Data-driven adaptive curve speed planning control method, system and vehicle

Abstract

The embodiment of the present invention discloses a data-driven adaptive curve speed planning control method, system and vehicle, which relates to the field of intelligent driving technology and is invented to better reflect the intelligence and anthropomorphism of autonomous driving. The method includes: taking the path point trace output by the path planning module of the autonomous driving system as input, determining the curvature value of each path point; traversing all path points, selecting the path point with the largest curvature as the anchor point, and taking the longitudinal position of the path point as the anchor distance; determining the corresponding lateral acceleration according to the curvature value and anchor distance of the anchor point; online learning the device lateral acceleration corresponding to different curvature radius sections when the vehicle is turning, obtaining the deviation value between the expected lateral acceleration under different curvature radii and the preset reference lateral acceleration, and correcting the lateral acceleration under different curvature radii according to the deviation value. The embodiment of the present invention is suitable for intelligent driving control of curves.

Description

Technical Field

[0001] The present invention relates to the technical field of path planning and vehicle dynamic control, and in particular to a data-driven adaptive curve speed planning control method, system and vehicle. Background Art

[0002] With the continuous advancement of autonomous driving technology, driving experience has gradually become the research focus of autonomous driving. For vehicles with longitudinal autonomous driving control functions, the road conditions are complex and changeable. Among them, curved roads with different curvatures are important scenarios for testing autonomous driving performance. The safety, smoothness and driving efficiency of driving on roads with large curvatures directly affect the driver's driving experience. The vehicle's cornering performance reflects the quality of the autonomous driving algorithm to a certain extent.

[0003] Existing autonomous driving algorithms generally lack interaction with drivers and cannot deeply reflect the intelligence and anthropomorphism of autonomous driving. Summary of the invention

[0004] In view of the above problems, the present invention provides a data-driven adaptive curve speed planning control method, system and vehicle, which can better reflect the intelligence and anthropomorphism of autonomous driving.

[0005] According to one aspect of the present invention, a data-driven adaptive curve speed planning control method is provided, comprising:

[0006] Step S1, taking the path point trace output by the path planning module of the autonomous driving system as input, determining the curvature value of each path point to eliminate the influence of the instability of the perception input and the incomplete match between the vehicle's driving path and the lane centerline;

[0007] Step S2, traversing all path points, selecting the path point with the largest curvature as the anchor point, and taking the longitudinal position of the path point as the anchor distance; determining the corresponding lateral acceleration according to the curvature value of the anchor point and the anchor distance;

[0008] Step S3, online learning the lateral acceleration corresponding to sections with different curvature radii when the driver drives the vehicle through a curve, obtaining the deviation value between the driver's expected lateral acceleration and the preset reference lateral acceleration under different curvature radii, and correcting the lateral acceleration under different curvature radii according to the deviation value.

[0009] In an optional manner, determining the curvature value of each path point further includes:

[0010] Select three consecutive points A, B, and C from the path points, where B is the path point whose curvature is to be calculated, and A and C are the previous point and the next point of point B, respectively;

[0011] Calculate the lengths of the three sides a, b, and c of triangle ABC, where a is the length of AB, b is the length of BC, and c is the length of AC;

[0012] Calculate the sine value of angle B using the cosine theorem and trigonometric functions. According to the sine value of angle B and the length of BC, obtain the radius r of the circumscribed circle and the corresponding curvature K.

[0013] The curvature K is taken as the curvature value of the current path point B.

[0014] In an optional manner, selecting the path point with the largest curvature as the anchor point further includes:

[0015] According to the current speed of the vehicle and the preset reaction deceleration, the safe distance required for the vehicle to slow down from the current speed to a standstill at the reaction deceleration is calculated;

[0016] Traverse all discrete points on the path and take each discrete point as a potential candidate anchor point;

[0017] Extending the candidate anchor point forward and backward by a safe distance to determine a candidate range, and taking all path points within the candidate range as key points affecting the vehicle driving decision;

[0018] The curvature value of each path point within the candidate range is calculated, and the path point with the largest curvature value is selected as the anchor point.

[0019] In an optional manner, the step S3 further comprises:

[0020] Ensure that the driving scene meets the trigger conditions for collecting user driving data. When the trigger conditions are met, continue to collect driving data at a preset frequency until the radius of curvature is greater than a set threshold, the lateral acceleration is 0, or the maximum data collection time is reached; wherein the driving data includes the radius of curvature of the road and the lateral acceleration of the vehicle;

[0021] Through the initial curvature radius table and the corresponding default reference lateral acceleration table, each collected curvature radius is linearly interpolated to obtain the corresponding default reference lateral acceleration;

[0022] All collected data are grouped according to the curvature radius, and the acceleration differences in each group of data are weighted averaged to obtain the average deviation value between the driver's expected lateral acceleration and the preset reference lateral acceleration under the curvature radius;

[0023] The lateral accelerations at different curvature radii are corrected according to the average deviation value.

[0024] In an optional manner, the method further includes

[0025] For each lateral acceleration corresponding to the radius of curvature in the initial lateral acceleration table, add the lateral acceleration to the corresponding difference in the difference table to obtain an updated lateral acceleration; wherein the difference table includes the expected lateral acceleration and the reference lateral acceleration;

[0026] The difference table is updated according to the collected new cornering driving data and the learning rate, wherein the updating formula of the difference table is:

[0027]

[0028] in, The table value updated for the last study; is the difference obtained by current learning; is the learning rate.

[0029] In this embodiment, before step S3, first-order linear interpolation is performed on the anchor point curvature radius in the curvature radius table according to the given curvature radius table and the default reference lateral acceleration table to obtain the comfortable lateral acceleration at the corresponding anchor point;

[0030] After completing the collection of driving data and online learning of lateral acceleration, Fig.10 As shown, according to the updated curvature radius table and the updated reference lateral acceleration table, the first-order linear interpolation is performed again on the anchor point curvature radius in the curvature radius table to obtain the updated comfortable lateral acceleration at the corresponding anchor point;

[0031] According to the given curvature radius and the comfortable lateral acceleration, a curvature speed limit is obtained to guide the speed control of the automatic driving system in the curve;

[0032] The expected acceleration is obtained according to the curvature speed limit, the anchor distance and the current vehicle speed.

[0033] In an optional manner, the anchor distance is calculated as follows:

[0034] ;

[0035] in, is the current vehicle speed; is the reaction deceleration.

[0036] In an optional manner, the method further includes:

[0037] Determining whether the expected acceleration is less than a preset threshold, and if so, setting a status bit to true to indicate that deceleration is required to enter a curve;

[0038] According to the expected acceleration table and the corresponding acceleration change rate table, the absolute value of the expected acceleration is subjected to first-order linear interpolation to obtain the absolute value of the corresponding comfortable acceleration change rate;

[0039] limiting the acceleration request within a preset range according to the absolute value of the comfort acceleration change rate to avoid excessive acceleration / deceleration;

[0040] After exiting the curve with large curvature, the state bit is set to false, and the cornering action and acceleration request are stopped.

[0041] In an optional manner, the calculation formula for the expected acceleration is:

[0042] ;

[0043] in, is the current vehicle speed; is the anchoring distance; is the reference speed.

[0044] According to another aspect of the present invention, there is provided a data-driven adaptive curve speed planning control system, comprising:

[0045] A path curvature calculation module is used to take the path points output by the path planning module of the autonomous driving system as input and determine the curvature value of each path point to eliminate the effects of sensor input instability and incomplete matching of the vehicle's driving path with the lane centerline;

[0046] The anchor point selection and acceleration calculation module is used to traverse all path points, select the path point with the largest curvature as the anchor point, and use the longitudinal position of the path point as the anchor distance; determine the corresponding lateral acceleration according to the curvature value of the anchor point and the anchor distance;

[0047] The lateral acceleration learning and correction module is used to online learn the lateral acceleration corresponding to sections with different curvature radii when the driver drives the vehicle through a curve, obtain the deviation value between the driver's expected lateral acceleration and the preset reference lateral acceleration under different curvature radii, and correct the lateral acceleration under different curvature radii according to the deviation value.

[0048] An embodiment of the present invention further provides a vehicle, on which an adaptive curve speed planning control system is configured, and the adaptive curve speed planning control system is used to execute the data-driven adaptive curve speed planning control method described in any of the aforementioned embodiments.

[0049] According to the data-driven adaptive curve speed planning control method, system and vehicle provided by the present invention, the path point trace output by the path planning module of the automatic driving system is used as input to determine the curvature value of each path point; all path points are traversed, and the path point with the largest curvature is selected as the anchor point, and the longitudinal position of the path point is used as the anchor distance; the corresponding lateral acceleration is determined according to the curvature value and anchor distance of the anchor point; the lateral acceleration corresponding to the road sections with different curvature radii when the driver drives the vehicle through a curve is learned online, and the deviation value between the driver's expected lateral acceleration and the preset reference lateral acceleration under different curvature radii is obtained, and the lateral acceleration under different curvature radii is corrected according to the deviation value. In this way, the interaction with the driver can be enhanced by learning driving behavior online, and the intelligence and anthropomorphism of automatic driving can be better reflected. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A schematic diagram showing a flow chart of a data-driven adaptive curve speed planning control method according to an embodiment of the present invention;

[0051] Figure 2 A schematic diagram of calculating the curvature of a three-point circumscribed circle according to an embodiment of the present invention is shown;

[0052] Figure 3 A schematic diagram of a single frame data processing process according to an embodiment of the present invention is shown;

[0053] Figure 4 A schematic diagram of online learning and optimization results in an embodiment of the present invention is shown;

[0054] Figure 5 A schematic diagram showing a flow chart of curve speed planning control according to an embodiment of the present invention is shown;

[0055] Figure 6 A schematic diagram of driving into a highway main road via a ramp with a large curvature according to an embodiment of the present invention is shown;

[0056] Figure 7 A schematic diagram of driving from a highway main road into a ramp with a large curvature is shown in an embodiment of the present invention;

[0057] Figure 8 A schematic diagram of learning a single set of collected data according to an embodiment of the present invention is shown;

[0058] Fig. 9 A schematic diagram of a single learning process according to an embodiment of the present invention is shown;

[0059] Fig.10 A schematic diagram of updating lateral acceleration according to an embodiment of the present invention is shown;

[0060] Fig.11 A schematic diagram of a scene of a single large curvature ramp according to an embodiment of the present invention is shown;

[0061] Fig.12 A schematic diagram of a ramp interconnection scenario in an embodiment of the present invention is shown;

[0062] Fig.13 A schematic diagram of the framework of a data-driven adaptive curve speed planning control system according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0063] The exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present invention and to enable the scope of the present invention to be fully communicated to those skilled in the art.

[0064] Figure 1 FIG. 1 is a flow chart of a data-driven adaptive curve speed planning control method according to an embodiment of the present invention. Specifically, Figure 1 , Figure 5 As shown, the data-driven adaptive curve speed planning control method may include the following steps:

[0065] Step S1, taking the path point trace output by the path planning module of the autonomous driving system as input, and determining the curvature value of each path point.

[0066] The path planning module is used to generate a driving path from the starting point to the end point. Due to factors such as sensor noise and map accuracy, these paths may not be completely smooth or accurately match the centerline of the lane. The curvature of a path point describes the degree of curvature of the path at a certain point. It is mathematically defined as the rotation rate of the tangent direction angle of the curve at a certain point to the arc length. By determining the curvature value of each path point, the effects of sensor input instability and incomplete matching of the vehicle's driving path with the centerline of the lane can be eliminated.

[0067] In an optional manner, determining the curvature value of each path point further includes:

[0068] Select three consecutive points A, B, and C from the path points, where B is the path point whose curvature is to be calculated, and A and C are the previous point and the next point of point B, respectively;

[0069] Calculate the lengths of the three sides a, b, and c of triangle ABC, where a is the length of AB, b is the length of BC, and c is the length of AC;

[0070] Calculate the sine value of angle B using the cosine theorem and trigonometric functions. According to the sine value of angle B and the length of BC, obtain the radius r of the circumscribed circle and the corresponding curvature K.

[0071] The curvature K is taken as the curvature value of the current path point B.

[0072] In this embodiment, the path planned by the path planning module is a series of points, and each path point includes position information (x, y) in the vehicle body coordinate system, where x is the longitudinal position, that is, the longitudinal distance between the point and the vehicle, and y is the lateral position, that is, the lateral distance between the point and the vehicle. Figure 2 As shown, the curvature of each path point is calculated according to the method of calculating the curvature of three points, including the following steps:

[0073] Step 1: According to the relevant properties of the triangle circumscribed circle, the center O of the triangle circumscribed circle can be obtained by making the intersection of the perpendicular bisectors of the three sides. Among them, A, B, and C are three continuous discrete points of the reference line, and a, b, and c are their opposite sides.

[0074] Step 2: In △ABC, the law of cosines shows that:

[0075] ;

[0076] Step 3: Connect CO and extend the intersection circle at point D. Since circle O is the circumscribed circle of quadrilateral ABCD, according to the property that the angles of the circumscribed circle of a quadrilateral are complementary, we can get:

[0077] ;

[0078] Step 4: Arrange the above formula, the curvature can be expressed as:

[0079] ;

[0080] Step S2, traverse all path points, select the path point with the largest curvature as the anchor point, and use the longitudinal position of the path point as the anchor distance; determine the corresponding lateral acceleration according to the curvature value of the anchor point and the anchor distance.

[0081] In this embodiment, all path points are traversed, and within a certain distance dis, the path point with the largest curvature is used as the anchor point, and the longitudinal position of this path point is used as the anchor distance. The distance dis is calculated as follows:

[0082] ;

[0083] in, is the current speed of the vehicle, is the reaction deceleration, for example, its value is The physical meaning of this distance is that the vehicle deceleration to standstill The required distance. This distance limit means: if the curve is too far away, then at the current vehicle speed, there is no need to slow down and perform the following acceleration calculation.

[0084] In an optional manner, selecting the path point with the largest curvature as the anchor point further includes:

[0085] According to the current speed of the vehicle and the preset reaction deceleration, the safe distance required for the vehicle to slow down from the current speed to a standstill at the reaction deceleration is calculated;

[0086] Traverse all discrete points on the path and take each discrete point as a potential candidate anchor point;

[0087] Extending the candidate anchor point forward and backward by a safe distance to determine a candidate range, and taking all path points within the candidate range as key points affecting the vehicle driving decision;

[0088] The curvature value of each path point within the candidate range is calculated, and the path point with the largest curvature value is selected as the anchor point.

[0089] Step S3, online learning the lateral acceleration corresponding to sections with different curvature radii when the driver drives the vehicle through a curve, obtaining the deviation value between the driver's expected lateral acceleration and the preset reference lateral acceleration under different curvature radii, and correcting the lateral acceleration under different curvature radii according to the deviation value.

[0090] In this embodiment, firstly, according to regulatory requirements and previous project engineering experience, the corresponding comfortable lateral accelerations under different curvature radii are summarized and represented by the curvature radius table RTable and the corresponding default reference lateral acceleration table ALatTableInit.

[0091] Then, the driver's cornering driving data is gradually collected to learn his driving behavior and to adaptively adjust the algorithm parameters to suit the driver's specific driving preferences. The algorithm mainly analyzes the lateral acceleration corresponding to different curvature radius sections when the driver is driving the vehicle through a corner, learns the deviation between the driver's expected lateral acceleration under different curvature radii and the default reference lateral acceleration initially set by the algorithm, and records a series of lateral acceleration deviation values ​​in the table DltALatTable; finally, the deviation values ​​in DltALatTable are added one by one to the default reference lateral acceleration values ​​in ALatTableInit set by the algorithm, so as to correct and optimize the ideal lateral acceleration under different curvature radii, so that the driving behavior of the autonomous driving system is more in line with the driver's specific driving style.

[0092] In an optional manner, the step S3 further comprises:

[0093] Determine whether the driving scene meets the trigger condition for collecting the user's driving data, and when the trigger condition is met, continue to collect driving data at a preset frequency until the curvature radius is greater than a set threshold, the lateral acceleration is 0, or the maximum data collection time is reached; wherein the driving data includes the curvature radius of the road and the lateral acceleration of the vehicle;

[0094] Through the initial curvature radius table and the corresponding default reference lateral acceleration table, each collected curvature radius is linearly interpolated to obtain the corresponding default reference lateral acceleration;

[0095] All collected data are grouped according to the curvature radius, and the acceleration differences in each group of data are weighted averaged to obtain the average deviation value between the driver's expected lateral acceleration and the preset reference lateral acceleration under the curvature radius;

[0096] The lateral accelerations at different curvature radii are corrected according to the average deviation value.

[0097] In this embodiment, the driving scene meets the following triggering conditions for collecting user driving data:

[0098] a) The driver is the driving subject: the driver controls the vehicle instead of the autonomous driving system; or the autonomous driving system only controls the longitudinal movement of the vehicle (longitudinal speed / acceleration control), and the lateral movement of the vehicle (lateral speed / acceleration control) is controlled by the driver.

[0099] b) Road curvature meets the condition: There are continuous sections with a curvature radius in the range of [10m, 600m] on the road ahead. This feature is used to indicate that there is a curve ahead. The minimum value means that roads with a curvature radius less than 10m are not within the scope of the algorithm, and there are basically no roads with a curvature radius less than 10m; in addition, roads with a curvature radius greater than 600m can be approximately regarded as straight roads, and there is no need to adjust their speed / acceleration.

[0100] c) No interference from the vehicle ahead: There is no vehicle ahead, or there is a vehicle ahead but it is far away or driving at a high speed, which will not affect the speed of the vehicle and force the driver to slow down. This condition excludes potential factors that affect vehicle speed control except for curves.

[0101] After the above trigger conditions are met, driving data will be collected at a certain frequency, for example, once per second, and the data includes the radius of curvature of the road and the lateral acceleration of the vehicle. This continues until the radius of curvature of the road ahead is greater than the set threshold RThreshold. The setting of this threshold needs to consider the impact of the curvature of the road on the automatic driving control. For example, a road with a radius of curvature greater than this threshold is considered to be a straight road by the automatic driving system, and no speed / acceleration adjustment is required; or the lateral acceleration of the vehicle is 0; or the time is greater than the set maximum data collection time, such as 60s.

[0102] The total amount of driving data collected in each group needs to be greater than a certain time, such as 5s, to ensure that this group of data includes (most of) the cornering process. There must be no abnormal data in the data set, such as obvious sudden changes in the radius of curvature. The driver's driving behavior must be legal, compliant and reasonable, and cannot drive on the lane line / across the lane line. The above conditions can ensure the rationality and effectiveness of the driver's driving data. If any of the above conditions is not met, the data will be invalid. That is, the autonomous driving system is not allowed to learn this illegal behavior to ensure the safety of the algorithm and avoid dangerous control of the autonomous driving system.

[0103] The road curvature radius information in the collected driving data cannot completely correspond to the data in the curvature radius table RTable, and further processing is required. Figure 3 As shown:

[0104] In an online learning process, a total of M pairs of data are collected, and the curvature radius of the i-th pair of data is , the lateral acceleration is .

[0105] The algorithm's initial curvature radius table RTable includes the curvature radius parameters The default reference lateral acceleration table ALatTableInit corresponds to the lateral accelerations: .

[0106] According to RTable, ALatTableInit, the collected curvature radius Perform linear interpolation to obtain the default reference lateral acceleration .

[0107] The difference between the actual lateral acceleration and the reference lateral acceleration is =-0.05.

[0108] Approximately, the radius of curvature can be considered The difference between the driver's desired and default reference lateral acceleration at The corresponding difference at:

[0109]

[0110]

[0111] The above approximate processing is performed on each pair of collected data during this learning process. Finally, the M processing results are grouped by curvature radius, and each group is weighted averaged to obtain the difference table between the driver's expected lateral acceleration and the default reference lateral acceleration corresponding to each curvature radius in RTable. (As shown in Table 1):

[0112]

[0113] like Figure 8 , Fig. 9 As shown, a total of M pairs of data were collected in this online learning process, and the curvature radius of I pairs of data was between , By taking the weighted average of the I acceleration differences corresponding to the A pair of data, we can get the curvature radius: , The learning results of the expected and reference lateral accelerations at : = = ；

[0114] In an optional manner, the method further includes

[0115] For each lateral acceleration corresponding to the radius of curvature in the initial lateral acceleration table, add the lateral acceleration to the corresponding difference in the difference table to obtain an updated lateral acceleration; wherein the difference table includes the expected lateral acceleration and the reference lateral acceleration;

[0116] The difference table is updated according to the collected new cornering driving data and the learning rate, wherein the updating formula of the difference table is:

[0117]

[0118] in, The table value updated for the last study; is the difference obtained by current learning; is the learning rate.

[0119] In this embodiment, the expected lateral acceleration and reference lateral acceleration difference table DltALatTable is the learning result of the driver's driving behavior. The next time the automatic driving system controls the cornering, the value in this table is added to the value in the initial lateral acceleration table ALatTableInit to achieve parameter update and obtain AlatTableFinal:

[0120] That is, the initial learning update of the algorithm is realized. As the driver's driving mileage increases, the driver's cornering driving data will be continuously and automatically collected and updated at a certain learning rate. The table DltALatTable is continuously updated to gradually improve the adaptation to the driver's specific driving preferences. The update algorithm is as follows:

[0121]

[0122] in, The table value updated for the last study; is the difference obtained by current learning; is the learning rate, The value of is related to weather conditions, lighting conditions (day, night), etc., and is usually limited to .

[0123] Each expected and reference lateral acceleration difference in DltALatTable is limited to a certain range to prevent the algorithm from learning occasional excessive driving behavior, thereby ensuring the safety of the autonomous driving system.

[0124] Finally, ALatTableFinal is further obtained based on the updated DltALatTable to update the lateral acceleration of the vehicle during cornering.

[0125] In an optional manner, before step S3, the method further includes:

[0126] According to the given curvature radius table and the default reference lateral acceleration table, the first-order linear interpolation is performed on the anchor point curvature radius in the curvature radius table to obtain the comfortable lateral acceleration at the corresponding anchor point;

[0127] After completing the collection of driving data and the online learning of lateral acceleration, first-order linear interpolation is performed again on the anchor point curvature radius in the curvature radius table according to the updated curvature radius table and the updated reference lateral acceleration table to obtain the updated comfortable lateral acceleration at the corresponding anchor point;

[0128] According to the given curvature radius and the comfortable lateral acceleration, a curvature speed limit is obtained to guide the speed control of the automatic driving system in the curve;

[0129] The expected acceleration is obtained according to the curvature speed limit, the anchor distance and the current vehicle speed.

[0130] In this embodiment, the online learning and optimization results are as follows: Figure 4 As shown, before collecting the driver's driving data and learning the driver's driving preferences, the anchor point curvature radius is calculated based on RTable and the default reference lateral acceleration table ALatTableInit. The first-order linear interpolation can be used to obtain the comfortable lateral acceleration corresponding to the anchor point: ; After online learning and updating of the lateral acceleration, the first-order linear interpolation of the anchor point curvature radius is performed according to RTable and the updated ALatTableFinal to obtain the comfortable lateral acceleration corresponding to the anchor point. .

[0131] The radius of curvature R and the lateral acceleration Substitute into the formula The square value of the comfortable reference speed can be obtained , this reference speed is the curvature speed limit.

[0132] The reference speed calculated above , Anchor distance and current vehicle speed , Substitute into the formula The expected acceleration can be obtained .

[0133] In an optional manner, the method further includes:

[0134] Determining whether the expected acceleration is less than a preset threshold, and if so, setting a status bit to true to indicate that deceleration is required to enter a curve;

[0135] According to the expected acceleration table and the corresponding acceleration change rate table, the absolute value of the expected acceleration is subjected to first-order linear interpolation to obtain the absolute value of the corresponding comfortable acceleration change rate;

[0136] limiting the acceleration request within a preset range according to the absolute value of the comfort acceleration change rate to avoid excessive acceleration / deceleration;

[0137] After exiting the curve with large curvature, the state bit is set to false, and the cornering action and acceleration request are stopped.

[0138] In this embodiment, the prerequisite for the status bit isValid to be true is the expected acceleration Less than threshold Only when this condition is met, it is considered necessary to slow down when entering the corner, and further calculate the final requested acceleration. This condition exists to ensure the consistency and smoothness of the system acceleration request, and to a certain extent avoid the sudden change of the longitudinal acceleration request caused by the input fluctuation of the perception and front module.

[0139] Determine the comfortable acceleration rate by linear interpolation Based on previous project engineering experience and a large amount of driving data, the expected acceleration change rate corresponding to different expected accelerations is summarized, represented by the expected acceleration table ATable and the corresponding acceleration change rate table JTable. The values ​​of both tables are positive. According to ATable and JTable, the absolute value of the expected acceleration The first-order linear interpolation can obtain the absolute value of the corresponding comfortable acceleration change rate: .

[0140] Limit acceleration requests based on the rate of change of acceleration :

[0141]

[0142] in, The algorithm represented by the above formula means that, based on the expected acceleration At the same time, the acceleration request of the current frame Limit acceleration requests to the previous frame A certain range of variation To ensure driving stability.

[0143] The acceleration request Limit it to a certain range to avoid excessive acceleration / deceleration and ensure driving safety and driving comfort:

[0144] ;

[0145] The threshold , Usually determined based on system requirements and engineering experience.

[0146] After exiting the curve with large curvature, the status bit isValid is set to false, that is, the longitudinal direction is no longer affected by the curve with large curvature, and the turning action is stopped. Please accelerate No longer valid, keep the default value. The conditions for judging the end of the turn are:

[0147] && ;

[0148] The above formula shows that the requested acceleration of the previous frame is equal to the expected acceleration calculated in the current frame. Already greater than , that is, the current acceleration has returned to near No need to continue action.

[0149] In an optional manner, the calculation formula for the expected acceleration is:

[0150] ;

[0151] in, is the current vehicle speed; is the anchoring distance; is the reference speed.

[0152] In this embodiment, after the curvature information and position information of the anchor point are determined, the comfortable lateral acceleration corresponding to the curvature is determined by looking up the table. (The specific values ​​of the curvature radius-lateral acceleration table are obtained based on project engineering experience); the curvature of the anchor point and lateral acceleration Substitute into the formula The corresponding comfortable reference speed can be obtained ; Set the reference speed , Current vehicle speed , Anchor distance Substitute into the formula The expected acceleration can be obtained This calculation method is more accurate and more practical than directly setting the acceleration value in the prior art.

[0153] According to the solution provided by the present invention, step S1, taking the path point trace output by the path planning module of the automatic driving system as input, determining the curvature value of each path point, so as to eliminate the influence of the instability of the perceived input and the incomplete matching of the vehicle's driving path with the center line of the lane; step S2, traversing all path points, selecting the path point with the largest curvature as the anchor point, and taking the longitudinal position of the path point as the anchor distance; determining the corresponding lateral acceleration according to the curvature value and the anchor distance of the anchor point; step S3, online learning the lateral acceleration corresponding to the road sections with different curvature radii when the driver drives the vehicle through a curve, obtaining the deviation value between the driver's expected lateral acceleration and the preset reference lateral acceleration under different curvature radii, and correcting the lateral acceleration under different curvature radii according to the deviation value. The present invention enables the automatic driving vehicle to safely and smoothly pass through the large curvature road without the influence of the preceding vehicle, and solves the problem of insufficient intelligence and anthropomorphism in the large curvature road scene due to lack of driver interaction through online learning of driving behavior.

[0154] It should be noted that the applicable scenarios of this embodiment include but are not limited to the following: Figure 6 , Figure 7 , Fig.11 , Fig.12 The scene shown. In the above scene, the curvature in the ramp is large, and the radius of curvature at the curve is about 250m. Before entering the curve, the vehicle speed is large and the acceleration is non-negative. In this embodiment, when the vehicle is at a certain position before the starting point of the curve, the vehicle is controlled to gradually reduce the acceleration; until it passes the end of the curve, the acceleration of the vehicle is controlled to gradually return to positive. The acceleration control of the vehicle can make the vehicle turn smoothly, and the acceleration change curve of the entire turning process is similar to a "V" shape. By continuously learning the driver's driving behavior when turning, the system will gradually adapt to the driver's driving preferences while ensuring the smooth control of the vehicle through the corner, so as to achieve further intelligence and anthropomorphism.

[0155] Compared with the prior art, the advantages of the present invention are as follows:

[0156] 1. The method of calculating the path curvature and determining the anchor point according to the vehicle control path reflects the superiority of the horizontal and vertical coordination to make the vehicle control trajectory smoother.

[0157] 2. The calculation method of expected acceleration is more physical and more in line with actual driving scenarios.

[0158] 3. The constraints for longitudinal speed regulation on roads with large curvature reflect the consistency of planning and control and the robustness of the system to disturbances.

[0159] 4. Adapting to the driver's driving preferences largely reflects the intelligence and anthropomorphism of the system, making the autonomous driving system better meet the driver's driving needs.

[0160] Fig.13 The structure diagram of the data-driven adaptive curve speed planning control system according to an embodiment of the present invention is shown. The data-driven adaptive curve speed planning control system comprises:

[0161] A path curvature calculation module 131 is used to take the path point trace output by the path planning module of the autonomous driving system as input and determine the curvature value of each path point to eliminate the influence of the instability of the perception input and the incomplete match between the vehicle's driving path and the lane centerline;

[0162] Anchor point selection and acceleration calculation module 132, used to traverse all path points, select the path point with the largest curvature as the anchor point, and use the longitudinal position of the path point as the anchor distance; determine the corresponding lateral acceleration according to the curvature value of the anchor point and the anchor distance;

[0163] The lateral acceleration learning and correction module 133 is used to online learn the lateral acceleration corresponding to the road sections with different curvature radii when the driver drives the vehicle through a curve, obtain the deviation value between the driver's expected lateral acceleration and the preset reference lateral acceleration under different curvature radii, and correct the lateral acceleration under different curvature radii according to the deviation value.

[0164] The data-driven adaptive curve speed planning control system provided by the embodiment of the present invention can realize the data-driven adaptive curve speed planning control method described in any of the above embodiments, and its implementation principle and technical effects are basically the same. In particular, according to the data-driven adaptive curve speed planning control system provided by the present invention, the path point trace output by the path planning module of the automatic driving system is used as input to determine the curvature value of each path point to eliminate the influence of the instability of the perceived input and the incomplete matching of the vehicle driving path with the lane centerline; traverse all path points, select the path point with the largest curvature as the anchor point, and use the longitudinal position of the path point as the anchor distance; determine the corresponding lateral acceleration according to the curvature value and anchor distance of the anchor point; online learn the lateral acceleration corresponding to the road sections with different curvature radii when the driver drives the vehicle through the curve, obtain the deviation value between the driver's expected lateral acceleration and the preset reference lateral acceleration under different curvature radii, and correct the lateral acceleration under different curvature radii according to the deviation value. The present invention enables an autonomous driving vehicle to safely and smoothly pass a road with a large curvature without the influence of a preceding vehicle, and solves the problem of lack of intelligence and anthropomorphism due to lack of driver interaction in scenes with large curvature roads by learning driving behaviors online.

[0165] An embodiment of the present invention further provides a vehicle, on which an adaptive curve speed planning control system is configured, and the adaptive curve speed planning control system is used to execute the data-driven adaptive curve speed planning control method described in any of the aforementioned embodiments.

[0166] The vehicle provided in the embodiment of the present invention can implement the data-driven adaptive curve speed planning control method described in any of the aforementioned embodiments. The implementation principle and technical effects thereof are basically the same and will not be repeated here.

[0167] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and further may be divided into multiple submodules or subunits or subcomponents. All features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed in this manner may be combined in any combination, except that at least some of such features and / or processes or units are mutually exclusive. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature that provides the same, equivalent or similar purpose. In addition, those skilled in the art will appreciate that, although some embodiments herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present invention and to form different embodiments. For example, in the following claims, any one of the claimed embodiments may be used in any combination. The present invention may be implemented by means of hardware including several different elements and by means of a suitably programmed computer. In a unit claim that lists several means, several of these means may be embodied by the same hardware item. The steps in the above embodiments should not be understood as limiting the execution order unless otherwise specified.

Claims

1. A data-driven adaptive curve speed planning control method, characterized in that: include: Step S1, taking the path point trace output by the path planning module of the autonomous driving system as input, and determining the curvature value of each path point; Step S2, traversing all path points, selecting the path point with the largest curvature as the anchor point, and taking the longitudinal position of the path point as the anchor distance; determining the corresponding lateral acceleration according to the curvature value of the anchor point and the anchor distance; Step S3, online learning the lateral acceleration corresponding to the road sections with different curvature radii when the driver drives the vehicle through a curve, obtaining the deviation value between the driver's expected lateral acceleration and the preset reference lateral acceleration under different curvature radii, and correcting the lateral acceleration under different curvature radii according to the deviation value; Wherein, the step S3 further includes: Determine whether the driving scene meets the trigger condition for collecting the user's driving data, and when the trigger condition is met, continue to collect driving data at a preset frequency until the curvature radius is greater than a set threshold, the lateral acceleration is 0, or the maximum data collection time is reached; wherein the driving data includes the curvature radius of the road and the lateral acceleration of the vehicle; Through the initial curvature radius table and the corresponding default reference lateral acceleration table, each collected curvature radius is linearly interpolated to obtain the corresponding default reference lateral acceleration; All collected data are grouped according to the curvature radius, and the acceleration differences in each group of data are weighted averaged to obtain the average deviation value between the driver's expected lateral acceleration and the preset reference lateral acceleration under the curvature radius; Correcting the lateral acceleration at different curvature radii according to the average deviation value; The method further comprises: For each lateral acceleration corresponding to the radius of curvature in the initial lateral acceleration table, add the lateral acceleration to the corresponding difference in the difference table to obtain an updated lateral acceleration; wherein the difference table includes the expected lateral acceleration and the reference lateral acceleration; The difference table is updated according to the collected new cornering driving data and the learning rate, wherein the updating formula of the difference table is: ； in, The table value updated for the last study; is the difference obtained by current learning; is the learning rate, The value of depends on weather conditions and light conditions and is limited to .

2. The data-driven adaptive curve speed planning control method according to claim 1, characterized in that: Determining the curvature value of each path point further comprises: Select three consecutive points A, B, and C from the path points, where B is the path point whose curvature is to be calculated, and A and C are the previous point and the next point of point B, respectively; Calculate the lengths of the three sides a, b, and c of triangle ABC, where a is the length of AB, b is the length of BC, and c is the length of AC; Calculate the sine value of angle B using the cosine theorem and trigonometric functions. According to the sine value of angle B and the length of BC, obtain the radius r of the circumscribed circle and the corresponding curvature K. The curvature K is taken as the curvature value of the current path point B.

3. The data-driven adaptive curve speed planning control method according to claim 1 or 2, characterized in that: The selecting the path point with the largest curvature as the anchor point further comprises: According to the current speed of the vehicle and the preset reaction deceleration, a safe distance required for the vehicle to slow down from the current speed to a standstill at the reaction deceleration is calculated; Traverse all discrete points on the path and take each discrete point as a potential candidate anchor point; Extending the candidate anchor point forward and backward by a safe distance to determine a candidate range, and taking all path points within the candidate range as key points affecting the vehicle driving decision; The curvature value of each path point within the candidate range is calculated, and the path point with the largest curvature value is selected as the anchor point.

4. The data-driven adaptive curve speed planning control method according to claim 1, characterized in that: Before step S3, the method further includes: According to the given curvature radius table and the default reference lateral acceleration table, the first-order linear interpolation is performed on the anchor point curvature radius in the curvature radius table to obtain the comfortable lateral acceleration at the corresponding anchor point; After completing the collection of driving data and the online learning of lateral acceleration, first-order linear interpolation is performed again on the anchor point curvature radius in the curvature radius table according to the updated curvature radius table and the updated reference lateral acceleration table to obtain the updated comfortable lateral acceleration at the corresponding anchor point; According to the given curvature radius and the comfortable lateral acceleration, a curvature speed limit is obtained to guide the speed control of the automatic driving system in the curve; The expected acceleration is obtained according to the curvature speed limit, the anchor distance and the current vehicle speed.

5. The data-driven adaptive curve speed planning control method according to claim 4 is characterized in that: The method further comprises: Determining whether the expected acceleration is less than a preset threshold, and if so, setting a status bit to true to indicate that deceleration is required to enter a curve; According to the expected acceleration table and the corresponding acceleration change rate table, the absolute value of the expected acceleration is subjected to first-order linear interpolation to obtain the absolute value of the corresponding comfortable acceleration change rate; limiting the acceleration request within a preset range according to the absolute value of the comfort acceleration change rate to avoid excessive acceleration / deceleration; After exiting the curve with large curvature, the state bit is set to false, and the cornering action and acceleration request are stopped.

6. The data-driven adaptive curve speed planning control method according to claim 5, characterized in that: The calculation formula of the expected acceleration is: ； in, is the current vehicle speed; is the anchoring distance; is the reference speed.

7. A data-driven adaptive curve speed planning control system, characterized in that: The system is used to implement the data-driven adaptive curve speed planning control method according to any one of claims 1 to 6, and comprises: A path curvature calculation module is used to take the path points output by the path planning module of the autonomous driving system as input and determine the curvature value of each path point to eliminate the effects of sensor input instability and incomplete matching of the vehicle's driving path with the lane centerline; The anchor point selection and acceleration calculation module is used to traverse all path points, select the path point with the largest curvature as the anchor point, and use the longitudinal position of the path point as the anchor distance; determine the corresponding lateral acceleration according to the curvature value of the anchor point and the anchor distance; The lateral acceleration learning and correction module is used to online learn the lateral acceleration corresponding to sections with different curvature radii when the driver drives the vehicle through a curve, obtain the deviation value between the driver's expected lateral acceleration and the preset reference lateral acceleration under different curvature radii, and correct the lateral acceleration under different curvature radii according to the deviation value.

8. A vehicle, characterized in that: The vehicle is equipped with an adaptive curve speed planning control system, and the adaptive curve speed planning control system is used to execute the data-driven adaptive curve speed planning control method described in any one of claims 1 to 6.

Images