Simulation traffic flow generation method and system based on markov model

Abstract

The application provides a simulation traffic flow generation method and system based on a Markov model, comprising: collecting traffic flow information in selected map information; classifying the traffic flow information, dividing each mobile object in the traffic flow information into a corresponding category, and obtaining data corresponding to each mobile object; training a Markov model for each mobile object according to the classified category, and obtaining a Markov model corresponding to each category; and generating a simulation environment traffic flow according to the Markov model corresponding to each category and a self-defined traffic flow parameter. The application is based on an algorithm, can automatically and quickly generate a large number of scene libraries, reduces labor costs, improves the accuracy of simulation test results, and defines a new scene in a configuration mode, can self-define the number of each object in the traffic flow according to specific test environment requirements, does not need to collect traffic flow again, and improves the efficiency of simulation testing.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic driving simulation testing, in particular to a simulation traffic flow generation method and system based on a Markov model. BACKGROUND

[0002] An automatic driving system refers to a train operation system with full automation and high centralized control of the work performed by a train driver. Automatic driving simulation testing mainly digitizes the application scenarios of automatic driving in a mathematical modeling manner, establishes a system model as close to the real world as possible, and realizes the testing and verification of the automatic driving system through software simulation. To some extent, simulation testing can replace real vehicle road testing. A difficulty of automatic driving simulation is how to generate a test scenario similar to the actual road traffic flow

[0003] Patent document CN110765605A discloses a traffic flow simulation scene generation method and device. The method includes: based on the natural driving data in the driving process of the ego vehicle, obtaining the relative information between the ego vehicle and the target vehicle, the GPS information of the ego vehicle and the OSM map information; based on the relative information between the ego vehicle and the target vehicle, the GPS information of the ego vehicle and the OSM map information, generating a traffic flow simulation scene. The method and device provided in the embodiment of the present application obtain the relative information between the ego vehicle and the target vehicle, the GPS information of the ego vehicle and the OSM map information based on the natural driving data in the driving process of the ego vehicle; generate a traffic flow simulation scene based on the relative information between the ego vehicle and the target vehicle, the GPS information of the ego vehicle and the OSM map information. The method automatically generates a traffic flow simulation scene through natural driving data.

[0004] The above patent automatically generates a traffic flow simulation scene through natural driving data, which requires a manually driven vehicle to collect a large amount of data, and then generates a traffic flow for playback. However, in the generated simulation scene, each object is fixed and cannot reflect the real scene environment, so that the test result is not accurate. SUMMARY

[0005] In view of the defects in the prior art, the purpose of the present application is to provide a simulation traffic flow generation method and system based on a Markov model.

[0006] According to the simulation traffic flow generation method based on the Markov model provided by the present application, the method comprises the following steps:

[0007] Step S1: collecting traffic flow information in selected map information;

[0008] Step S2: classifying and processing the traffic flow information, dividing each moving object in the traffic flow information into a corresponding category, and obtaining data corresponding to each moving object;

[0009] Step S3: training Markov models respectively according to the data corresponding to each moving object as input according to the classified categories, to obtain Markov models corresponding to each category;

[0010] Step S4: generating a simulation environment traffic flow according to the Markov models corresponding to each category and self-defined traffic flow parameters.

[0011] Preferably, the categories include pedestrians, cars and bicycles in motion.

[0012] The training of Markov models respectively according to the data corresponding to each moving object as input according to the classified categories includes pedestrian Markov models, car Markov models and bicycle Markov models.

[0013] Preferably, the data corresponding to each moving object includes sampling time, grid coordinates of the map position where the object is located and object moving speed.

[0014] Preferably, the step S3 includes:

[0015] Step S3.1: inputting the object data into the corresponding model for training;

[0016] Step S3.2: establishing an object state collection at t time points according to the object state value S t =(W t ,V t ) and the traffic light state value L t =(R t ,G t ), denoted as {S1, S2, S3...S t}, wherein W t and V t represent the position coordinates and speed vector of the pedestrian respectively, R t represents a red light and G t represents a green light.

[0017] Step S3.3: calculating the state transition probability at the current time point according to the object state collection, to obtain the overall transition probability corresponding to the object, and the calculation formula is as follows:

[0018] p(S t ,S t-1 ,S t-2 ,…,S0)=p(S t |S t-1 )×p(S t-1 |S t-2 )×…×p(S0)

[0019] p(S t |S t-1) = N(S t-1 ,S t ) / N(S t-1 )

[0020] wherein p(S t ,S t-1 ,S t-2 ,…,S0) represents the current state probability, N represents the number of occurrences of each state; N(S t-1 ,S t ) represents the number of occurrences of S t-1 ,S t at the same time within the entire information statistics; and N(S t-1 ) represents the number of occurrences of S t-1 within the entire information statistics.

[0021] Preferably, step S4 comprises:

[0022] Step S4.1: selecting a map and configuration information, the configuration information comprising the number of objects;

[0023] Step S4.2: initializing the initial state of the objects, the initial state comprising the position coordinates and velocity vectors of the objects;

[0024] Step S4.3: generating the sequence state of the objects corresponding to the Markov model corresponding to the object category;

[0025] Steps S4.2 to S4.3 are repeatedly executed until the number of objects matched with the configuration information is generated.

[0026] According to the simulation traffic flow generation system based on the Markov model provided by the application, the system comprises:

[0027] Module M1: collecting traffic flow information in the selected map information;

[0028] Module M2: classifying and processing the traffic flow information, dividing each mobile object in the traffic flow information into a corresponding category, and obtaining the data corresponding to each mobile object;

[0029] Module M3: training a Markov model for each mobile object according to the classified category, and obtaining the Markov model corresponding to each category;

[0030] Module M4: generating a simulation environment traffic flow according to the Markov model corresponding to each category and a self-defined traffic flow parameter.

[0031] Preferably, the categories comprise moving pedestrians, cars and bicycles.

[0032] The Markov model is respectively trained by taking the data corresponding to each moving object as input according to a classification category, including a pedestrian Markov model, a car Markov model and a bicycle Markov model.

[0033] Preferably, the data corresponding to each moving object includes a sampling time, a grid coordinate of a map position where the object is located and a moving speed of the object.

[0034] Preferably, the module M3 includes:

[0035] Module M3.1: inputting the object data into the corresponding model for training;

[0036] Module M3.2: establishing an object state set at t time points according to the object state value S t =(W t ,V t ) and the traffic light state value L t =(R t ,G t ), denoted as {S1, S2, S3...S t}, wherein W t and V t represent the position coordinate and the speed vector of the pedestrian respectively, R t represents 0 or 1 for a red light, and G t represents 0 or 1 for a green light;

[0037] Module M3.3: calculating the state transition probability at the current time according to the object state set, and then obtaining the overall transition rate corresponding to the object, with the calculation formula as follows:

[0038] p(S t ,S t-1 ,S t-2 ,…,S0)=p(S t |S t-1 )×p(S t-1 |S t-2 )×…×p(S0)

[0039] p(S t |S t-1 )=N(S t-1 ,S t ) / N(S t-1 )

[0040] wherein p(S t ,S t-1 ,S t-2 ,…,S0) represents the state probability at the current time, and N represents the number of occurrences of each state; N(S t-1 ,S t ) represents the number of occurrences of St-1 , S t The number of times of simultaneous occurrence; N(S t-1 ) represents the number of times of occurrence of S t-1 in the entire information statistics.

[0041] Preferably, the module M4 comprises:

[0042] Module M4.1: selected map and configuration information, the configuration information comprising the number of objects;

[0043] Module M4.2: initialization of the initial state of the object, the initial state comprising the position coordinates and the velocity vector of the object;

[0044] Module M4.3: generating the corresponding sequence state of the object through the Markov model corresponding to the object category;

[0045] Repeating the modules M4.2 to M4.3 until the number of objects matched with the configuration information is generated.

[0046] Compared with the prior art, the present application has the following beneficial effects:

[0047] 1. The present application is based on software algorithm, automatically and quickly generates a large number of scene libraries without manual editing, reducing the labor cost.

[0048] 2. The present application uses real data as model input, and the output scene is close to the real situation, so that the simulation test is close to the real scene, and the accuracy of the simulation test result is improved.

[0049] 3. The present application uses the configuration method to define new scenes, and can customize the number of each object in the traffic flow according to the specific test environment requirements, without the need to re-collect the traffic flow, improving the efficiency of simulation test. BRIEF DESCRIPTION OF DRAWINGS

[0050] Other features, objects and advantages of the present application will become more apparent from the following detailed description of non-limiting embodiments with reference to the attached drawings:

[0051] Figure 1 The flowchart of the present application. DETAILED DESCRIPTION

[0052] The present application will be described in detail below with reference to specific embodiments. The following embodiments will help those skilled in the art to further understand the present application, but do not limit the present application in any form. It should be noted that for those skilled in the art, without departing from the concept of the present application, a number of changes and improvements can be made. These all belong to the protection scope of the present application.

[0053] Example one

[0054] According to the application, a simulation traffic flow generation method based on a Markov model is provided, as shown in the accompanying drawings, comprising: Figure 1

[0055] Step S1: collecting traffic flow information in selected map information. It should be noted that the traffic flow information is collected in the real world. Specifically, a limited range of physical map is selected, and then a camera is placed in the selected range to record video and obtain traffic information within a period of time, such as setting up a camera to record video at both sides of the road, intersection, etc.

[0056] Step S2: classifying the traffic flow information, dividing each moving object in the traffic flow information into a corresponding category, and obtaining data corresponding to each moving object. The category includes pedestrians, cars and bicycles in motion. The data corresponding to each moving object includes sampling time, grid coordinates of the map position where the object is located, and object moving speed. Specifically, for the traffic information recorded in the video, a corresponding data table is output for each moving object. For example, the moving object is a pedestrian, and the traffic information contains 4 pedestrians. Therefore, a corresponding data table is output for each pedestrian in the traffic information. The corresponding data table of pedestrian a is shown in Table 1, the corresponding data table of pedestrian b is shown in Table 2, and so on to obtain the corresponding data table of pedestrian c and pedestrian d:

[0057] Table 1

[0058] Time point Grid coordinates of map location Speed 8:15:01 (x a1 ,y a1 )]]> (v xa1 ,v ya1 )]]> 8:15:02 (x a2 ,y a2 )]]> (v xa2 ,v ya2 )]]> …… …… ……

[0059] Table 2

[0060] Time point Grid coordinates of map location Speed 8:15:01 (x b1 ,y b1 )]]> (v xb1 ,v yb1 )]]> 8:15:02 (x b2 ,y b2 )]]> (v xb2 ,v yb2 )]]> …… …… ……

[0061] Wherein, the time is sampled and counted in seconds. The object is a moving person or object such as a pedestrian, a car or a bicycle observed at this moment, which is a basic element of the scene. The position is to grid the map, so that any x, y coordinates will belong to a certain grid. For example, a 100*100 map can be divided into 100 10*10 small grids, each grid has a number from 0 to 99, and any object on the map will belong to a certain grid. The speed is a two-dimensional map speed vector, v x , v y represent the speed components of x and y axes respectively.

[0062] Similarly, for each moving object in the video, that is, each pedestrian and each car outputs a corresponding data table, and one data table represents the moving track of the moving object.

[0063] ​The present application has a variation, which is that the time is sampled and counted in seconds; more variations, which is that the statistics are automatically sampled at intervals.

[0064] Step S3: training the Markov model according to the classification category and the data corresponding to each mobile object as input, obtaining the Markov model corresponding to each category. The Markov model includes a pedestrian Markov model, a car Markov model, and a bicycle Markov model. The step S3 includes:

[0065] Step S3.1: inputting the object data into the corresponding model for training.

[0066] Step S3.2: establishing the object state set at t time according to the object state value S t =(W t ,V t ) and the traffic light state value L t =(R t ,G t ), denoted as {S1, S2, S3...S t}, wherein W t and V t represent the position coordinates and velocity vectors of pedestrians, respectively, R t represents a red light, and G t represents a green light.

[0067] Step S3.3: calculating the state transition probability at the current time according to the object state set, and then obtaining the overall transition probability corresponding to the object, the calculation formula is as follows:

[0068] p(S t ,S t-1 ,S t-2 ,…,S0)=p(S t |S t-1 )×p(S t-1 |S t-2 )×…×p(S0)

[0069] p(S t |S t-1 )=N(S t-1 ,S t ) / N(S t-1 )

[0070] wherein p(S t ,S t-1 ,S t-2 ,…,S0) represents the state probability at the current time, and N represents the number of occurrences of each state; N(S t-1 ,St S represents the entire information statistics. t-1 S t The number of times they occur simultaneously; N(S) t-1 S represents the entire information statistics. t-1 Number of times it appears.

[0071] Specifically, taking the pedestrian Markov model as an example, the method for training the model is described in detail below:

[0072] Input all collected pedestrian data tables into the model for training.

[0073] Define the pedestrian's state value S at time t. t =(W t V t Traffic light status L t =(R t G t Establish a pedestrian state sequence t = 0, 1, ..., t. The set of pedestrian states is {S1, S2, S3...S}. t}, where W t V t L represents the pedestrian's position coordinates and velocity vector, respectively; t =(R t G t The red / green status indicates whether the light is currently red or green. t Indicates that the red light takes either 0 or 1, G t This indicates that the green light is either 0 or 1. For example, R t A value of 0 indicates that the red light is off. t A value of 1 indicates the red light is on, G t A value of 0 indicates that the green light is off. t A value of 1 indicates that the green light is on.

[0074] Markov formula: p(S) t ,S t-1 ,S t-2 ,…,S0)=p(S t |S t-1 )×p(S t-1 |S t-2 )×…×p(S0), where p(S t |S t-1 ), p(S t-1 |S t-2 ...p(S0) can all be obtained through statistical analysis of the collected data tables;

[0075] According to statistical information p(S) t |S t-1 )=N(S t-1 ,St ) / N(S t-1 Let N(S) represent the state transition probability, indicating the probability that the state at the next moment can be obtained given the current state. Here, N represents the number of times each state occurs; N(S) t-1 ,S t S represents the entire information statistics. t-1 ,S t The number of times they occur simultaneously; N(S) t-1 S represents the entire information statistics. t-1 The number of times it appears. And so on, we can obtain p(S) t-1 |S t-2 ), p(S t-2 |S t-3 ...p(S0). And so on, generating models for vehicles and bicycles respectively.

[0076] Step S4: Generate the traffic flow in the simulation environment based on the Markov model corresponding to each type and the user-defined traffic flow parameters. Step S4 includes:

[0077] Step S4.1: Select the map and configuration information, including the number of objects. This number can be customized according to the needs of simulation training.

[0078] Step S4.2: Initialize the initial state of the object, which includes the object's position coordinates and velocity vector.

[0079] Step S4.3: Generate the sequence state corresponding to the object using the Markov model corresponding to the object type.

[0080] Repeat steps S4.2 to S4.3 until the number of objects matching the configuration information is generated.

[0081] This invention combines configuration requirements and randomization to generate new maps and scenes. In other words, it uses configuration to define new scenes, allowing for customization of the number of objects in the traffic flow based on specific test environment requirements, eliminating the need to re-collect traffic flow data and improving simulation testing efficiency. This invention aims to solve the problem of generating a large number of traffic flows for simulation tests. Previous methods involved manually editing traffic flows or collecting and replaying traffic flows from existing scenes. This invention, based on software algorithms, automatically and quickly generates a large library of scenes without manual editing.

[0082] Example 2

[0083] The present invention also provides a simulated traffic flow generation system based on a Markov model. Those skilled in the art can implement the simulated traffic flow generation system based on a Markov model by executing the steps of the simulated traffic flow generation method based on a Markov model. That is, the simulated traffic flow generation method based on a Markov model can be understood as a preferred embodiment of the simulated traffic flow generation system based on a Markov model.

[0084] A traffic flow simulation generation system based on a Markov model, according to the present invention, includes:

[0085] Module M1: Collects traffic flow information from the selected map information.

[0086] Module M2: Classifies the traffic flow information, categorizing each moving object in the traffic flow information into a corresponding type, and acquiring data corresponding to each moving object. The types include moving pedestrians, cars, and bicycles. The data corresponding to each moving object includes the sampling time, the gridded coordinates of the object's location on the map, and the object's moving speed.

[0087] Module M3: Trains a Markov model for each moving object category by using its corresponding data as input, resulting in a Markov model for each category. This training includes pedestrian Markov models, car Markov models, and bicycle Markov models. Module M3 includes:

[0088] Module M3.1: Inputs object data into the corresponding model for training.

[0089] Module M3.2: Based on object state value S t =(W t V t ) and traffic light status value L t =(R t G t Establish a set of object states at time t, denoted as {S1, S2, S3...S}. t}, where W t , and V t R represents the pedestrian's position coordinates and velocity vector, respectively. t Indicates that the red light takes either 0 or 1, G t This indicates that the green light is either 0 or 1.

[0090] Module M3.3: Calculates the current state transition probability based on the set of object states, and then obtains the overall transition rate corresponding to the object. The calculation formula is as follows:

[0091] p(S t,S t-1 ,S t-2 ,…,S0)=p(S t |S t-1 )×p(S t-1 |S t-2 )×…×p(S0)

[0092] p(S t |S t-1 )=N(S t-1 ,S t ) / N(S t-1 )

[0093] Wherein, p(S) t ,S t-1 ,S t-2 S0, ..., S0) represent the probabilities of the current state, and N represents the number of times each state occurs; N(S0, ..., S0) t-1 ,S t S represents the entire information statistics. t-1 S t The number of times they occur simultaneously; N(S) t-1 S represents the entire information statistics. t-1 Number of times it appears.

[0094] Module M4: Generates traffic flow in the simulation environment based on the Markov model corresponding to each type and the user-defined traffic flow parameters. Module M4 includes:

[0095] Module M4.1: Select the map and configuration information, including the number of objects. Module M4.2: Initialize the initial states of the objects, including their position coordinates and velocity vectors. Module M4.3: Generate the sequence states corresponding to the object type using the Markov model corresponding to that object. Modules M4.2 to M4.3 are triggered repeatedly until the number of objects matching the configuration information is generated.

[0096] Those skilled in the art will understand that, in addition to implementing the system, apparatus, and their modules provided by this invention in purely computer-readable program code, the same program can be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system, apparatus, and their modules provided by this invention can be considered a hardware component, and the modules included therein for implementing various programs can also be considered structures within the hardware component; alternatively, modules for implementing various functions can be considered both software programs implementing the method and structures within the hardware component.

[0097] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A method for generating simulated traffic flow based on a Markov model, characterized in that, include: Step S1: Collect traffic flow information from the selected map information; Step S2: Classify the traffic flow information, divide each moving object in the traffic flow information into a corresponding category, and obtain the data corresponding to each moving object; Step S3: Train the Markov model separately according to the data corresponding to each moving object according to the classification category, and obtain the Markov model corresponding to each category; Step S3 includes: Step S3.1: Input the object data into the corresponding model for training; Step S3.2: Based on the object's state value S t = (W t V t ) and traffic light status value L t = (R t G t Establish a set of object states at time t, denoted as . W t , and V t R represents the pedestrian's position coordinates and velocity vector, respectively. t Indicates a red light, G t Represents a green light; define the pedestrian's state value at time t. ; Step S3.3: Calculate the current state transition probability based on the set of object states, and then obtain the overall transition probability corresponding to the object. The calculation formula is as follows: p(S t , S t 1, S t 2, …, S0) = p(S t |S t 1) × p(S t 1|S t 2) × … × p(S0) p(S t |S t 1)=N(S t 1, S t ) / N(S t 1) is the state transition probability, which represents the probability that the state at the next moment can be obtained given the current state; Wherein, p(S) t , S t 1, S t (2, …, S0) represents the probability of the current state, and N represents the number of times each state occurs; N(S t 1, S t S represents the entire information statistics. t 1, S t The number of times they occur simultaneously; N(S) t 1) Represents the entire information statistics within S t The number of times 1 appears; Step S4: Generate the traffic flow of the simulation environment based on the Markov model corresponding to each type and the custom traffic flow parameters; The data corresponding to each moving object includes the sampling time, the gridded coordinates of the object's location on the map, and the object's moving speed; Step S4 includes: Step S4.1: Select the map and configuration information, wherein the configuration information includes the number of objects; Step S4.2: Initialize the initial state of the object, which includes the object's position coordinates and velocity vector; Step S4.3: Generate the sequence state corresponding to the object using the Markov model corresponding to the object type; Repeat steps S4.2 to S4.3 until the number of objects matching the configuration information is generated.

2. The method for generating simulated traffic flow based on a Markov model according to claim 1, characterized in that, The categories include pedestrians, cars, and bicycles in motion; The process involves training Markov models, including pedestrian Markov models, car Markov models, and bicycle Markov models, by using the data corresponding to each moving object as input according to the classification category.

3. A traffic flow simulation generation system based on a Markov model, characterized in that, include: Module M1: Collects traffic flow information from the selected map data; Module M2: Classifies the traffic flow information, categorizes each moving object in the traffic flow information into a corresponding category, and obtains the data corresponding to each moving object; Module M3: Train Markov models by taking the data corresponding to each moving object as input according to the classification category, and obtain the Markov model corresponding to each category; The module M3 includes: Module M3.1: Inputs object data into the corresponding model for training; Module M3.2: Based on object state value S t = (W t V t ) and traffic light status value L t = (R t G t Establish a set of object states at time t, denoted as . W t , and V t R represents the pedestrian's position coordinates and velocity vector, respectively. t Indicates that the red light takes either 0 or 1, G t The green light indicates a value of 0 or 1; the pedestrian's state value at time t is defined. ; Module M3.3: Calculates the current state transition probability based on the set of object states, and then obtains the overall transition rate corresponding to the object. The calculation formula is as follows: p(S t , S t 1, S t 2, …, S0) = p(S t |S t 1) × p(S t 1|S t 2) × … × p(S0) p(S t |S t 1)=N(S t 1, S t ) / N(S t 1) is the state transition probability, which represents the probability that the state at the next moment can be obtained given the current state; Wherein, p(S) t , S t 1, S t (2, …, S0) represents the probability of the current state, and N represents the number of times each state occurs; N(S t 1, S t S represents the entire information statistics. t 1, S t The number of times they occur simultaneously; N(S) t 1) Represents the entire information statistics within S t The number of times 1 appears; Module M4: Generates traffic flow in the simulation environment based on the Markov model corresponding to each type and the user-defined traffic flow parameters; The data corresponding to each moving object includes the sampling time, the gridded coordinates of the object's location on the map, and the object's moving speed; Module M4 includes: Module M4.1: Select map and configuration information, including the number of objects; Module M4.2: Initializes the initial state of the object, which includes the object's position coordinates and velocity vector; Module M4.3: Generates the sequence state corresponding to the object using the Markov model corresponding to the object type; Repeatedly trigger modules M4.2 through M4.3 until the number of objects matching the configuration information is generated.

4. The Markov model-based traffic flow generation system according to claim 3, characterized in that, The categories include pedestrians, cars, and bicycles in motion; The process involves training Markov models, including pedestrian Markov models, car Markov models, and bicycle Markov models, by using the data corresponding to each moving object as input according to the classification category.

Images