Automatic driving trajectory evaluation method, device, equipment and storage medium

Abstract

The present invention relates to the field of autonomous driving, and discloses an autonomous driving trajectory evaluation method, device, equipment and storage medium, the method comprising: collecting a driving data set and a corresponding actual driving trajectory; dividing the driving data set based on a preset time interval to obtain a plurality of scene driving data packets; simulating the scene driving data packets to obtain a simulated driving trajectory; comparing the difference between the simulated driving trajectory and the actual driving trajectory, and performing similarity calculations based on the difference to obtain the trajectory similarity corresponding to each scene driving data packet. The method collects a large amount of human driving data, and then plays back the data in a simulation environment, allowing the autonomous driving system to run in the simulation environment, compares the difference between the autonomous driving trajectory and the human trajectory, and gives a comprehensive score in combination with map information. The score can quantitatively, objectively and accurately evaluate the similarity between the autonomous driving and human drivers, and can be used to optimize the output of the autonomous driving algorithm.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving, and in particular to an autonomous driving trajectory evaluation method, device, equipment and storage medium. Background Art

[0002] With the rapid development of information and control technology, autonomous driving technology is gradually being accepted by automobile manufacturers and users. Autonomous driving can not only minimize the danger of driving, but also reduce the heavy driving tasks of users. Therefore, autonomous driving is also a major trend in the future development of automobiles.

[0003] In the process of applying autonomous driving technology, the scenarios to be handled are very complex and diverse, so the autonomous driving algorithm needs to be continuously iterated to adapt to different road scenarios. In order to accurately evaluate the autonomous driving algorithm before it is put into use, the autonomous driving algorithm can be pre-tested using simulation before real road testing. At this stage, the testing content of the autonomous driving algorithm is mainly to determine whether the driving behavior of the autonomous driving system is similar to that of human drivers, but there is a lack of quantitative and accurate judgment methods for the degree of similarity. Summary of the invention

[0004] The main purpose of the present invention is to solve the technical problem that the existing autonomous driving trajectory evaluation method lacks a quantitative and accurate similarity judgment method.

[0005] A first aspect of the present invention provides an automatic driving trajectory evaluation method, comprising: collecting a driving data set and a corresponding actual driving trajectory; dividing the driving data set based on a preset time interval to obtain a plurality of scene driving data packets; simulating the scene driving data packets to obtain a simulated driving trajectory; comparing the difference between the simulated driving trajectory and the actual driving trajectory, and performing similarity calculations based on the differences to obtain trajectory similarities corresponding to each scene driving data packet.

[0006] Optionally, in a first implementation manner of the first aspect of the present invention, comparing the difference between the simulated driving trajectory and the actual driving trajectory, and performing similarity calculations based on the differences to obtain the trajectory similarity corresponding to each scenario driving data packet includes: obtaining a first set of intersections passed by the actual driving trajectory corresponding to each scenario driving data packet and a second set of intersections passed by the corresponding simulated driving trajectory; comparing the difference between the first set of intersections and the second set of intersections, and judging whether there are serious differences between the actual driving trajectory and the simulated driving trajectory corresponding to each scenario driving data packet based on the differences; using the scenario driving data packet with serious differences as the first driving data packet, and setting the trajectory similarity of the first driving data packet to zero; using the scenario driving data packet with no serious differences as the second driving data packet, and performing trajectory similarity calculations on the actual driving trajectory and the simulated driving trajectory corresponding to the second driving data packet.

[0007] Optionally, in a second implementation method of the first aspect of the present invention, comparing the difference between the first intersection set and the second intersection set, and judging whether there are serious differences between the actual driving trajectory and the simulated driving trajectory corresponding to the driving data package of each scenario based on the difference includes: judging whether the first intersection set can contain the second intersection set; if the first intersection set does not contain the second intersection set, judging whether the second intersection set can contain the first intersection set; if the second intersection set does not contain the first intersection set, determining that there are serious differences between the corresponding actual driving trajectory and the simulated driving trajectory; if the first intersection set can contain the second intersection set or the second intersection set can contain the first intersection set, determining that there are no serious differences between the corresponding actual driving trajectory and the simulated driving trajectory.

[0008] Optionally, in a third implementation of the first aspect of the present invention, the scene driving data packet includes displacement data;

[0009] The method of taking the scenario driving data packet without serious disagreement as the second driving data packet and performing trajectory similarity calculation on the actual driving trajectory and the simulated driving trajectory corresponding to the second driving data packet includes: taking the scenario driving data packet without serious disagreement as the second driving data packet; calculating, based on the displacement data corresponding to the second driving data packet, a first probability value that the scene corresponding to the second driving data packet is a static scene; calculating, based on the displacement data in the second driving data packet and a preset motion scene scoring algorithm, a motion scene score corresponding to the second driving data packet; calculating, based on the displacement data in the second driving data packet and a preset static scene scoring algorithm, a static scene score corresponding to the second driving data packet; and calculating a trajectory similarity score corresponding to the second driving data packet based on the motion scene score, the static scene score and the first probability value.

[0010] Optionally, in a fourth implementation manner of the first aspect of the present invention, calculating the motion scene score corresponding to the second driving data packet based on the displacement data in the second driving data packet and a preset motion scene score algorithm includes: calculating the displacement difference value between the actual driving trajectory and the simulated driving trajectory in the scene corresponding to the second driving data packet based on the displacement data in the second driving data packet; normalizing the displacement difference value, and projecting it along the lane line direction and the direction perpendicular to the lane line, respectively, to obtain a first displacement difference in the lane line direction and a second displacement difference perpendicular to the lane line direction; calculating a first motion score in the lane line direction based on the first displacement difference; calculating a second motion score perpendicular to the lane line direction based on the second displacement difference; and calculating the motion scene score corresponding to the second driving data packet based on the first motion score and the second motion score.

[0011] Optionally, in a fifth implementation manner of the first aspect of the present invention, calculating the stationary scene score corresponding to the second driving data packet based on the displacement data in the second driving data packet and a preset stationary scene score algorithm includes: calculating a first stationary score in the lane line direction based on the first displacement difference; calculating a second stationary score perpendicular to the lane line direction based on the second displacement difference; and calculating the stationary scene score corresponding to the second driving data packet based on the first stationary score and the second stationary score.

[0012] Optionally, in a sixth implementation manner of the first aspect of the present invention, it is characterized in that the calculation of the trajectory similarity score corresponding to the second driving data packet based on the motion scene score, the static scene score and the first probability value includes: calculating a second probability value that the scene corresponding to the second driving data packet is a motion scene based on the first probability value; adding the product of the first probability value and the static scene score and the product of the second probability value and the motion scene score to obtain the trajectory similarity score corresponding to the second driving data packet.

[0013] The second aspect of the present invention provides an automatic driving trajectory evaluation device, comprising: a data acquisition module, used to collect a driving data set and a corresponding actual driving trajectory; a slicing module, used to slice the driving data set based on a preset time interval to obtain a plurality of scene driving data packets; a simulation module, used to simulate the scene driving data packet to obtain a simulated driving trajectory; a calculation module, used to compare the difference between the simulated driving trajectory and the actual driving trajectory, and perform similarity calculations based on the differences to obtain the trajectory similarity corresponding to each scene driving data packet.

[0014] Optionally, in a first implementation manner of the second aspect of the present invention, the calculation module includes: a set acquisition unit, used to acquire a first set of intersections passed by the actual driving trajectory corresponding to each scenario driving data packet and a second set of intersections passed by the corresponding simulated driving trajectory; a judgment unit, used to compare the difference between the first intersection set and the second intersection set, and judge whether there are serious differences between the actual driving trajectory and the simulated driving trajectory corresponding to each scenario driving data packet based on the difference; a first similarity unit, used to take the scenario driving data packet with serious differences as the first driving data packet, and set the trajectory similarity of the first driving data packet to zero; a second similarity unit, used to take the scenario driving data packet without serious differences as the second driving data packet, and perform trajectory similarity calculation on the actual driving trajectory and the simulated driving trajectory corresponding to the second driving data packet.

[0015] Optionally, in a second implementation method of the second aspect of the present invention, the judgment unit is specifically used to: judge whether the first intersection set can contain the second intersection set; if the first intersection set does not contain the second intersection set, judge whether the second intersection set can contain the first intersection set; if the second intersection set does not contain the first intersection set, determine that there are serious differences between the corresponding actual driving trajectory and the simulated driving trajectory; if the first intersection set can contain the second intersection set or the second intersection set can contain the first intersection set, determine that there are no serious differences between the corresponding actual driving trajectory and the simulated driving trajectory.

[0016] Optionally, in a third implementation manner of the second aspect of the present invention, the scene driving data packet includes displacement data; the second similarity unit specifically includes: a setting subunit, used to take the scene driving data packet without serious differences as the second driving data packet; a probability calculation subunit, used to calculate a first probability value that the scene corresponding to the second driving data packet is a static scene based on the displacement data corresponding to the second driving data packet; a motion calculation subunit, used to calculate the motion scene score corresponding to the second driving data packet based on the displacement data in the second driving data packet and a preset motion scene scoring algorithm; a static calculation subunit, used to calculate the static scene score corresponding to the second driving data packet based on the displacement data in the second driving data packet and a preset static scene scoring algorithm; a similarity calculation subunit, used to calculate the trajectory similarity score corresponding to the second driving data packet based on the motion scene score, the static scene score and the first probability value.

[0017] Optionally, in a fourth implementation manner of the second aspect of the present invention, the motion calculation subunit is specifically used to: calculate the displacement difference value between the actual driving trajectory and the simulated driving trajectory in the scene corresponding to the second driving data packet according to the displacement data in the second driving data packet; normalize the displacement difference value, and project it along the lane line direction and the direction perpendicular to the lane line, respectively, to obtain a first displacement difference in the lane line direction and a second displacement difference perpendicular to the lane line direction; calculate a first motion score in the lane line direction based on the first displacement difference; calculate a second motion score perpendicular to the lane line direction based on the second displacement difference; and calculate a motion scene score corresponding to the second driving data packet based on the first motion score and the second motion score.

[0018] Optionally, in a fifth implementation manner of the second aspect of the present invention, the static calculation subunit is specifically used to: calculate a first static score in the lane line direction based on the first displacement difference; calculate a second static score perpendicular to the lane line direction based on the second displacement difference; and calculate a static scene score corresponding to the second driving data packet based on the first static score and the second static score.

[0019] Optionally, in a sixth implementation manner of the second aspect of the present invention, the similarity calculation subunit is specifically used to: calculate, based on the first probability value, a second probability value that the scene corresponding to the second driving data packet is a motion scene; add the product of the first probability value and the static scene score and the product of the second probability value and the motion scene score to obtain a trajectory similarity score corresponding to the second driving data packet.

[0020] A third aspect of the present invention provides an autonomous driving trajectory evaluation device, comprising: a memory and at least one processor, wherein instructions are stored in the memory, and the memory and the at least one processor are interconnected through lines; the at least one processor calls the instructions in the memory so that the autonomous driving trajectory evaluation device performs the steps of the above-mentioned autonomous driving trajectory evaluation method.

[0021] A fourth aspect of the present invention provides a computer-readable storage medium, which stores instructions that, when executed on a computer, enable the computer to execute the steps of the above-mentioned autonomous driving trajectory evaluation method.

[0022] In the technical solution of the present invention, a driving data set and a corresponding actual driving trajectory are collected; the driving data set is divided based on a preset time interval to obtain multiple scene driving data packets; the scene driving data packets are simulated to obtain simulated driving trajectories; the differences between the simulated driving trajectories and the actual driving trajectories are compared, and similarity calculations are performed based on the differences to obtain the trajectory similarity corresponding to each scene driving data packet. This method collects a large amount of human driving data, then plays back the data in a simulation environment, allows the autonomous driving system to run in a simulation environment, compares the differences between the autonomous driving trajectory and the human trajectory, and gives a comprehensive score in combination with map information. The score can quantitatively, objectively and accurately evaluate the similarity between autonomous driving and human drivers, and can be used to optimize the output of the autonomous driving algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a schematic diagram of a first embodiment of an automatic driving trajectory evaluation method in an embodiment of the present invention;

[0024] Figure 2 Schematic diagram of a second embodiment of the automatic driving trajectory evaluation method in an embodiment of the present invention;

[0025] Figure 3 Schematic diagram of a third embodiment of the automatic driving trajectory evaluation method in an embodiment of the present invention;

[0026] Figure 4 A schematic diagram of an embodiment of an automatic driving trajectory evaluation device in an embodiment of the present invention;

[0027] Figure 5 It is a schematic diagram of another embodiment of the automatic driving trajectory evaluation device in an embodiment of the present invention;

[0028] Figure 6 Schematic diagram of an embodiment of an automatic driving trajectory evaluation device in an embodiment of the present invention.

[0029] Figure 7 A schematic diagram of a trajectory with serious divergence in an embodiment of the present invention;

[0030] Figure 8 is a schematic diagram of a function between the distance traveled by a human driver in a small time period before the end of a scene and the probability that the scene is a static scene in an embodiment of the present invention;

[0031] Fig. 9 Schematic diagram of a function between the difference in driving distance between the autonomous driving and the human driver in the direction perpendicular to the lane line and the first motion score in an embodiment of the present invention;

[0032] Fig.10A schematic diagram of a function between the time required for the automatic driving final position to travel to the final position of the human driver and the second motion score in an embodiment of the present invention;

[0033] Fig.11 Schematic diagram of a function between the difference in driving distance along the lane line between the autonomous driving and the human driver and the second stationary score in an embodiment of the present invention;

[0034] Fig.12 Schematic diagram of a function between the difference in driving distance between the autonomous driving and the human driver in the direction perpendicular to the lane line and the first stationary score in an embodiment of the present invention. DETAILED DESCRIPTION

[0035] The embodiments of the present application provide an autonomous driving trajectory evaluation method, apparatus, device and storage medium, which are used to solve the technical problem that the existing autonomous driving trajectory evaluation method lacks a quantitative and accurate similarity judgment method.

[0036] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0037] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , a first embodiment of the automatic driving trajectory evaluation method in an embodiment of the present invention includes:

[0038] 101. Collect driving data sets and corresponding actual driving trajectories;

[0039] It is understandable that the execution subject of the present invention may be an automatic driving trajectory evaluation device, or a terminal or a server, which is not limited here. The embodiment of the present invention is described by taking a server as the execution subject as an example.

[0040] In this embodiment, the positioning module, perception module and data acquisition module of the automatic driving system are turned on. The driver drives the car in various scenarios. The data acquisition module records the passing position, the vehicle speed, acceleration and surrounding obstacle information in real time, and uses this information as a driving data set. By connecting the passing positions of the vehicle recorded by the data acquisition module, the corresponding actual driving trajectory can be obtained.

[0041] 102. Segment the driving data set based on a preset time interval to obtain multiple scene driving data packets;

[0042] In this embodiment, if a long simulation is performed, the driving trajectory of the autonomous driving vehicle and the human driver in the original data may be quite different, resulting in the deviation of the relationship between the autonomous driving vehicle and the surrounding obstacles from reality. Therefore, the collected driving data needs to be segmented based on a preset time interval, wherein the time interval can be set according to actual needs. In this embodiment, taking half a minute as an example, the driving data set is segmented into data packets of each half a minute, and in order to ensure the integrity of a scene during segmentation, a certain scene slicing rule can be set. For example, after the driving data set is time-sliced, the scene of each data packet is identified, wherein a data packet is identified to be in the middle of a certain intersection at the end of the slice time, and the scene should correspond to the vehicle driving through the intersection. However, after slicing, the vehicle at the end of the data packet drives to the middle of the intersection, and the integrity of the scene cannot be guaranteed. Here, the slicing time of the corresponding driving scene data packet can be extended for a certain time to ensure that the vehicle in the scene corresponding to the driving data packet completely passes through the intersection. After slicing the driving data set, simulation is performed on each segmented data packet.

[0043] 103. Simulate the scene driving data packet to obtain a simulated driving trajectory;

[0044] In this embodiment, the collected driving data packets are played back in the simulation environment, and the autonomous driving software is started in the simulation environment. The autonomous driving system receives the positioning and perception information in the data, and outputs control instructions to control the driving of the autonomous driving vehicle in the simulation to obtain a simulated driving trajectory.

[0045] 104. Compare the differences between the simulated driving trajectory and the actual driving trajectory, and perform similarity calculations based on the differences to obtain the trajectory similarity corresponding to the driving data packets of each scenario.

[0046] In this embodiment, each driving data packet is divided into time slices based on about half a minute. The simulated driving trajectory and the actual driving trajectory corresponding to each driving data packet are different. The intersections passed by the simulated driving trajectory corresponding to each driving data packet and the intersections passed by the actual driving trajectory are counted respectively. The above difference is the difference in the intersections passed by the two. For scene driving data packets with large differences between the two, the corresponding trajectory similarity is directly set to zero. For scene driving data packets with small differences between the two, further calculations are performed based on the data in the scene driving data packets.

[0047] In this embodiment, a driving data set and a corresponding actual driving trajectory are collected; the driving data set is divided based on a preset time interval to obtain multiple scene driving data packets; the scene driving data packets are simulated to obtain simulated driving trajectories; the differences between the simulated driving trajectories and the actual driving trajectories are compared, and similarity calculations are performed based on the differences to obtain the trajectory similarity corresponding to each scene driving data packet. This method collects a large amount of human driving data, then plays back the data in a simulation environment, allows the autonomous driving system to run in a simulation environment, compares the differences between the autonomous driving trajectory and the human trajectory, and gives a comprehensive score in combination with map information. The score can quantitatively, objectively and accurately evaluate the similarity between autonomous driving and human drivers, and can be used to optimize the output of the autonomous driving algorithm.

[0048] See also Figure 2 , a second embodiment of the automatic driving trajectory evaluation method in an embodiment of the present invention includes:

[0049] 201. Collecting driving data sets and corresponding actual driving trajectories;

[0050] 202. Segment the driving data set based on a preset time interval to obtain a plurality of scene driving data packets;

[0051] 203. Simulate the scene driving data packet to obtain a simulated driving trajectory;

[0052] Steps 201-203 in this embodiment are similar to steps 101-103 in the first embodiment, and are not described again here.

[0053] 204. Obtain a set of first intersections passed by an actual driving trajectory corresponding to each scenario driving data packet and a set of second intersections passed by a corresponding simulated driving trajectory;

[0054] In this embodiment, the GPS positioning system can be used to obtain map information corresponding to the coordinates of the vehicle during driving, obtain the intersections passed during driving, and record the numbers of the intersections passed by the driver in the actual driving trajectory into a first intersection set. By simulating the scene driving data packet, a simulated driving trajectory is obtained, the intersections passed by the simulated driving trajectory are obtained, and the intersection numbers are recorded in a second intersection set.

[0055] 205. Determine whether the first intersection set can include the second intersection set;

[0056] 206. If the first intersection set does not include the second intersection set, determining whether the second intersection set can include the first intersection set;

[0057] 207. If the second intersection set does not include the first intersection set, it is determined that there is a serious disagreement between the corresponding actual driving trajectory and the simulated driving trajectory;

[0058] In this embodiment, if the first intersection set cannot contain the second intersection set and the second intersection set cannot contain the first intersection set, it means that the autonomous driving vehicle in the simulation and the human driver in the actual situation drove to different intersections, and the trajectories of the two are seriously divergent. Figure 7 As shown, the rectangle for straight driving is the simulated driving trajectory of the autonomous driving vehicle and the rectangle for arc driving is the actual driving trajectory of the human driver. They go to intersection A and intersection B respectively. The corresponding first intersection set cannot contain the second intersection set and the second intersection set cannot contain the first intersection set, which meets the condition of serious trajectory divergence.

[0059] 208. If the first intersection set can contain the second intersection set or the second intersection set can contain the first intersection set, then determining that there is no serious disagreement between the corresponding actual driving trajectory and the simulated driving trajectory;

[0060] 209. Using the scene driving data packet with serious disagreement as the first driving data packet, and setting the trajectory similarity of the first driving data packet to zero;

[0061] In this embodiment, for scene driving data packets with serious disagreements, the corresponding trajectory similarity is directly set to zero, and for scene driving data packets without serious disagreements, further calculation is performed based on the data in the scene driving data packets.

[0062] 210. The scenario driving data packet without serious disagreement is used as the second driving data packet, and trajectory similarity calculation is performed on the actual driving trajectory and the simulated driving trajectory corresponding to the second driving data packet.

[0063] Based on the previous embodiment, this embodiment describes in detail the process of comparing the difference between the simulated driving trajectory and the actual driving trajectory, and performing similarity calculations according to the differences to obtain the trajectory similarity corresponding to each scene driving data packet. By obtaining the first intersection set through which the actual driving trajectory corresponding to each scene driving data packet passes and the second intersection set through which the corresponding simulated driving trajectory passes; comparing the difference between the first intersection set and the second intersection set, judging whether there is a serious difference between the actual driving trajectory and the simulated driving trajectory corresponding to each scene driving data packet based on the difference; taking the scene driving data packet with serious differences as the first driving data packet, and setting the trajectory similarity of the first driving data packet to zero; taking the scene driving data packet without serious differences as the second driving data packet, and performing trajectory similarity calculations on the actual driving trajectory and the simulated driving trajectory corresponding to the second driving data packet. This method collects a large amount of human driving data, and then plays back the data in a simulation environment, allowing the automatic driving system to run in a simulation environment, comparing the difference between the automatic driving trajectory and the human trajectory, and giving a comprehensive score in combination with map information. The score can quantitatively, objectively and accurately evaluate the similarity between automatic driving and human drivers, and can be used to optimize the output of the automatic driving algorithm.

[0064] See also Figure 3 , a third embodiment of the automatic driving trajectory evaluation method in the embodiment of the present invention includes:

[0065] 301. Collecting driving data sets and corresponding actual driving trajectories;

[0066] 302. Segment the driving data set based on a preset time interval to obtain a plurality of scene driving data packets;

[0067] 303. Simulate the scene driving data packet to obtain a simulated driving trajectory;

[0068] 304. Obtain a set of first intersections passed by an actual driving trajectory corresponding to each scenario driving data packet and a set of second intersections passed by a corresponding simulated driving trajectory;

[0069] 305. Compare the difference between the first intersection set and the second intersection set, and determine whether there is a serious disagreement between the actual driving trajectory and the simulated driving trajectory corresponding to the driving data package of each scenario based on the difference;

[0070] 306. Using the scene driving data packet with serious disagreement as the first driving data packet, and setting the trajectory similarity of the first driving data packet to zero;

[0071] Steps 301 - 306 in this embodiment are similar to steps 201 - 206 in the second embodiment and are not described again here.

[0072] 307. Using the scenario driving data packet without serious disagreement as the second driving data packet;

[0073] 308. Calculate, according to the displacement data corresponding to the second driving data packet, a first probability value that the scene corresponding to the second driving data packet is a stationary scene;

[0074] In this embodiment, for the second driving data packet, the driver's displacement s0 within a small time period before the corresponding scene ends is calculated, wherein the small time period is a time less than a preset time interval, such as 5s.

[0075] In this embodiment, the smaller s0 is, the more likely the corresponding scene is a scene that should be stationary, such as a traffic jam ahead where the vehicle needs to stop and queue up, or a stop at a parking spot where passengers need to get on and off. As s0 increases, the possibility that the scene is a scene that should be moving gradually increases. By statistically analyzing a large number of collected scenes, the probability P of a scene being a scene that should be stationary is determined as follows: Figure 8 As shown, the function form is y1=1 / (1.0+exp(a1*(x1+b1))), where x1 is the driver's displacement s0 in a small time period before the end of the corresponding scene, y1 is the probability that the corresponding scene is a scene that should be stationary, a1 and b1 are constants calculated based on a large amount of collected data, and the corresponding constants may be different based on different driving data.

[0076] 309. Calculate, according to the displacement data in the second driving data packet, a displacement difference value between the actual driving trajectory and the simulated driving trajectory in the scene corresponding to the second driving data packet;

[0077] 310. Normalize the displacement difference value, and project it along the lane line direction and in a direction perpendicular to the lane line, respectively, to obtain a first displacement difference in the lane line direction and a second displacement difference in the direction perpendicular to the lane line, respectively;

[0078] 311. Calculate a first motion score in the lane line direction based on the first displacement difference;

[0079] 312. Calculate a second motion score perpendicular to the lane line direction based on the second displacement difference;

[0080] In this embodiment, when the difference of the trajectory perpendicular to the lane line (lateral direction) exceeds about 3 meters during the driving process, the vehicle is one lane away, while in the direction along the lane line (longitudinal direction), the difference of about 3 meters is a small difference. Therefore, when calculating the score, it is necessary to simulate the formal trajectory and the actual driving trajectory and project them along the lane line and perpendicular to the lane line respectively to obtain the displacement difference in two directions, namely, the displacement difference l along the lane line direction (longitudinal direction) s and perpendicular to the lane line direction (lateral direction). According to statistics of a large number of collected scenes, the change of the first motion score move_l_score with the displacement difference is determined as follows: Fig. 9 As shown, move_l_score adopts the function form of y2=a2*log(1 / (b2*x2+0.5)-1)+1, where x2 is the difference in driving distance between the autonomous driving and human driver trajectories in the direction perpendicular to the lane line, y2 is the first movement score, a2 and b2 are constants calculated based on a large amount of collected data, and the corresponding constants may be different based on different driving data.

[0081] In this embodiment, since the calculation is for a scene that should be moving, in addition to considering the difference in driving distance between the automatic driving and the human driver, the speed of the vehicle in the scene also needs to be considered when calculating the longitudinal score. If the longitudinal driving distance of the automatic driving is greater than the longitudinal driving distance of the human driver, the second movement score move_s_score is set to 1.0. If it is less than the longitudinal driving distance of the human driver, the average speed v of the human driver in the last short period of time in the scene is calculated, and the distance traveled by the human driver in the longitudinal direction more than the automatic driving is divided by v to obtain the time t required to travel from the final position of the automatic driving to the final position of the human driver. According to statistics of a large number of collected scenes, the relationship between move_s_score and t is obtained, as shown in FIG. Fig.10 As shown, the functional relationship is in the form of y3=1 / (1.0+exp(a3*(x3+b3))), where x3 is the time required for the autonomous driving to travel from the final position to the final position of the human driver, y3 is the second motion score, and a3 and b3 are constants calculated based on a large amount of collected data. The corresponding constants may be different based on different driving data.

[0082] 313. Calculate a motion scene score corresponding to the second driving data packet according to the first motion score and the second motion score;

[0083] In this embodiment, the motion scene score is calculated based on the first motion score and the second motion score, and the calculation formula is move_score=max(0.0, move_s_score+move_l_score-1.0).

[0084] 314. Calculate a first stationary score in the lane line direction based on the first displacement difference;

[0085] 315. Calculate a second stationary score perpendicular to the lane line direction based on the second displacement difference;

[0086] In this embodiment, the calculation of static_l_score is similar to the calculation of move_l_score, which is calculated by the difference in lateral distance traveled by the autonomous driving and the human driver, but the coefficients are different. According to statistics of a large number of collected scenes, the second static score static_s_score and the first static score static_l_score are determined according to the change of the displacement difference as follows: Fig.11 and Fig.12 As shown, static_s_score adopts the function form of y3=1 / (1.0+exp(a3*(x3+b3))), wherein x3 is the difference in driving distance between the autonomous driving and human driver trajectories in the direction along the lane line, y3 is the second static score, a3 and b3 are constants calculated based on a large amount of collected data, and the corresponding constants may be different based on different driving data. Static_l_score adopts the function form of y4=a4*log(1 / (b4*x4+0.5)-1)+1, wherein x4 is the difference in driving distance between the autonomous driving and human driver trajectories in the direction perpendicular to the lane line, y4 is the first static score, a4 and b4 are constants calculated based on a large amount of collected data, and the corresponding constants may be different based on different driving data, wherein the human driver trajectory is the actual driving trajectory.

[0087] 316. Calculate a stationary scene score corresponding to the second driving data packet according to the first stationary score and the second stationary score;

[0088] In this embodiment, the still scene score is calculated based on the first still score and the second still score, and the calculation formula is move_score=max(0.0,move_s_score+move_l_score-1.0).

[0089] 317. Calculate, based on the first probability value, a second probability value that the scene corresponding to the second driving data packet is a motion scene;

[0090] In this embodiment, the first probability value is P, and the corresponding second probability value of the scene being a motion scene is 1-P.

[0091] 318. Add the product of the first probability value and the static scene score and the product of the second probability value and the motion scene score to obtain a trajectory similarity score corresponding to the second driving data packet.

[0092] In this embodiment, progress_score=P*static_score+(1-P)*move_score. For example, in a lane change scenario, assume that the longitudinal and lateral distances traveled by the autonomous driving are 24.3 meters and -0.4 meters respectively (the offset to the left hand side of the driving direction is negative), and the longitudinal and lateral distances traveled by the human driver are 24.0 meters and 3.7 meters respectively, that is, the longitudinal driving distances of the autonomous driving and the human driver are similar, and the human driver completes a lane change to the right. Assume that the distance traveled by the human driver in the last 5 seconds of the scene is 12.4 meters. The calculated probability that the scene is a scene that should be stationary is 0, move_s_score is 1.0, move_l_score is 0.73, move_score is 0.73, and the calculated progress_score is 0.73.

[0093] For another example, in a scenario where the vehicle stops to pick up a passenger, assume that the longitudinal and lateral distances traveled by the autonomous driving are 50.1 meters and -0.3 meters respectively, and the longitudinal and lateral distances traveled by the human driver are 51.1 meters and 1.1 meters respectively, that is, the human driver parks the vehicle more towards the right side of the road. Assume that the displacement of the human driver in the last 5 seconds of the scenario is 0.1 meters. The probability that the scenario is a static one is 1.0, the static_s_score is 1.0, the static_l_score is 0.78, the static_score is 0.78, and the progress_score is 0.78.

[0094] Based on the previous embodiment, this embodiment describes in detail the process of using a scene driving data packet without serious disagreement as the second driving data packet, and calculating the trajectory similarity between the actual driving trajectory and the simulated driving trajectory corresponding to the second driving data packet, by using a scene driving data packet without serious disagreement as the second driving data packet; calculating the first probability value that the scene corresponding to the second driving data packet is a static scene according to the displacement data corresponding to the second driving data packet; calculating the motion scene score corresponding to the second driving data packet according to the displacement data in the second driving data packet and the preset motion scene scoring algorithm; calculating the static scene score corresponding to the second driving data packet according to the displacement data in the second driving data packet and the preset static scene scoring algorithm; and calculating the trajectory similarity score corresponding to the second driving data packet based on the motion scene score, the static scene score and the first probability value. By collecting a large amount of human driving data, and then playing back the data in a simulation environment, allowing the automatic driving system to run in the simulation environment, comparing the difference between the automatic driving trajectory and the human trajectory, and combining the map information to give a comprehensive score, the score can quantitatively, objectively and accurately evaluate the similarity between the automatic driving and human drivers, and can be used to optimize the output of the automatic driving algorithm.

[0095] The above describes the automatic driving trajectory evaluation method in the embodiment of the present invention. The following describes the automatic driving trajectory evaluation device in the embodiment of the present invention. Figure 4 , an embodiment of the automatic driving trajectory evaluation device in the embodiment of the present invention includes:

[0096] The data collection module 401 is used to collect driving data sets and corresponding actual driving trajectories;

[0097] A slicing module 402, configured to slice the driving data set based on a preset time interval to obtain a plurality of scene driving data packets;

[0098] A simulation module 403 is used to simulate the scenario driving data packet to obtain a simulated driving trajectory;

[0099] The calculation module 404 is used to compare the difference between the simulated driving trajectory and the actual driving trajectory, and perform similarity calculations according to the differences to obtain trajectory similarities corresponding to each scene driving data package.

[0100] In an embodiment of the present invention, the automatic driving trajectory evaluation device runs the automatic driving trajectory evaluation method, and the automatic driving trajectory evaluation device collects a driving data set and a corresponding actual driving trajectory; divides the driving data set based on a preset time interval to obtain multiple scene driving data packets; simulates the scene driving data packets to obtain simulated driving trajectories; compares the difference between the simulated driving trajectory and the actual driving trajectory, and performs similarity calculations based on the difference to obtain the trajectory similarity corresponding to each scene driving data packet. This method collects a large amount of human driving data, and then plays back the data in a simulation environment, allowing the automatic driving system to run in a simulation environment, compares the difference between the automatic driving trajectory and the human trajectory, and gives a comprehensive score in combination with map information. The score can quantitatively, objectively and accurately evaluate the similarity between the automatic driving and human drivers, and can be used to optimize the output of the automatic driving algorithm.

[0101] See also Figure 5 , a second embodiment of the automatic driving trajectory evaluation device in the embodiment of the present invention includes:

[0102] The data collection module 401 is used to collect driving data sets and corresponding actual driving trajectories;

[0103] A slicing module 402, configured to slice the driving data set based on a preset time interval to obtain a plurality of scene driving data packets;

[0104] A simulation module 403 is used to simulate the scenario driving data packet to obtain a simulated driving trajectory;

[0105] The calculation module 404 is used to compare the difference between the simulated driving trajectory and the actual driving trajectory, and perform similarity calculations according to the differences to obtain trajectory similarities corresponding to each scene driving data package.

[0106] In this embodiment, the calculation module 404 includes: a set acquisition unit 4041, which is used to obtain a first set of intersections passed by the actual driving trajectory corresponding to each scene driving data packet and a second set of intersections passed by the corresponding simulated driving trajectory; a judgment unit 4042, which is used to compare the difference between the first set of intersections and the second set of intersections, and judge whether there are serious differences between the actual driving trajectory and the simulated driving trajectory corresponding to each scene driving data packet based on the difference; a first similarity unit 4043, which is used to use the scene driving data packet with serious differences as the first driving data packet, and set the trajectory similarity of the first driving data packet to zero; a second similarity unit 4044, which is used to use the scene driving data packet without serious differences as the second driving data packet, and perform trajectory similarity calculation on the actual driving trajectory and the simulated driving trajectory corresponding to the second driving data packet.

[0107] In this embodiment, the judgment unit 4042 is specifically used to: judge whether the first intersection set can contain the second intersection set; if the first intersection set does not contain the second intersection set, judge whether the second intersection set can contain the first intersection set; if the second intersection set does not contain the first intersection set, determine that there are serious differences between the corresponding actual driving trajectory and the simulated driving trajectory; if the first intersection set can contain the second intersection set or the second intersection set can contain the first intersection set, determine that there are no serious differences between the corresponding actual driving trajectory and the simulated driving trajectory.

[0108] In this embodiment, the scene driving data packet includes displacement data; the second similarity unit 4044 specifically includes: a setting subunit 40441, used to use the scene driving data packet without serious disagreement as the second driving data packet; a probability calculation subunit 40442, used to calculate the first probability value that the scene corresponding to the second driving data packet is a static scene according to the displacement data corresponding to the second driving data packet; a motion calculation subunit 40443, used to calculate the motion scene score corresponding to the second driving data packet according to the displacement data in the second driving data packet and a preset motion scene scoring algorithm; a static calculation subunit 40444, used to calculate the static scene score corresponding to the second driving data packet according to the displacement data in the second driving data packet and a preset static scene scoring algorithm; a similarity calculation subunit 40445, used to calculate the trajectory similarity score corresponding to the second driving data packet based on the motion scene score, the static scene score and the first probability value.

[0109] In this embodiment, the motion calculation subunit 40443 is specifically used to: calculate the displacement difference value between the actual driving trajectory and the simulated driving trajectory in the scene corresponding to the second driving data packet according to the displacement data in the second driving data packet; normalize the displacement difference value, and project it along the lane line direction and the direction perpendicular to the lane line, respectively, to obtain a first displacement difference in the lane line direction and a second displacement difference perpendicular to the lane line direction; calculate a first motion score in the lane line direction based on the first displacement difference; calculate a second motion score perpendicular to the lane line direction based on the second displacement difference; and calculate a motion scene score corresponding to the second driving data packet according to the first motion score and the second motion score.

[0110] In this embodiment, the static calculation subunit 40444 is specifically used to: calculate a first static score in the lane line direction based on the first displacement difference; calculate a second static score perpendicular to the lane line direction based on the second displacement difference; and calculate a static scene score corresponding to the second driving data packet based on the first static score and the second static score.

[0111] In this embodiment, the similarity calculation subunit 40445 is specifically used to: calculate, based on the first probability value, a second probability value that the scene corresponding to the second driving data packet is a motion scene; add the product of the first probability value and the static scene score and the product of the second probability value and the motion scene score to obtain a trajectory similarity score corresponding to the second driving data packet.

[0112] Based on the previous embodiment, this embodiment describes in detail the specific functions of each module and the unit structure of some modules. By collecting a large amount of human driving data and then playing back the data in a simulation environment, the autonomous driving system is allowed to run in a simulation environment, and the differences between the autonomous driving trajectory and the human trajectory are compared. A comprehensive score is given in combination with map information. This score can quantitatively, objectively and accurately evaluate the similarity between autonomous driving and human drivers, and can be used to optimize the output of the autonomous driving algorithm.

[0113] above Figure 4 and Figure 5 The autonomous driving trajectory evaluation device in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The autonomous driving trajectory evaluation device in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0114] Figure 6 6 is a schematic diagram of the structure of an automatic driving trajectory evaluation device provided by an embodiment of the present invention. The automatic driving trajectory evaluation device 600 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 610 (for example, one or more processors) and a memory 620, and one or more storage media 630 (for example, one or more mass storage devices) storing application programs 633 or data 632. Among them, the memory 620 and the storage medium 630 may be short-term storage or permanent storage. The program stored in the storage medium 630 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the automatic driving trajectory evaluation device 600. Furthermore, the processor 610 may be configured to communicate with the storage medium 630, and execute a series of instruction operations in the storage medium 630 on the automatic driving trajectory evaluation device 600 to implement the steps of the above-mentioned automatic driving trajectory evaluation method.

[0115] The autonomous driving trajectory evaluation device 600 may also include one or more power supplies 640, one or more wired or wireless network interfaces 650, one or more input and output interfaces 660, and / or one or more operating systems 631, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will appreciate that Figure 6 The structure of the autonomous driving trajectory evaluation device shown does not constitute a limitation on the autonomous driving trajectory evaluation device provided by the present application, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0116] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the autonomous driving trajectory evaluation method.

[0117] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, device, or unit can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here.

[0118] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0119] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An automatic driving trajectory evaluation method, characterized in that: The automatic driving trajectory evaluation method comprises: Collect driving data sets and corresponding actual driving trajectories; Segmenting the driving data set based on a preset time interval to obtain a plurality of scene driving data packets; Simulating the scenario driving data packet to obtain a simulated driving trajectory; Comparing the difference between the simulated driving trajectory and the actual driving trajectory, and performing similarity calculations according to the differences to obtain trajectory similarities corresponding to driving data packets of each scenario; The comparing the difference between the simulated driving trajectory and the actual driving trajectory, and performing similarity calculations according to the differences to obtain trajectory similarities corresponding to driving data packets of each scenario includes: Obtain a set of first intersections passed by an actual driving trajectory corresponding to each scenario driving data packet and a set of second intersections passed by a corresponding simulated driving trajectory; Comparing the difference between the first intersection set and the second intersection set, and judging whether there is a serious disagreement between the actual driving trajectory and the simulated driving trajectory corresponding to the driving data package of each scenario based on the difference; The scene driving data packet with serious disagreement is taken as the first driving data packet, and the trajectory similarity of the first driving data packet is set to zero; The scenario driving data packet without serious disagreement is used as the second driving data packet, and the trajectory similarity calculation is performed on the displacement difference projected along the lane line direction and the displacement difference projected perpendicular to the lane line direction for the actual driving trajectory and the simulated driving trajectory corresponding to the second driving data packet.

2. The automatic driving trajectory evaluation method according to claim 1, characterized in that: The comparing the difference between the first intersection set and the second intersection set, and judging whether there is a serious disagreement between the actual driving trajectory and the simulated driving trajectory corresponding to the driving data package of each scenario based on the difference comprises: Determining whether the first intersection set includes the second intersection set; If the first intersection set does not include the second intersection set, determining whether the second intersection set can include the first intersection set; If the second intersection set does not include the first intersection set, it is determined that there is a serious disagreement between the corresponding actual driving trajectory and the simulated driving trajectory; If the first intersection set can include the second intersection set or the second intersection set can include the first intersection set, it is determined that there is no serious disagreement between the corresponding actual driving trajectory and the simulated driving trajectory.

3. The automatic driving trajectory evaluation method according to claim 2, characterized in that: The scenario driving data packet includes displacement data; The method of using the scene driving data packet without serious disagreement as the second driving data packet, and calculating the trajectory similarity of the displacement difference projected along the lane line direction and the displacement difference projected perpendicular to the lane line direction on the actual driving trajectory and the simulated driving trajectory corresponding to the second driving data packet, comprises: The scenario driving data package without serious disagreement is used as the second driving data package; calculating, according to the displacement data corresponding to the second driving data packet, a first probability value that the scene corresponding to the second driving data packet is a stationary scene; Calculating a motion scene score corresponding to the second driving data packet according to the displacement data in the second driving data packet and a preset motion scene scoring algorithm; Calculating a stationary scene score corresponding to the second driving data packet according to the displacement data in the second driving data packet and a preset stationary scene score algorithm; A trajectory similarity score corresponding to the second driving data packet is calculated based on the motion scene score, the static scene score and the first probability value.

4. The automatic driving trajectory evaluation method according to claim 3, characterized in that: The calculating the motion scene score corresponding to the second driving data packet according to the displacement data in the second driving data packet and a preset motion scene score algorithm comprises: Calculating, according to the displacement data in the second driving data packet, a displacement difference value between an actual driving trajectory and a simulated driving trajectory in a scene corresponding to the second driving data packet; The displacement difference value is normalized, and projected along the lane line direction and perpendicular to the lane line direction, respectively, to obtain a first displacement difference in the lane line direction and a second displacement difference perpendicular to the lane line direction; Calculating a first motion score in the lane line direction based on the first displacement difference; Calculating a second motion score perpendicular to the lane line direction based on the second displacement difference; A motion scene score corresponding to the second driving data packet is calculated according to the first motion score and the second motion score.

5. The automatic driving trajectory evaluation method according to claim 4, characterized in that: The calculating the static scene score corresponding to the second driving data packet according to the displacement data in the second driving data packet and a preset static scene score algorithm comprises: Calculating a first stationary score in the lane line direction based on the first displacement difference; Calculating a second stationary score perpendicular to the lane line direction based on the second displacement difference; A stationary scene score corresponding to the second driving data packet is calculated according to the first stationary score and the second stationary score.

6. The automatic driving trajectory evaluation method according to any one of claims 3 to 5, characterized in that: The calculating the trajectory similarity score corresponding to the second driving data packet based on the motion scene score, the static scene score and the first probability value comprises: Based on the first probability value, calculating a second probability value that the scene corresponding to the second driving data packet is a motion scene; The product of the first probability value and the static scene score and the product of the second probability value and the motion scene score are added to obtain a trajectory similarity score corresponding to the second driving data packet.

7. An automatic driving trajectory evaluation device, characterized in that: The automatic driving trajectory evaluation device comprises: A data collection module, used to collect driving data sets and corresponding actual driving trajectories; A slicing module, used for slicing the driving data set based on a preset time interval to obtain a plurality of scene driving data packets; A simulation module, used to simulate the scenario driving data packet to obtain a simulated driving trajectory; A calculation module, used for comparing the difference between the simulated driving trajectory and the actual driving trajectory, and performing similarity calculations according to the differences, so as to obtain trajectory similarities corresponding to driving data packets of each scenario; The calculation module comprises: A set acquisition unit, used to acquire a set of first intersections passed by an actual driving trajectory corresponding to each scenario driving data packet and a set of second intersections passed by a corresponding simulated driving trajectory; A judgment unit, configured to compare the difference between the first intersection set and the second intersection set, and judge whether there is a serious divergence between the actual driving trajectory and the simulated driving trajectory corresponding to the driving data package of each scenario based on the difference; A first similarity unit, used to take the scene driving data packet with serious disagreement as the first driving data packet, and set the trajectory similarity of the first driving data packet to zero; The second similarity unit is used to use the scene driving data packet without serious disagreement as the second driving data packet, and to calculate the trajectory similarity of the actual driving trajectory and the simulated driving trajectory corresponding to the second driving data packet, in terms of the displacement difference projected along the lane line direction and the displacement difference projected perpendicular to the lane line direction.

8. The automatic driving trajectory evaluation device according to claim 7, characterized in that: The judgment unit is specifically used to: judge whether the first intersection set can contain the second intersection set; if the first intersection set does not contain the second intersection set, judge whether the second intersection set can contain the first intersection set; if the second intersection set does not contain the first intersection set, determine that there are serious differences between the corresponding actual driving trajectory and the simulated driving trajectory; if the first intersection set can contain the second intersection set or the second intersection set can contain the first intersection set, determine that there are no serious differences between the corresponding actual driving trajectory and the simulated driving trajectory.

9. An automatic driving trajectory evaluation device, characterized in that: The autonomous driving trajectory evaluation device comprises: a memory and at least one processor, the memory storing instructions, the memory and the at least one processor being interconnected via a line; The at least one processor calls the instructions in the memory so that the autonomous driving trajectory evaluation device performs the steps of the autonomous driving trajectory evaluation method according to any one of claims 1 to 6.

10. 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 autonomous driving trajectory assessment method according to any one of claims 1 to 6 are implemented.

Images