Intersection vehicle trajectory correction method and system combined with signal light perception

Abstract

The application discloses a kind of intersection vehicle trajectory correction methods combined with signal light perception, it includes the following steps: (1) based on high-precision map information and the position information of ego vehicle, obtain the current lane where ego vehicle is located;(2) based on high-precision map information and the position information of target vehicle, obtain the current lane where target vehicle is located;(3) based on the current lane where ego vehicle is located and the signal light image data collected by ego vehicle and high-precision map information, determine the signal light corresponding to the current lane where ego vehicle is located, and collect signal light color state information;(4) based on the direction difference angle of the current lane information where ego vehicle is located and the current lane where target vehicle is located, and the color state information of the signal light corresponding to the current lane where ego vehicle is located, determine the signal light color state information of target vehicle;(5) based on the signal light color state information of target vehicle and the road network topology information extracted from high-precision map, the initial predicted trajectory of target vehicle is corrected.

Description

TECHNICAL FIELD

[0001] The present application relates to a trajectory correction method and system, in particular to a vehicle trajectory correction method and system. BACKGROUND

[0002] In recent years, with the rapid development and continuous growth of the automobile industry, people's requirements for the functionality and safety of vehicles are becoming higher and higher. In order to meet the current market and consumer demand, more and more automobile manufacturers are working on improving the intelligent degree of the automobile, and automatic driving assistance is becoming the basic demand of people for future vehicles.

[0003] Among them, driving assistance is a more basic stage in the intelligent driving stage, but it is also an extremely important module. It is the basis of all high-order intelligent driving and an important indicator of vehicle safety and comfort for drivers and passengers.

[0004] At present, in the designed intelligent driving assistance, one of the most important links is the prediction of the surrounding vehicle driving trajectory. High accuracy of target vehicle prediction will directly lead to the accuracy and rationality of the behavior decision of the ego vehicle. However, due to the uncertainty of the target vehicle, the incompleteness of the ego vehicle observation data, and the suddenness of environmental factors, there are considerable challenges to the trajectory prediction of the target vehicle. Therefore, the current common and feasible method is to obtain the relative position, relative speed, and relative heading of the target vehicle based on the ego vehicle perception, and then learn the predicted trajectory of the target vehicle through big data network. However, this method is extremely dependent on the current observation data of the vehicle, and the observation error and false detection have certain interference on the trajectory prediction. Therefore, after using the conventional network model for prediction, it is particularly important to implement further post-processing of the trajectory based on lane lines and road network topology information.

[0005] In view of the current complex road environment in China, the trajectory prediction of the target vehicle at the intersection is still one of the popular processing problems in intelligent driving vehicle prediction. How to effectively combine the traffic lights at the intersection and correct the predicted trajectory of the target vehicle based on the intersection road network topology information so that it can more accurately fit the future trajectory of the target vehicle will be the problem to be solved by the algorithm technology mentioned in the present application. SUMMARY

[0006] One of the purposes of the present application is to provide a signal light perception combined intersection vehicle trajectory correction method, which can effectively combine the color state information of the signal light at the intersection, and based on the intersection road network topology information, the initial predicted trajectory of the target vehicle is corrected, thereby improving the accuracy of the target vehicle predicted trajectory, obtaining a more accurate future trajectory of the target vehicle, and improving the accuracy of subsequent automatic driving judgment and driving comfort.

[0007] In order to achieve the above-mentioned purpose, the present application provides a signal light perception combined intersection vehicle trajectory correction method, which comprises the following steps:

[0008] 100: based on the collected high-precision map information and the position information of the ego vehicle, obtaining the current lane where the ego vehicle is located;

[0009] 200: based on the collected high-precision map information and the position information of the target vehicle, obtaining the current lane where the target vehicle is located;

[0010] 300: based on the current lane where the ego vehicle is located and the signal light image data and high-precision map information collected by the ego vehicle, determining the signal light corresponding to the current lane where the ego vehicle is located, and collecting the color state information of the signal light;

[0011] 400: based on the direction difference angle between the current lane information of the ego vehicle and the current lane of the target vehicle, and the color state information of the signal light corresponding to the current lane of the ego vehicle, determining the color state information of the signal light of the target vehicle;

[0012] 500: based on the color state information of the signal light of the target vehicle and the road network topology information extracted from the high-precision map, correcting the obtained initial predicted trajectory of the target vehicle.

[0013] Research shows that in the current automatic driving field, the vehicle prediction module performs poorly in predicting the trajectory of the vehicle in the intersection scene, mainly in the following aspects:

[0014] (1) The existing vehicle prediction algorithm is completely based on the observed parameters of the target vehicle, including the position, speed, acceleration and orientation of the target vehicle, but does not consider traffic rules and road structure.

[0015] (2) The existing model-based vehicle prediction algorithm can effectively handle the target vehicle behavior prediction in the lane scene at non-intersection, but the designed algorithm has a large deviation in predicting the target vehicle in the complex scene of multiple possibilities at the intersection, especially at the intersection.

[0016] Therefore, the present application expects to improve the difficulties existing in the current vehicle prediction module, which aims to combine the data that can be perceived by the sensors generally possessed by the existing automatic driving vehicle and the high-precision map, maximize the use degree of each signal source, correct the initial prediction trajectory of the target vehicle based on a rule and logical judgment algorithm technology without high algorithm, improve the accuracy of the target vehicle prediction trajectory, and thus improve the accuracy of subsequent automatic driving judgment and the comfort of driving.

[0017] Further, in the intersection vehicle trajectory correction method of the present application, in step 100:

[0018] Based on the collected high-precision map information and the position information of the ego vehicle, the absolute position of the ego vehicle in the high-precision map is obtained;

[0019] Based on the collected high-precision map information and the position information of the ego vehicle, all lane centerlines within a set range of the surrounding area of the ego vehicle are obtained;

[0020] The distance d l from the absolute position to all lane centerlines is calculated l ;

[0021] The shortest distance d l，mim is selected from the distance d l , and the lane where the corresponding lane centerline is located is taken as the current lane of the ego vehicle.

[0022] Further, in the intersection vehicle trajectory correction method of the present application, in step 200:

[0023] Based on the collected high-precision map information and the position information of the target vehicle, the absolute position of the target vehicle in the high-precision map is obtained;

[0024] Based on the collected high-precision map information and the position information of the target vehicle, all lane centerlines within a set range of the surrounding area of the target vehicle are obtained;

[0025] The distance d l from the absolute position to all lane centerlines is calculated l ;

[0026] The shortest distance d l，mim is selected from the distance d l , and the lane where the corresponding lane centerline is located is taken as the current lane of the target vehicle.

[0027] The intersection of the lane centerline of the current lane of the target vehicle and the target vehicle is taken as the starting point of the prediction trajectory.

[0028] Further, in the intersection vehicle trajectory correction method of the present application, in step 300:

[0029] Collecting image data of all signal lights in the range of the area that can be perceived by the dual camera on the vehicle, and obtaining the relative positions of all perceived signal lights and the vehicle;

[0030] Based on the relative positions, the positions of all perceived signal lights in the high-precision map are obtained through coordinate conversion;

[0031] Based on the current lane where the vehicle is located, the position of the signal light corresponding to the vehicle in the high-precision map is determined;

[0032] The distances h t between all perceived signal lights and the signal light corresponding to the vehicle in the high-precision map are calculated

[0033] The minimum distance h t，min is selected from the distances d t between all perceived signal lights and the signal light corresponding to the vehicle in the high-precision map, and the perceived signal light is taken as the signal light corresponding to the current lane where the vehicle is located, and the color state information of the signal light is collected.

[0034] Further, in the intersection vehicle trajectory correction method described in the present application, in step 400, the color state information of the signal light of the target vehicle is determined according to the following rules:

[0035] If the direction difference angle ΔAng l satisfies: ΔAng l < 45° or ΔAng l ≥ 135°, the color state information of the signal light corresponding to the target vehicle is consistent with that of the ego vehicle;

[0036] If the direction difference angle ΔAng l satisfies: ΔAng l ≥ 45° or ΔAng l < 135°, the color state information of the signal light corresponding to the target vehicle is opposite to that of the ego vehicle.

[0037] Further, in the intersection vehicle trajectory correction method described in the present application, in step 500:

[0038] If the color state of the signal light of the target vehicle is yellow or red, the predicted trajectory of the target vehicle is corrected to the current position to the stop line at the intersection;

[0039] If the color state of the signal light of the target vehicle is green, and according to the road network topology information, it is determined that the current lane where the target vehicle is located is a single-direction lane, the predicted trajectory of the target vehicle is corrected to continue moving along the topology relationship of the current lane where the target vehicle is located;

[0040] If the color state of the signal light of the target vehicle is green, and it is determined according to the road network topology information that the current lane where the target vehicle is located is a non-single direction lane, compare all possible topology lane lines of the current lane where the target vehicle is located with the initial prediction trajectory: if the similarity between all possible topology lane lines and the initial prediction trajectory is greater than or equal to a set threshold, select the topology lane line closest to the initial prediction trajectory from all possible topology lane lines as the corrected target vehicle prediction trajectory; if the similarity between all possible topology lane lines and the initial prediction trajectory is lower than the set threshold, select the topology lane line closest to the initial prediction trajectory, and interpolate the initial prediction trajectory to obtain the corrected target vehicle prediction trajectory.

[0041] Correspondingly, another object of the present application is to provide a signal light perception combined intersection vehicle trajectory correction system which can effectively implement the intersection vehicle trajectory correction method of the present application, and can also correct the initial prediction trajectory of the target vehicle, thereby improving the accuracy of the target vehicle prediction trajectory, and improving the accuracy of subsequent automatic driving judgment and driving comfort.

[0042] In order to achieve the above-mentioned objects, the present application provides a signal light perception combined intersection vehicle trajectory correction system, which comprises:

[0043] A data acquisition device acquires high-precision map information, position information of the ego vehicle and position information of the target vehicle;

[0044] A correction module performs the following steps:

[0045] 100: based on the acquired high-precision map information and the position information of the ego vehicle, obtaining the current lane where the ego vehicle is located;

[0046] 200: based on the acquired high-precision map information and the position information of the target vehicle, obtaining the current lane where the target vehicle is located;

[0047] 300: based on the current lane where the ego vehicle is located, and the signal light image data and high-precision map information collected by the ego vehicle, determining the signal light corresponding to the current lane where the ego vehicle is located, and collecting the color state information of the signal light;

[0048] 400: based on the direction difference angle between the current lane information of the ego vehicle and the current lane where the target vehicle is located, and the color state information of the signal light corresponding to the current lane where the ego vehicle is located, determining the color state information of the signal light of the target vehicle;

[0049] 500: correcting the obtained initial prediction trajectory of the target vehicle based on the color state information of the signal light of the target vehicle and the road network topology information extracted from the high-definition map.

[0050] Further, in the intersection vehicle trajectory correction system, the correction module performs the following steps in step 100:

[0051] obtaining the absolute position of the ego vehicle in the high-definition map based on the collected high-definition map information and the position information of the ego vehicle;

[0052] obtaining all lane centerlines within a set range of the surrounding area of the ego vehicle based on the collected high-definition map information and the position information of the ego vehicle;

[0053] calculating the distance d l from the absolute position to all lane centerlines;

[0054] selecting the shortest distance d l from the distances d l，mim to all lane centerlines, and regarding the lane where the corresponding lane centerline is located as the current lane of the ego vehicle;

[0055] The correction module performs the following steps in step 200:

[0056] obtaining the absolute position of the target vehicle in the high-definition map based on the collected high-definition map information and the position information of the target vehicle;

[0057] obtaining all lane centerlines within a set range of the surrounding area of the target vehicle based on the collected high-definition map information and the position information of the target vehicle;

[0058] calculating the distance d l from the absolute position to all lane centerlines;

[0059] selecting the shortest distance d l from the distances d l，mim to all lane centerlines, and regarding the lane where the corresponding lane centerline is located as the current lane of the target vehicle;

[0060] taking the intersection of the lane centerline of the current lane of the target vehicle and the target vehicle as the starting point of the prediction trajectory.

[0061] Further, in the intersection vehicle trajectory correction system, the correction module performs the following steps in step 300:

[0062] Based on the image data of all traffic lights within the sensed area, the relative positions of all sensed traffic lights and the vehicle are obtained;

[0063] Based on the relative positions, the positions of all sensed traffic lights in the high-precision map are obtained through coordinate transformation;

[0064] Based on the current lane where the vehicle is located, determine the position of the traffic light corresponding to the vehicle in the high-precision map;

[0065] Calculate the distance h between all detected traffic lights and the traffic light corresponding to the vehicle in the high-precision map. t ;

[0066] The distance d between all detected traffic lights and the traffic light corresponding to the vehicle in the high-precision map. t Select the minimum distance h t，min The detected traffic light is used as the traffic light corresponding to the current lane of the vehicle, and the color status information of the traffic light is collected.

[0067] Furthermore, in the intersection vehicle trajectory correction system of the present invention, in step 400, the correction module determines the color status information of the target vehicle's traffic lights according to the following rules:

[0068] If the direction difference angle ΔAng l Satisfy: ΔAng l <45° or ΔAng l If the angle is ≥135°, the color status information of the traffic light corresponding to the target vehicle is consistent with that of the vehicle itself.

[0069] If the direction difference angle ΔAng l Satisfy: ΔAng l ≥45° or ΔAng l If the angle is less than 135°, the traffic light color status information of the target vehicle is opposite to that of the vehicle itself; and / or,

[0070] The correction module performs the following steps in step 500:

[0071] If the target vehicle's traffic light is yellow or red, then the predicted trajectory of the target vehicle will be corrected to the current position to the stop line at the intersection.

[0072] If the target vehicle's traffic light is green, and the road network topology information indicates that the target vehicle's current lane is a one-way lane, then the target vehicle's predicted trajectory will be corrected to continue along the topological relationship of the target vehicle's current lane.

[0073] If the color state of the signal light of the target vehicle is green, and it is determined according to the road network topology information that the current lane where the target vehicle is located is a non-single direction lane, compare all possible topology lane lines of the current lane where the target vehicle is located with the initial predicted trajectory: if the similarity between all possible topology lane lines and the initial predicted trajectory is greater than or equal to a set threshold value, select the topology lane line closest to the initial predicted trajectory from all possible topology lane lines as the corrected target vehicle predicted trajectory; if the similarity between all possible topology lane lines and the initial predicted trajectory is less than the set threshold value, select the topology lane line closest to the initial predicted trajectory, and interpolate the initial predicted trajectory to obtain the corrected target vehicle predicted trajectory.

[0074] Compared with the prior art, the intersection vehicle trajectory correction method and system combined with signal light perception have the following advantages:

[0075] The intersection vehicle trajectory correction method combined with signal light perception fully combines and applies the complex but regular mandatory feature information of the intersection, thereby correcting and optimizing the predicted trajectory of the target vehicle at the intersection, avoiding the prediction deviation risk caused by sensor data errors, greatly reducing the risk of prediction errors to the ego vehicle control, and improving the smoothness, accuracy and comfort of automatic driving.

[0076] In the present application, the intersection vehicle trajectory correction method designed fully utilizes the color state information of the signal light at the intersection, which can filter many unnecessary vehicle trajectory judgments through simple traffic light logic judgment, thereby performing more efficient automatic driving control.

[0077] In actual use of the intersection vehicle trajectory correction method designed in the present application, the color state information of the signal light at the intersection can be effectively combined, and the initial predicted trajectory of the target vehicle is corrected based on the intersection road network topology information, thereby improving the accuracy of the predicted trajectory of the target vehicle, obtaining a future trajectory that is more consistent with the target vehicle, and improving the accuracy of subsequent automatic driving judgment and the comfort of driving.

[0078] Correspondingly, the intersection vehicle trajectory correction system designed in the present application can effectively implement the intersection vehicle trajectory correction method of the present application, and also has the above advantages and beneficial effects. BRIEF DESCRIPTION OF DRAWINGS

[0079] Figure 1 The intersection vehicle trajectory correction system combined with signal light perception according to the present application is shown in the step flowchart in one embodiment.

[0080] Figure 2The intersection vehicle trajectory correction method according to the present application is schematically shown to acquire the method of the corresponding lane where the vehicle is located based on the collected high-precision map information and the position information of the vehicle in one embodiment.

[0081] Figure 3 The intersection vehicle trajectory correction method according to the present application is schematically shown to determine the method of the signal light corresponding to the current lane where the ego vehicle is located in one embodiment.

[0082] Figure 4 The intersection vehicle trajectory correction method according to the present application is schematically shown to determine the color state information of the signal light of the target vehicle in one embodiment.

[0083] Figure 5 The vehicle driving schematic diagram under one single driving logic is schematically shown.

[0084] Figure 6 The vehicle driving schematic diagram under another single driving logic is schematically shown.

[0085] Figure 7 The vehicle driving schematic diagram under one of the multiple driving logics is schematically shown.

[0086] Figure 8 The vehicle driving schematic diagram under another of the multiple driving logics is schematically shown. DETAILED DESCRIPTION

[0087] The intelligent vehicle active suspension adjustment system according to the present application will be further explained and described below in combination with the drawings and specific embodiments, however, the explanation and description do not constitute undue limitations on the technical solutions of the present application.

[0088] In the present application, the intersection vehicle trajectory correction system combined with signal light perception is designed, which specifically comprises a data acquisition device and a correction module. The intersection vehicle trajectory correction system can correct the initial predicted trajectory of the target vehicle at the intersection, thereby improving the accuracy of the predicted trajectory of the target vehicle, obtaining a future trajectory that is more suitable for the target vehicle, and improving the accuracy of subsequent automatic driving judgment and the comfort of driving.

[0089] Figure 1 The step flow chart of the intersection vehicle trajectory correction system combined with signal light perception according to the present application in one embodiment is shown.

[0090] In the present application, the data acquisition device in the intersection vehicle trajectory correction system can effectively collect high-precision map information, ego vehicle information, and target vehicle information.

[0091] AsFigure 1 As shown in the embodiment, the collected high-precision map information can specifically include: lane line position, lane line topology, and signal light (traffic light) position; the collected target vehicle information can specifically include: position information, speed information, acceleration information, and orientation information of the target vehicle; and the collected ego vehicle information can specifically include: position information, speed information of the ego vehicle, and color state information of the signal light (traffic light) of the current lane where the ego vehicle is located. The signal light mentioned in the application can be specifically understood as Figure 1 a traffic light.

[0092] Based on the collected information, the correction module in the intersection vehicle trajectory correction system designed by the application can specifically perform the following steps 100-500:

[0093] 100: Obtain the current lane where the ego vehicle is located based on the collected high-precision map information and the position information of the ego vehicle.

[0094] 200: Obtain the current lane where the target vehicle is located based on the collected high-precision map information and the position information of the target vehicle.

[0095] It should be noted that in the above steps 100 and 200 of the application, the current lane where the corresponding vehicle is located is obtained based on the collected high-precision map information and the position information of the corresponding vehicle. The specific obtaining method can be referred to in Figure 2 .

[0096] Figure 2 The method for obtaining the lane where the corresponding vehicle is located based on the collected high-precision map information and the position information of the vehicle in an embodiment of the intersection vehicle trajectory correction method of the application is schematically shown.

[0097] As Figure 2 shown in the embodiment, when in the step 100 of the application and the current lane where the ego vehicle is located needs to be obtained, it can specifically include the following steps:

[0098] (1) Obtain the absolute position (x, y) of the ego vehicle in the high-precision map based on the collected high-precision map information and the position information of the ego vehicle; (2) based on the collected high-precision map information and the position information of the ego vehicle, search for all lane center line information in the range of the 20m area around the ego vehicle with the absolute position (x, y) of the ego vehicle in the high-precision map as the center and counterclockwise from north; (3) calculate the distance d l from the absolute position (x, y) of the ego vehicle to all lane center lines; (4) select the shortest distance d l from the distance d l，mim from all lane center lines, and the lane where the corresponding lane center line is located is taken as the current lane where the ego vehicle is located.

[0099] For example, with Figure 2 In the example of the implementation shown, the lane where the candidate center lane line 3 is located is the current lane where the vehicle is located.

[0100] Accordingly, see Figure 2 In step 200 of this invention, when it is necessary to obtain the current lane where the target vehicle is located, it may specifically include the following steps:

[0101] (1) Obtain the absolute position (x, y) of the target vehicle in the high-precision map based on the collected high-precision map information and the position information of the target vehicle; (2) Based on the collected high-precision map information and the position information of the target vehicle, search counterclockwise from north to find the center line information of all lanes within the area around the target vehicle with a radius of 20m, using the absolute position (x, y) of the target vehicle in the high-precision map as the center; (3) Calculate the distance d from the absolute position (x, y) of the target vehicle to all lane center lines. l (4) Distance d from the center line of all lanes l Select the shortest distance d from the options. l，mim The lane where the center line of the corresponding lane is located is taken as the current lane where the target vehicle is located.

[0102] It should be noted that in step 200, the intersection of the center line of the current lane where the target vehicle is located and the target vehicle can be specifically used as the starting point of the predicted trajectory.

[0103] 300: Based on the current lane where the vehicle is located, as well as the traffic light image data and high-precision map information collected by the vehicle, determine the traffic light corresponding to the current lane where the vehicle is located, and collect the color status information of the traffic light.

[0104] In this technical solution designed by the present invention, in step 300, it is necessary to determine the traffic light corresponding to the current lane where the vehicle is located, and to collect the color status information of the traffic light. The specific collection method can be found below. Figure 3 The example shown.

[0105] Figure 3 The diagram illustrates a method for determining the traffic light corresponding to the current lane of a vehicle in an intersection, as described in one embodiment of the intersection vehicle trajectory correction method of the present invention.

[0106] like Figure 3 As shown in this example, in order to determine the traffic light corresponding to the current lane of the vehicle and collect the color status information of the traffic light, step 300 of the present invention is specifically designed with the following steps:

[0107] (1) using the binocular camera set on the ego vehicle to collect image data of all signal lights (such as signal lights 1, 2, and 3 shown in Figure 3 ) in the range that can be perceived, and obtaining the relative positions of all perceived signal lights and the ego vehicle E; (2) based on the relative positions of the perceived signal lights and the ego vehicle E, obtaining the positions of all perceived signal lights in the high-definition map through coordinate conversion; (3) based on the current lane in which the ego vehicle E is located, determining the position of the signal light corresponding to the ego vehicle in the high-definition map (such as the signal light 4 shown in Figure 3 ); (4) calculating the distances h t (such as h1, h2, and h3 shown in Figure 3 ) between all perceived signal lights and the signal light corresponding to the ego vehicle in the high-definition map; (5) selecting the minimum distance h t from all distances h t，min between all perceived signal lights and the signal light corresponding to the ego vehicle in the high-definition map, and taking the perceived signal light as the signal light corresponding to the current lane in which the ego vehicle is located, and collecting the color state information of the signal light.

[0108] 400: based on the direction difference angle between the current lane information of the ego vehicle and the current lane in which the target vehicle is located, and the color state information of the signal light corresponding to the current lane in which the ego vehicle is located, determining the color state information of the signal light of the target vehicle.

[0109] In the present application, after obtaining the color state information of the signal light corresponding to the current lane in which the ego vehicle is located, it is necessary to determine the color state information of the signal light of the target vehicle according to the direction difference angle between the current lane information of the ego vehicle and the current lane in which the target vehicle is located, and the color state information of the signal light corresponding to the current lane in which the ego vehicle is located. The judgment logic of the color state information is described below. Figure 4 .

[0110] Figure 4 The logical judgment flowchart of the intersection vehicle trajectory correction method according to the present application in one embodiment is schematically shown.

[0111] Referring to Figure 4 It can be seen that in the present embodiment, the color state information of the signal light of the target vehicle can be determined according to the following rules in step 400 of the present application:

[0112] If the direction difference angle ΔAng l between the current lane information of the ego vehicle and the current lane in which the target vehicle is located satisfies: ΔAng l < 45° or ΔAng l≥ 135°, the signal light color state information corresponding to the target vehicle is consistent with the ego vehicle;

[0113] If the direction difference angle ΔAng between the current lane information of the ego vehicle and the current lane of the target vehicle is l satisfies: ΔAng l ≥ 45° or ΔAng l < 135°, the signal light color state information corresponding to the target vehicle is opposite to the ego vehicle.

[0114] 500: based on the color state information of the signal light of the target vehicle and the road network topology information extracted from the high-precision map, the obtained initial prediction trajectory of the target vehicle is corrected to improve the accuracy of the prediction trajectory of the target vehicle, and a future trajectory more consistent with the target vehicle is obtained.

[0115] In step 500 of the present application, the obtained initial prediction trajectory of the target vehicle needs to be corrected according to the color state information of the signal light of the vehicle obtained in step 400 and the road network topology information extracted from the high-precision map. The initial prediction trajectory of the target vehicle is a known quantity in this technical solution, which is usually learned based on a neural network model in the prior art.

[0116] It should be noted that when correcting the initial prediction trajectory of the target vehicle, it is necessary to consider the possibility that the target vehicle has only one driving logic (i.e. single driving logic). Moreover, when the target vehicle has more than one driving logic, it is also necessary to consider whether at least one of the multiple possible driving trajectories is similar (similarity > 80%) to the initial prediction trajectory of the target vehicle.

[0117] Therefore, in the present application, the logic for correcting the initial prediction trajectory of the target vehicle is as follows:

[0118] (1) Single driving logic:

[0119] 1.1 If the color state of the signal light of the target vehicle is yellow or red, the prediction trajectory of the target vehicle is corrected to the current position to the stop line at the intersection.

[0120] Figure 5 A vehicle driving schematic diagram under a single driving logic is schematically shown.

[0121] As Figure 5 shown, in this way, the color state of the signal light corresponding to the ego vehicle E (Ego Vehicle) is red, and according to the signal light logic, it is determined that the color state of the signal light of the target vehicle O (Object Vehicle) is also red. Therefore, the corrected prediction trajectory of the target vehicle O is: driving from the current position to the stop line at the intersection, as shown in Figure 5The arrowed dashed line in the middle.

[0122] If the target vehicle prediction trajectory obtained by simply relying on the model predicts that the behavior trajectory of the target vehicle O will intersect with the ego vehicle E due to the orientation of the target vehicle O, and if the predicted collision point is beyond the stop line at the intersection, the ego vehicle E will stop far away from the stop line in order to avoid collision with the target vehicle. However, in reality, due to the red light, the target vehicle O will not travel along the trajectory predicted by the model, but will stop in front of the stop line at the intersection. Therefore, based on the corrected prediction trajectory according to the present application, the ego vehicle E can travel to the stop line at the intersection and then stop, thereby avoiding the poor automatic driving experience of the ego vehicle caused by the error of the target vehicle prediction trajectory.

[0123] 1.2 If the color state of the signal light of the target vehicle is green, and it is determined according to the road network topology information that the current lane where the target vehicle is located is a single-direction lane, the target vehicle prediction trajectory is corrected to continue moving along the topology relationship of the current lane where the target vehicle is located.

[0124] Figure 6 A vehicle driving schematic diagram under another single driving logic is schematically shown.

[0125] As shown in Figure 6 , in this mode, the color state of the signal light corresponding to the ego vehicle E is green, and according to the signal light logic, it is determined that the color state of the signal light of the target vehicle O is also green. At the same time, the lane where the target vehicle O is located has only one driving possibility, which is a single left-turn lane as shown in the figure, and therefore the corrected prediction trajectory of the target vehicle O is the topology lane line of this lane, as shown in the middle Figure 6 of the figure.

[0126] (2) Multiple driving logics:

[0127] If the color state of the signal light of the target vehicle is green, and it is determined according to the road network topology information that the current lane where the target vehicle is located is a non-single-direction lane, the current lane where the target vehicle is located is compared with the initial prediction trajectory obtained based on the information such as current speed, acceleration, and heading angle after learning by the neural network:

[0128] If there is a lane line in all possible topology lane lines that is greater than or equal to one lane line with a similarity greater than a set threshold to the initial prediction trajectory, the lane line that is most similar to the initial prediction trajectory in all possible topology lane lines is selected as the corrected target vehicle prediction trajectory.

[0129] If the similarity of all possible topological lane lines to the initial predicted trajectory is lower than a set threshold, the topological lane line with the highest similarity to the initial predicted trajectory is selected, and the initial predicted trajectory is interpolated to obtain a corrected target vehicle predicted trajectory.

[0130] It should be noted that in the present application Figure 1 In this embodiment, the threshold of the similarity can be set to 80%, and the following Figure 7 and Figure 8 The vehicle driving schematic diagram in another case of the plurality of driving logics is schematically shown.

[0131] Figure 7 The vehicle driving schematic diagram in another case of the plurality of driving logics is schematically shown.

[0132] As shown in Figure 7 In this way, the target vehicle O(Object Vehicle) is in a current lane with multiple driving possibilities, i.e., the topological lane line 1 (continuing straight in the current lane) and the topological lane line 2 (turning left at the intersection) can be selected.

[0133] In this case with multiple driving logics, after comparing all possible topological lane lines of the current lane where the target vehicle O is located with the initial predicted trajectory (dashed line A), it is found that the similarity of all candidate topological lane lines to the initial predicted trajectory (dashed line A) is lower than 80%.

[0134] Figure 8 The vehicle driving schematic diagram in another case of the plurality of driving logics is schematically shown.

[0135] As shown in Figure 8 In this way, the target vehicle O(Object Vehicle) is in a current lane with multiple driving possibilities, i.e., the topological lane line 1 (continuing straight in the current lane) and the topological lane line 2 (turning left at the intersection) can be selected.

[0136] In this case with multiple driving logics, after comparing all possible topological lane lines of the current lane where the target vehicle O is located with the initial predicted trajectory (dashed line A), it is found that the similarity of all candidate topological lane lines to the initial predicted trajectory (dashed line A) is lower than 80%.

[0137] Thus, the lane (topology lane line 1) with the highest similarity to the initial predicted trajectory (dotted line A) in the candidate topology lane line of the current lane where the target vehicle O is located is interpolated with the initial predicted trajectory (dotted line A) to obtain the corrected target vehicle predicted trajectory (dotted line B).

[0138] In summary, the intersection vehicle trajectory correction system designed in the application can effectively correct the initial predicted trajectory of the target vehicle and output the corrected vehicle predicted trajectory information of the intersection, which is mainly based on the red light and lane information rules of the intersection. The prediction accuracy of the vehicle that obeys the traffic regulations will be greatly improved.

[0139] In addition, the combination of the technical features in the application is not limited to the combination of the claims or the combination of the embodiments. All the technical features disclosed in the application can be freely combined or combined in any way, unless they are contradictory to each other.

[0140] It should also be noted that the above-mentioned embodiments are only specific embodiments of the application. Obviously, the application is not limited to the above-mentioned embodiments, and similar changes or modifications made by those skilled in the art based on the disclosure of the application are directly derived or easily conceived, and all should belong to the protection scope of the application.

Claims

1. A method for intersection vehicle trajectory correction in conjunction with signal light perception, characterized in that, The method comprises the following steps: 100: obtaining a current lane in which a self vehicle is located based on collected high-precision map information and position information of the self vehicle; 200: obtaining a current lane in which a target vehicle is located based on collected high-precision map information and position information of the target vehicle; 300: determining a signal lamp corresponding to the current lane in which the self vehicle is located based on the current lane in which the self vehicle is located, signal lamp image data collected by the self vehicle and high-precision map information, and collecting color state information of the signal lamp; 400: determining color state information of a signal lamp of the target vehicle based on a direction difference angle between the current lane information of the self vehicle and the current lane in which the target vehicle is located and the color state information of the signal lamp corresponding to the current lane in which the self vehicle is located; the color state information of the signal lamp of the target vehicle is determined according to the following rules: If the direction difference angle ΔAng l satisfies: ΔAng l < 45° or ΔAng l ≥ 135°, the signal light color state information of the target vehicle is consistent with that of the ego vehicle. If the direction difference angle ΔAng l satisfies: ΔAng l ≥ 45° or ΔAng l < 135°, the signal light color state information of the target vehicle is opposite to the ego vehicle. 500: correcting an obtained initial prediction trajectory of the target vehicle based on the color state information of the signal lamp of the target vehicle and road network topology information extracted from the high-precision map; wherein: if the color state of the signal lamp of the target vehicle is yellow or red, the prediction trajectory of the target vehicle is corrected to a current position to a stop line at an intersection; if the color state of the signal lamp of the target vehicle is green and the current lane in which the target vehicle is located is a single-direction lane according to the road network topology information, the prediction trajectory of the target vehicle is corrected to continue driving along the topological relationship of the current lane in which the target vehicle is located; if the color state of the signal lamp of the target vehicle is green and the current lane in which the target vehicle is located is a non-single-direction lane according to the road network topology information, all possible topological lane lines of the current lane in which the target vehicle is located are compared with the initial prediction trajectory: if the similarity between all possible topological lane lines and the initial prediction trajectory is greater than a set threshold, the topological lane line closest to the initial prediction trajectory is selected as the corrected prediction trajectory of the target vehicle; if the similarity between all possible topological lane lines and the initial prediction trajectory is less than the set threshold, the topological lane line closest to the initial prediction trajectory is selected, and an interpolation is performed on the initial prediction trajectory to obtain the corrected prediction trajectory of the target vehicle.

2. The intersection vehicle trajectory revision method of claim 1, wherein, In step 100: an absolute position of the self vehicle in the high-precision map is obtained based on the collected high-precision map information and the position information of the self vehicle; all lane center lines in a set range around the self vehicle are obtained based on the collected high-precision map information and the position information of the self vehicle; calculating the distance d of the absolute position to all lane center lines l ; The shortest distance d is selected from the distances d to all lane centerlines l The shortest distance d is selected from the distances d to all lane centerlines l，mim The lane in which the corresponding lane centerline is located is taken as the current lane in which the ego vehicle is located.

3. The intersection vehicle trajectory revision method of claim 1, wherein, In step 200: an absolute position of the target vehicle in the high-precision map is obtained based on the collected high-precision map information and the position information of the target vehicle; all lane center lines in a set range around the target vehicle are obtained based on the collected high-precision map information and the position information of the target vehicle; calculating the distance d of the absolute position to all lane center lines l ; The shortest distance d is selected from the distances d to all lane centerlines l The shortest distance d is selected from the distances d to all lane centerlines l，mim The lane where the corresponding lane centerline is located is taken as the current lane where the target vehicle is located an intersection of the lane center line of the current lane in which the target vehicle is located and the target vehicle is taken as a starting point of the prediction trajectory.

4. The intersection vehicle trajectory revision method of claim 1, wherein, In step 300: image data of all signal lamps in a range that can be perceived is collected by using a binocular camera arranged on the self vehicle, and relative positions of all perceived signal lamps and the self vehicle are obtained; Based on the relative position, the positions of all perceived traffic lights in the high-definition map are obtained through coordinate conversion; Based on the current lane where the ego vehicle is located, the position of the traffic light corresponding to the ego vehicle in the high-definition map is determined; calculating distances d of all perceived traffic lights from a traffic light corresponding to the ego vehicle in the high-definition map t ; All perceived signal lights and the distance d of the signal light corresponding to the ego vehicle in the high-definition map t The minimum distance d is selected t，min The perceived signal light is taken as the signal light corresponding to the current lane where the ego vehicle is located, and the color state information of the signal light is collected.

5. A system for intersection vehicle trajectory correction in conjunction with signal light perception, the system comprising: Comprise: Data acquisition device, which acquires high-definition map information, position information of the ego vehicle and position information of the target vehicle; Correction module, which performs the following steps: 100: Based on the acquired high-definition map information and the position information of the ego vehicle, the current lane where the ego vehicle is located is obtained; 200: Based on the acquired high-definition map information and the position information of the target vehicle, the current lane where the target vehicle is located is obtained; 300: Based on the current lane where the ego vehicle is located, the signal lamp corresponding to the current lane where the ego vehicle is located is determined based on the signal lamp image data collected by the ego vehicle and the high-definition map information, and the color state information of the signal lamp is collected; 400: Based on the direction difference angle between the current lane information of the ego vehicle and the current lane of the target vehicle, and the color state information of the signal lamp corresponding to the current lane where the ego vehicle is located, the color state information of the signal lamp of the target vehicle is determined; 500: Based on the color state information of the signal lamp of the target vehicle and the road network topology information extracted from the high-definition map, the obtained initial prediction trajectory of the target vehicle is corrected; Wherein, in step 400, the correction module determines the color state information of the signal lamp of the target vehicle according to the following rules: If the direction difference angle ΔAng l satisfies: ΔAng l <45° or ΔAng l ≥135°, the signal light color state information of the target vehicle is consistent with that of the ego vehicle. if the direction difference angle ΔAng l satisfies: ΔAng l ≥ 45° or ΔAng l < 135°, the signal light color state information corresponding to the target vehicle is opposite to the ego vehicle; and / or, In step 500, the correction module performs the following steps: If the color state of the signal lamp of the target vehicle is yellow or red, the prediction trajectory of the target vehicle is corrected to stop at the current position to the stop line of the intersection; If the color state of the signal lamp of the target vehicle is green, and according to the road network topology information, it is determined that the current lane where the target vehicle is located is a single direction lane, then the prediction trajectory of the target vehicle is corrected to continue driving along the topological relationship of the current lane where the target vehicle is located; If the color state of the signal lamp of the target vehicle is green, and according to the road network topology information, it is determined that the current lane where the target vehicle is located is a non-single direction lane, then all possible topological lane lines of the current lane where the target vehicle is located are compared with the initial prediction trajectory: if there is a lane line in all possible topological lane lines with a similarity greater than a set threshold with the initial prediction trajectory, then the lane line closest to the initial prediction trajectory in all possible topological lane lines is selected as the corrected prediction trajectory of the target vehicle; If the similarity of all possible topological lane lines to the initial prediction trajectory is lower than the set threshold, then the topological lane line closest to the initial prediction trajectory is selected, and the initial prediction trajectory is interpolated to obtain the corrected prediction trajectory of the target vehicle.

6. The intersection vehicle trajectory revision system of claim 5, wherein, In step 100, the correction module performs the following steps: Based on the acquired high-definition map information and the position information of the ego vehicle, the absolute position of the ego vehicle in the high-definition map is obtained; Based on the acquired high-definition map information and the position information of the ego vehicle, all lane center lines within a set range around the ego vehicle are obtained; calculating the distance d of the absolute position to all lane center lines l ; The shortest distance d is selected from the distances d to all lane centerlines l The shortest distance d is selected from the distances d to all lane centerlines l，mim The lane where the corresponding lane centerline is located is taken as the current lane where the ego vehicle is located In step 200, the correction module performs the following steps: Based on the collected high-precision map information and the position information of the target vehicle, the absolute position of the target vehicle in the high-precision map is obtained; Based on the collected high-precision map information and the position information of the target vehicle, all lane center lines within the set target vehicle surrounding area range are obtained; calculating the distance d of the absolute position to all lane center lines l ; The shortest distance d is selected from the distances d to all lane centerlines l The shortest distance d is selected from the distances d to all lane centerlines l，mim The lane where the corresponding lane centerline is located is taken as the current lane where the target vehicle is located. The intersection of the lane center line of the current lane where the target vehicle is located and the target vehicle is taken as the starting point of the predicted trajectory.

7. The intersection vehicle trajectory revision system of claim 5, wherein, Also included is a binocular camera provided on the ego vehicle, which collects image data of all signal lights within the perceived area range and transmits it to the correction module, which performs the following steps in step 300: Based on the image data of all signal lights within the perceived area range, the relative positions of all perceived signal lights and the ego vehicle are obtained; Based on the relative positions, the positions of all perceived signal lights in the high-precision map are obtained through coordinate conversion; Based on the current lane where the ego vehicle is located, the positions of the signal lights corresponding to the ego vehicle in the high-precision map are determined; calculating distances d of all perceived traffic lights from a traffic light corresponding to the ego vehicle in the high-definition map t ; all perceived signal lights and the distance d of the signal light corresponding to the ego vehicle in the high-definition map t selecting the minimum distance d t，min take the perceived signal light as the signal light corresponding to the current lane where the ego vehicle is located, and collect the color state information of the signal light.

Images