state is randomly selected to transform the speed state chain into a sub-scenario chain.

[0143] like Figure 11 As shown, the complete synthesized scene is output.

[0144] The present invention designs and develops a data-driven vehicle economy simulation test scenario generation method, which uses the Markov state transition probability matrix to extract the dynamic transfer characteristics of the actual vehicle speed, thereby improving the similarity between the simulation test scenario and the real world; compared with the vehicle economy simulation test based on working conditions, the randomness and selectivity are stronger, making the simulation test results closer to the actual results; the sub-scenario reproduction method for speed fragments greatly reduces the combination of scenario elements and improves the efficiency of scenario generation; with the scenario as the information basis and the vehicle model as the material basis, a more complete information environment is provided for hardware-in-the-loop and driver-in-the-loop tests, thereby improving the level of vehicle economy testing.

[0145] Although the embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the description and implementation methods. They can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and embodiments shown and described herein.

Claims

1. A data-driven vehicle economy simulation test scenario generation method, characterized in that: The steps include: Step 1: Collect the vehicle's driving speed and divide it into speed segments: When v≤1.5km / h, it is the idling segment; When v>1.5km / h and a>0.15m / s 2 For accelerated segments; When v>1.5km / h and a≤0.15m / s 2 For the deceleration segment; When v>1.5km / h and -0.15m / s 2 <a≤0.15m / s 2 When is a uniform speed segment; Among them, v is the vehicle speed and a is the vehicle acceleration; Step 2: Calculate the state transition probability matrix of the speed segment and perform sub-scene reproduction on the speed segment; Among them, the state transition probability matrix is: Where P is the state transition probability matrix, p ij is the transition probability from state i to state j, where state i and state j refer to any state in the n speed states, and i = 1, 2, ···, n, j = 1, 2, ···, n. The speed state refers to the average speed of each speed segment divided into state 1, state 2, state 3, …, state n at intervals of 5 km / h. Step 3: Output a speed state chain based on the state transition probability matrix, randomly select a sub-scenario corresponding to the speed state, and convert the speed state chain into a sub-scenario chain to obtain a complete simulation test scenario; Outputting a speed state chain based on the state transition probability matrix specifically includes: Select the initial speed state, generate a random number uniformly distributed in (0,1) using the Monte Carlo method, determine the next speed state, and thus obtain the speed state chain; The sub-scene refers to a scene segment reproduced for a certain speed segment, and is the smallest unit for scene synthesis.

2. The data-driven vehicle economy simulation test scenario generation method according to claim 1, characterized in that: The step one further comprises: When the speed segment is less than 5 seconds and the speed segment types before and after the speed segment are the same, the type of the speed segment is kept consistent with both sides.

3. The data-driven vehicle economy simulation test scenario generation method according to claim 2, characterized in that: Calculating the state transition probability matrix of the speed segment in step 2 specifically includes: Step 1: Calculate the average speed of each speed segment, define a state based on the average speed of each speed segment at intervals of 5 km / h, and divide the speed states into state 1, state 2, state 3, ..., state n; Step 2: All velocity states constitute the state space; Step 3: Calculate the transition probability between different speed states: Where p ij is the transition probability from state i to state j, N ij is the frequency of the segment from state i to state j, where state i and state j refer to any state in n speed states, ∑ j N ij is the frequency of the fragments of all states that state i can transfer to; Step 4. Calculate the state transition probability matrix:

4. The data-driven vehicle economy simulation test scenario generation method according to claim 3, characterized in that: Performing sub-scene reproduction on the speed segment includes: Semantic description of keywords and selection of scene elements.

5. The data-driven vehicle economy simulation test scenario generation method according to claim 4, characterized in that: The semantic description of the keyword includes sub-scene distance, sub-scene termination condition and scene elements; Wherein, the sub-scene distance is the travel distance of the speed segment; The sub-scenario termination condition is to terminate according to the sub-scenario termination time.

6. The data-driven vehicle economy simulation test scenario generation method according to claim 5, characterized in that: The sub-scenario termination time satisfies: Where, L senario is the sub-scene distance, is the average speed of the speed segment, i is the speed state to which the speed segment belongs, and k is the number of the speed segment in the speed segment set under the speed state i.

7. The data-driven vehicle economy simulation test scenario generation method according to claim 6, characterized in that: The scene elements include static scene elements, dynamic scene elements, meteorological environment elements and measured vehicle elements.

8. The data-driven vehicle economy simulation test scenario generation method according to claim 7, characterized in that: The static scene elements include: Road types, traffic facilities, geographic information and static obstacles; The dynamic scene elements include: Dynamic traffic signs and traffic participants; The meteorological environment factors include: precipitation, light, temperature, humidity, and climate; The tested vehicle elements include: The initial state of the vehicle, the vehicle's driving goals and behavioral elements.

9. The data-driven vehicle economy simulation test scenario generation method according to claim 8, characterized in that: The sub-scenario corresponding to the randomly selected speed state specifically includes: A speed segment is selected from state 1 with a speed range of [0,5) as the beginning of the speed state chain, and the speed difference between adjacent speed segments does not exceed 0.5 km / h.

Images