Reference line correction method and device in vehicle-road cooperation environment and storage medium

Abstract

The invention discloses a reference line correction method and device in a vehicle-road cooperation environment and a storage medium, and the method comprises the steps: obtaining a reference driving track set of unmanned equipment based on the high-precision map information and the driving information of the unmanned equipment in a Cartesian coordinate system, carrying out the smoothing of the reference driving track set, and carrying out the correction of a reference line according to a value function, and screening out an optimal reference driving track, determining a reference line loss function form and a to-be-calibrated weight coefficient on the basis of the optimal reference driving track, solving the weight coefficient of the reference line loss function according to the optimal reference driving track, and correcting the reference line.

Description

Technical Field

[0001] The present invention relates to the field of intelligent traffic control, and in particular to a reference line correction method, device and storage medium in a vehicle-road collaborative environment. Background Art

[0002] The reference line is the basis of autonomous driving trajectory planning. By comprehensively considering factors such as safety, driving efficiency, road geometry and lane changes, it provides a feasible and optimized driving path for unmanned driving equipment. A reasonable reference line loss function can ensure that a reference line that ensures the safe and efficient driving of unmanned vehicles is designed in a complex traffic environment, thereby ensuring safe interaction with other road participants and providing passengers with a stable and comfortable driving experience. Current research in the field of intelligent traffic control mainly uses fixed weight coefficients to determine the reference line loss function. This method can ensure the accuracy and stability of the planned trajectory to a certain extent.

[0003] Fixed calibration coefficients are often unable to adapt to real-time changing traffic conditions and road conditions when faced with complex and dynamically changing traffic environments. For example, in the face of sudden traffic jams, obstacles, and uncertain interactions between vehicles, the loss function with fixed coefficients cannot be flexibly adjusted, thus affecting the response speed and driving efficiency of the driverless vehicle. Therefore, more and more studies have begun to explore adaptive reference line correction, by introducing a dynamic adjustment mechanism to optimize the loss function based on real-time traffic data.

[0004] The adaptive loss function can dynamically adjust the weight coefficient according to the changes in the environment around the driverless vehicle, road conditions and the vehicle's own status, making the generated reference line more flexible and adaptable. For example, when the vehicle needs to avoid an emergency obstacle, the safety weight can be automatically increased, while on relatively open roads with less traffic, the weight of driving efficiency may be increased accordingly. Such a design can significantly improve the ability of driverless vehicles to cope with complex environments and further enhance the stability and comfort of the driving experience. Summary of the invention

[0005] Purpose of the invention: In order to overcome the shortcomings of the above-mentioned prior art, the purpose of the present invention is to propose a reference line correction method, device and storage medium in a vehicle-road collaborative environment, based on the high-precision map information in the Cartesian coordinate system and the driving information of the unmanned driving equipment, to obtain the reference driving trajectory set of the unmanned driving equipment and smooth it, and then determine the form of the value function, screen out the optimal reference driving trajectory, and on this basis determine the form of the reference line loss function and its weight coefficient to be calibrated, and solve the weight coefficient of the reference line loss function according to the optimal reference driving trajectory.

[0006] Technical solution, in order to solve the above technical problems, the present invention proposes a reference line correction method in a vehicle-road cooperative environment, the method comprising the following steps:

[0007] (1) Obtaining map information and unmanned driving equipment driving information in a Cartesian coordinate system through roadside sensing equipment;

[0008] (2) Calculate the reference driving trajectory set of the unmanned driving device at time t based on the map information and the driving information of the unmanned driving device;

[0009] (3) Smoothing all reference driving trajectories in the set;

[0010] (4) determining the form of the value function, and selecting the optimal reference driving trajectory from all the reference driving trajectories in step (3) according to the value function;

[0011] (5) Determine the form of the reference line loss function and its weight coefficient to be calibrated, and solve the weight coefficient of the reference line loss function according to the optimal reference driving trajectory;

[0012] (6) Using the solved reference line loss function and the calibrated weight coefficient, the reference line is corrected.

[0013] Furthermore, the specific method of step (1) is as follows:

[0014] When a vehicle passes by on the road, the roadside sensing device will obtain map information and unmanned driving equipment driving information. The map information includes: road width W, road curvature k r , the center line collection point set Line{x l_1 ,y l_1 , x l_2 ,y l_2 , …, x l_j’ ,y l_j’ , …, x l_m ,y l_m}, where (x l_j’ ,y l_j’ ) is the coordinate of the lane centerline collection point, 1≤j'≤m, m is the total number of lane centerline collection points, the road obstacle boundary information set OBS{obs_1,obs_2,…,obs_i,…,obs_q}, 1≤i≤q, q is the total number of obstacles, the boundary of the i-th obstacle obs_i is (x ob1 ,y ob1 , x ob2 ,y ob2 , …, x obj ,y obj , …, x obm’ ,y obm’ ), where (xobj ,y obj ) is the boundary coordinate of the obstacle obs_i collected by the roadside sensing device, 1≤j≤m', m' is the total number of boundary coordinates of the ith obstacle obs_i; the unmanned driving equipment driving information includes the center of mass plane coordinates (x t ,y t ), vehicle path planning end point coordinates (x p ,y p ).

[0015] Furthermore, in step (2), a reference driving trajectory set of the unmanned driving device at time t is calculated based on the map information and the driving information of the unmanned driving device. The specific process is as follows:

[0016] ① The plane coordinates of the vehicle center of mass collected at time t (x t ,y t ) as the initial point of path planning The vehicle path planning end point coordinates (x p ,y p ) as the target point The initial point As the root node of the node tree path_tree_j(t), the collected road width W, road curvature k r , coordinates(x t ,y t ) and coordinates (x p ,y p ) as boundary constraints to generate planning space R;

[0017] ②In the planning space R, randomly obtain a sampling point x rand , with sampling point x rand As the goal, traverse the existing nodes on the node tree, calculate the Euclidean distance from each node to the sampling point, and select the node with the smallest distance as the parent node When there is only the root node in the node tree Directly select this point as the parent node );

[0018] ③From the parent node To a random point x rand Move with a step length of ΔL to get the leaf node x new_1 , judge x new_1 Does it overlap with the range of the obstacle set OBS in the map environment? If so, remove x new_1 , and re-search for a new leaf node x new_1 ; If not, then x new_1 As a parent node Add to the node tree path_tree_j(t) and start looking for leaf node x new_2 ;

[0019] ④ Repeat steps ②-③ until the parent node Expand to target point Within the set neighborhood, the target point Add to the node tree;

[0020] ⑤From the target point Start by searching for the parent node in reverse order until you reach the root node. Get the j-th node tree path_tree_j(t), which consists of the parent node The curve composed of the connecting lines between all adjacent parent nodes in the node tree is the corresponding reference driving trajectory trajectory j(t);

[0021] ⑥ Repeat operations ①-⑤ to obtain a set of n node trees PATHTREE(t){path_tree_1(t), path_tree_2(t),…,path_tree_n(t)}, and the corresponding reference driving trajectory set TRA(t){trajectory 1(t), trajectory 2(t),…,trajectory n(t)}.

[0022] Furthermore, the smoothing method in step (3) is as follows:

[0023] The reference driving trajectory at time t is smoothed using the Bezier curve. The Bezier curve equation is as follows:

[0024]

[0025] Among them, P j (u, t) represents the smoothed reference driving trajectory, q represents the number of parent nodes in the jth node tree, L i j (t) represents the i-th parent node of the j-th node tree, which is also the control point of the Bezier curve. u represents a parameter in the process of generating the parameterized curve, which is used to describe the progress of the curve between the control points, 0≤u≤1.

[0026] Furthermore, in step (4), according to the value function, the optimal reference driving trajectory P is selected from the reference driving trajectory set TRA(t) {trajectory 1(t), trajectory 2(t), ..., trajectory n(t)} at time t. optimal (t), the value function is as follows:

[0027]

[0028] in, is the horizontal coordinate of the kth node in the jth node tree path_tree_j at time t, k r is the road curvature. The trajectory with the smallest value is found through the above formula, which is also the optimal reference driving trajectory.

[0029] Furthermore, in step (5), the reference line loss function at time t is expressed as:

[0030]

[0031] Where n′ is the total number of points on the optimal reference driving trajectory, is the optimal reference driving trajectory P optimal (t) is the coordinate of the kth point on the s ,ω d ,ω t is the coefficient to be calibrated; where ω s The reference line smoothness is taken into account to ensure that the trajectory does not have abrupt direction changes; ω d The reference line offset is taken into account to prevent the reference line from exceeding the planned space; ω t The uniformity of the reference line is taken into account to prevent the vehicle from colliding with obstacles during driving. The minimum value obtained at time t-1 is used as the reference value of the loss function cost at time t. The coefficient of the loss function is calibrated using the least squares method, and the calibrated coefficient is recorded as

[0032] Furthermore, the specific method of step (6) is as follows:

[0033] The initial reference line is set as the center line of the lane where the vehicle is located. Using the reference line loss function and the calibrated coefficients, the collection points on the reference line are gradually adjusted through the gradient descent method until the loss function is minimized. The loss function used to correct the reference line is as follows:

[0034]

[0035] Among them, (x l_k ,y l_k ) is the coordinate of the kth acquisition point on the initial reference line, To calibrate the coefficients, the update rules for the collection points on the reference line are as follows:

[0036]

[0037] Among them, η is the learning rate, which controls the step size of each update. is the updated coordinate point, is the current coordinate point. Through continuous iteration, the coordinates of all reference line collection points are updated, and the value of the loss function is gradually reduced until the change in the loss function is less than the set threshold ε. When the change in the loss function is less than the set threshold ε, the collection points (x l_k ,y l_k ) will be adjusted to a smooth and optimal position, and the smoothed collection points will obtain the corrected reference line.

[0038] In addition, the present invention proposes a reference line correction device in a vehicle-road collaborative environment, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any of the methods when executing the computer program.

[0039] In addition, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of any of the methods when executed by a processor.

[0040] Beneficial effects: Compared with the prior art, the technical solution of the present invention has the following beneficial technical effects:

[0041] The present invention discloses a reference line correction method, device and storage medium in a vehicle-road collaborative environment. By acquiring high-precision map information and equipment driving information, the reference driving trajectory and reference line loss function of the unmanned vehicle are designed. The method can achieve smooth processing and optimal selection of the reference driving trajectory, and obtain efficient and safe reference line correction by determining the weight coefficients of the value function and the reference line loss function. This will improve the driving accuracy and reliability of unmanned vehicles, promote vehicle-road collaboration, realize a higher level of automatic driving system, and bring important technological progress and application prospects to the fields of intelligent transportation and traffic safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0043] (1) Obtain high-precision map information and unmanned driving equipment driving information in the Cartesian coordinate system through roadside sensing equipment. The traffic status information of this embodiment is designed as follows: road width W = 3.75m, road curvature k r = 0, road boundary {(x, y)|0<x<150, y=0or 7.5}, road obstacle coordinate set {(x, y)|40<x<44, 0.9<y<2.9}, vehicle center of mass plane coordinates in Cartesian coordinate system (x i ,y i )=(0,1.92), vehicle path planning end point coordinates (100,1.92).

[0044] (2) Obtaining a reference driving trajectory set of the unmanned driving device based on high-precision map information and driving information;

[0045] Set Step = 10m, n = 5, PATHTREE {path_tree1, path_tree2, ..., path_tree 5}, where the control nodes of each trajectory are: path_tree1 = {(0, 1.92), (10.04, 3.42), (20.32, 4.04), (31.34, 4.42), (41.43, 5), (51.43, 5.77), (58.85, 5.62), (71.32, 5.43), (80.01, 4.32), (90.32, 3.21), (99.98, 1.9)}; path_tree2 = {(0, 1.92), (10.9436, 3.078), (19.9136, 3.636), (32.280, 3 .978), (39.7728,4.5), (51.9443,5.193), (63.558,5.058), (64.188,4.887), (80.8101,3.888), (98.4488,2.889), (107.9784,1.71)}; path_tree3={(0,1.92),(9.8643,2.736),(20.6248,3.232),(31.10495,3.536),(42.2586,4),(51.9443,4.616),(57.37875,4.496),( 70.0719,4.344),(78.209775,3.456),(88.062,2.568),(98.23035,1.52)}; path_tree4={(0,1.92),(10.1404,3.762),(18.288,4.444 ),(31.6534,4.862),(42.6729,5.5),(54.0015,6.347),(64.1465,6.182),(67.754,5.973),(82.4103,4.752),(96.6424,3.531),(99. 98,2.09)}; path_tree5={(0,1.92),(9.973066667,3.42),(19.77813333,4.1208),(31.75786667,4.5747),(41.2919,5.15),(50.5728 3333,5.68345),(60.41933333,5.5638),(70.13133333,5.5386),(81.8769,4.5144),(87.30933333,3.1137),(102.3128667,1.919)}.

[0046] (3) All reference driving trajectories in the set are smoothed to obtain a smoothed Bezier curve; the horizontal and vertical coordinates of each point in the curve are as follows:

[0047]

[0048]

[0049] (4) Determine the value function form and select the optimal reference driving trajectory;

[0050] The curvatures at different trajectory control nodes are as follows:

[0051]

[0052] According to the curvature values ​​in the above table, the value of different trajectories is calculated:

[0053]

[0054] Therefore, Path4 is the optimal reference driving trajectory.

[0055] (5) Determine the reference line loss function form as:

[0056]

[0057] The weight coefficient to be calibrated is ω s ,ω d and ω t .

[0058] (6) The weight coefficient of the reference line loss function is solved according to the optimal reference driving trajectory, and this weight coefficient is used as the basis for online rapid optimization of the reference line. The process of solving the weight coefficient of the reference line loss function is:

[0059]

[0060] in,

[0061] Since Path4 is the optimal reference driving trajectory, according to Path4 Select the weight coefficient. The selection rules are:

[0062]

[0063] Therefore, the recommended weight coefficient for this example is ω s =0.41,ω d =0.02,ω t =0.57. At this time, the cost of each path is as follows:

[0064]

[0065] The Cost value of the weight coefficient recommended in this embodiment verifies that Path4 is the optimal reference driving trajectory.

[0066] It should be emphasized that the examples described in the present invention are illustrative rather than restrictive, and therefore the present invention is not limited to the examples described in the specific embodiments. Any other embodiments derived by those skilled in the art based on the technical solution of the present invention that do not depart from the purpose and scope of the present invention, whether modified or replaced, also fall within the scope of protection of the present invention.

Claims

1. A reference line correction method in a vehicle-road cooperative environment, characterized in that: The method comprises the following steps: (1) Obtaining map information and unmanned driving equipment driving information in a Cartesian coordinate system through roadside sensing equipment; (2) Calculate the reference driving trajectory set of the unmanned driving device at time t based on the map information and the driving information of the unmanned driving device; (3) Smoothing all reference driving trajectories in the set; (4) determining the form of the value function, and selecting the optimal reference driving trajectory from all the reference driving trajectories in step (3) according to the value function; (5) Determine the form of the reference line loss function and its weight coefficient to be calibrated, and solve the weight coefficient of the reference line loss function according to the optimal reference driving trajectory; (6) Using the solved reference line loss function and the calibrated weight coefficient, the reference line is corrected.

2. The reference line correction method in a vehicle-road cooperative environment according to claim 1, characterized in that: The specific method of step (1) is as follows: When a vehicle passes by on the road, the roadside sensing device will obtain map information and unmanned driving equipment driving information. The map information includes: road width W, road curvature k r , the center line collection point set Line{x l_1 ,y l_1 , x l_2 ,y l_2 , …, x l_j’ ,y l_j’ , …, x l_m ,y l_m }, where (x l_j’ ,y l_j’ ) is the coordinate of the lane centerline collection point, 1≤j'≤m, m is the total number of lane centerline collection points, the road obstacle boundary information set OBS{obs_1,obs_2,…,obs_i,…,obs_q}, 1≤i≤q, q is the total number of obstacles, the boundary of the i-th obstacle obs_i is (x ob1 ,y ob1 , x ob2 ,y ob2 , …, x obj ,y obj , …, x obm’ ,y obm’ ), where (x obj ,y obj ) is the boundary coordinate of the obstacle obs_i collected by the roadside sensing device, 1≤j≤m', m' is the total number of boundary coordinates of the ith obstacle obs_i; the unmanned driving equipment driving information includes the center of mass plane coordinates (x t ,y t ), vehicle path planning end point coordinates (x p ,y p ).

3. The reference line correction method in a vehicle-road cooperative environment according to claim 2, characterized in that: In step (2), the reference driving trajectory set of the unmanned driving device at time t is calculated based on the map information and the driving information of the unmanned driving device. The specific process is as follows: ① The plane coordinates of the vehicle center of mass collected at time t (x t ,y t ) as the initial point of path planning The vehicle path planning end point coordinates (x p ,y p ) as the target point The initial point As the root node of the node tree path_tree_j(t), the collected road width W, road curvature k r , coordinates(x t ,y t ) and coordinates (x p ,y p ) as boundary constraints to generate planning space R; ②In the planning space R, randomly obtain a sampling point x rand , with sampling point x rand As the goal, traverse the existing nodes on the node tree, calculate the Euclidean distance from each node to the sampling point, and select the node with the smallest distance as the parent node When there is only the root node in the node tree Directly select this point as the parent node ); ③From the parent node To a random point x rand Move with a step length of ΔL to get the leaf node x new_1 , judge x new_1 Does it overlap with the range of the obstacle set OBS in the map environment? If so, remove x new_1 , and re-search for a new leaf node x new_1 ; If not, then x new_1 As a parent node Add to the node tree path_tree_j(t) and start looking for leaf node x new_2 ; ④ Repeat steps ②-③ until the parent node Expand to target point Within the set neighborhood, the target point Add to the node tree; ⑤From the target point Start by searching for the parent node in reverse order until you reach the root node. Get the j-th node tree path_tree_j(t), which consists of the parent node The curve composed of the connecting lines between all adjacent parent nodes in the node tree is the corresponding reference driving trajectory trajectory j(t); ⑥ Repeat operations ①-⑤ to obtain a set of n node trees PATHTREE(t){path_tree_1(t), path_tree_2(t),…,path_tree_n(t)}, and the corresponding reference driving trajectory set TRA(t){trajectory 1(t), trajectory 2(t),…,trajectory n(t)}.

4. The reference line correction method in a vehicle-road cooperative environment according to claim 3, characterized in that: The smoothing method in step (3) is as follows: The reference driving trajectory at time t is smoothed using the Bezier curve. The Bezier curve equation is as follows: Among them, P j (u, t) represents the smoothed reference driving trajectory, q represents the number of parent nodes in the jth node tree, Represents the i-th parent node of the j-th node tree, which is also the control point of the Bezier curve. u represents a parameter in the process of generating the parameterized curve, which is used to describe the progress of the curve between the control points. 0≤u≤1.

5. The reference line correction method in a vehicle-road cooperative environment according to claim 4, characterized in that: In step (4), according to the value function, the optimal reference driving trajectory P is selected from the reference driving trajectory set TRA(t) {trajectory1(t), trajectory2(t), ..., trajectoryn(t)} at time t. optimal (t), the value function is as follows: in, is the horizontal coordinate of the kth node in the jth node tree path_tree_j at time t, k r is the road curvature. The trajectory with the smallest value(t) is found through the above formula, which is also the optimal reference driving trajectory.

6. The reference line correction method in a vehicle-road cooperative environment according to claim 5, characterized in that: In step (5), the reference line loss function at time t is expressed as: Where n′ is the total number of points on the optimal reference driving trajectory, is the optimal reference driving trajectory P optimal (t) is the coordinate of the kth point on the s ,ω d ,ω t is the coefficient to be calibrated; the minimum value obtained at time t-1 is used as the reference value of the loss function cost(t) at time t, and the loss function coefficient is calibrated using the least squares method, and the calibrated coefficient is recorded as 7. The reference line correction method in a vehicle-road cooperative environment according to claim 5, characterized in that: The specific method of step (6) is as follows: The initial reference line is set as the center line of the lane where the vehicle is located. Using the reference line loss function and the calibrated coefficients, the collection points on the reference line are gradually adjusted through the gradient descent method until the loss function is minimized. The loss function used to correct the reference line is as follows: Among them, (x l_k ,y l_k ) is the coordinate of the kth acquisition point on the initial reference line, To calibrate the coefficients, the update rules for the collection points on the reference line are as follows: Among them, η is the learning rate, which controls the step size of each update. is the updated coordinate point, is the current coordinate point. Through continuous iteration, the coordinates of all reference line collection points are updated, and the value of the loss function is gradually reduced until the change in the loss function is less than the set threshold ε. When the change in the loss function is less than the set threshold ε, the collection points (x l_k ,y l_k ) will be adjusted to a smooth and optimal position, and the smoothed collection points will obtain the corrected reference line.

8. A reference line correction device in a vehicle-road cooperative environment, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images