A vehicle path planning method for intersections in an autonomous driving environment

Abstract

The present invention discloses a vehicle path planning method for an intersection in an autonomous driving environment, and belongs to the field of intelligent transportation and traffic planning. For a typical plane intersection, first, the space of the conflict area inside the intersection is rasterized; secondly, without considering the lane function division of the entrance lane of the intersection, all driving paths inside the intersection are determined according to the combination of the entrance lanes and exit lanes in each direction; thirdly, the grids passed by each path are determined; vehicle status information is collected, and an autonomous driving vehicle path planning model is established with the goal of maximizing the intersection space utilization and minimizing the weighted sum of conflicts. This method does not optimize the moment when the vehicle enters the intersection, but only plans the best driving path of the vehicle inside the intersection according to the vehicle's location and steering. The goal of this method is not to eliminate all conflicts, but to reduce the conflicts of vehicles inside the intersection as much as possible by planning the path while improving the spatial utilization of the intersection.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent transportation and traffic planning, and relates to the field of automatic driving intersection path planning, and more specifically, to a method for intersection vehicle path planning in an automatic driving environment. Background Art

[0002] Autonomous driving technology provides new opportunities for traffic control. In an autonomous driving environment, intersections can abandon the existing signal light control method and use autonomous driving vehicles to communicate and collaborate with each other to pass through intersections. However, when multiple vehicles are cooperating to pass at the same time, the vehicles need to perceive all potential conflicting vehicles around them in real time and avoid conflicts. This real-time perception and control will greatly increase the computing load of the vehicle control center and requires powerful computing power to achieve. How to reduce conflicts between vehicles by planning the driving path inside the intersection before the vehicle enters the intersection, thereby reducing the computing load of the autonomous driving vehicle control system, is an urgent problem to be solved.

[0003] Existing research on intersection management for autonomous driving can be divided into two categories based on whether the actual size of the vehicle is taken into account. One category does not consider the size of the vehicle and treats the vehicle as a particle. For example, the paper "Erasing Lane Changes from Roads: A Design of Future Road Intersections" treats the vehicle as a particle and proposes the concept of All-Direction Turn Lanes (ADTL). The paper "Development and Evaluation of a Cooperative Vehicle Intersection Control Algorithm Under the Connected Vehicles Environment" establishes a Cooperative Vehicle Intersection Control (CVIC) system with the goal of minimizing the overlap of all trajectories. However, the system does not have a lane division function for the entrance lane. The control method of treating vehicles as particles can provide theoretical guidance for intersection control, but it cannot be directly applied in practice.

[0004] The other is to process intersections in a grid manner, taking vehicle size into consideration. The paper "A Multiagent Approach to Autonomous Intersection Management" proposed earlier to process intersections in a grid manner, but the intersection did not use full turn lanes, and each entrance lane was still divided into lane functions. After that, there were many studies on the grid processing of intersections, such as "Cooperative intersection management: A survey", but most of the studies focused on the order of vehicles at intersections, mainly by adjusting the vehicle entry time to avoid conflicts within the intersection, such as First Come First Served (FCFS), Global Optimum GO, and vehicle platoons (platoon-based method). There is a lack of research directly targeting vehicle path planning at grid-processed intersections.

[0005] A large number of existing studies have shown that in a fully automatic driving environment, at an automatic driving intersection without signal control, the efficiency of automatic driving vehicles interweaving with each other is higher. Under the condition of no signal control, conflicts are mainly avoided by discretizing the intersection conflict space and based on conflict points. In the conflict space discretization method, the time when the vehicle enters the intersection is mainly controlled so that the time when the vehicle passes through each grid does not overlap, ensuring driving safety. Even if the driving path of the vehicle inside the intersection is optimized, the goal is to ensure that the time when the vehicle passes through the grid does not overlap. The present invention does not consider controlling the time when the vehicle enters the intersection. It only plans the driving path of the vehicle inside the intersection to achieve the goal of minimizing the overlap of the paths of each vehicle passing through the intersection and maximizing the space utilization of the intersection, so as to reduce the internal conflicts of the intersection as much as possible and improve the traffic efficiency of the intersection. The minimum path overlap is quantified by the minimum number of vehicles passing through each grid, and the maximum space utilization is quantified by the maximum number of vehicles passing through all grids. Summary of the invention

[0006] Technical problem: Most existing studies focus on the order of vehicles passing through intersections, and lack vehicle path planning methods directly for grid-processed intersections. The purpose of this invention is to provide a vehicle path planning method for intersections in an autonomous driving environment, which does not consider the time when the vehicle enters the intersection, but only plans the best driving path of the vehicle inside the intersection based on the vehicle's location and steering. Improve the spatial utilization of intersections and reduce conflicts between vehicles inside intersections.

[0007] Technical solution: To solve the above technical problems, the path planning method for an autonomous driving vehicle at a flat intersection of the present invention comprises the following steps:

[0008] Step 1: Modeling the internal conflict area of ​​the autonomous driving intersection. For typical plane intersections, including cross intersections, T-intersections and Y-intersections. First, the internal area is gridded and divided into several small grids. The side length of the grid is determined according to the width of the import lane. The grid is represented by g, where g∈G, G represents the set of all grids. A rectangular coordinate system is established inside the intersection, and the boundary equation of each grid is determined according to the established rectangular coordinate system, and each grid is numbered.

[0009] Step 2: Determine all driving paths inside the intersection. Number the entry and exit lanes separately. At the autonomous driving intersection, the lane function division of the intersection entrance lane is not considered, that is, the left-turn, straight-ahead and right-turn lanes are not divided, and the vehicle can turn left, go straight or turn right at any entrance lane. According to the combination of the entrance lanes and exit lanes in each direction, all driving paths inside the intersection are determined and numbered, and r is used to represent the path, r∈R, and R represents the set of all paths in each direction; considering the physical size of the vehicle, the boundary of each path is determined, where the turning path is characterized by an elliptical curve or a transition curve, and the straight path is characterized by a point-slope straight line equation; the grid boundary equation determined in step 1 is combined to determine whether the grid is on the path. Indicates whether the grid g passed by vehicle i is on the path r in the import direction o and the exit direction d. It is a 0-1 variable. , it means that vehicle i passes through grid g on path r in the direction of import o and exit d, otherwise

[0010] Step 3: Collect vehicle information before the autonomous vehicle enters the intersection, including the number of vehicles, represented by I; the current entrance direction of vehicle i, represented by o i Represents, where i∈I, o∈O, import lane l i , exit direction d i and exit lane L i information.

[0011] Step 4: With the goal of maximizing intersection space utilization and minimizing conflicts, a vehicle path planning model is established, including the following steps:

[0012] Step 41: The vehicle can only choose one path to pass through the intersection, satisfying the constraints shown in formula (1):

[0013]

[0014] In the formula, Φ i,r is a 0-1 variable, indicating whether vehicle i passes the intersection via path r. i,r =1, it means that vehicle i passes through the intersection from path r. When Φ i,rWhen =0, it means that vehicle i does not pass through the intersection from path r; r represents the path, r∈R, R represents the set of all paths in all directions; i represents the vehicle, I represents the set of vehicle numbers, i∈I.

[0015] After optimizing the route, the vehicle selects the best entrance road using l i ′ represents the best exit path, L i ', calculated by formulas (2)-(3) respectively:

[0016]

[0017]

[0018] In the formula, A r Indicates the import channel number corresponding to path r, B r Indicates the exit number corresponding to path r;

[0019] The grid that the vehicle passes through after optimizing the path is calculated by formula (4):

[0020]

[0021] In the formula, σ i,g is a 0-1 variable, indicating whether vehicle i passes through grid g. i,g =1 means vehicle i has passed through grid g, otherwise it has not passed through; g represents grid, g∈G, G represents the set of all grids; Indicates whether the grid g passed by vehicle i is on the path r in the direction of o import and d export. is a 0-1 variable. , it means that vehicle i passes through grid g on path r in the direction of import o and exit d, otherwise o i represents the import direction of vehicle i, d i represents the exit direction of vehicle i, o∈O, d∈D, O and D represent the import direction and the exit direction set respectively;

[0022] The number of vehicles passing through grid g among I vehicles is calculated by formula (5):

[0023]

[0024] In the formula, Ω g represents the number of vehicles passing through grid g among all vehicles I;

[0025] Step 42: The objective function is the weighted sum of the minimum number of vehicles passing each grid and the maximum number of vehicles passing all grids. The maximum number of vehicles passing all grids is calculated by formula (6):

[0026] Ω=maxg∈G Ω g (6)

[0027] Where Ω represents the maximum number of vehicles in all grids;

[0028] Objective function 1: the number of vehicles passing through each grid is the minimum, that is, to ensure that as few vehicles as possible pass through the same grid, which can ensure that the conflicts between vehicles in the intersection are the minimum, as shown in formula (7):

[0029] minmax g∈G Ω g (7)

[0030] Objective function 2, the number of grids passed by all vehicles is the largest, which can ensure the maximum space utilization of the intersection, as shown in formula (8):

[0031] max∑ i∈I ∑ g∈G σ i,g (8)

[0032] The objective function 1 contains the max function, which is a nonlinear constraint. After linearization, we can obtain formulas (9)-(10):

[0033] minΩ (9)

[0034]

[0035] By introducing the weight coefficient, objective function 1 and objective function 2 are combined to obtain the combined objective function as shown in formula (11):

[0036] min(10·ω·Ω-(1-ω)·∑ i∈I ∑ g∈G σ i,g ) (11)

[0037] In the formula, ω is the weight coefficient, 0≤ω≤1, and the coefficient 10 in formula (11) is the priority coefficient; the number of vehicles passing through each grid is minimized, which can ensure that the conflict between vehicles inside the intersection is minimized; the number of all vehicles passing through the grid is maximized, which can ensure that the space utilization rate of the internal area of ​​the intersection is maximized. Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0038] The present invention only plans the vehicle path with the goal of maximizing the intersection space utilization and minimizing the weighted sum of conflicts based on the location information such as the import lane where the vehicle is located. While avoiding conflicts as much as possible, the spatial utilization of the intersection is improved. After planning the best vehicle driving path, it can be determined on which grids the vehicle has potential conflicts, and some conflicts can be avoided. For some conflicts that may not be avoided, common first-come-first-served, optimization and other methods are used to allocate intersection rights of way to vehicles. The present invention reduces the computational load of the autonomous driving vehicle control system by avoiding conflicts as much as possible. This method has great scalability and can be used in conjunction with the right-of-way allocation model and the lane change model on the road section to optimize the vehicle path in the optimized intersection area. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is a flow chart of the method of the present invention;

[0040] Figure 2 A schematic diagram of the research object of the method of the present invention;

[0041] Figure 3 This is a grid diagram of the left-turn path from the east entrance direction 3 entrance road to the south exit direction 4 exit road;

[0042] Figure 4 A schematic diagram of the center lines of the 16 paths turning left from the east direction. DETAILED DESCRIPTION

[0043] In conjunction with the accompanying drawings and embodiments, the technical solution of the present invention is described in detail as follows:

[0044] Figure 1 is a flow chart of the method of the present invention;

[0045] Figure 2 Schematic diagram of the research object of the method of the present invention. The research objects of the present invention include plane cross intersections, Y-type intersections and T-type intersections, and the number of import lanes and exit lanes in each direction is not necessarily the same. It is necessary to rasterize the internal conflict area according to different types of intersections, and determine the driving path according to the import lanes and exit lanes. The driving path can be determined by commonly used elliptical curves, transition curves, straight lines, etc. By establishing a rectangular coordinate system and determining the grid on the path according to the grid and path equation, since the plane cross intersection is the most representative in urban roads, Figure 2In the description, a two-way four-lane flat cross intersection is used as an example, but the method of the present invention is not limited to this type of intersection. The entrance lanes and exit lanes in the four directions of the two-way four-lane flat cross intersection are numbered 1-4 respectively, and the four directions of east, west, south and north are represented by the English letters E, W, S, and N, that is, the entrance direction set O = {E, W, S, N}, the exit direction set D = {E, W, S, N}, and the directions are numbered as E-1, W-2, S-3, and N-4. The grid side length is consistent with the lane side length. The intersection conflict area is divided into 64 grids, numbered with numbers 1-64, and the numbers correspond to the grid positions as shown in Figure 2 As shown, a rectangular coordinate system is established inside the intersection.

[0046] Figure 3 This is a schematic diagram of the grids that the left-turn path from the east entrance direction 3 entrance road to the south exit direction 4 exit road passes through. In order to ensure the safety of vehicle traffic, the width of the vehicle's driving path is consistent with the lane width. The inner and outer boundaries of the driving path are determined according to the elliptical curve to determine the vehicle's driving path at the intersection. The grids that the path passes through are shown in the gray grids in the figure, and a total of 24 grids are passed through. Similarly, the number of grids passed by all other paths and the number of each grid can be determined.

[0047] Figure 4 The centerline diagram of the 16 paths for turning left from the east direction. Since the entrance road does not need channelization or lane division, each entrance road can turn left. Figure 4 As shown in the figure, among the 4 entrance roads in the east (E-1) direction, each entrance road has a corresponding left turn to the 4 exit roads in the south (S-3) direction, so there are 4*4=16 left turn paths. For the convenience of display, each path is represented by only the center line of the path. Without considering right turns, there are a total of 16*3*4=192 paths for the two-way four-lane intersection.

[0048] Example: The present invention can calculate the optimal path for vehicle travel for typical plane intersections, including cross intersections, T-intersections, and Y-intersections. The embodiment takes the most representative two-way four-lane plane cross intersection as an example for illustration. In step 2, O, D = {E, W, S, N}, where E, W, S, N represent the four directions of east, south, and north, respectively. According to the Highway Line Design Specifications (JTG D20-2016), the width of the entrance and exit lanes in each direction of the intersection is uniformly set to 3.5m. According to step 1, the side length of the grid is determined according to the width of the import lane, and the side length of the grid is 3.5m. The entrance and exit lanes in each direction of the intersection are numbered, such as Figure 2 As shown, according to Figure 2 The entrance and exit lane numbers corresponding to path r can be summarized as shown in Table 1:

[0049] Table 1 Summary of the import and export lane numbers corresponding to the routes

[0050] Path number r 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 <![CDATA[Import lane number A r > 1 1 1 1 2 2 2 2 3 3 3 3 4 4 4 4 <![CDATA[Exit Lane Number B r > 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4

[0051] According to step 2, at a two-way four-lane intersection without considering the functional division of the entrance lanes, there are a total of 4*4*3*4=192 paths, there are 4 entrance directions, each direction has 4 entrance lanes, each lane has 3 turns: left turn, straight go, and right turn, and each turn corresponds to 4 exit lanes. Therefore, without considering right turns, there are a total of 192 paths.

[0052] Take the left turn path from the east entrance 3 to the south exit 4 as an example. The grids of the path are as follows: Figure 3 shown.

[0053] Taking the left turn of each import lane in the east import direction as an example, the center line is used instead of the path, such as Figure 4 As shown in the figure, the 16 paths of left turn are taken as an example, and all the paths passing through the grid are summarized as shown in Table 2. In Table 2, the path with path number 5, import lane number 2, and exit lane number 1 is taken as an example. If vehicle 3 passes through this path, then It indicates that the vehicle 3 passes through the grid 4 on the path 5 in the direction of o entry and d exit; This indicates that the vehicle 3 does not pass through the grid 6 when traveling on the path 5 in the o entrance direction and the d exit direction.

[0054] Table 2 Summary of grids for 16 left-turn paths in the east import direction

[0055]

[0056] According to step 3, the location information of 36 vehicles is randomly generated, I=36, including the import direction o of vehicle i i , Import Lane i , exit direction d i and exit lane L i , as shown in Table 3:

[0057] Table 3 Basic information of 36 vehicles

[0058]

[0059]

[0060] According to step 4, the paths of 36 vehicles are planned, and the weight coefficient ω = 0.8 is taken to obtain the optimal entrance path l for each vehicle. i ′ and the best exit path L i 'As shown in Table 4:

[0061] Table 4 Comparison of the best import lane and best export lane for 36 vehicles before and after optimization when ω = 0.8

[0062]

[0063] At this time, the number of vehicles passing through each grid is shown in Table 5. It can be seen from Table 5 that the maximum number of vehicles passing through the grid is 9.

[0064] Table 5: Statistics of the number of vehicles passing through each grid when ω=0.8

[0065]

[0066] When ω=0.8, the vehicles pass through 540 grids in total, and the maximum number of vehicles passing through a grid is 9. When the driving path of 36 vehicles is not planned, the total number of grids passed is 448, and the maximum number of vehicles passing through a grid is 14, indicating that the intersection utilization rate can be improved by 17% and the path conflict can be reduced by 36%. When ω=0 is adjusted, according to formula (11) in step 4, it can be seen that the goal of minimizing the number of vehicles passing through each grid is not considered, and only the goal of maximizing the intersection space utilization is taken. At this time, the total number of vehicles passing through the grid is 653, and the maximum number of vehicles passing through the grid is 16.

[0067] Take ω = 0, and get the best entrance path l for each vehicle i ′ and the best exit path L i 'As shown in Table 6:

[0068] Table 6 Comparison of the best import lane and best export lane for 36 vehicles before and after optimization when ω=0

[0069]

Claims

1. A method for vehicle path planning at an intersection in an autonomous driving environment, characterized in that: The method comprises the following steps: Step 1: Model the internal conflict area of ​​the autonomous driving intersection. For a typical plane intersection, firstly grid the internal area and divide it into several small grids. The side length of the grid is determined according to the width of the entrance lane. A rectangular coordinate system is established inside the intersection. The boundary equation of each grid is determined according to the established rectangular coordinate system, and each grid is numbered. Step 2: Determine all driving paths inside the intersection; number the entrance and exit lanes respectively. At the autonomous driving intersection, the lane function division of the intersection entrance lane is not considered, that is, the intersection entrance lane is not divided into left-turn, straight-ahead and right-turn lanes, and the vehicle can complete the left turn, straight-ahead or right turn at any entrance lane; determine all driving paths inside the intersection and number them according to the combination of entrance lanes and exit lanes in each direction; determine the inner and outer boundary equations of each path under the condition of considering the physical size of the vehicle, where the inner and outer boundary equations of the turning path are characterized by elliptic curves, and the inner and outer boundary equations of the straight path are characterized by point-slope straight line equations; Combine the grid boundary equations determined in step 1 to determine whether each grid is on the path; Step 3: Before the autonomous driving vehicle enters the intersection, vehicle information is collected, including the number of vehicles, represented by I; The current import direction of vehicle i, indicated by o i Represents, where i∈I, o∈O, import lane l i , export direction d i and exit lane L i information; Step 4: Establish a vehicle path planning model with the goal of maximizing intersection space utilization and minimizing weighted conflicts.

2. The method for planning a vehicle path at an intersection in an autonomous driving environment according to claim 1, characterized in that: The step 4 comprises the following steps: Step 41: The vehicle can only choose one path to pass through the intersection, satisfying the constraints shown in formula (1): In the formula, Φ i，r is a 0-1 variable, indicating whether vehicle i passes the intersection via path r. i，r =1, it means that vehicle i passes through the intersection from path r. When Φ i，r = 0, it means that vehicle i does not pass the intersection from path r; r represents the path, r∈R, R represents the set of all paths in each direction; i represents the vehicle, I represents the number of vehicles, i∈I; After optimizing the route, the vehicle selects the best entrance road using l i ′ represents the best exit path, L i ', calculated by formulas (2)-(3) respectively: In the formula, A r Indicates the import channel number corresponding to path r, B r Indicates the exit number corresponding to path r; The grid that the vehicle passes through after optimizing the path is calculated by formula (4): In the formula, σ i，g is a 0-1 variable, indicating whether vehicle i passes through grid g. i，g =1 means vehicle i has passed through grid g, otherwise it has not passed through; g represents grid, g∈G, G represents the set of all grids; Indicates whether the grid g passed by vehicle i is on the path r in the direction of o import and d export. is a 0-1 variable. , it means that vehicle i passes through grid g on path r in the direction of import o and exit d, otherwise o i represents the import direction of vehicle i, d i represents the exit direction of vehicle i, o∈O, d∈D, O and D represent the import direction and the exit direction set respectively; The number of vehicles passing through grid g among I vehicles is calculated by formula (5): In the formula, Ω g represents the number of vehicles passing through grid g among all vehicles I; Step 42: The objective function is the weighted sum of the minimum number of vehicles passing each grid and the maximum number of vehicles passing all grids. The maximum number of vehicles passing all grids is calculated by formula (6): Ω=max g∈G Oh g (6) Where Ω represents the maximum number of vehicles in all grids; Objective function 1: the number of vehicles passing through each grid is the minimum, that is, to ensure that as few vehicles as possible pass through the same grid, which can ensure that the conflicts between vehicles in the intersection are minimal, as shown in formula (7): minmax g∈G Ω g (7) Objective function 2, the number of grids passed by all vehicles is the largest, which can ensure the maximum space utilization of the intersection, as shown in formula (8): max∑ i∈I ∑ g∈G s i，g (8) The objective function 1 contains the max function, which is a nonlinear constraint. After linearization, we can obtain formulas (9)-(10): minΩ (9) By introducing the weight coefficient, objective function 1 and objective function 2 are combined to obtain the combined objective function as shown in formula (11): min(10·ω·Ω-(1-ω)·∑ i∈I ∑ g∈G s i，g ) (11) Where ω is the weight coefficient, 0≤ω≤1, and the coefficient 10 in formula (11) is the priority coefficient. The minimum number of vehicles passing through each grid can ensure that the conflict between vehicles inside the intersection is minimized; the maximum number of vehicles passing through the grid can ensure that the space utilization rate of the internal area of ​​the intersection is maximized.

3. The method for vehicle path planning at an intersection in an autonomous driving environment according to claim 1, characterized in that: The typical plane intersection includes a cross intersection, a T-shaped intersection or a Y-shaped intersection.

Images