An intersection full-link traffic simulation method based on discrete grid structure

Abstract

The application relates to a kind of intersection full-link traffic simulation methods based on discrete grid structure, comprising the following steps: first, build the traffic simulation environment based on deep discrete grid structure, comprising: determining simulation time accuracy and discrete grid accuracy;Intersection space discretization processing;Define intersection space grid attribute;Determine data storage mode;Build vehicle discretization mapping and driving model, traffic individual generation model;Afterwards, establish intersection game decision model and trajectory planning model respectively, the game decision result of game decision model output is used as the prior input and constraint condition of trajectory planning model, and the trajectory planning and execution result output by the trajectory planning model is used as the input of game decision model, to build the full-link simulation process of game interaction decision-trajectory planning. Compared with the prior art, the application can realize high-precision simulation of complex environment and truly reproduce irregular motion of traffic individuals.

Description

Technical Field

[0001] This invention relates to the field of traffic simulation technology, and in particular to a method for simulating the entire traffic flow at intersections based on a discrete grid structure. Background Technology

[0002] In intersection spaces, due to the lack of clear lane rules, vehicle movement exhibits a distinctly non-lane-based and variable trajectory distribution. Simultaneously, various interaction patterns among traffic participants exist, including one-to-one, one-to-many, and even many-to-many interactions. Numerous studies have revealed several key characteristics of traffic flow within intersections: First, the distribution of traffic flow trajectories within the intersection is dynamically influenced by the intersection's geometry, traffic conflicts, and priority rules, resulting in a wide distribution range. Second, decision-making behavior during traffic interactions is affected by dynamic factors such as vehicle spacing, the speed of interacting parties, and the relative speeds of conflicting parties. Third, traffic participants do not necessarily strictly adhere to priority traffic rules, leading to ambiguity in priority. The underlying behavioral mechanisms behind these characteristics are of significant reference value for a deeper understanding of the essence of driving behavior and the achievement of more reliable traffic simulations; however, effective methods for simulating and modeling these mechanisms are currently lacking.

[0003] Traditional microscopic traffic simulation models mainly use lanes as the basic unit, focusing on describing the longitudinal (one-dimensional lane) movement of vehicles. They simplify or even ignore the lateral offset movement of individual traffic vehicles, and mainly focus on macroscopic operational indicators such as intersection capacity, queue length, and delays, neglecting the simulation of the complex microscopic behavioral mechanisms of individual traffic participants. Therefore, existing traffic simulation models are difficult to effectively reproduce the characteristics of real traffic operation.

[0004] Furthermore, the current research focus on intelligent connected environments places higher demands on high-precision and refined environmental background simulation. Virtual simulation testing for autonomous driving also requires microscopic simulation models to innovate and develop towards greater refinement and depth. Under these requirements, traffic simulation models need to be based on a more general underlying structure to realistically reproduce real road traffic environments, thereby effectively providing reliable data feedback for the research and testing phases of autonomous driving. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a full-link traffic simulation method for intersections based on discrete grid structure, which can achieve high-precision simulation of complex environments and realistically reproduce the irregular movement of individual traffic.

[0006] The objective of this invention can be achieved through the following technical solution: a method for simulating the entire traffic flow at intersections based on a discrete grid structure, comprising the following steps:

[0007] S1. Establish a traffic simulation environment based on a deep discrete mesh structure, including:

[0008] Determine the simulation time accuracy and the discrete mesh accuracy;

[0009] Intersection spatial discretization processing;

[0010] Define the grid properties of the intersection space;

[0011] Determine the data storage method;

[0012] Construct a vehicle discretization mapping and driving model, and a traffic individual generation model;

[0013] S2. Establish a crossroads game decision-making model and a trajectory planning model respectively. Use the game decision-making results output by the game decision-making model as the prior input and constraints of the trajectory planning model, and use the trajectory planning and execution results output by the trajectory planning model as the input of the game decision-making model. In this way, a full-link simulation process of game interaction decision-making and trajectory planning is constructed.

[0014] Furthermore, the simulation time accuracy and discrete mesh accuracy in step S1 are specifically as follows:

[0015]

[0016]

[0017] Among them, T s To ensure simulation timing accuracy, r g For discrete grid accuracy, Let b be the minimum speed corresponding to the micro-behavior b of traffic individual i. Let n be the maximum speed of the micro-behavior b corresponding to traffic individual i. i Let z be the maximum number of grid cells that traffic individual i can move within a simulation step. i N represents the maximum number of simulated steps required for traffic individual i to move one unit forward. * It is the set of natural numbers that do not contain 0.

[0018] Furthermore, the specific process of the intersection space discretization in step S1 is as follows:

[0019] A rectangular coordinate system is established with the lower left corner of the intersection's interior space as the origin. Lanes are numbered sequentially from the outermost edge of the approach lane, discretizing the intersection space into several grids of equal side length. The grid coordinates (x, y) range as follows:

[0020]

[0021]

[0022] Among them, lr Where p is the lane width, q is the grid number, and m is the number of lanes;

[0023] The intersection space grid attributes in step S1 include occupancy attributes and vehicle attributes, specifically:

[0024] Occupancy attribute = {Permanent occupancy = -1; Idle = 0; Temporary occupancy = 1}

[0025] Vehicle attributes = {x-axis; y-axis; heading angle ψ; velocity v; acceleration a; front wheel steering angle δ};

[0026] In step S1, the data storage method specifically adopts a three-dimensional hash table structure to store the data.

[0027] Furthermore, the specific process of constructing the vehicle discretization mapping and driving model in step S1 is as follows:

[0028] First, determine the set of grid space occupied by the vehicle. Assume the vehicle's center point coordinates are (x, y), and the angle between the vehicle's front orientation and the x-axis is θ. Then, the set of grid space occupied by the vehicle is the enclosed region Ω formed by connecting the coordinates of the vehicle's four vertices (x1, y1), (x2, y2), (x3, y3), and (x4, y4) in sequence. T :

[0029] (x1,y1)=(x+D·cos(θ+δ),y+D·sin(θ+δ))

[0030] (x2,y2)=(x+D·cos(θ+δ)+L·sin(θ),y+D·sin(θ+δ)-L·cos(θ))

[0031] (x3,y3)=(xD·cos(θ+δ)-L·sin(θ),yD·sin(θ+δ)+L·cos(θ))

[0032] (x4,y4)=(xD·cos(θ+δ),yD·sin(θ+δ))

[0033]

[0034]

[0035] Where H is the vehicle track width, L is the wheelbase, and D is the radius of the area where the vehicle turns;

[0036] Next, a vehicle driving model is established, targeting straight-ahead and left-turning vehicles within the intersection area, where the east-west straight-ahead vehicle path coordinates (x... ew ,y ew ), North-South Path Coordinates (xsn ,y sn Specifically:

[0037] x ew =x0+v·t

[0038] y sn =y0+v·t

[0039]

[0040] Where (x0, y0) represents the vehicle's current position coordinates, v represents the driving speed, t represents the simulation time starting from position (x0, y0), and m represents the simulation time. r The number of lanes for the entrance lane is μ, the lane number where the vehicle is located is W, and the lane width is W.

[0041] The specific movement path of the left-turning vehicle is as follows:

[0042]

[0043]

[0044]

[0045]

[0046] Where (x0,y0) are the coordinates of the starting point of the vehicle's turn, θ0 is the heading angle of the vehicle's starting position, ω is the vehicle's angular velocity when the vehicle turns at a linear velocity v, and t is the simulation time starting from the position (x0,y0).

[0047] Furthermore, the traffic individual generation model in step S1 is specifically as follows:

[0048]

[0049] Among them, t i+1 and t i λ represents the time when the (i+1)th and ith traffic individuals enter the simulation area, respectively; λ represents the average arrival rate of vehicles; and U represents a random number in the interval [0,1].

[0050] Furthermore, step S2 specifically involves establishing a game decision model based on the actual interaction process of left-turning and straight-going vehicles at an intersection by identifying game participants, designing game strategies, payoff functions, optimization objectives, and solution algorithms.

[0051] Furthermore, the game participants are specifically:

[0052]

[0053]

[0054] In the formula, For vehicle K in time period (tt) d ,t+t d The set of grids occupied within ) For vehicle S during time period (tt) d ,t+t d The set of grids occupied within the current simulation time, t is the current simulation time. d For safe headway;

[0055] The game strategy is as follows: the game strategy for left-turning vehicles is trajectory and speed. The trajectory strategy includes passing the nearest vehicle, driving normally, and driving around. The speed strategy includes accelerating, decelerating, and stopping.

[0056] The game strategy for vehicles traveling straight is only a speed strategy, namely, accelerating, decelerating, or stopping;

[0057] The specific profit function is as follows:

[0058]

[0059] α+β+γ=1

[0060]

[0061]

[0062]

[0063]

[0064]

[0065] Where f is the profit, t is the simulation time, j is the vehicle number, ΔT is the collision time difference, ΔV is the velocity difference, Δa is the acceleration difference, α, β, and γ are the weights, Δt is the decision time, and L c Let (x, y) be the distance from the vehicle to the point of conflict, (x, y) be the coordinates of the vehicle's center point, a be the acceleration, and v be the velocity. max This refers to the vehicle's maximum speed.

[0066] The specific optimization objective is as follows:

[0067]

[0068]

[0069] Among them, f K (Y K ,Y S ) and f S (YK ,Y S Y and S are the payoff functions for vehicles K and S respectively, given the acceleration of the vehicles in the game. K Y S The game strategies for vehicle K and vehicle S are respectively. For the Nash equilibrium of the game interaction between vehicle K and vehicle S, a min For the minimum acceleration of the vehicle, a max v is the maximum acceleration of the vehicle. min For the minimum speed of the vehicle, (x) K (t),y K (t) represents the position coordinates of vehicle K at time t, (x) S (t),y S (t) represents the position coordinates of vehicle S at time t, and R is the safe distance between the vehicles in the game.

[0070] The solution algorithm specifically employs a genetic algorithm, and updates the motion state of the game interaction object based on the obtained optimal acceleration.

[0071] Furthermore, the trajectory planning model in step S2 includes a path planning model and a velocity planning model. The path planning model is used to solve a local driving path from the current position to the target position, so that the vehicle can safely and quickly reach the target position. The velocity planning model is specifically a discrete spatiotemporal velocity planning model based on vehicle dynamics velocity constraints, acceleration sampling and velocity limit principles, used to add velocity and acceleration information to each path point.

[0072] Furthermore, the specific working process of the path planning model is as follows:

[0073] First, initialize the Open and Close tables. The Open table only includes the starting point, and the Close table is empty. Then, add the mesh containing the simulated pose to the Open table.

[0074] Take the grid with the smallest F value from the Open table, and calculate the G and H values ​​of the grid where the simulated pose is located to obtain its F value. Here, F value is the total decision cost, G value is the movement cost of the vehicle from the starting point to the specified grid, and H value is the estimated cost of the vehicle from the specified grid to the destination.

[0075] Calculate the mesh corresponding to the simulated pose with the minimum cost function value, and move it to the Close table, which contains the predetermined optimal trajectory points;

[0076] Repeat the above process until the vehicle reaches its destination;

[0077] Specifically, the expansion of the grid containing the simulated pose is centered on the grid coordinates of the vehicle's center point, and its adjacent grids are evenly divided into... The grid is used to determine the range of wheel angle values ​​based on the curvature variation range of the actual left-turning vehicle's trajectory, and the values ​​are discretely taken within each range.

[0078] The formulas for calculating the F, H, and G values ​​are as follows:

[0079] F(n) = G(n) + H(n)

[0080]

[0081]

[0082] if nearby pathθ n ∈[-37°,-14°),

[0083] ifnormal pathθ n ∈[-14°, 14°],

[0084] ifdetour pathθ n ∈(-14°, 37°]

[0085]

[0086]

[0087] Where F(n) is the total cost of the decision, G(n) is the cost of the vehicle moving from the starting point to the specified grid, including the actual distance traveled, the difference in wheel angle, and the wheel angle, α, β, and γ are weight terms, H(n) is the estimated cost of the vehicle moving from the specified grid to the destination, D is the minimum cost of moving from one location to a nearby location, i.e., the radius of the area when the vehicle turns, θ is the wheel angle, (x,y) are the coordinates of the vehicle's center point, and (goal.x,goal.y) are the coordinates of the destination.

[0088] Furthermore, the specific working process of the speed planning model is as follows:

[0089] First, determine the possible set of accelerations:

[0090]

[0091] Among them, a a For the set of positive accelerations, a d For the negative acceleration selection set, v d v represents the desired driving speed, and v represents the current driving speed.

[0092] Then, the set of accelerations is traversed, and the expected driving speed v is calculated for each acceleration. εe and the time t required to reach the conflict point εe With the velocity parameter v εe Predicted location parameters of planned and interacting vehicles:

[0093]

[0094] v εe =v s +a ε ·t εe

[0095]

[0096]

[0097]

[0098]

[0099]

[0100] in, Let ρ be the x-coordinate of the interactive vehicle grid after the planning point in step ρ. Let ρ be the ordinate of the interactive vehicle grid after the planning point, and t be the coordinate of the grid. εe_ρ Let ρ be the travel time required for the vehicle to travel at the given acceleration and initial velocity after the planning point ρ. cur d pre These represent the distances between the vehicle's current location, its predicted location, and the predicted interactive vehicle, respectively.

[0101] Finally, the cost function is set to select the acceleration parameter:

[0102]

[0103]

[0104]

[0105]

[0106]

[0107]

[0108] Where c1 is the rate of change of velocity index, when v eWhen the curvature and maximum longitudinal acceleration constraints are not met, this term is infinite; c2 is the acceleration change rate index, with different weights set according to different driver types. When the longitudinal acceleration does not meet the constraints, this term is infinite; c3 is the distance between the vehicle and the interactive object at the current moment; c4 represents the distance after ρ-step prediction, and the criterion is whether the planned vehicle can safely stop after prediction using the current speed parameters and whether the distance between it and the interactive vehicle is greater than the minimum safe distance d. min If the safe parking requirements are not met, the cost function for this acceleration option is infinite; κ is the curvature, w1, w2, w3, and w4 are the weights of its components, and d safe This is the safe distance threshold.

[0109] Compared with existing technologies, this invention proposes a full-link traffic simulation scheme for intersections based on a deep discrete grid structure. By adopting the idea of ​​discretized grids, the road space and dynamic traffic participants in microscopic analysis are represented by discrete grids, and the occupied grids are assigned corresponding attributes to construct a simulation structure based on a deep discrete grid space. Compared with the traditional one-dimensional description method based on "lanes", this overcomes the shortcomings of traditional traffic simulation and can achieve high-precision simulation of complex environments. It lays the foundation for the simulation structure of two-dimensional traffic simulation based on two-dimensional surface domain motion models and the realization of two-dimensional traffic simulation of mixed traffic scenarios. At the same time, this invention proposes a full-link simulation framework based on game decision-making and trajectory planning, which can realistically reflect and reveal the complex interactive behavior patterns between left-turning vehicles and straight-going vehicles at intersections. Attached Figure Description

[0110] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0111] Figure 2 This is a schematic diagram illustrating the application process of an example.

[0112] Figure 3 This is a diagram of the three-level hash table data structure in this invention;

[0113] Figure 4a and Figure 4b This is a schematic diagram of the vehicle mesh mapping in the embodiment;

[0114] Figure 5 This is a flowchart of the game theory model algorithm in this invention;

[0115] Figure 6 A schematic diagram of potential conflict points within the intersection space;

[0116] Figure 7 This is a flowchart of the path planning algorithm in this invention;

[0117] Figure 8 This is a schematic diagram of path planning node expansion in this invention;

[0118] Figure 9 This is a schematic diagram of the intersection of empirical and simulation verification in the embodiment;

[0119] Figure 10 This is a flowchart of the simulation logic in this invention;

[0120] Figure 11 This is the extracted trajectory space coverage map in the embodiment;

[0121] Figure 12 This is a schematic diagram of the travel time distribution of left-turning vehicles in the simulation verification of the embodiment;

[0122] Figure 13 This is a schematic diagram comparing the post-intrusion time of the simulation results of left-turning vehicles in the embodiment with empirical data;

[0123] Figure 14a and Figure 14b This is a heatmap showing the coverage of left-turning vehicle trajectories in the simulation verification of the embodiment. Detailed Implementation

[0124] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0125] Example

[0126] like Figure 1 As shown, a method for simulating the entire traffic flow at intersections based on a discrete grid structure includes the following steps:

[0127] S1. Establish a traffic simulation environment based on a deep discrete mesh structure, including:

[0128] Determine the simulation time accuracy and the discrete mesh accuracy;

[0129] Intersection spatial discretization processing;

[0130] Define the grid properties of the intersection space;

[0131] Determine the data storage method;

[0132] Construct a vehicle discretization mapping and driving model, and a traffic individual generation model;

[0133] S2. Establish a crossroads game decision-making model and a trajectory planning model respectively. Use the game decision-making results output by the game decision-making model as the prior input and constraints of the trajectory planning model, and use the trajectory planning and execution results output by the trajectory planning model as the input of the game decision-making model. In this way, a full-link simulation process of game interaction decision-making and trajectory planning is constructed.

[0134] This embodiment applies the above technical solution and verifies its effectiveness, such as... Figure 2 As shown, the main process is as follows:

[0135] First, a traffic simulation environment based on a deep discrete mesh structure is built: the principles for determining simulation time accuracy and discrete mesh accuracy, intersection spatial discretization, mesh attribute definition method and data storage method, vehicle discretization mapping and driving model, and traffic individual generation method;

[0136] Based on this, a game-theoretic interaction model between left-turning and straight-going vehicles at an intersection and a trajectory planning model based on discrete grids are established respectively. The organic connection and closed-loop data flow process between the two are constructed to realize the full-link traffic simulation process of game decision-making and trajectory planning. This breaks through the one-dimensional simulation method of existing micro traffic models based on lanes and can effectively reflect the irregular movement of traffic individuals based on two-dimensional surface domains.

[0137] Finally, the full-link traffic simulation method was validated based on actual trajectory data from the intersection.

[0138] The main processes include the following:

[0139] 1) Construct a traffic simulation environment based on a deep discrete mesh structure, including: principles for determining simulation time accuracy and discrete mesh accuracy, intersection spatial discretization, mesh attribute definition methods and data storage methods, vehicle discretization mapping and driving models, and traffic individual generation methods. The specific methods are as follows:

[0140] (i) The principles for determining simulation time accuracy and discrete mesh accuracy are as shown in Equation (1);

[0141] One of the core objectives of this invention is to achieve more realistic simulations of microscopic traffic behavior. Therefore, the accuracy of the discrete mesh and the simulation time accuracy should be determined based on the motion capabilities of the individual traffic entities being analyzed. Since the motion speeds of individual traffic entities vary significantly across different microscopic behaviors, speed is used as the standard for determining the accuracy of the discrete mesh.

[0142]

[0143]

[0144] Among them, T s To ensure simulation timing accuracy, r g For discrete grid accuracy, Let b be the minimum speed corresponding to the micro-behavior b of traffic individual i. Let n be the maximum speed of the micro-behavior b corresponding to traffic individual i. i Let z be the maximum number of grid cells that traffic individual i can move within a simulation step. i N represents the maximum number of simulated steps required for traffic individual i to move one unit forward. *It is the set of natural numbers that do not contain 0.

[0145] (ii) Simulation time precision and mesh precision allow the simulation environment to be divided at entity and time thresholds. The divided mesh units should be assigned different attribute definitions based on different traffic elements. Furthermore, considering the data access efficiency of fine-grained meshes, this invention designs a fast data reading and storage model. The methods for intersection space discretization, attribute definition, and data storage are as follows:

[0146] a. Intersection spatial discretization

[0147] A rectangular coordinate system is established with the lower left corner of the intersection space as the origin. The lanes are numbered sequentially from the outermost side of the approach lane, and the intersection space is discretized into several grids with the same side length. The range of grid coordinates (x,y) is shown in equations (2) to (3):

[0148]

[0149]

[0150] Among them, l r p represents the lane width, q represents the grid number, and m represents the number of lanes.

[0151] b. Attributed mesh

[0152] Road traffic environment information constitutes the basic information of traffic simulation models. Based on different environmental information attributes, it can be divided into static attribute information and dynamic attribute information, each with different impacts on the occupancy status of the grid road network. Static attribute information permanently occupies road grid space, while dynamic attribute information only temporarily changes the grid occupancy status. This invention divides the grid state into three state spaces: permanent occupancy (-1), temporary occupancy (1), and free (0). The grid attributes occupied by motor vehicles also include heading angle, speed, acceleration, front wheel steering angle, and other attribute information, as shown in Table 1. Furthermore, the permanent occupancy status can also be subdivided into specific categories based on different static objects and marked with different numbers to determine the grid occupancy status at any given time. Dynamic traffic individuals such as non-motorized vehicles and pedestrians can also define the occupancy of grids according to their motion attributes.

[0153] Table 1 Mesh Attribute Definitions

[0154] Data layer Attribute value Attribute meaning Occupied Attributes -1，0，1 Permanently occupied, idle, temporarily occupied Vehicle attributes x,y,ψ,v,a,δ x-axis, y-axis, heading angle, velocity, acceleration, front wheel steering angle

[0155] c. Discretized data storage

[0156] Taking an intersection scenario as an example, assuming the internal space of the intersection is 50m*50m, if a grid precision of 0.4m*0.4m is used, a 125*125 matrix will be generated to store road network-related data. If a regular two-dimensional array is used for storage, there will be inefficiency in accessing and manipulating the data. Therefore, this invention uses a hash table to store the discrete intersection data matrix, mapping key attributes to their positions in the table for direct data access, such as... Figure 3 As shown, the three-layer table structure is used to store different attributes required for simulation, namely the bottom-layer road network occupancy attribute, vehicle attribute, and road network static attribute.

[0157] (iii) After determining the road network environment, dynamic traffic element grid occupancy and attribute information need to be loaded into the simulation model. Taking motor vehicles as an example, the vehicle discretization grid mapping method and driving model are explained. It should be noted that other dynamic traffic participants can also be studied in the same way, such as pedestrians and non-motorized vehicles.

[0158] a. Motor vehicle discretization mesh mapping method

[0159] While the actual shape of a motor vehicle is irregular, it can be approximated in practical applications by representing its length and width as a rectangle. Assuming a grid accuracy of 0.4m x 0.4m, a motor vehicle requires multiple grid spaces to represent it, such as... Figure 4a and Figure 4b As shown. The set of grid spaces occupied by motor vehicles is:

[0160] Assuming the vehicle's center point coordinates are (x, y) and the angle between the vehicle's front orientation and the x-axis is θ, then the grid set occupied by the vehicle is the region Ω enclosed by the coordinates of the four vertices of the vehicle as shown in equations (4) to (7). T .

[0161] (x1,y1)=(x+D·cos(θ+δ),y+D·sin(θ+δ)) (4)

[0162] (x2,y2)=(x+D·cos(θ+δ)+L·sin(θ),y+D·sin(θ+δ)-L·cos(θ)) (5)

[0163] (x3,y3)=(xD·cos(θ+δ)-L·sin(θ),yD·sin(θ+δ)+L·cos(θ)) (6)

[0164] (x4,y4)=(xD·cos(θ+δ),yD·sin(θ+δ)) (7)

[0165]

[0166]

[0167] Where (x,y) are the coordinates of the vehicle's center point, θ is the angle between the vehicle's front orientation and the x-axis, H is the vehicle's track width, L is the wheelbase, and D is the radius of the area where the vehicle is turning.

[0168] b. Vehicle driving model

[0169] This invention focuses on the interaction behavior of through-traffic and left-turning vehicles within an intersection area. Therefore, it analyzes the driving trajectory and range using left-turning and through-traffic vehicles as examples. During the simulation, the through-traffic vehicle's path is set to a straight line, so one of its path coordinates remains constant. Taking an through-traffic vehicle from the east entrance as an example, its x-coordinate during its passage through the intersection... ew The vertical coordinate y changes with the driving speed and initial position. sn It remains constant. Therefore, the coordinates of the east-west straight-ahead vehicle path (x...) ew ,y ew ), North-South Path Coordinates (x sn ,y sn The calculation method is shown in equations (10) to (12).

[0170] x ew =x0+v·t (10)

[0171] y sn =y0+v·t (11)

[0172]

[0173] In the formula, (x0, y0) represents the current position coordinates of the vehicle, v represents the driving speed, t represents the simulation time, x and y represent the boundary coordinates of the straight-moving vehicle's driving area, 0 represents the lower boundary, and 1 represents the upper boundary; E, S, W, and N represent the four directions, namely east, south, west, and north; m r μ represents the number of lanes on the approach road, μ represents the lane number where the vehicle is located, and W represents the lane width.

[0174] During the interaction between straight-going and left-turning vehicles, the actual trajectory of a left-turning vehicle is an irregular curve. Considering vehicle motion constraints, this invention selects an Archimedean spiral to represent the motion path of a left-turning vehicle in order to reproduce the actual left-turn trajectory. The left-turn trajectory is considered as a smooth curve y with a certain width. l =h(x l Establish the spatiotemporal trajectory function of left-turning vehicles when they travel within the intersection, and the path equations are shown in equations (13) to (14).

[0175]

[0176]

[0177]

[0178] In the formula, (x0,y0) are the coordinates of the starting point of the vehicle's turn, θ0 is the heading angle of the starting position of the vehicle's turn, and when the vehicle turns at a linear velocity v, ω is the angular velocity of the vehicle passing any position on the grid. The angle traveled by the vehicle in the turning direction is shown in formula (15).

[0179] c. After determining the traffic individual grid mapping and driving model, determine the traffic individual generation model based on actual traffic data. When generating traffic individuals and their information, it is necessary to determine the arrival time intervals between traffic individuals, which is described using a headway distribution model.

[0180] According to the research results of classical traffic flow theory, if the arrival of vehicles follows a Poisson distribution, then the arrival time of vehicles follows a negative exponential distribution, as shown in equation (16).

[0181]

[0182] Where λ is the average arrival rate of vehicles (Veh / s), which can be determined based on actual traffic survey data, and N h Let F(t) represent hourly traffic volume. Since F(t) is a continuously increasing function, the inverse transform method is used to obtain time t. Let F... -1 Let F denote the inverse function of F, and U be a random number on the interval [0,1]. Then we have equation (17).

[0183]

[0184] It can be deduced that for a time-discrete traffic simulation system, the arrival time of the vehicle is as shown in equation (18).

[0185]

[0186] Among them, t i+1 and t i These are the times when the (i+1)th and ith traffic individuals enter the simulation area, respectively.

[0187] 2) Based on the actual interaction process of left-turning and straight-going vehicles at a real intersection, the game participants are identified, game strategies are designed, optimization objectives and solution algorithms are established, and a game decision model is built. The game decision results are used as prior inputs and constraints for the trajectory planning model. The trajectory planning is decoupled into path planning and speed planning, which are optimized and executed separately. The trajectory execution results are used as one of the inputs to the game model. Thus, a full-link simulation process of game interaction decision-making and trajectory planning is constructed, mainly including the following two models:

[0188] (i) Interactive game decision-making model for left-turning and straight-going vehicles at intersections;

[0189] (ii) Trajectory planning model;

[0190] 3) The game-theoretic decision-making model for left-turning and straight-going vehicles at intersections includes the following steps:

[0191] 31) The process of game decision-making algorithm is as follows: Figure 5 As shown, it specifically includes the following steps:

[0192] 311. Determine if there are any potential interaction objects in the vehicle. If not, proceed to step 314; otherwise, proceed to step 312.

[0193] 312. Predict the grid occupied by the expected trajectory of the potential interaction object, and determine whether there is an intersection with the planned trajectory grid of this vehicle. If the two trajectory grids do not intersect, return to 311; if the trajectory grids intersect, determine whether the time for the two vehicles to pass through the intersection is the same. If they are the same, engage in a game interaction with the potential interaction object and proceed to 313; otherwise, return to 312.

[0194] 313. In each cycle, the multi-objective optimization problem is solved using a genetic algorithm to obtain the Nash equilibrium solution. Simultaneously, the motion states of the two players in the game interaction are updated based on the obtained optimal acceleration, and then proceed to 315;

[0195] 314. If either party in the game interaction leaves the conflict zone, the algorithm ends; otherwise, return to 313 to continue the game interaction.

[0196] 315. Compare the payoffs of both sides when left-turning vehicles choose different driving routes (nearest, conventional, detour), select the optimal payoff combination, and generate the optimal game strategy.

[0197] 32) In an intersection with multiple vehicles, considering the oncoming straight-ahead vehicle to which a left-turning vehicle is facing as a potential target, if there is a potential conflict point between the expected trajectories of the vehicles within a safe spatial and temporal range, then interactive behavior is considered to exist between them, such as... Figure 6 As shown.

[0198] From the perspective of left-turning motor vehicles, considering the uncertainties of vehicle size and left-turn trajectory, a conflict zone between left-turning and straight-going vehicles is constructed, and trajectory conflict zone judgment equations are established as shown in equations (19) to (20). If the system of equations has a solution, the current traffic participant is determined to be the interaction object; otherwise, it is considered that there is no interaction between the two.

[0199]

[0200]

[0201] in, For vehicle K in time period (tt) d ,t+t d The set of grids occupied within ) For vehicle S during time period (tt) d ,t+t d The set of grids occupied within the simulation is calculated as shown in equations (4) to (9), where t is the current simulation time. d For safe headway.

[0202] 33) Game Theory Strategies

[0203] After conducting interactive analysis of actual intersection video data, the driving trajectories of left-turning vehicles can be divided into three types: passing the nearest intersection, driving normally, and detouring. Speed ​​changes can be divided into acceleration, deceleration, and stopping. The driving trajectory of straight-going vehicles is approximately a straight line, so the key decision variable for straight-going vehicles is only speed change.

[0204] 34) During the interaction process, the main factors influencing the vehicle's game interaction behavior decision-making are safety, efficiency, and comfort. Therefore, the payoff function of the game participants set in this invention consists of three parts: safety, efficiency, and comfort. The game payoff calculation method is shown in equation (21):

[0205]

[0206] α+β+γ=1 (22)

[0207]

[0208]

[0209]

[0210]

[0211]

[0212] Where f is the profit, t is the simulation time, j is the vehicle number, ΔT is the collision time difference, ΔV is the velocity difference, Δa is the acceleration difference, α, β, and γ are the weights, Δt is the decision time, and L c Let (x, y) be the distance from the vehicle to the point of conflict, (x, y) be the coordinates of the vehicle's center point, a be the acceleration, and v be the velocity. max This is the vehicle's maximum speed.

[0213] 35) In a finite number of repeated games, there is usually at least one multiple Nash equilibrium; the key is to select the best one. This invention transforms the multiple Nash equilibrium problem into a multi-objective optimization problem, solving the Nash equilibrium of the model to maximize the payoff of each player. Seeking the Nash equilibrium of game interactions. Right now The optimization objective and its constraints are shown in equations (28) to (29).

[0214]

[0215]

[0216] Among them, f K (Y K ,Y S ) and f S (Y K ,Y S Y and S are the payoff functions for vehicles K and S respectively, given the acceleration of the vehicles in the game. K Y S The game strategies for vehicle K and vehicle S are respectively. For the Nash equilibrium of the game interaction between vehicle K and vehicle S, a min For the minimum acceleration of the vehicle, a max v is the maximum acceleration of the vehicle. min For the minimum speed of the vehicle, (x) K (t),y K (t) represents the position coordinates of vehicle K at time t, (x) S (t),y S (t) represents the position coordinates of vehicle S at time t, and R represents the safe distance between vehicles in the game.

[0217] 4) The game-theoretic decision-making model provides semantic-level decision instructions to the vehicle, while the trajectory left-turn vehicle motion planning model outputs the specific driving parameters of the vehicle. The decision result from step 3) is used as one of the prior inputs to the trajectory planning model, and trajectory planning is decoupled into path planning and speed planning, which are solved separately. Specifically, this includes the following parts:

[0218] 41) Path planning. The vehicle path planning problem is to find a local travel path from the current location to the target location, enabling the vehicle to reach the target location safely and quickly. The path planning model of this invention mainly includes the following steps:

[0219] 411. Initialize the Open and Close tables. The Open table only includes the starting point, and the Close table is empty. Add the mesh containing the simulated pose to the Open table.

[0220] 412. Take the grid with the smallest F value from the Open table, and calculate the G and H values ​​of the grid where the simulated pose is located to obtain its F value.

[0221] 413. Calculate the mesh corresponding to the simulated pose with the minimum cost function value, and move it to the Close table, which contains the predetermined optimal trajectory points.

[0222] 414. Repeat the above process until the vehicle reaches the destination. If the Open table becomes empty before reaching the destination, it means there is no reachable path between the starting position and the destination position. The algorithm flow is as follows: Figure 7 As shown.

[0223] 42) The method for expanding the grid where the simulated pose is located is to divide the adjacent grids evenly into grids centered on the grid coordinates of the vehicle's center point. The grid is used to determine the range of wheel angle values ​​based on the curvature variation range of the actual left-turning vehicle's trajectory, and the values ​​are discretely taken within each range.

[0224] Based on statistical data of left-turning vehicles at actual intersections, when a vehicle turns, the curvature κ exceeds the maximum curvature limit κ. max At that time, it was stipulated that detour behavior was required. Statistical analysis showed that k max The value is 0.05; at this time, the curvature of the bypass trajectory ranges from κ. d ∈(k max According to the relationship between curvature, turning radius, and wheel turning angle θ, θ = arcsin(k·L), where L is the vehicle length, when k0 = 0.05, θ = 18°. Considering the actual maximum turning angle of the vehicle's wheels, with the vehicle's direction of travel as the 0° coordinate, the range of the vehicle's turning angle for the detour trajectory is θ. detour The nearest trajectory corresponds to the detour trajectory within the range of (18°, 40°). Therefore, the range of the turning angle for the vehicle on the nearest trajectory is θ. nearby ∈(-18°,-40°], the normal trajectory lies between these two values, and the vehicle turning angle ranges from θ. normal ∈[-18°, 18°], considering the feasible range of the vehicle in the grid space, the feasible wheel angle range of the three trajectories is corrected to: the nearest trajectory θ nearby ∈(-14°,-37°], conventional trajectory θ normal ∈[-14°, 14°], orbital trajectory θ detour ∈(14°, 37°). Therefore, the path is constrained by the wheel turning angle. The feasible space diagram of the planned path is shown below. Figure 8 As shown.

[0225] 43) The calculation methods for F, H, and G values ​​are shown in equations (30) to (32):

[0226] F(n)=G(n)+H(n) (30)

[0227]

[0228]

[0229]

[0230]

[0231] Where F(n) is the total cost of the decision, G(n) is the cost of the vehicle moving from the starting point to the specified grid, including the actual distance traveled, the difference in wheel angle, and the wheel angle, and α, β, and γ are weight terms; H(n) is the estimated cost of the vehicle moving from the specified grid to the destination, D is the minimum cost of moving from one location to a nearby location, i.e., the radius of the area when the vehicle turns, θ is the wheel angle, (x,y) are the coordinates of the vehicle's center point, and (goal.x,goal.y) are the coordinates of the destination.

[0232] 44) The task of speed planning is to incorporate speed and acceleration information at each path point along the trajectory planning path, considering lower-level execution constraints and behavioral decision results. This invention proposes a discrete-time speed planning model based on vehicle dynamics speed constraints, acceleration sampling, and speed limit principles. Since the speed planning process for left-turning vehicles is more complex when interactive objects are involved, a detailed description of the left-turning vehicle speed planning method is provided, specifically including the following steps:

[0233] 441. Determine the set of accelerations as shown in equation (35).

[0234]

[0235] In the absence of interactive objects, when the driving speed is less than the desired speed v d When, the acceleration is chosen as a a set, a a All options for concentrated acceleration are positive, indicating that the algorithm expects the planned vehicle to reach the desired speed by using acceleration. When the driving speed equals the desired speed v... d When the acceleration is 0, it indicates that the planned vehicle is expected to pass smoothly at the current speed; when the speed is greater than the expected speed v, the acceleration is zero. d When, the acceleration is chosen as a d set, a d All options for concentrated acceleration are negative, indicating that it is desired to achieve the desired speed for the planned vehicle by employing deceleration actions. Where a... a and a dThe selection range can be obtained by referring to the actual statistical data of the intersection.

[0236] When there are interactive objects, the acceleration set is selected according to different game decision instructions. When the decision instruction is acceleration, the acceleration set is a. a When the decision instruction is deceleration or stopping, the set of possible accelerations is a. d .

[0237] 442. Then, iterate through the set of accelerations and calculate the expected driving speed v for each. εe and the time t required to reach the conflict point εe With the velocity parameter v εe Predict the position parameters of the planned vehicle and the interacting vehicle. First, select an acceleration set based on the game interaction results, and calculate the velocity v at the next path point for each element in this acceleration state set. εe and time t εe ε is the number of acceleration sequences in this set. Using the currently calculated v... εe and t εe After calculating the ρ-step planning point, the travel time t required for the planned vehicle to travel at the given acceleration and initial velocity is calculated. εe_ρ Similarly, calculate the position S of the interactive vehicle. obs_pre The distance d between the vehicle's current position and its predicted position and the predicted interactive vehicle. cur d pre As shown in equations (36) to (42).

[0238]

[0239] v εe =v s +a ε ·t εe (37)

[0240]

[0241]

[0242]

[0243]

[0244]

[0245] In the formula, Let ρ be the x-coordinate of the interactive vehicle grid after the planning point in step ρ. Let ρ be the ordinate of the interactive vehicle grid after the planning point, and t be the coordinate of the grid. εe_ρ Let ρ be the travel time required for the vehicle to travel at the given acceleration and initial velocity after the planning point ρ.cur d pre These represent the distances between the vehicle's current location, its predicted location, and the distances between the vehicles that will interact with it after the prediction.

[0246] 443. Setting the cost function and selecting acceleration parameters. In the autonomous driving speed planning problem, safety and comfort are the two most important factors, with safety taking precedence over comfort. The goal is to find an optimal set of speed control parameters for the planned vehicle. Therefore, the cost function of this invention is C(a i The selection principle is to meet the requirements of diverse driving behaviors while ensuring safety and comfort, as shown in equation (43).

[0247] The current distance between the two workshops is d cur Greater than the safe distance threshold d safe At this point, the planned vehicle and the interacting vehicle are considered to have no potential collision risk; therefore, the cost function for the current driving phase prioritizes comfort and smoothness. The current distance d between the two vehicles... cur If the value is below the safety threshold, it indicates a potential collision risk between the planned vehicle and the interacting vehicle, and safety judgment standards need to be increased to avoid collisions.

[0248]

[0249]

[0250]

[0251]

[0252]

[0253]

[0254] Where c1 represents the rate of change of velocity index, as shown in equation (46), when v e When the curvature and maximum longitudinal acceleration constraints are not met, this term is infinite; c2 represents the acceleration change rate index, as shown in equation (47). This term can be set with different weights according to different driver types. When the longitudinal acceleration does not meet the constraints, this term is infinite; c3 represents the distance between the current time and the interactive object; c4 represents the distance after n steps of prediction. The judgment criterion is whether the planned vehicle can safely stop after prediction using the current speed parameters and whether the distance between it and the interactive vehicle is greater than the minimum safe distance. If the safe stopping requirement is not met, the cost function of this acceleration option is infinite; k is the curvature, and w1, w2, w3, and w4 are the weights of its components, respectively.

[0255] 5) This embodiment uses trajectory data of straight-through and left-turning vehicles at the intersection of Xianxia West Road and Jianhe Road in Shanghai. A simulation environment and game model are established using Python programming software to verify the game-based interactive simulation model based on a discrete grid simulation structure. The intersection diagram is shown below. Figure 9 As shown, the simulation logic is as follows: Figure 10 As shown.

[0256] The verification focused on traffic flow during the period from 16:00 to 17:40 on November 20, 2017. Data was collected by recording vehicle movement via video and extracting trajectory parameters using the high-precision video processing tool George, including global time, position coordinates, vertical and horizontal components of velocity and acceleration, spatial distance, and curvature, among other vehicle-related status information. During the observation period, 210 left-turning vehicle trajectories and 801 straight-going vehicle trajectories were collected. Of the 210 left-turning vehicle trajectories, 43 passed through the intersection independently without interaction with other traffic participants. The spatial coverage map of these trajectories is shown below. Figure 11 As shown, the trajectories of various traffic participants at the intersection do not strictly follow the lanes, but are roughly distributed in a strip-like pattern within the two-dimensional space inside the intersection. This invention takes this intersection as the research object to realize the full-link traffic simulation process of left-turning and straight-going vehicles.

[0257] The simulation mesh accuracy is 0.4m*0.4m, the time accuracy is 0.1s, the vehicle horizontal distance arrival rate λ is set to 0.5, the total simulation time is 4800s, the vehicle size is set to 6m*2m, and the vehicle operation parameters are shown in Table 2.

[0258] Table 2. Set of parameters and values ​​for the steering vehicle simulation model

[0259] Model parameters Parameter values Expected speed <![CDATA[16.7m / s 2 ]]> Expected acceleration during acceleration phase <![CDATA[1m / s 2 ]]> Desired acceleration during deceleration phase <![CDATA[-2m / s 2 ]]> Maximum longitudinal acceleration <![CDATA[3m / s 2 ]]> Maximum lateral acceleration <![CDATA[1.5m / s 2 ]]> Minimum parking distance <![CDATA[2m / s 2 ]]> Expected time interval between front and rear vehicles cs Minimum turning radius 6m

[0260] The present invention selects the similarity of game decision results, the travel time distribution of left-turning vehicles, the distribution of the time of subsequent intrusion, and the spatial coverage of turning trajectory as verification indicators. Among them, the similarity of game decision results can be calculated according to formula (49), the travel time of left-turning vehicles can be calculated according to formula (50), the time of subsequent intrusion can be calculated according to formula (51), and the spatial coverage of trajectory can be calculated according to formula (52).

[0261]

[0262] in, Z represents the number of simulated decisions for a certain type of game; Z represents the actual number of such decisions.

[0263] T D =t e -t s (50)

[0264] Among them, T D t represents the travel time for vehicles turning left. e t is the time when the vehicle leaves the intersection. s The time when the vehicle enters the intersection.

[0265] T in =t l -t f (51)

[0266] Among them, T in For the time after intrusion, t l t is the moment when the previous player leaves the conflict zone. f This refers to the moment when the next player in the game arrives at the conflict zone.

[0267]

[0268] Among them, Q t For the spatial coverage of the turning trajectory, Ω r Ω represents the number of grid cells occupied by the actual trajectory. s The number of grid cells used to generate the trajectory for simulation.

[0269] 51. Conduct an analysis of the interaction behavior between left-turning and straight-going vehicles. There are nine interaction behaviors between left-turning and straight-going vehicles: nearest-deceleration, nearest-acceleration, nearest-stop, detour-deceleration, detour-acceleration, detour-stop, normal-deceleration, normal-acceleration, and normal-stop. In the game theory model, the weights α, β, and γ of the safety index, efficiency index, and comfort index in the payoff function are 0.34, 0.33, and 0.33, respectively. Initially, vehicle 1 turns left from south to west through the intersection with an initial speed v1 = 40 km / h and an initial acceleration a1 = 0 m / s². 2 Vehicle 2 is traveling straight from north to south through the intersection with an initial speed of v2 = 30 km / h and an initial acceleration of a2 = 0 m / s². 2 The initial values ​​for the genetic algorithm are shown in Table 3. The simulation results of the left-turn / straight-going vehicle game decision-making are shown in Table 4. It can be seen that the similarity between the actual data and the selection results of the game interaction model is greater than 75%, indicating that the game model can effectively reproduce the actual behavioral selection results.

[0270] Table 3 Genetic Algorithm Parameter Settings

[0271] Parameter name Parameter values Evolutionary Algebra 10 Population size 20 Crossover probability 0.6 Mutation probability 0.01 Encoded binary string length 6 Chromosome length 12 Acceleration accuracy 0.1

[0272] Table 4 Simulation results of game interaction in conventional vehicles

[0273] frequency Real data Simulation results Similarity (%) Nearest - slow down 4 4 100 Nearest - slow down 10 8 80 Nearest - slow down 1 1 100 Nearest - slow down 85 90 94 Nearest - slow down 10 8 80 Nearest - slow down 13 15 84 Nearest - slow down 4 5 75 Nearest - slow down 32 27 84 Nearest - slow down 8 10 75

[0274] 52. Conduct a travel time distribution analysis. Figure 12 The simulation results show the travel time distribution for left-turning vehicles. The mean and standard deviation of the travel time in the interactive model are 12.8s and 6.6s, respectively, while the actual data are 12.92s and 6.44s. Since the sample size does not show a clear normal distribution, a z-test is used to check the consistency between the two sets of data. The z-test statistic is 0.05, which is less than the 95% confidence threshold of 1.96. Therefore, the simulation results and the actual data are considered to have no significant difference in travel time and are highly similar. Generally, traditional one-dimensional models may have high accuracy in travel time. However, the movement of turning vehicles is simply planned according to a preset path, the trajectory is only distributed within a predefined lane, and the interactive behavior is described too simply. The model proposed in this invention not only has high accuracy in travel time and trajectory distribution but can also simulate safety performance under complex conditions.

[0275] 53. Perform post-intrusion time analysis. Figure 13 This study compares the simulated intrusion time of left-turning vehicles with real-world data. The intrusion time refers to the time difference between the departure of the previous straight-going vehicle (left-turning vehicle) from the conflict zone and the arrival of the next left-turning vehicle (straight-going vehicle), reflecting the time difference between the two vehicles passing through the conflict zone during the interaction. The results show that the mean and standard deviation of the intrusion time in the simulation were 3.92s and 2.18s, respectively, while the real-world values ​​were 4.13s and 2.28s, respectively. The z-test statistic was 0.88. This indicates that the mean and variance of the simulation results are very close to the real values, demonstrating that the model can accurately reproduce the operational patterns between left-turning and straight-going vehicles at a real intersection.

[0276] 54. Conduct trajectory coverage analysis. The method used is to quantitatively evaluate the simulated trajectory. Trajectory coverage is the ratio of the number of grids occupied by both the simulation model and the actual trajectory to the number of grids occupied by the actual trajectory. The simulation results are shown in the figure. If a vehicle trajectory falls within a grid at an intersection, that grid is assigned a value. Figure 14a and Figure 14b The heatmap shows the trajectory coverage; the more times a grid is occupied, the darker its color. Simulation results show that the model's trajectory coverage is 95.45%, indicating that the simulation process can accurately represent most of the real trajectories.

[0277] In summary, this technical solution proposes a full-link traffic simulation scheme for intersections based on a deep discrete mesh structure. On one hand, this invention proposes a method for discretizing road space in microscopic traffic simulation, enabling refined simulation and safety analysis of real traffic scenarios, while providing an accurate simulation environment for future mobile internet and autonomous driving traffic scenarios. On the other hand, based on the underlying structure of this discrete mesh, the proposed full-link traffic simulation process based on a game-theoretic decision-making model and trajectory planning model can more realistically reflect the actual interactive behavior of intersections, thereby enabling the analysis and improvement of the real traffic operating environment. This technical solution breaks through the limitations of traditional microscopic traffic simulation, realizing a new approach to microscopic analysis of traffic systems. It breaks away from the traditional one-dimensional simulation structure based on lanes, realizing the movement process of individual traffic objects with two degrees of freedom, and can more realistically reproduce the irregular movements of individual traffic objects.

Claims

1. A method for intersection all-link traffic simulation based on discrete grid structure, characterized in that, The method comprises the following steps: S1, building a traffic simulation environment based on a deep discrete grid structure, comprising: determining simulation time accuracy and discrete grid accuracy; crossing space discretization processing; defining intersection space grid attributes; determining data storage mode; constructing vehicle discretization mapping and driving model, traffic individual generation model; The simulation time accuracy and discrete grid accuracy are specifically: ， ， wherein is the simulation time precision, is the discrete grid precision, is the traffic individual is the corresponding microscopic behavior is the minimum speed, is the traffic individual is the corresponding microscopic behavior is the maximum speed, is the traffic individual is the maximum number of grids the traffic individual can advance within one simulation step, is the traffic individual is the maximum number of simulation steps the traffic individual needs to advance one grid, is the set of natural numbers without 0; S2, respectively establishing an intersection game decision model and a trajectory planning model, taking the game decision result output by the game decision model as the prior input and constraint condition of the trajectory planning model, taking the trajectory planning and execution result output by the trajectory planning model as the input of the game decision model, and thus constructing a game interaction decision-trajectory planning full-link simulation process.

2. The intersection full-link traffic simulation method based on discrete grid structure according to claim 1, characterized in that, The specific process of the intersection space discretization processing in step S1 is: Establish a rectangular coordinate system with the lower left corner of the intersection's interior space as the origin. Number the lanes sequentially from the outermost edge of the approach lane, discretizing the intersection space into several grids of equal side length. Grid coordinates ( , The range is: ， ， wherein, is the lane width, is the grid number, is the number of lanes; The intersection space grid attributes in step S1 include occupancy attributes and vehicle attributes, specifically: ， Vehicle attributes ; The data storage mode in step S1 is specifically to store data by using a three-dimensional hash table structure.

3. The intersection full-link traffic simulation method based on discrete grid structure according to claim 2, characterized in that, The specific process of constructing the vehicle discretization mapping and driving model in step S1 is: Firstly, the grid space set occupied by the vehicle is determined, assuming that the vehicle center point coordinates are , the vehicle head direction is the angle between the axis and the axis, and the vehicle occupied grid set is the four vertex coordinates of the vehicle , , , . The surrounding areas of the pairs are connected in sequence to form a surrounding area : ， ， ， ， ， ， wherein, is the wheel track of the vehicle, is the wheel base of the vehicle, is the radius of the area when the vehicle is turning; Then the vehicle driving model is established. For the straight vehicles and left-turn vehicles in the intersection range, the east-west straight vehicle path coordinates (x ), and the north-south path coordinates (y ) are specifically as follows: ， ， ， wherein, is the current position coordinate of the vehicle, is the driving speed, is the simulation time from the start position, is the number of lanes of the entrance ramp, is the lane number in which the vehicle is located, is the lane width; The left-turn vehicle motion path is specifically: ， ， ， ， wherein, is the coordinate of the start point of the vehicle turn, is the heading angle of the start position of the vehicle turn, when the vehicle is moving with a linear velocity while turning, is the angular velocity of the vehicle, is the simulation time from the start position.

4. The intersection full-link traffic simulation method based on discrete grid structure according to claim 3, characterized in that, The traffic individual generation model in step S1 is specifically: ， wherein, and are the times of the first and the second traffic individual entering the simulation area, is the average arrival rate of vehicles, is a random number on the interval .

5. The intersection full-link traffic simulation method based on discrete grid structure according to claim 1, characterized in that, Step S2 is specifically to establish a game decision model according to the real intersection left-turn-straight vehicle interaction process by determining game participants, designing game strategies, revenue functions, optimization objectives and solving algorithms.

6. The intersection full-link traffic simulation method based on discrete grid structure according to claim 5, characterized in that, The game participants are specifically: ， ， wherein is a vehicle occupies a grid cell during a time period, is a vehicle occupies a grid cell during a time period, is a current simulation time instant, is a safety headway. The game strategies are specifically: the left-turn vehicle game strategy is trajectory and speed, the trajectory strategy includes near crossing, regular driving, and round driving, and the speed strategy includes acceleration, deceleration, and parking; The straight vehicle game strategy is only the speed strategy, i.e. acceleration, deceleration, and parking; The revenue function is specifically: ， ， ， ， ， ， ， wherein, is the benefit, is the simulation time, is the vehicle number, is the collision time difference, is the speed difference, is the acceleration difference, is the weight, is the decision time, is the vehicle distance to the conflict point, is the vehicle center point coordinate, is the acceleration, is the speed, is the vehicle maximum driving speed; The optimization objective is specifically: ， ， wherein, and are the payoff functions of the vehicle and the vehicle respectively under given game vehicle acceleration, , are the game strategies of the vehicle and the vehicle respectively, is the Nash equilibrium of the game interaction of the vehicle and the vehicle , is the minimum acceleration of the vehicle, is the maximum acceleration of the vehicle, is the minimum driving speed of the vehicle, is the position coordinate of the vehicle K at the time t , is the position coordinate of the vehicle S at the time t , is the safety distance between the game vehicles. The solving algorithm specifically adopts a genetic algorithm, and updates the motion state of the game interaction object according to the obtained optimal acceleration.

7. The intersection full-link traffic simulation method based on discrete grid structure according to claim 6, characterized in that, The trajectory planning model in step S2 includes a path planning model and a speed planning model, the path planning model is used to solve a local driving path from the current position to the target position, so that the vehicle can safely and quickly reach the target position; the speed planning model is specifically a discrete space-time speed planning model based on vehicle dynamics speed constraint, acceleration sampling and speed limit principle, which is used to add speed and acceleration information to each path point.

8. The intersection full-link traffic simulation method based on discrete grid structure according to claim 7, characterized in that, The specific working process of the path planning model is: First, initialize the Open table and the Close table, wherein the Open table only includes the starting point, the Close table is empty, and the grid where the simulation pose is located is added to the Open table; Take the grid with the minimum F value from the Open table, and calculate the G value and H value of the grid where the simulation pose is located to obtain its F value, wherein the F value is the total decision cost, the G value is the moving cost of the vehicle from the starting point to the specified grid, and the H value is the estimated cost of the vehicle moving from the specified grid to the terminal point; Calculate the grid corresponding to the simulation pose with the minimum cost function value, and move it to the Close table, which is the determined optimal trajectory point in the Close table; Repeat the above process until the vehicle reaches the end point; The extension of the grid where the simulated pose is located is specifically to uniformly split the adjacent grids of the grid where the center point of the vehicle is located as the center into According to the range of the curvature change of the actual left-turn vehicle driving track, the wheel rotation angle value range is determined, and the wheel rotation angle is discretely taken in each range. The calculation formula of the F, H, G values is: ， ， ， ， ， wherein, is the total cost of the decision, is the movement cost of the vehicle from the start point to the specified grid, including the actual distance term, the wheel angle difference term and the wheel angle term, , , is the weight term, is the estimated cost of the vehicle from the specified grid to the end point, is the minimum cost of moving from one location to the adjacent location, i.e. the area radius when the vehicle turns, is the wheel angle, is the vehicle center point coordinate, is the end point coordinate.

9. The intersection full-link traffic simulation method based on discrete grid structure according to claim 8, characterized in that, The specific working process of the speed planning model is: Firstly, the acceleration optional set is determined: ， wherein is a positive acceleration selectable set, is a negative acceleration selectable set, is a desired driving speed, is a current driving speed; After that, the acceleration set is traversed and the expected travel speed and the time needed to reach the conflict point are calculated for each speed parameter The position parameters of the planning vehicle and the interacting vehicle are predicted: ， ， ， ， ， ， ， wherein, is is the lateral grid coordinate of the vehicle after the step of planning the point of interaction, is is the longitudinal grid coordinate of the vehicle after the step of planning the point of interaction, is is the time of travel needed for the vehicle to travel with the acceleration and the initial velocity after the step of planning the point of interaction, , is the distance between the current position of the vehicle and the predicted position after the interaction with the predicted vehicle, respectively. Finally, the cost function is set to select the acceleration parameter: ， ， ， ， ， ， wherein, is a speed change rate index, when does not satisfy the curvature and maximum longitudinal acceleration constraints, the value of is infinite; is an acceleration change rate index, different weights are set according to different driver types, when the longitudinal acceleration does not satisfy the constraint, the value of is infinite; is the distance between the current time and the interactive object; represents the distance after step prediction, the judgment standard is whether the planning vehicle can safely stop after prediction and the distance between the planning vehicle and the interactive vehicle is greater than the minimum safety distance , if the safe stopping requirement is not satisfied, the value of is infinite; is the curvature, , , , are the weights of the components, respectively, is a safety distance threshold.

Images