Path planning method and device, automatic driving equipment and readable storage medium

Abstract

The invention discloses a path planning method and device, automatic driving equipment and a readable storage medium. The path planning method comprises the steps that path information and the starting point position of automatic driving equipment are acquired, the path information comprises a path reference line, and a to-be-optimized path point is determined based on the path reference line; determining a Frenet coordinate system based on the starting point position and the path reference line; based on the to-be-optimized path point, a cost function is established under a Frenet coordinate system, the cost function is a quadratic function, a first penalty term is introduced into the cost function, and the first penalty term is used for representing the deviation between the to-be-optimized path point and the transversely-biased path reference line; and solving the cost function based on a preset constraint condition, and obtaining a target path corresponding to the minimum cost function. According to the invention, when the distance between the travelable boundary and the path reference line is small due to the fact that the obstacle partially occupies the travelable area, the planned target path can be far away from the obstacle, so that the driving experience is improved.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent driving, and particularly to a path planning method, device, autonomous driving device, and readable storage medium. Background Art

[0002] Vehicles equipped with autonomous driving technology need to perform path planning autonomously in some scenarios. According to the degree of mastery of environmental information, the vehicle can perform global path planning based on prior complete information, and then perform local path planning based on sensor information to find a collision-free path from the starting point to the target point in an environment with obstacles.

[0003] For structured road scenarios, when static obstacles appear in the current lane or the target lane where the vehicle is traveling, it is usually planned based on numerical optimization to provide a smooth and safe obstacle avoidance path. However, in the common local path quadratic programming method, in order to avoid the influence of too small a feasible region on the solution, the convex space boundary only adds a small safety distance to the outer contour of the original obstacle. Then, in the case where the obstacle occupies the lane, the planned path will be too close to the obstacle, affecting the driving experience. Summary of the Invention

[0004] Embodiments of the present application provide a path planning method, device, autonomous driving device, and readable storage medium to improve the problem that the locally planned path is too close to the obstacle and enhance the driving experience.

[0005] To solve the above technical problems, embodiments of the present application disclose the following technical solutions:

[0006] In a first aspect, a path planning method is provided, which is applied to an autonomous driving device. The method includes:

[0007] Obtain path information and the starting position of the autonomous driving device. The path information includes a path reference line, and determine a path point to be optimized based on the path reference line;

[0008] Determine a Frenet coordinate system based on the starting position and the path reference line;

[0009] Establish a cost function in the Frenet coordinate system based on the path point to be optimized. The cost function is a quadratic function, and a first penalty term is introduced into the cost function. The first penalty term is configured to represent the deviation between the path point to be optimized and the laterally offset path reference line;

[0010] Solve the cost function based on preset constraint conditions to obtain the target path corresponding to the minimum value of the cost function.

[0011] In a second aspect, a path planning device is provided, which is configured in an autonomous driving device. The device includes:

[0012] An information acquisition unit, configured to acquire path information and the starting position of the autonomous driving device. The path information includes a path reference line, and based on the path reference line, to determine path points to be optimized;

[0013] A coordinate system establishment unit, configured to determine a Frenet coordinate system based on the starting position and the path reference line;

[0014] A cost function construction unit, configured to establish a cost function in the Frenet coordinate system based on the path points to be optimized. The cost function is a quadratic function, and the cost function introduces a first penalty term, which is configured to characterize the deviation of the path points to be optimized from the laterally offset path reference line;

[0015] A target path acquisition unit, configured to solve the cost function based on preset constraint conditions and acquire the target path corresponding to the minimum value of the cost function.

[0016] In a third aspect, a path planning device for an autonomous driving device is provided, including:

[0017] A memory, configured to store a program;

[0018] A processor, configured to execute the program stored in the memory;

[0019] When the program stored in the memory is executed, the processor executes the path planning method according to any one of the first aspect.

[0020] In a fourth aspect, an autonomous driving device is provided, including the path planning device according to the second aspect.

[0021] In a fifth aspect, a computer-readable storage medium is provided. The computer-readable medium stores instructions for a computing device to execute. When the computing device executes the instructions, the method according to any one of the first aspect is implemented.

[0022] One of the above technical solutions has the following advantages or beneficial effects:

[0023] Compared with the prior art, a path planning method of the present application includes: obtaining path information and the starting position of an autonomous driving device, where the path information includes a path reference line, and determining path points to be optimized based on the path reference line; determining a Frenet coordinate system based on the starting position and the path reference line; establishing a cost function in the Frenet coordinate system based on the path points to be optimized, the cost function being a quadratic function, and the cost function introducing a first penalty term, where the first penalty term is used to characterize the deviation between the path points to be optimized and the laterally offset path reference line; solving the cost function based on preset constraint conditions to obtain a target path corresponding to the minimum value of the cost function. The path planning method provided by the present application introduces the first penalty term into the cost function, which can make the planned target path far away from obstacles when the obstacle partially occupies the drivable area and the distance between the drivable boundary and the path reference line is small, thereby improving the driving experience.

[0024] The path planning device provided by the present application can make the planned target path far away from obstacles when the obstacle partially occupies the drivable area and the distance between the drivable boundary and the path reference line is small, thereby improving the driving experience.

[0025] The path planning device of the autonomous driving device provided by the present application can make the planned target path far away from obstacles when the obstacle partially occupies the drivable area and the distance between the drivable boundary and the path reference line is small, thereby improving the driving experience.

[0026] The autonomous driving device provided by the present application can plan a smooth path far away from obstacles, thereby having a better driving experience

[0027] The computer-readable storage medium provided by the present application can make the planned target path far away from obstacles when the obstacle partially occupies the drivable area and the distance between the drivable boundary and the path reference line is small, thereby improving the driving experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present application. For those skilled in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0029] Figure 1 It is a schematic diagram of the system architecture of the autonomous driving device provided by the embodiment of the present application;

[0030] Figure 2 It is a schematic diagram of the overall principle framework of the path planning of the autonomous driving device provided by the embodiment of the present application;

[0031] Figure 3a Schematic diagram of one of the application scenarios of the embodiments of the present application;

[0032] Figure 3b Schematic diagram of the second application scenario of the embodiments of the present application;

[0033] Figure 3c Schematic diagram of the third application scenario of the embodiments of the present application;

[0034] Figure 4 Overall flowchart of the path planning method of the embodiments of the present application;

[0035] Figure 5 Schematic diagram of the Frenet coordinate system of the embodiments of the present application;

[0036] Figure 6 Schematic diagram of an example of constructing a first penalty term when the obstacle and the autonomous driving device are in a first positional relationship in the embodiments of the present application;

[0037] Figure 7 Schematic diagram of an example of constructing a first penalty term when the obstacle and the autonomous driving device are in a second positional relationship in the embodiments of the present application;

[0038] Figure 8 Schematic diagram of the model established for the autonomous driving device in the embodiments of the present application;

[0039] Figure 9a Schematic diagram of the target path planned by the path planning method of the embodiments of the present application in one of the application scenarios;

[0040] Figure 9b Schematic diagram of the target path planned by the path planning method of the embodiments of the present application in the second application scenario;

[0041] Figure 9c Schematic diagram of the target path planned by the path planning method of the embodiments of the present application in the third application scenario;

[0042] Figure 10 Schematic diagram of the structure of the path planning device of the embodiments of the present application;

[0043] Figure 11 Schematic diagram of the structure of the path planning device of the autonomous driving device of the embodiments of the present application.

[0044] Reference numerals:

[0045] 100 - Planning system; 110 - Lane - changing decision - maker; 120 - Reference line smoother; 130 - Path constraint decision - maker; 140 - Path quality optimizer; 141 - Information acquisition unit; 142 - Coordinate system establishment unit; 143 - Cost function construction unit; 144 - Target path acquisition unit; 150 - Path determiner; 1101 - Memory; 1102 - Processor. Detailed implementation manners

[0046] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0047] In the description of the present application, it should be understood that in the description of the present application, the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. It is only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be understood as a limitation to the present application. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more features. In the description of the present application, "a plurality" means two or more, and "at least one" means one, two, or more, unless otherwise specifically and clearly defined.

[0048] Please refer to Figure 1 , Figure 1Schematically shows the system architecture of the autonomous driving device provided by the embodiments of the present application. Devices equipped with autonomous driving technology, such as vehicles with L2+ or L3 level autonomous driving technology, usually may include NUC PC1 (the first compact computing device), NUC PC2 (the second compact computing device), CAR PC (in-vehicle computer), ASDM (Automated System Development and Management), Autobox3 (Autoboxing), and Host Vehicle CAN (host vehicle local area network communication). Among them, NUCPC1 receives the positioning information transmitted by RTK, converts it, and then transmits it to NUC PC2 for use as the input for planning; ASDM receives the information of four short-range millimeter waves (SRR), one long-range millimeter wave (MRR), and one front camera, fuses them, and gives the fused result to NUC PC2 for the perception of the surrounding environment. At the same time, ASDM gives the information to DVTOOL (visualization tool) in the CAR PC (in-vehicle computer) for recording; after receiving the positioning and fusion information, NUC PC2 performs path planning, and finally gives the planned path to Autobox3. Autobox3 runs the relevant internal modules, generates control signals and status signals, and sends each signal to the actual vehicle through Host Vehicle CAN to achieve the control of the actual vehicle.

[0049] Please refer to Figure 2 , Figure 2 which schematically shows the overall principle framework of the path planning of the autonomous driving device provided by the embodiments of the present application. At the software level, vehicles equipped with autonomous driving technology usually may include a perception system, a planning system 100, and a control system (for example: Autobox3 in Figure 1 can be adopted). Among them, the planning system 100 (for example: can adopt Figure 1The NUCPC2) in it may include a Lane Change Decider 110, a Reference line smoother 120, a Path bound decider 130, a Path QP optimizer 140, and a Path decider 150. Among them, the Lane Change Decider 110 is used to evaluate the lane change risk, determine the lane change state, and determine the path reference line according to the lane change state; the Reference line smoother 120 is used to perform quadratic programming smoothing on the path reference line to ensure that the path reference line is relatively smooth; the Path bound decider 130 is used to determine the drivable boundary and provide a convex space for path planning; the Path QP optimizer 140 is used to perform quadratic programming within the convex space provided upstream to optimize an optimal target path; the Path decider 150 makes a decision calibration setting for the obstacle based on the relative relationship between the planned path result and the obstacle for subsequent speed decision planning as a reference. Each device in the planning system 100 works together and finally sends the planned path to the control system for use.

[0050] Please refer to Figures 3a to 3c , Figure 3a illustrates one of the application scenarios of the embodiments of the present application. Figure 3b illustrates a second application scenario of the embodiments of the present application. Figure 3c illustrates a third application scenario of the embodiments of the present application. Exemplarily, one of the application scenarios is the lane follow scenario, that is, the autonomous driving device (Host) drives on the road in the direction and position of the current lane B 1 and the obstacle (Obstacle) only occupies a partial boundary of the current lane B 1 and does not occupy the driving path of the autonomous driving device. The second application scenario is the lane borrow scenario, that is, the autonomous driving device (Host) drives on the road in the direction and position of the current lane B 1 and the obstacle (Obstacle) occupies the current driving path reference line of the autonomous driving device. The third application scenario is the lane change scenario, that is, the autonomous driving device (Host) changes from the current lane B 1 to the target lane B 2 and drives, and the obstacle (Obstacle) occupies a partial boundary of the target lane B 2 .

[0051] For the above-mentioned structured road scenarios, the path quality optimizer 140 generally can adopt two types of local path planning methods to provide a smooth path that can avoid obstacles safely. One method is to plan based on a discrete sampling method, that is, by discretizing the structured road, determining candidate search paths through state sampling, and screening out the optimal path from the candidate paths through objective functions and constraint conditions. Although this method is relatively simple and has a low computational cost for simple scenarios, since the shape of the path mainly depends on the starting and ending states of sampling, the discrete state sampling resolution, and range, the path flexibility is limited. Moreover, for complex scenarios, in many cases, the calculated path curve is not optimal. Another method is to plan based on a numerical optimization method, that is, based on an optimization method, iteratively solve the objective function under the designed constraint conditions to find the optimal path. Although this method has high flexibility, when static obstacles appear in the lane where the vehicle is traveling or in the target lane, since the convex space boundary only superimposes a small safety distance on the outer contour of the original obstacle, specifically, refer to when the autonomous driving device is in Figures 3a to 3c different scenarios, the planned target paths P 1 , P 2 and P 3 . If the planned path only considers the reference line and the smoothness penalty term, the path planned in the case where the obstacle occupies the lane will be too close to the convex space boundary, affecting the driving experience. In addition, if only the lateral safety distance is enlarged as a mandatory constraint, the feasible region will be reduced, increasing the possibility of solving failure and making the iterative solution slower; if the penalty term in the center of the convex space is increased, it will be impossible to quickly reach the path reference line in the lane change or lane borrowing scenario, and the unevenness of the convex space boundary will lead to the unevenness of the planned path.

[0052] In view of this, the embodiments of the present application provide an autonomous driving device. When performing local path planning, within the standard framework of quadratic programming, by introducing a first penalty term representing the deviation between the path point to be optimized and the path reference line of the lateral offset into the cost function, the cost function is optimized so that when the obstacle partially occupies the drivable area and the distance between the drivable boundary and the path reference line is small, the planned target path can achieve the effect of being far from the obstacle, thereby improving the driving experience and thus solving at least part of the above technical problems.

[0053] Please continue to refer to Figure 2 , the autonomous driving device of the embodiments of the present application includes a path quality optimizer 140, and the path quality optimizer 140 is used to adopt the path planning method of the embodiments of the present application to generate the optimal target path in various scenarios.

[0054] The path planning method of the embodiments of the present application will be introduced below with reference to the accompanying drawings.

[0055] Please refer to Figure 4 , Figure 4 , which illustrates the overall process of the path planning method according to an embodiment of the present application. The path planning method is applied to an autonomous driving device, and specifically includes the following steps:

[0056] Step 401: Obtain path information and the starting position of the autonomous driving device. The path information includes a path reference line, and determine the path points to be optimized based on the path reference line.

[0057] Exemplarily, the path information may further include obstacles. The path reference line is usually the center line of the lane.

[0058] In some examples, along the extension direction of the path reference line, the path points to be optimized can be determined at a fixed interval d s . Exemplarily, the fixed interval d s can be 1 meter (m).

[0059] Step 402: Determine the Frenet coordinate system based on the starting position and the path reference line.

[0060] Please refer to Figure 5 , Figure 5 , which illustrates the Frenet coordinate system according to an embodiment of the present application. In some embodiments, the corresponding position of the starting position of the autonomous driving device (the position in the xy coordinates of the Cartesian coordinate system) on the path reference line P 0 can be determined as the origin O of the Frenet coordinate system, the tangent direction of the path reference line P 0 is determined as the longitudinal direction s of the Frenet coordinate system, and the normal direction of the path reference line P 0 is determined as the lateral direction l of the Frenet coordinate system.

[0061] By determining the Frenet coordinate system in the above manner, the unified adaptability to structured roads with different shapes and curvatures can be ensured.

[0062] Step 403: Based on the path points to be optimized, establish a cost function in the Frenet coordinate system. The cost function is a quadratic function, and the cost function introduces a first penalty term, which is configured to characterize the deviation between the path points to be optimized and the laterally offset path reference line.

[0063] In some examples, considering that the planned path can adapt to complex scenarios of multiple environmental participants, the planned path is described by discrete points, and considering the curvature continuity constraint of vehicle kinematics, a cubic polynomial is used to describe between every two discrete path points to be optimized s i . Therefore, in the Frenet coordinate system, the path points to be optimized s iIt can be configured with a first variable l i , a second variable l' i and a third variable l'' i . The first variable l i is used to represent the abscissa of the path point s to be optimized i in the Frenet coordinate system. The second variable l' i is the first derivative of the first variable l i . The third variable l'' i is the second derivative of the first variable l i . Since the discrete i-th path point s to be optimized i and the (i + 1)-th path point s to be optimized i+1 are described by a cubic polynomial, so (l'' i+1 - l'') / d i is a constant s .

[0064] If the number of path points s to be optimized i is n, then the variable x to be optimized in the embodiment of the present application can be represented by the following formula (1):

[0065] x = {l 0 , l' 0 , l'' 0 , l 1 , l' 1 , l'' 1 , …, l n-1 , l' n-1 , l'' n-1} (1)

[0066] In formula (1), l n-1 is the first variable of the n-th path point s to be optimized n , l' n-1 is the second variable of the n-th path point s to be optimized n , and l'' n-1 is the third variable of the n-th path point s to be optimized n .

[0067] In the embodiment of the present application, after defining the variable x to be optimized, a cost function needs to be established. The cost function in the embodiment of the present application is a quadratic form function, and its standard form is represented by the following formula (2):

[0068]

[0069] s.t. l ≤ Ax ≤ u

[0070]

[0071] In formula (2), Q is x2 The coefficient matrix of c T is the coefficient matrix of x, min is to take the minimum value of the given parameter, s.t. is the constraint condition, A is the constraint matrix, l is the lower bound vector of the constraint, u is the upper bound vector of the constraint, and R n 、 are the corresponding feasible ranges respectively.

[0072] In some embodiments, the cost function can be established in the Frenet coordinate system through the following steps:

[0073] Step 1, based on the second variable and the third variable, construct the second penalty term, and the second penalty term is used to characterize the path smoothness.

[0074] Specifically, the second penalty term can be constructed by the following formula (3):

[0075]

[0076] In formula (3), cost 2 is the second penalty term, and w 1 、w 2 、w 3 represent the weights of each smoothing penalty term respectively, i is an integer greater than or equal to 1 and less than or equal to n, n is the number of path points to be optimized, l′ i is the second variable of the i-th path point to be optimized, l″ i is the third variable of the i-th path point to be optimized, and l″′ i is the third derivative of the first variable of the i-th path point to be optimized.

[0077] Step 2, based on the first penalty term and the second penalty term, construct the cost function.

[0078] Specifically, the cost function can be represented by the following formula (4):

[0079] min f = cost 1 + cost 2 (4)

[0080] In formula (4), f is the cost function, cost 1 is the first penalty term, and cost 2 is the second penalty term.

[0081] In some embodiments, the first penalty term cost 1 can be constructed through the following steps:

[0082] Step 1, based on the obstacles, determine the drivable boundary.

[0083] Specifically, considering that obstacle avoidance in the planned path is usually ensured by the convex space boundary, the obstacle can be expanded by a certain margin of safety distance along the transverse direction l. The safety distance should not be set too large, because an overly large transverse safety distance will narrow the transverse feasible driving area within the lane, making it difficult to solve the forced constraint conditions, and the narrowing of the feasible region will slow down the search for feasible solutions during the optimization iteration process, increasing the iteration solving time. Therefore, under the action of only the cost functions of smoothness and reference line distance, the planned path usually gets too close to the obstacle, reducing safety and comfort.

[0084] Please refer to Figure 6 and Figure 7 , Figure 6 which illustrates an example of constructing the first penalty term when the obstacle and the autonomous driving device are in the first positional relationship in the embodiment of the present application. Figure 7 which illustrates an example of constructing the first penalty term when the obstacle and the autonomous driving device are in the second positional relationship in the embodiment of the present application. In some examples, the drivable boundary may include a first boundary l min and a second boundary l max that are oppositely arranged along the transverse direction l of the Frenet coordinate system. At least one of the first boundary l min and the second boundary l max has a third boundary protruding towards the other boundary, and the third boundary is used to avoid the obstacle Obstacle. The third boundary has a boundary line l b parallel to the longitudinal direction s of the Frenet coordinate system. The boundary line l b has a first dimension L 1 along the longitudinal direction s, and the obstacle Obstacle has a second dimension L 2 along the longitudinal direction s, satisfying L 1 > L 2 . Exemplarily, it can further satisfy L 1 - L 2 > ΔL, where ΔL is a preset longitudinal distance threshold. It can be understood that the first dimension L b of the boundary line l 1 along the longitudinal direction s in the embodiment of the present application can be greatly expanded according to actual needs.

[0085] It can be understood that the third boundary belongs to a part of the first boundary l min or the second boundary l max . As Figure 6 shown, when the obstacle occupies a part of the first boundary l min , the first boundary l min has a third boundary protruding towards the second boundary l max . Correspondingly, asFigure 7 As shown, when the obstacle occupies a partial area of the second boundary l max , the second boundary l max has a third boundary protruding towards the first boundary l min . In addition, if both the first boundary l min and the second boundary l max are occupied by obstacles, then both the first boundary l min and the second boundary l max have third boundaries protruding towards the other boundary, which will not be elaborated here.

[0086] In the above way, the convex space boundary of the static obstacle in the longitudinal direction can be expanded, which helps to further improve the smoothness of the planned path.

[0087] Step 2: Based on the drivable boundary and the path reference line, construct the reference line offset parameter corresponding to the path point to be optimized.

[0088] In some examples, the reference line offset parameter l i corresponding to the path point s to be optimized can be constructed in the following way: d (s):

[0089] First step, if the distance between the drivable boundary corresponding to the current position of the path point s to be optimized and the path reference line P i corresponding to the current position is less than the preset safety distance l 0 , based on the lateral coordinate of the drivable boundary corresponding to the current position, the current pose information of the autonomous driving device, and the preset safety distance l s , obtain the reference line offset parameter l s corresponding to the path point s to be optimized. i (s). d (s).

[0090] Among them, the preset safety distance ls can be the ideal interval between the obstacle boundary and the autonomous driving device boundary.

[0091] Please continue to refer to Figure 6 and Figure 7 , for example, the current pose information may include the width w of the autonomous driving device along the lateral direction l of the Frenet coordinate system.

[0092] Specifically, the reference line offset parameter l i corresponding to the path point s to be optimized can be obtained through the following formula (5): d (s):

[0093]

[0094] In formula (5), l d(s) is the path point s to be optimized i The corresponding reference line offset parameter, (l min (s), l max (s)) is the path point s to be optimized i The lateral coordinate range of the drivable boundary corresponding to the current position of, w is the width of the autonomous driving device, l s Is the preset safety distance.

[0095] Second, if the distance between the drivable boundary corresponding to the current position and the path reference line P corresponding to the current position 0 Is greater than or equal to the preset safety distance l s , the reference line offset parameter l i Corresponding to the path point s to be optimized d (s) is set to zero.

[0096] Specifically, formula (5) can be further supplemented as formula (6) below:

[0097]

[0098] It can be understood that the reference line offset parameter l d (s) is a function of the path point s to be optimized i .

[0099] In Figure 6 In the shown example, the obstacle occupies a part of the first boundary l min At the current position of the path point s to be optimized i The abscissa of the corresponding first boundary l min Is less than or equal to -l s In this case, at the current position of the path point s to be optimized i The abscissa of the corresponding first boundary l min And the path reference line P corresponding to the current position 0 The distance between them is greater than or equal to l s , then the reference line offset parameter l d (s) is set to zero, that is to say, this part of the path reference line P 0 Does not need to be offset. At the current position of the path point s to be optimized i The abscissa of the corresponding first boundary l min Is greater than -l s In this case, at the current position of the path point s to be optimized i The abscissa of the corresponding first boundary l min And the path reference line P corresponding to the current position 0 The distance between them is less than l s , then the reference line offset parameter l d (s) is set to the first boundary lmin The abscissa of is the sum of w / 2 and l s , that is, this part of the path reference line P 0 needs to be offset to the first local offset reference line A 1 position.

[0100] Exemplarily, the first local offset reference line A 1 has a third dimension L along the longitudinal direction s 3 , satisfying L 3 ≥L 1 (the first dimension of the boundary line l of the third boundary b ).

[0101] In Figure 7 the example shown, the obstacle occupies a partial area of the second boundary l max , at the path point s to be optimized i when the abscissa of the second boundary l corresponding to the current position is greater than or equal to l max , at this time, the distance between the second boundary l corresponding to the current position of the path point s to be optimized s and the path reference line P corresponding to the current position is greater than or equal to l i , then the reference line offset parameter l max (s) is set to zero, that is to say, this part of the path reference line P 0 does not need to be offset. When the abscissa of the second boundary l corresponding to the current position of the path point s to be optimized s is less than l d , at this time, the distance between the second boundary l corresponding to the current position of the path point s to be optimized 0 and the path reference line P corresponding to the current position is less than l i , then the reference line offset parameter l max is set to the difference between the abscissa of the second boundary l and the sum of w / 2 and l s , that is, this part of the path reference line P i needs to be offset to the second local offset reference line A max position. 0 s d max s (s) is set to the difference between the abscissa of the second boundary l and the sum of w / 2 and l max , that is, this part of the path reference line P s needs to be offset to the second local offset reference line A 0 position. 2 position.

[0102] Exemplarily, the second local offset reference line A 2 has a fourth dimension L along the longitudinal direction s 4 , satisfying L 4 ≥L 1 (the first dimension of the boundary line l of the third boundary b ).

[0103] In the above manner, through the path point s to be optimized i the lateral coordinate range (l min (s), l max (s)) of the drivable boundary at the path reference line P 0 is used to determine whether there is a sufficient ideal set distance between the autonomous driving device and the drivable boundary, so as to determine the offset parameter. When the distance between the drivable boundary and the path reference line P 0 is small, local offset is performed on the path reference line P 0 to make the planned path achieve the effect of staying away from obstacles.

[0104] Step 3: Based on the first variable corresponding to each path point to be optimized and the reference line offset parameter, construct a first penalty term.

[0105] Specifically, the first penalty term cost 1 can be expressed as the following formula (7):

[0106]

[0107] In formula (7), l i is the first variable corresponding to the i-th path point to be optimized, l d (s i ) is the reference line offset parameter corresponding to the i-th path point to be optimized, and n is the number of path points to be optimized.

[0108] In some embodiments, a dynamic weight w 1 is configured for the first penalty term cost 4 , and the dynamic weight w 4 is configured to vary based on the longitudinal distance between each path point to be optimized and the obstacle along the longitudinal direction of the Frenet coordinate system.

[0109] In some examples, the dynamic weight w 4 can be constructed through the following steps:

[0110] First step: If the longitudinal distance between the path point to be optimized and the obstacle along the longitudinal direction exceeds the preset range, set the dynamic weight to the first preset weight value.

[0111] Exemplarily, the first preset weight value can be set to zero.

[0112] Second step: If the longitudinal distance is within the preset range, obtain the dynamic weight based on the first preset weight value, the second preset weight value, the position of the path point to be optimized along the longitudinal direction, and the position of the obstacle along the longitudinal direction.

[0113] Specifically, the dynamic weight w can be obtained through the following formula (8): 4 (s i )

[0114]

[0115] In formula (8), w 4 (s i ) is the dynamic weight, w r is the first preset weight value, w d is the second preset weight value, s i is the position of the path point to be optimized along the longitudinal direction, s l is the lower limit of the preset range, s u is the upper limit of the preset range, s c is the position of the obstacle along the longitudinal direction, where the position s c of the obstacle along the longitudinal direction can be based on the center point of the obstacle.

[0116] Taking the example shown Figure 6 as an example, when the longitudinal position s i of the path point s to be optimized i is less than the lower limit s l of the preset range, the path point s i to be optimized is located on the left side of the obstacle. At this time, the longitudinal distance between the path point to be optimized and the obstacle along the longitudinal direction exceeds the preset range, and the dynamic weight w 4 (s i ) can be set to the first preset weight value w r . When the longitudinal position s i of the path point s i to be optimized is between the lower limit s l of the preset range and the longitudinal position s c of the obstacle, the path point s i to be optimized is located on the left side of the obstacle. At this time, the longitudinal distance between the path point to be optimized and the obstacle along the longitudinal direction is within the preset range, then the dynamic weight w 4 (s i ) can be set to gradually increase.

[0117] When the longitudinal position s i of the path point s i to be optimized is between the longitudinal position s c of the obstacle and the upper limit s u of the preset range, the path point s i to be optimized is located on the right side of the obstacle. At this time, the longitudinal distance between the path point to be optimized and the obstacle along the longitudinal direction is within the preset range, then the dynamic weight w 4 (s i) is gradually reduced. At the path point to be optimized s i The vertical position s i Greater than the upper limit of the preset range u In the case of i Located on the right side of the obstacle, the longitudinal distance between the path point to be optimized and the obstacle along the longitudinal direction exceeds the preset range, and the dynamic weight w 4 (s i ) can be restored to the first preset weight value w r .

[0118] In the above manner, because the transition between the local bias reference line and the original path reference line is not smooth, the dynamic weight is set as a linear function of the distance between the path point to be optimized and the obstacle, so that the path can transition smoothly in the process of approaching the local bias reference line, thereby improving the smoothness of the planned path, thereby further improving driving comfort while staying away from obstacles.

[0119] In some embodiments, the first penalty term may be constructed by the following formula (9):

[0120]

[0121] In formula (9), cost 1 is the first penalty term, w 4 (s i ) is the dynamic weight corresponding to the i-th path point to be optimized, l i is the first variable corresponding to the i-th path point to be optimized, l d (s i ) is the reference line bias parameter corresponding to the i-th path point to be optimized.

[0122] Then based on the above formula (3) and formula (9), the cost function f represented by formula (4) of the embodiment of the present application can be expressed by the following formula (10):

[0123]

[0124] In this way, by constructing a cost function in the manner of an embodiment of the present application, a reference line bias penalty item may be added when the obstacle partially occupies the boundary of the drivable area, resulting in a small distance between the boundary of the convex space and the reference line, so that the planned path can achieve the effect of being away from the obstacle. At the same time, since the changes of the local biased reference line and the original reference line are not smooth, the boundary of the obstacle is expanded along the longitudinal direction s and a dynamic weight is introduced, which can make the planned path smoother and further meet the driving comfort requirements.

[0125] Step 404: Solve the cost function based on the preset constraint conditions to obtain the target path corresponding to the minimum cost function.

[0126] In some examples, the preset constraint conditions include equality constraints and inequality constraints, and the inequality constraints include convex space boundary constraints and kinematic limit constraints.

[0127] Exemplarily, for the equality constraints in the embodiments of the present application, cubic polynomial fitting is adopted between every two path points to be optimized to achieve the curvature continuity condition of the path relative to the path reference line. Therefore, the equality constraint shown in the following formula (11) is satisfied between adjacent points:

[0128]

[0129] In formula (11), l″′ i→i+1 is the third-order term between two adjacent path points s i and s i+1 , l″ i+1 is the third variable of the path point s to be optimized i+1 , l″ i is the third variable of the path point s to be optimized i , d s is the longitudinal difference between two adjacent path points s to be optimized i and s i+1 in the frenet coordinate system, l′ i+1 is the second variable of the path point s to be optimized i+1 , l′ i is the second variable of the path point s to be optimized i , l i+1 is the first variable of the path point s to be optimized i+1 , l i is the first variable of the path point s to be optimized i .

[0130] Exemplarily, for the convex space boundary constraints in the embodiments of the present application, the autonomous driving device needs to ensure driving within the road boundary and at the same time needs to satisfy avoiding collisions with obstacles. Therefore, for the autonomous driving device, it is necessary to satisfy the constraints considering the convex space boundary of the lane and obstacles. Then, different convex space boundaries will be generated in different scenarios, such as scenarios of lane keeping, lane changing, or borrowing a lane to bypass, etc. For the scenario of borrowing a lane to bypass, reference can be made to Figure 6 and Figure 7 for the convex space boundary (l min (s), l max (s)) shown.

[0131] Please refer to Figure 8 , Figure 8Schematically shows the model established for the autonomous driving device in the embodiments of the present application. The width of the autonomous driving device is w, and the distance from the centroid of the autonomous driving device to the center of the front bumper (a protective device installed at the front end of the autonomous driving device, such as a front bumper) is d 1 , and the distance from the centroid of the autonomous driving device to the midline of the rear bumper (a protective device installed at the rear end of the autonomous driving device, such as a rear bumper) is d 2 , for the four corner points p 1 、p 2 、p 3 、p 4 convex space boundary constraints are performed, that is, the constraints of l. Let the coordinates of the centroid of the obstacle in the frenet coordinate system be (s, l), and the angle between the unit tangent vector of the projection of the centroid of the obstacle on the reference line and the heading of the autonomous driving device be θ. Correspondingly, the lateral coordinates of the four corner points p 1 、p 2 、p 3 、p 4 can be determined by the following formula (12):

[0132]

[0133] In formula (12), are respectively the lateral coordinates of the four corner points p 1 、p 2 、p 3 、p 4 , l is the lateral coordinate of the centroid of the obstacle, d 1 is the distance from the centroid to the center of the front bumper, d 2 is the distance from the centroid to the center of the rear bumper, θ is the angle between the unit tangent vector of the projection of the centroid on the reference line and the heading of the autonomous driving device, and w is the width of the autonomous driving device.

[0134] Since the trigonometric functions involved in the calculation of the corner points are non-linear, and approximately sin(θ)≈l′, cos(θ)≈1, then formula (12) can be transformed into the following formula (13):

[0135]

[0136] Find the maximum value ub i - d 2 , s i + d 1 ) of the lateral coordinate l of the centroid of the obstacle and the minimum value lb i , and finally obtain the convex space boundary constraint shown in the following formula (14): i

[0137] ​

[0138] Exemplarily, for the kinematic constraint in the embodiments of the present application, according to the state of the autonomous driving device, it is necessary to consider the kinematic constraint limitations of the lateral speed, lateral acceleration, and lateral acceleration change rate of the autonomous driving device, and it is necessary to correspondingly constrain l′, l″, and l″′. Among them, for the constraint of the third variable l″, the curvature κ of the path reference line needs to be considered r , according to the conversion formula between the natural coordinate system and the Cartesian coordinate system, there is an expression shown in the following formula (15):

[0139]

[0140] In formula (15), the third variable l″ is the lateral acceleration of the autonomous driving device, κ′ r is the change rate of the curvature κ of the path reference line, the second variable l′ is the lateral speed of the autonomous driving device, the first variable l is the lateral coordinate of the autonomous driving device, θ is the heading angle of the autonomous driving device, θ r is the heading angle of the path reference line, and κ is the planned curvature of the autonomous driving device.

[0141] Assume that the trajectory is planned almost along the path reference line, then Δθ≈0, and the planned curvature is a small quantity. Therefore, κ·κ r ≈0, and formula (15) can be simplified to the following formula (16):

[0142]

[0143] Therefore, there is a constraint on the third variable l″ as shown in the following formula (17):

[0144] -κ max -κ r ≤l″≤κ max +κ r (17)

[0145] In formula (17), κ max is the maximum curvature of the vehicle motion trajectory.

[0146] According to the vehicle kinematic model, the relationship between the maximum front wheel steering angle δ max and the maximum curvature κ of the vehicle motion trajectory is as shown in the following formula (18): max

[0147]

[0148] In formula (18), L is the wheelbase of the vehicle.

[0149] In summary, for each fixed path point s to be optimized i ​​, there are kinematic limit constraints as shown in the following formula (19):

[0150]

[0151] After setting the preset constraint conditions of the embodiments of the present application for the above method, a solver can be used to solve the cost function.

[0152] Specifically, the quadratic term and the coefficient of the linear term in the cost function of quadratic programming can be respectively converted into matrix Q and vector c, the equality constraints and inequality constraints can be converted into matrix A, and the upper and lower bounds of the constraints can be converted into vectors l and u, so as to be substituted into the solver for solution. Commonly used open-source solvers for quadratic programming numerical optimization include OSQP (Operator Splitting Quadratic Program) and qpOASES. Among them, OSQP uses the ADMM (Alternating Direction Method of Multipliers) method for solution, and this solver has better solution efficiency for problems with large scale and containing a large number of equality or inequality constraints. The solver qpOASES (Quadratic Programming Optimization and Simulation Engine) uses the feasible region method for solution, and for small-scale problems with fewer constraints, qpOASES is faster in solution. Since the path planning in the embodiments of the present application involves large-scale equality constraints and inequality constraints, OSQP can be selected to solve this quadratic programming problem.

[0153] To more clearly illustrate the method of the embodiments of the present application, please refer to Figures 9a to 9c , Figure 9a which schematically shows the target path planned by the path planning method of the embodiments of the present application in one of the application scenarios, Figure 9b which schematically shows the target path planned by the path planning method of the embodiments of the present application in the second application scenario, Figure 9c which schematically shows the target path planned by the path planning method of the embodiments of the present application in the third application scenario.

[0154] The target path T planned by the path planning method of the embodiments of the present application in the first application scenario 1 , within the drivable boundary (l min ~l max ), due to the local offset of the path reference line P 0 , when the autonomous driving device travels near the obstacle Obstacle, it will travel along the first local offset reference line A 1 , compared with directly traveling along the path reference line P 0 (for example Figure 3a the target path P shown)1 ) can be further away from obstacles, with a smoother path and a better driving experience.

[0155] Under Application Scenario 2, the target path T planned by the path planning method of the embodiment of the present application 2 , within the drivable boundary (l min ~l max ), due to the local offset of the path reference line P 0 , when the autonomous driving device travels near the obstacle Obstacle, it will travel along the second local offset reference line A 2 . Compared with traveling along the target path planned by the traditional method (for example Figure 3b the target path P shown 2 ), it can be further away from obstacles, with a smoother path and a better driving experience.

[0156] Under Application Scenario 3, the target path T planned by the path planning method of the embodiment of the present application 3 , within the drivable boundary (l min ~l max ), due to the local offset of the path reference line P 0 , when the autonomous driving device travels near the obstacle Obstacle, it will travel along the third local offset reference line A 3 . Compared with traveling along the path planned by the traditional method (for example Figure 3c the target path P shown 3 ), it can be further away from obstacles, with a smoother path and a better driving experience.

[0157] It can be understood that the path planning method provided by the embodiment of the present application introduces a first penalty term into the cost function. When the obstacle partially occupies the drivable area and the distance between the drivable boundary and the path reference line is small, the planned target path can achieve the effect of being far away from the obstacle. Since the change between the offset part and the original reference line is not smooth, the boundary in the longitudinal direction of the obstacle is extended and a dynamic weight is introduced, making the planned path smoother and meeting the comfort requirements of driving, thereby improving the driving experience.

[0158] Correspondingly, please refer to Figure 10 , Figure 10 which shows the structural diagram of the path planning device of the embodiment of the present application. The path planning device provided by the embodiment of the present application can be a path quality optimizer 140, which is specifically configured in the autonomous driving device. The path planning device includes an information acquisition unit 141, a coordinate system establishment unit 142, a cost function construction unit 143, and a target path acquisition unit 144.

[0159] An information acquisition unit 141 is configured to acquire path information and a starting position of an autonomous driving device. The path information includes a path reference line, and a path point to be optimized is determined based on the path reference line.

[0160] A coordinate system establishment unit 142 is configured to determine a Frenet coordinate system based on the starting position and the path reference line.

[0161] A cost function construction unit 143 is configured to establish a cost function in the Frenet coordinate system based on the path point to be optimized. The cost function is a quadratic function, and the cost function introduces a first penalty term, where the first penalty term is used to characterize the deviation between the path point to be optimized and the laterally offset path reference line.

[0162] A target path acquisition unit 144 is configured to solve the cost function based on a preset constraint condition to obtain a target path corresponding to the minimum value of the cost function.

[0163] In some embodiments, the path point to be optimized is configured with a first variable, where the first variable is used to characterize the abscissa of the path point to be optimized in the Frenet coordinate system. The path information further includes obstacles.

[0164] The cost function construction unit 143 constructs the first penalty term in the following manner:

[0165] Based on the obstacles, a drivable boundary is determined.

[0166] Based on the drivable boundary and the path reference line, a reference line offset parameter corresponding to the path point to be optimized is constructed.

[0167] Based on the first variable and the reference line offset parameter corresponding to each path point to be optimized, the first penalty term is constructed.

[0168] In some embodiments, the cost function construction unit 143 is specifically configured to:

[0169] If the distance between the drivable boundary corresponding to the current position of the path point to be optimized and the path reference line corresponding to the current position is less than a preset safety distance, based on the lateral coordinate of the drivable boundary corresponding to the current position, the current pose information of the autonomous driving device, and the preset safety distance, a reference line offset parameter corresponding to the path point to be optimized is obtained.

[0170] If the distance between the drivable boundary corresponding to the current position and the path reference line corresponding to the current position is greater than or equal to the preset safety distance, the reference line offset parameter corresponding to the path point to be optimized is set to zero.

[0171] In some embodiments, the current pose information includes the width of the autonomous driving device along the lateral direction of the Frenet coordinate system.

[0172] The cost function construction unit 143 is specifically configured to:

[0173] Obtain the reference line offset parameter corresponding to the path point to be optimized through the following formula:

[0174]

[0175] where l d (s) is the reference line offset parameter corresponding to the path point s to be optimized, (l i (s), l min (s)) is the lateral coordinate range of the drivable boundary corresponding to the current position of the path point s to be optimized, w is the width of the autonomous driving device, and l max is the preset safety distance. i s

[0176] In some embodiments, a dynamic weight is configured for the first penalty term, and the dynamic weight is configured to vary based on the longitudinal distance between each path point to be optimized and the obstacle along the longitudinal direction of the Frenet coordinate system.

[0177] In some embodiments, the cost function construction unit 143 is specifically configured to construct the dynamic weight in the following manner:

[0178] If the longitudinal distance between the path point to be optimized and the obstacle along the longitudinal direction exceeds the preset range, set the dynamic weight to the first preset weight value.

[0179] If the longitudinal distance is within the preset range, obtain the dynamic weight based on the first preset weight value, the second preset weight value, the position of the path point to be optimized along the longitudinal direction, and the position of the obstacle along the longitudinal direction.

[0180] In some embodiments, the cost function construction unit 143 is specifically configured to:

[0181] Obtain the dynamic weight through the following formula:

[0182]

[0183] where w 4 (s i ) is the dynamic weight, w r is the first preset weight value, w d is the second preset weight value, s i is the position of the path point to be optimized along the longitudinal direction, s l is the lower limit of the preset range, s u is the upper limit of the preset range, and s c is the position of the obstacle along the longitudinal direction.

[0184] In some embodiments, the cost function construction unit 143 is specifically configured to:

[0185] Construct a first penalty term through the following formula:

[0186]

[0187] where cost 1 is the first penalty term, w 4 (s i ) is the dynamic weight corresponding to the i-th path point to be optimized, l i is the first variable corresponding to the i-th path point to be optimized, l d (s i ) is the reference line offset parameter corresponding to the i-th path point to be optimized.

[0188] In some embodiments, the variable to be optimized is further configured with a second variable and a third variable. The second variable is the first derivative of the first variable, and the third variable is the second derivative of the first variable.

[0189] The cost function construction unit 143 is specifically configured to:

[0190] Construct a second penalty term based on the second variable and the third variable, and the second penalty term is used to characterize path smoothness.

[0191] Construct a cost function based on the first penalty term and the second penalty term.

[0192] In some embodiments, the drivable boundary includes a first boundary and a second boundary that are oppositely arranged in the lateral direction of the Frenet coordinate system. At least one of the first boundary and the second boundary has a third boundary that protrudes toward the other boundary, and the third boundary is used to avoid obstacles.

[0193] The third boundary has a boundary line parallel to the longitudinal direction of the Frenet coordinate system, and the boundary line has a first dimension L 1 , the obstacle has a second dimension L 2 , and it satisfies L 1 > L 2 .

[0194] In some embodiments, the preset constraint conditions include equality constraints and inequality constraints, and the inequality constraints include convex space boundary constraints and kinematic limit constraints.

[0195] It can be understood that the path planning device provided in this application introduces a first penalty term in the cost function, which can make the planned target path reach the effect of staying away from obstacles when the obstacle partially occupies the drivable area and the distance between the drivable boundary and the path reference line is small, thereby improving the driving experience.

[0196] Correspondingly, please refer to Figure 11 , Figure 11 which schematically shows the structure of the path planning device of the automatic driving device according to the embodiment of the present application. The embodiment of the present application also provides a path planning device for an automatic driving device, including a memory 1101 and a processor 1102. The memory 1101 is used to store programs. The processor 1102 is used to execute the programs stored in the memory 1101. When the programs stored in the memory 1101 are executed, the processor 1102 executes the path planning method of the foregoing embodiment of the present application.

[0197] Correspondingly, the automatic driving device provided by the embodiment of the present application can plan a smooth path away from obstacles, thus having a better driving experience.

[0198] Correspondingly, the embodiment of the present application also provides a computer-readable storage medium, which stores instructions for a computing device to execute. When the computing device executes these instructions, the path planning method as described in the foregoing embodiment of the present application is implemented.

[0199] The above has introduced in detail a path planning method, device, automatic driving device and readable storage medium provided by the embodiments of the present application. Specific examples are used herein to elaborate on the principle and implementation manner of the present application. The description of the above embodiments is only used to help understand the technical solution and its core idea of the present application; those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A path planning method, characterized in that, applied to an autonomous driving device, the method includes: Obtain path information and the starting position of the autonomous driving device, the path information includes a path reference line, and determine the path points to be optimized based on the path reference line; Determine the Frenet coordinate system based on the starting position and the path reference line; Based on the path points to be optimized, establish a cost function in the Frenet coordinate system, the cost function is a quadratic function, and the cost function introduces a first penalty term, and the first penalty term is configured to characterize the deviation between the path points to be optimized and the laterally offset path reference line; Solve the cost function based on preset constraint conditions to obtain the target path corresponding to the minimum value of the cost function.

2. A path planning method according to claim 1, characterized in that, The path points to be optimized are configured with a first variable, and the first variable is used to characterize the abscissa of the path points to be optimized in the Frenet coordinate system.

3. A path planning method according to claim 2, characterized in that, The path information further includes obstacles, and the first penalty term is constructed in the following manner: Based on the obstacles, determine the drivable boundary; based on the drivable boundary and the path reference line, construct the reference line offset parameter corresponding to the path points to be optimized; Based on the first variable and the reference line offset parameter corresponding to each path point to be optimized, construct the first penalty term.

4. A path planning method according to claim 3, characterized in that, Based on the drivable boundary and the path reference line, constructing the reference line offset parameter corresponding to the path points to be optimized includes: If the distance between the drivable boundary corresponding to the current position of the path point to be optimized and the path reference line corresponding to the current position is less than a preset safety distance, based on the lateral coordinate of the drivable boundary corresponding to the current position, the current pose information of the autonomous driving device, and the preset safety distance, obtain the reference line offset parameter corresponding to the path point to be optimized; If the distance between the drivable boundary corresponding to the current position and the path reference line corresponding to the current position is greater than or equal to the preset safety distance, set the reference line offset parameter corresponding to the path point to be optimized to zero.

5. A path planning method according to claim 4, characterized in that, The current pose information includes the width of the autonomous driving device along the lateral direction of the Frenet coordinate system; Based on the lateral coordinate of the drivable boundary corresponding to the current position, the current pose information of the autonomous driving device, and the preset safety distance, obtaining the reference line offset parameter corresponding to the path point to be optimized includes: Obtain the reference line offset parameter corresponding to the path point to be optimized through the following formula: Among them, l d (s) is the path point s to be optimized i The corresponding reference line bias parameter, (l min (s), l max (s)) is the path point s to be optimized i The horizontal coordinate range of the drivable boundary corresponding to the current position of the automatic driving device, w is the width of the automatic driving device, l s The preset safety distance.

6. A path planning method according to claim 3, characterized in that, A dynamic weight is configured for the first penalty term, and the dynamic weight is configured to vary based on the longitudinal distance between each of the path points to be optimized and the obstacle along the longitudinal direction of the Frenet coordinate system.

7. A path planning method according to claim 6, wherein, the dynamic weight is constructed in the following manner: If the longitudinal distance between the path point to be optimized and the obstacle along the longitudinal direction exceeds a preset range, the dynamic weight is set to a first preset weight value; If the longitudinal distance is within the preset range, the dynamic weight is obtained based on the first preset weight value, the second preset weight value, the position of the path point to be optimized along the longitudinal direction, and the position of the obstacle along the longitudinal direction.

8. A path planning method according to claim 7, wherein, obtaining the dynamic weight based on the first preset weight value, the second preset weight value, the position of the path point to be optimized along the longitudinal direction, and the position of the obstacle along the longitudinal direction, includes: obtaining the dynamic weight through the following formula: Among them, w 4 (s i ) is the dynamic weight, w r is the first preset weight value, w d is the second preset weight value, s i is the position of the path point to be optimized along the longitudinal direction, s l is the lower limit of the preset range, s u is the upper limit of the preset range, s c is the position of the obstacle along the longitudinal direction.

9. A path planning method according to claim 6, wherein, constructing the first penalty term based on the first variable corresponding to each of the path points to be optimized and the reference line offset parameter, includes: constructing the first penalty term through the following formula: Among them, cost 1 is the first penalty term, w 4 (s i ) is the dynamic weight corresponding to the i-th path point to be optimized, l i is the first variable corresponding to the i-th path point to be optimized, l d (s i ) is the reference line offset parameter corresponding to the i-th path point to be optimized.

10. A path planning method according to claim 3, wherein, the drivable boundary includes a first boundary and a second boundary oppositely arranged along the lateral direction of the Frenet coordinate system, and at least one of the first boundary and the second boundary has a third boundary protruding towards the other boundary, and the third boundary is used to avoid the obstacle; The third boundary has a boundary line parallel to the longitudinal direction of the Frenet coordinate system, and the boundary line has a first dimension L along the longitudinal direction 1 , and the obstacle has a second dimension L along the longitudinal direction 2 , satisfying L 1 > L 2 .

11. A path planning method according to claim 2, wherein, the variable to be optimized is further configured with a second variable and a third variable, the second variable is the first derivative of the first variable, and the third variable is the second derivative of the first variable; establishing a cost function in the Frenet coordinate system based on the path point to be optimized, includes: constructing a second penalty term based on the second variable and the third variable, and the second penalty term is used to characterize the path smoothness; constructing the cost function based on the first penalty term and the second penalty term.

12. A path planning method according to claim 1, wherein, the preset constraint conditions include equality constraints and inequality constraints, and the inequality constraints include convex space boundary constraints and kinematic limit constraints.

13. A path planning device, wherein, configured in an autonomous driving device, the device includes: an information acquisition unit, configured to acquire path information and the starting position of the autonomous driving device, the path information includes a path reference line, and determine path points to be optimized based on the path reference line; a coordinate system establishment unit, configured to determine a Frenet coordinate system based on the starting position and the path reference line; A cost function construction unit, configured to establish a cost function in the Frenet coordinate system based on the path points to be optimized, where the cost function is a quadratic function, and a first penalty term is introduced into the cost function, and the first penalty term is configured to characterize the deviation between the path points to be optimized and a path reference line with a lateral offset; A target path acquisition unit, configured to solve the cost function based on preset constraint conditions and acquire a target path corresponding to the minimum value of the cost function.

14. A path planning device for an autonomous driving device, characterized in that, it includes: a memory, configured to store a program; a processor, configured to execute the program stored in the memory; when the program stored in the memory is executed, the processor executes the path planning method according to any one of claims 1-12.

15. An autonomous driving device, characterized in that, it includes the path planning device according to claim 13.

16. A computer-readable storage medium, characterized in that, the computer-readable medium stores instructions for a computing device to execute, and when the computing device executes the instructions, the method according to any one of claims 1-12 is implemented.

Images