Lateral control method and device based on vehicle dynamic prediction model

Abstract

The application provides a lateral control method and device based on a vehicle dynamic prediction model, and the method comprises the following steps: inputting a first steering wheel steering angle in a target time window obtained in advance into the vehicle dynamic prediction model; performing convolution on the input first steering wheel steering angle in the target time window and a unit impulse response of the vehicle dynamic prediction model, to obtain a predicted yaw rate; and obtaining a predicted steering wheel steering angle based on the predicted yaw rate and a pre-established cost function. The application improves the accuracy of the predicted steering wheel steering angle, reduces the adjustment of an actuator while ensuring the accuracy, improves the durability of the actuator, and reduces the total cost of ownership (TCO); in addition, a large amount of operation data is used to iteratively optimize the constraint conditions in the algorithm, so that the automatic driving behavior is more accurate.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and in particular to a lateral control method and device based on a vehicle dynamic prediction model. Background Technology

[0002] With the increasing popularity of autonomous vehicles, they can be used as taxis or public transportation. When using an autonomous vehicle, passengers need to input their destination. The autonomous vehicle will then generate a route based on the current location and destination and drive along the generated route.

[0003] Currently, most systems employ planning and control systems to generate driving routes and use control modules to track those routes. As a necessary and sufficient condition for safe vehicle operation, planning and control systems provide customers with five key benefits: safe driving, precise control, comfortable ride, economical use, and good durability.

[0004] However, there are some mutually restrictive indicators in the control module. For example, if you pursue control precision too much, the controlled object will inevitably need to be adjusted more frequently, thereby reducing the durability of the actuator. Summary of the Invention

[0005] This invention provides a lateral control method and device based on a vehicle dynamic prediction model, which solves the defect in the prior art where the mutual constraints between indicators reduce the durability of the actuator while ensuring accuracy. Under the premise of ensuring safety boundaries, the method reduces the adjustment of the actuator, thereby improving the durability of the actuator.

[0006] This invention provides a lateral control method based on a vehicle dynamic prediction model, comprising: inputting a first steering wheel angle within a pre-acquired target time window into a vehicle dynamic prediction model; convolving the first steering wheel angle within the input target time window with the unit impulse response of the vehicle dynamic prediction model to obtain a predicted yaw rate; and obtaining a predicted steering wheel angle based on the predicted yaw rate and a pre-established cost function, so as to control the vehicle according to the predicted steering wheel angle.

[0007] According to the present invention, a lateral control method based on a vehicle dynamic prediction model is provided, wherein obtaining the predicted steering wheel angle based on the predicted yaw rate and a pre-established cost function includes: obtaining multiple costs and a second steering wheel angle corresponding to each cost based on the predicted yaw rate and the pre-established cost function; and selecting the corresponding second steering wheel angle as the predicted steering wheel angle based on the minimum cost and preset constraints.

[0008] According to the present invention, a lateral control method based on a vehicle dynamic prediction model is provided. The method involves obtaining multiple costs and corresponding second steering wheel angles based on the predicted yaw rate and a pre-established cost function. The method includes: obtaining a yaw rate offset error based on the predicted yaw rate and a pre-acquired inferred yaw rate; obtaining the heading angle, sideslip angle, and lateral position of a trajectory point corresponding to the first steering wheel angle within a target time window based on the first steering wheel angle corresponding to the predicted yaw rate; obtaining a heading angle offset error based on the lateral offset error, the heading angle, and the sideslip angle; obtaining a lateral position offset error based on the yaw rate offset error, the heading angle offset error, and the lateral position; and obtaining multiple costs and corresponding second steering wheel angles based on the yaw rate offset error, the heading angle offset error, the lateral position offset error, and the pre-established cost function.

[0009] According to a lateral control method based on a vehicle dynamic prediction model provided by the present invention, before obtaining the yaw rate offset error based on the predicted yaw rate and the pre-acquired inferred yaw rate, the method includes: acquiring trajectory points within a target time window; obtaining the curvature corresponding to each trajectory point based on the trajectory points; and obtaining the inferred yaw rate based on the curvature of each trajectory point and the vehicle speed corresponding to the trajectory point.

[0010] According to a lateral control method based on a vehicle dynamic prediction model provided by the present invention, the step of selecting a corresponding second steering wheel angle based on minimum cost and preset constraints to obtain a predicted steering wheel angle includes: determining lateral position error constraints and steering wheel speed constraints based on the target driving area; selecting a corresponding second steering wheel angle according to the lateral position error constraints and the steering wheel speed constraints; and selecting the second steering wheel angle with the minimum cost as the predicted steering wheel angle based on the cost of the selected second steering wheel angle.

[0011] According to a lateral control method based on a vehicle dynamic prediction model provided by the present invention, before inputting the first steering wheel angle within a pre-acquired target time window into the vehicle dynamic prediction model, the method includes: acquiring the first steering wheel angle and the first yaw rate within the target time window; and normalizing the first steering wheel angle within the target time window.

[0012] After obtaining the predicted yaw rate, the process includes: updating the vehicle dynamic prediction model based on the predicted yaw rate, the first yaw rate within the target time window, and the normalized first steering wheel angle.

[0013] According to the lateral control method based on a vehicle dynamic prediction model provided by the present invention, after obtaining the first steering wheel angle and the first yaw rate within the target time window, the method further includes: obtaining the corresponding average steering wheel angle and average yaw rate based on the first steering wheel angle and the first yaw rate within the target time window; reducing each first steering wheel angle within the target time window based on the average steering wheel angle; and reducing each first yaw rate within the target time window based on the average yaw rate.

[0014] The present invention also provides a lateral control device based on a vehicle dynamic prediction model, comprising: a yaw rate prediction module, which inputs a first steering wheel angle within a pre-acquired target time window into the vehicle dynamic prediction model, wherein the vehicle dynamic prediction model convolves the first steering wheel angle within the input target time window with the unit impulse response of the vehicle dynamic prediction model to obtain a predicted yaw rate; and an angle prediction module, which obtains a predicted steering wheel angle based on the predicted yaw rate and a pre-established cost function, so as to control the vehicle according to the predicted steering wheel angle.

[0015] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the lateral control method based on the vehicle dynamic prediction model as described above.

[0016] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the lateral control method based on the vehicle dynamic prediction model as described above.

[0017] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the lateral control method based on the vehicle dynamic prediction model as described above.

[0018] The lateral control method and apparatus based on a vehicle dynamic prediction model provided by this invention obtains the predicted yaw rate by convolving the first steering wheel angle within the input target time window with the unit impulse response using the vehicle dynamic prediction model. Based on the predicted yaw rate and a pre-established cost function, the predicted steering wheel angle is obtained, thereby improving the accuracy of the predicted steering wheel angle. While ensuring accuracy, it reduces the need for actuator adjustments, thus improving actuator durability and reducing the total cost of ownership (TCO). Furthermore, by using a large amount of operational data to iterate the optimal constraints in the algorithm, the autonomous driving behavior becomes more precise, improving the adaptability of the control method and the vehicle dynamic prediction model, thereby achieving better control performance. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating the lateral control method based on a vehicle dynamic prediction model provided by the present invention.

[0021] Figure 2 This is a schematic diagram of the lateral control device based on a vehicle dynamic prediction model provided by the present invention.

[0022] Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0024] Figure 1 This invention illustrates a lateral control method based on a vehicle dynamic prediction model, the method comprising:

[0025] S11, the first steering wheel angle within the target time window is pre-acquired and input into the vehicle dynamic prediction model. The vehicle dynamic prediction model convolves the first steering wheel angle within the target time window with the unit impulse response of the vehicle dynamic prediction model to obtain the predicted yaw rate.

[0026] S12, based on the predicted yaw rate and the pre-established cost function, obtains the predicted steering wheel angle, so as to control the vehicle according to the predicted steering wheel angle.

[0027] It should be noted that S1N in this specification does not represent the order of the lateral control methods based on the vehicle dynamic prediction model. The lateral control method based on the vehicle dynamic prediction model of this invention is described in detail below.

[0028] Step S11: Input the first steering wheel angle within the target time window obtained in advance into the vehicle dynamic prediction model. The vehicle dynamic prediction model convolves the first steering wheel angle within the input target time window with the unit impulse response of the vehicle dynamic prediction model to obtain the predicted yaw rate.

[0029] In an optional embodiment, before inputting the first steering wheel angle within the pre-acquired target time window into the vehicle dynamic prediction model, the method includes: acquiring the first steering wheel angle within the target time window. Specifically, this includes: acquiring trajectory points within the target time window; and acquiring the first steering wheel angle corresponding to each trajectory point. It should be noted that the length of the target time window can be set according to actual design requirements, such as 3s, 6s, 10s, etc., without further limitation here. For example, if the length of the target time window is 3 seconds, then the first steering wheel angle within the corresponding historical 3 seconds is acquired based on the target time window. If the system sampling frequency is 50Hz, then the first steering wheel angles of 150 trajectory points are taken as model input, where the first steering wheel angle of the first point is the model input at the current time, and the first steering wheel angles of the 2nd to 150th points are the model input within the historical 0.02 to historical 3 seconds. By acquiring the first steering wheel angle within the target time window, data can be acquired online in real time, thereby facilitating the prediction of the yaw rate based on the currently acquired first steering wheel angle within the target time window.

[0030] After obtaining the first steering wheel angle within the target time window, the first steering wheel angle within the target time window is input into the vehicle dynamic prediction model to obtain the predicted yaw rate. In this embodiment, the predicted yaw rate is expressed as:

[0031]

[0032] Where, ω k h represents the predicted yaw rate at time k. i h represents the unit impulse response of the vehicle dynamics prediction model. When the vehicle dynamics prediction model uses an FIR model, h i The discrete FIR function can be used to discretly convolve the first steering wheel angle within the target time window with the unit impulse response of the vehicle dynamic prediction model to obtain the predicted yaw rate; m represents the length of the vehicle dynamic prediction model, u k-i Let represent the first steering wheel angle at time ki within the target time window, and b represent the offset of the yaw rate. The FIR model possesses advantages that traditional vehicle models do not have, allowing this control method to be perfectly adapted to the FIR model, thereby achieving better control performance.

[0033] It should be noted that the predicted yaw rate output by the vehicle dynamic prediction model can form a dataset, represented as [ω].k ,ω k+1 ,...,ω k+n Let ], where n represents the number of trajectory points within the target time window. For example, if the sampling time is 10 Hz, meaning the target time window is shifted once every 0.1 seconds to obtain one trajectory point, then when t = 0.1, the corresponding k = 1; similarly, when t = 0.2, the corresponding k = 2. If the FIR model length is 3 seconds, then m = 30. The first steering wheel angle of 30 trajectory points needs to be obtained by shifting the target time window as input to the vehicle dynamic prediction model, and then convolved with the vehicle dynamic prediction model to obtain the predicted yaw rate output by the vehicle dynamic prediction model.

[0034] In one optional embodiment, before inputting the first steering wheel angle within a pre-acquired target time window into the vehicle dynamic prediction model, the method includes: acquiring the first steering wheel angle and the first yaw rate within the target time window; and normalizing the first steering wheel angle within the target time window. Correspondingly, after obtaining the predicted yaw rate, the method includes: updating the vehicle dynamic prediction model based on the predicted yaw rate, the first yaw rate within the target time window, and the normalized first steering wheel angle.

[0035] It should be noted that the first steering wheel angle and the first yaw rate within the target time window can be obtained simultaneously or not simultaneously. Obtaining the first steering wheel angle within the target time window only needs to be done before inputting the pre-obtained first steering wheel angle within the target time window into the vehicle dynamics prediction model. Similarly, obtaining the first yaw rate within the target time window only needs to be done before updating the vehicle dynamics prediction model based on the predicted yaw rate, the first yaw rate within the target time window, and the normalized first steering wheel angle. Furthermore, the method for obtaining the first yaw rate within the target time window can be referred to the section on obtaining the first steering wheel angle within the target time window above, and will not be repeated here.

[0036] Specifically, the first steering wheel angle within the target time window is normalized, including: obtaining the torque corresponding to the first steering wheel angle within the target time window; obtaining the median steering wheel torque based on the first steering wheel angle within the target time window, the torque corresponding to the first steering wheel angle, and the pre-established first impulse response model; and obtaining the normalized first steering wheel angle based on the median steering wheel torque and the corresponding first steering wheel angle.

[0037] In this embodiment, the first impulse response model is represented as follows:

[0038] y1=h1 T *P+b

[0039] Where y1 represents the output of the first impulse response model, i.e., the first steering wheel angle; P represents the input of the first impulse response model, i.e., the torque corresponding to the first steering wheel angle within the target time window, represented as a row vector; h1 represents the unit impulse response of the first impulse response model, h1 T Let represent the transpose of h1. When the torque is 0, 'b' represents the center of the steering wheel torque. It should be noted that since h1 is a column vector, it needs to be transposed into its corresponding row vector h1 for easier calculation. T .

[0040] It should be noted that the first impulse response model can adopt an online updated FIR model. During vehicle operation, both h1 and b will be updated according to the input first steering wheel angle and the torque corresponding to the first steering wheel angle. For details, please refer to the update method of the vehicle dynamic prediction model below, which will not be described here.

[0041] Because the steering wheel has a dead zone—a certain angular range within the center of the steering wheel where turning the steering wheel will not cause vehicle movement—it is necessary to first determine the median steering wheel torque. Then, based on the distance between the first steering wheel angle and the median steering wheel torque, the influence of the dead zone on the steering wheel angle is determined, resulting in the steering wheel angle affected by the dead zone, i.e., the normalized first steering wheel angle mentioned above. Further, the normalized first steering wheel angle is obtained based on the median steering wheel torque and the corresponding first steering wheel angle, including: determining the degree of influence of the dead zone on the steering wheel angle based on the median steering wheel torque and the corresponding first steering wheel angle, combined with vehicle weight and a preset dead zone sensitivity; and obtaining the normalized first steering wheel angle based on the preset dead zone width, the first steering wheel angle, and the degree of influence of the dead zone on the steering wheel angle.

[0042] Additionally, the normalized first steering wheel angle is represented as:

[0043]

[0044] Where y2 represents the normalized first steering wheel angle, u1 represents the first steering wheel angle obtained within the target time window, c represents the median steering wheel torque, m represents the vehicle weight, k1 represents the dead zone width, and k2 represents the dead zone sensitivity. This indicates the degree to which the dead zone affects the steering wheel angle. For example, there exists... When y2 = y3 = d, we obtain the corresponding u1 and u2, where k1 represents the distance between u1 and u2, and k2 represents the steepness of the y2 curve at point (c,0). It should be noted that the tanh function changes rapidly near the point (c,0) and approaches plus or minus 1 when the input is relatively large. Therefore, subtracting a value of k1*1 for large angles ensures that the angle change is not within the dead zone. For small angles, the value obtained by the tanh function is closer to the original small angle; subtracting it brings it close to 0, thus linking the dead zone to the center of the steering wheel torque.

[0045] In an optional embodiment, since some types of vehicles, such as trucks, exhibit zero bias and nonlinearity, after obtaining the first steering wheel angle and first yaw rate within the target time window, the method further includes: preprocessing the first steering wheel angle and first yaw rate within the target time window. Specifically, this includes: obtaining the corresponding average steering wheel angle and average yaw rate based on the first steering wheel angle and first yaw rate within the target time window; reducing each first steering wheel angle within the target time window based on the average steering wheel angle; and reducing each first yaw rate within the target time window based on the average yaw rate.

[0046] It should be noted that, based on the average steering wheel angle, each first steering wheel angle within the target time window is reduced, that is, each first steering wheel angle within the target time window is subtracted from the average steering wheel angle, in order to achieve preprocessing of the first steering wheel angle within the acquired target time window.

[0047] Similarly, based on the average yaw rate, each first yaw rate within the target time window is reduced, which means subtracting the average yaw rate from each first yaw rate within the target time window.

[0048] It should be noted that the preprocessing of the first steering wheel angle and the first yaw rate within the target time window can be performed before or after the normalization of the first steering wheel angle within the target time window, and no further restrictions are imposed here.

[0049] It should be noted that if the first steering wheel angle and first yaw rate within the target time window are preprocessed before normalization, then the preprocessed first steering wheel angle will be normalized during the normalization process. Similarly, if the first steering wheel angle and first yaw rate within the target time window are preprocessed after normalization, then the normalized first steering wheel angle will be preprocessed after the preprocessing of the first steering wheel angle and first yaw rate within the target time window.

[0050] Furthermore, after obtaining the predicted yaw rate, the process includes: updating the vehicle dynamic prediction model based on the predicted yaw rate, the first yaw rate within the target time window, and the normalized first steering wheel angle. Specifically, this includes: obtaining the yaw rate error based on the predicted yaw rate and the first yaw rate within the target time window; updating the vehicle dynamic prediction model based on the normalized first steering wheel angle, a first preset value, and the yaw rate error; determining the response strength of the updated vehicle dynamic prediction model, where the response strength is the sum of all weight values ​​of the updated vehicle dynamic prediction model; determining whether the response strength is within a preset interval, and determining the updated vehicle dynamic model based on the determination result.

[0051] Furthermore, based on the normalized first steering wheel angle, the first preset value, and the yaw rate error, the unit impulse response of the vehicle dynamic prediction model is updated, thereby updating the vehicle dynamic prediction model. The updated unit impulse response of the vehicle dynamic prediction model is expressed as:

[0052] h new (k)=h(k)+Δh

[0053]

[0054] Among them, h new (k) represents the unit impulse response of the updated vehicle dynamic prediction model, h(k) represents the unit impulse response of the vehicle dynamic prediction model before the update, k3 represents the first preset value, δ represents the second preset value, e represents the yaw rate error, and u represents the first steering wheel angle after preprocessing and normalization within the target time window. It should be noted that the first and second preset values ​​can be set according to actual design requirements or prior experience, and are not further limited here.

[0055] In addition, the vehicle dynamic update model is determined based on the judgment result, including: adjusting the first preset value based on the response intensity being outside the preset range; using the adjusted first preset value, combined with the normalized first steering wheel angle and yaw rate error, to re-update the vehicle dynamic prediction model; redetermining the response intensity based on the re-updated vehicle dynamic prediction model; and re-judging whether the re-determined response intensity is within the preset range, so as to redetermine the vehicle dynamic update model based on the judgment result.

[0056] It should be noted that the system does not depend on vehicle specifications, the mass and mass distribution of the trailer, or the coefficient of adhesion of the ground. It automatically updates the vehicle dynamic model based solely on the vehicle's attributes, avoiding the need for individual parameter adjustments for each vehicle. This allows it to adapt to various driving environments and is suitable for large-scale production deployment.

[0057] Specifically, based on the response intensity being outside a preset range, the first preset value is adjusted, including: decreasing the first preset value based on the response intensity being greater than the maximum boundary value of the preset range; and increasing the first preset value based on the response intensity being less than the minimum boundary value of the preset range. By adjusting the magnitude of the first preset value, the response intensity of the updated vehicle dynamic prediction model is made to be within the preset range.

[0058] In an optional embodiment, determining the vehicle dynamic update model based on the judgment result further includes: determining the currently updated vehicle dynamic prediction model as the vehicle dynamic update model based on the response intensity being within a preset range.

[0059] Step S12: Based on the predicted yaw rate and the pre-established cost function, the predicted steering wheel angle is obtained to control the vehicle according to the predicted steering wheel angle.

[0060] In this embodiment, the predicted steering wheel angle is obtained based on the predicted yaw rate and a pre-established cost function, including: obtaining multiple costs and the second steering wheel angle corresponding to each cost based on the predicted yaw rate and the pre-established cost function; and selecting the corresponding second steering wheel angle as the predicted steering wheel angle based on the minimum cost and preset constraints.

[0061] Specifically, based on the predicted yaw rate and combined with a pre-established cost function, multiple costs and the corresponding second steering wheel angle for each cost are obtained, including:

[0062] First, the yaw rate offset error is obtained based on the predicted yaw rate and the pre-obtained inferred yaw rate.

[0063] It should be noted that before obtaining the yaw rate offset error based on the predicted yaw rate and the pre-acquired inferred yaw rate, the process includes: acquiring trajectory points within the target time window; obtaining the curvature of each trajectory point; and obtaining the inferred yaw rate based on the curvature of each trajectory point and the vehicle speed corresponding to that point. It should be added that the trajectory points can be acquired during the acquisition of the first steering wheel angle within the target time window, as described above, and will not be repeated here.

[0064] Furthermore, based on the trajectory points, the curvature of each corresponding point is obtained, including: calculating the first and second derivatives of the trajectory points; and obtaining the curvature of each trajectory point based on the first and second derivatives.

[0065] For example, if the trajectory points within the target time window form a trajectory line L, denoted as y = f(x), and f(x) has a second derivative, then the slope of the tangent line at any point M on the curve is y' = tanα.

[0066]

[0067]

[0068]

[0069] again Then the curvature K of the trajectory line L at point M is:

[0070]

[0071] Assume the trajectory line L is given by the parametric equation Then, by differentiating the parametric equations, we can obtain the curvature κ: It should be noted that (x, y) in the parametric equation represents the coordinates of the trajectory point at time t in the vehicle coordinate system. The positive x-axis in the vehicle coordinate system is the direction of the vehicle's front, and the positive y-axis is the direction perpendicular to the front of the vehicle and located to the right of the driver. The specific positive directions of the x-axis and y-axis can be determined according to the actual design. For example, the opposite direction of the vehicle's front can be used as the positive x-axis, and the direction perpendicular to the x-axis and located to the left of the driver can be used as the positive y-axis. No further restrictions are made here.

[0072] Furthermore, if the sampling frequency is 50 Hz, meaning the time interval between any two adjacent points on the trajectory is 0.02 seconds, then by combining the velocity of each point on the trajectory, the entire trajectory can be converted into a parametric equation with t as the independent variable: Where t is the distance from each point on the trajectory to the current position, the first derivative of that point in the x and y directions can be calculated by taking the coordinates of any two adjacent points:

[0073]

[0074]

[0075] Similarly, by taking the first derivative of any two adjacent points, the second derivative can be calculated:

[0076]

[0077]

[0078] The curvature κ at that point can then be calculated.

[0079] Correspondingly, the estimated yaw rate is expressed as: This facilitates the calculation of the yaw rate deviation error, Ω, based on the predicted yaw rate and the pre-obtained inferred yaw rate. e Represented as in, Indicates the inferred yaw rate The matrix formed by these two variables, Ω, represents the matrix formed by the predicted yaw rate ω obtained from the vehicle dynamic prediction model.

[0080] Secondly, based on the first steering wheel angle corresponding to the predicted yaw rate, the heading angle, sideslip angle, and lateral position of the trajectory point corresponding to the first steering wheel angle within the target time window are obtained. It should be noted that since the target time window includes multiple trajectory points, and each trajectory point corresponds to a first steering wheel angle, heading angle, sideslip angle, and lateral position, when the predicted yaw rate is obtained based on the first steering wheel angle, the corresponding trajectory point can be determined based on the first steering wheel angle corresponding to the predicted yaw rate, thereby obtaining the heading angle, sideslip angle, and lateral position of that trajectory point.

[0081] Subsequently, based on the lateral offset error, heading angle, and sideslip angle, the heading angle offset error is obtained. In this embodiment, the heading angle offset error is expressed as:

[0082]

[0083] Where, Θ e C1 represents the heading angle offset error, and C1 represents the preset first-order heading angle error matrix. Indicates the inferred yaw rate The matrix formed by Ω represents the matrix composed of the predicted yaw rate ω obtained from the vehicle dynamic prediction model, and Λ represents the preset scaling matrix; θ e Let β represent the heading angle and β represent the sideslip angle. It should be noted that when the curvature is large, scaling can be achieved based on a scaling matrix; when the curvature is small, an identity matrix can be used. The specific scaling matrix used can be determined based on the actual situation, and no further restrictions are made here.

[0084] Subsequently, based on the yaw rate offset error, the heading angle offset error, and the lateral position, the lateral position offset error is obtained. In this embodiment, the lateral position offset error is expressed as:

[0085]

[0086] Among them, Y e Θ represents the lateral position offset error, V represents the vehicle's longitudinal velocity, C1 represents the preset first-order heading angle error matrix, C2 represents the preset second-order heading angle error matrix, and Θ represents the lateral position offset error. e Indicates the heading angle offset error, Ω e This represents the yaw rate deviation error, y e Indicates lateral error. Indicates the inferred yaw rate The matrix formed by Ω represents the matrix composed of the predicted yaw rate ω obtained from the vehicle dynamic prediction model, Λ represents the preset scaling matrix, and θ represents the matrix composed of the predicted yaw rate ω.e β represents the heading angle error, and β represents the sideslip angle.

[0087] Finally, based on the yaw rate offset error, heading angle offset error, and lateral position offset error, and combined with the pre-established cost function, multiple costs and the corresponding second steering wheel angles are obtained.

[0088] In this embodiment, the cost function J is expressed as:

[0089]

[0090] in, The second derivative of the steering wheel angle represents the steering wheel angular acceleration corresponding to the second steering wheel angle. The first derivative of the second steering wheel angle, i.e., the angular velocity of the second steering wheel, k ω k θ k y k s k a These represent the weights of yaw rate deviation error, heading angle deviation error, lateral position deviation error, steering wheel speed, and steering wheel angular acceleration in the cost function, respectively, and are all positive numbers.

[0091] It should be noted that the yaw rate deviation error, heading angle deviation error, and lateral position deviation error determined according to the above method are substituted into the cost function J, and the above k is adjusted. ω k θ k y k s and k a Parameters such as these are used to obtain the corresponding cost.

[0092] Furthermore, based on minimum cost and preset constraints, a corresponding second steering wheel angle is selected to obtain the predicted steering wheel angle. This includes: determining lateral position error constraints and steering wheel speed constraints based on the target driving area; selecting a corresponding second steering wheel angle based on the lateral position error constraints and steering wheel speed constraints; and selecting the second steering wheel angle with the minimum cost based on the cost of the selected second steering wheel angle, as the predicted steering wheel angle. Specifically, in actual practice, this is achieved by adjusting k... ω k θ k y k s and k a Parameters such as lateral position error and steering wheel speed are used, and the second steering wheel angle corresponding to the minimum cost J is selected as the predicted steering wheel angle under the constraints of lateral position error and steering wheel speed.

[0093] It should be noted that the lateral position error constraint is represented by Y. e,min ≤Ye ≤Y e,max This means that the lateral position offset error is located within the left and right boundaries of the preset lateral target driving area. The left and right boundaries of the target driving area can be determined based on actual road and traffic conditions or prior experience; no further limitations are imposed here. Additionally, the steering wheel speed constraint indicates... This means that the second steering wheel angular velocity is within the preset range. The preset range is designed to allow for safe vehicle steering. The preset range can be determined based on actual driving conditions, road conditions, or prior experience, and is not further limited here.

[0094] It should be noted that by setting a target driving area, the lateral drivable area is used as a constraint for the model predictive control algorithm. This ensures that the algorithm prioritizes avoiding touching safety boundaries when using the calculation results, thereby guaranteeing the safety of the control results. Furthermore, when the lateral position error constraint and the steering wheel speed constraint cannot be satisfied simultaneously, the steering wheel speed constraint must be satisfied first to ensure that the speed is within the functional safety limits, thus guaranteeing that the actuator can respond successfully.

[0095] It should be noted that the predicted steering wheel angle obtained by the above method controls vehicle driving. In terms of lateral control, the average error distance is controlled within 5.5cm, which can ensure good control accuracy and guarantee vehicle driving safety.

[0096] In summary, this embodiment of the invention uses a vehicle dynamic prediction model to convolve the first steering wheel angle within the input target time window with the unit impulse response to obtain the predicted yaw rate. Based on the predicted yaw rate and a pre-established cost function, the predicted steering wheel angle is obtained, thereby improving the accuracy of the predicted steering wheel angle. While ensuring accuracy, it reduces the need for actuator adjustments, thereby improving actuator durability and reducing the total cost of ownership (TCO). In addition, by using a large amount of operational data to iterate the optimal constraints in the algorithm, the autonomous driving behavior becomes more precise.

[0097] The lateral control device based on the vehicle dynamic prediction model provided by the present invention will be described below. The lateral control device based on the vehicle dynamic prediction model described below can be referred to in correspondence with the lateral control method based on the vehicle dynamic prediction model described above.

[0098] Figure 2 A schematic diagram of a lateral control device based on a vehicle dynamic prediction model is shown. The device includes:

[0099] The yaw rate prediction module 21 inputs the first steering wheel angle within the target time window into the vehicle dynamic prediction model. The vehicle dynamic prediction model convolves the first steering wheel angle within the input target time window with the unit impulse response to obtain the predicted yaw rate.

[0100] The angle prediction module 22 obtains the predicted steering wheel angle based on the predicted yaw rate and a pre-established cost function, so as to control the vehicle according to the predicted steering wheel angle.

[0101] In an optional embodiment, the device further includes: a data acquisition module, which acquires the first steering wheel angle within the target time window before inputting the pre-acquired first steering wheel angle within the target time window into the vehicle dynamic prediction model. The data acquisition module includes: a trajectory point acquisition unit, which acquires trajectory points within the target time window; the data acquisition unit acquires the first steering wheel angle corresponding to each trajectory point.

[0102] In this embodiment, the yaw rate prediction module 21 includes: a data input unit, which inputs the first steering wheel angle within the target time window obtained in advance into the vehicle dynamic prediction model; and a speed prediction unit, which obtains the predicted yaw rate by convolving the first steering wheel angle within the input target time window with the unit impulse response through the vehicle dynamic prediction model.

[0103] In an optional embodiment, the data acquisition module further includes: acquiring a first yaw rate within a target time window. Correspondingly, the device further includes: a normalization processing module, which normalizes the first steering wheel angle within the target time window after the data acquisition module acquires the first steering wheel angle within the target time window; and a model update module, which updates the vehicle dynamic prediction model based on the predicted yaw rate, the first yaw rate within the target time window, and the normalized first steering wheel angle after obtaining the predicted yaw rate.

[0104] Specifically, the normalization processing module includes: a torque acquisition unit, which acquires the torque corresponding to the first steering wheel angle within the target time window; a median acquisition unit, which obtains the median steering wheel torque based on the first steering wheel angle within the target time window, the torque corresponding to the first steering wheel angle, and a pre-established first impulse response model; and a normalization processing unit, which obtains the normalized first steering wheel angle based on the median steering wheel torque and the corresponding first steering wheel angle.

[0105] Furthermore, the normalization processing unit includes: an influence degree determination subunit, which determines the influence degree of the dead zone on the steering wheel angle based on the median steering wheel torque and the corresponding first steering wheel angle, combined with the vehicle weight and the preset dead zone sensitivity; and a normalization processing subunit, which determines the normalized first steering wheel angle based on the preset dead zone width, the first steering wheel angle, and the influence degree of the dead zone on the steering wheel angle.

[0106] In an optional embodiment, the normalization processing module further includes: an impulse response model update unit, used to update the first impulse response model. It should be noted that the impulse response model update unit can be referred to as the model update module 34 below, and will not be further described here.

[0107] In an optional embodiment, since some types of vehicles, such as trucks, exhibit zero bias and nonlinearity, the device further includes a preprocessing module for preprocessing the first steering wheel angle and the first yaw rate within the target time window. Further, the preprocessing module includes: an average value acquisition unit for obtaining the corresponding average steering wheel angle and average yaw rate based on the first steering wheel angle and the first yaw rate within the target time window; and a preprocessing unit for reducing each first steering wheel angle within the target time window based on the average steering wheel angle, and reducing each first yaw rate within the target time window based on the average yaw rate.

[0108] It should be noted that the preprocessing module can perform preprocessing of the first steering wheel angle and the first yaw rate within the target time window before or after the normalization module performs normalization processing on the first steering wheel angle within the target time window; no further limitation is made here.

[0109] In addition, the model update module includes: a yaw rate error acquisition unit, which obtains the yaw rate error based on the predicted yaw rate and the first yaw rate within the target time window; a model update unit, which updates the vehicle dynamic prediction model based on the normalized first steering wheel angle, the first preset value, and the yaw rate error; an intensity determination unit, which determines the response intensity of the updated vehicle dynamic prediction model, where the response intensity is the sum of all weight values ​​of the updated vehicle dynamic prediction model; and a model determination unit, which determines whether the response intensity is within a preset interval and determines the vehicle dynamic update model based on the determination result.

[0110] Furthermore, the model determination unit includes: a judgment subunit, which judges whether the response intensity is within a preset range; a first model determination subunit, which, based on the response intensity being within the preset range, determines the currently updated vehicle dynamic prediction model as the vehicle dynamic update model; an adjustment subunit, which adjusts a first preset value based on the response intensity being outside the preset range; a model update subunit, which, using the adjusted first preset value and combining it with the normalized first steering wheel angle and yaw rate error, re-updates the vehicle dynamic prediction model; an intensity update subunit, which redetermines the response intensity based on the re-updated vehicle dynamic prediction model; and a second model determination subunit, which re-judges whether the re-determined response intensity is within the preset range, so as to redetermine the vehicle dynamic update model based on the judgment result.

[0111] The adjustment sub-unit includes: a first adjustment sub-unit, which decreases a first preset value based on the maximum boundary value of the response intensity being greater than the preset interval; and a second adjustment sub-unit, which increases the first preset value based on the minimum boundary value of the response intensity being less than the preset interval. By adjusting the magnitude of the first preset value, the response intensity of the updated vehicle dynamic prediction model is made to fall within the preset interval.

[0112] The angle prediction module 22 includes: an angle acquisition unit, which obtains multiple costs and the second steering wheel angle corresponding to each cost based on the predicted yaw rate and in combination with a pre-established cost function; and an angle prediction unit, which selects the corresponding second steering wheel angle as the predicted steering wheel angle based on the minimum cost and preset constraints.

[0113] Specifically, the angle acquisition unit includes: a first error acquisition subunit, which obtains the yaw rate offset error based on the predicted yaw rate and the pre-acquired inferred yaw rate; a data acquisition subunit, which obtains the heading angle, sideslip angle, and lateral position of the same trajectory point corresponding to the first steering wheel angle within the target time window based on the first steering wheel angle corresponding to the predicted yaw rate; a second error acquisition subunit, which obtains the heading angle offset error based on the lateral offset error, heading angle, and sideslip angle; a third error acquisition subunit, which obtains the lateral position offset error based on the yaw rate offset error, heading angle offset error, and lateral position; and an angle acquisition subunit, which obtains multiple costs and the second steering wheel angle corresponding to each cost based on the yaw rate offset error, heading angle offset error, lateral position offset error, and a pre-established cost function.

[0114] In an optional embodiment, the angle prediction module 22 further includes: a yaw rate estimation unit, which acquires the estimated yaw rate before obtaining the yaw rate offset error based on the predicted yaw rate and the pre-acquired estimated yaw rate. Specifically, the yaw rate estimation unit includes: a trajectory point acquisition subunit, which acquires trajectory points within the target time window; a curvature acquisition subunit, which obtains the curvature corresponding to each trajectory point based on the trajectory points; and a yaw rate estimation subunit, which obtains the estimated yaw rate based on the curvature of each trajectory point and the vehicle speed corresponding to the trajectory point.

[0115] Furthermore, the curvature acquisition sub-unit includes: a derivative-finding sub-unit, which calculates the corresponding first and second derivatives based on the trajectory points; and a curvature acquisition sub-unit, which obtains the curvature of each trajectory point based on the first and second derivatives.

[0116] In addition, the steering angle prediction unit includes: a constraint acquisition subunit, which determines the lateral position error constraint and the steering wheel speed constraint based on the target driving area; a steering angle selection subunit, which selects the corresponding second steering wheel angle according to the lateral position error constraint and the steering wheel speed constraint; and a steering angle prediction subunit, which selects the second steering wheel angle with the lowest cost according to the cost of the selected second steering wheel angle, as the predicted steering wheel angle.

[0117] In summary, this embodiment of the invention uses a yaw rate prediction module to convolve the first steering wheel angle within the input target time window with the unit impulse response based on the vehicle dynamic prediction model to obtain the predicted yaw rate. Then, an angle prediction module uses the predicted yaw rate and a pre-established cost function to obtain the predicted steering wheel angle, thereby improving the accuracy of the predicted steering wheel angle. While ensuring accuracy, this reduces the need for actuator adjustments, thus improving actuator durability and lowering the total cost of ownership (TCO). Furthermore, by using a large amount of operational data to iterate the optimal constraints in the algorithm, the autonomous driving behavior becomes more precise.

[0118] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3As shown, the electronic device may include a processor 31, a communication interface 32, a memory 33, and a communication bus 34. The processor 31, communication interface 32, and memory 33 communicate with each other via the communication bus 34. The processor 31 can call logical instructions in the memory 33 to execute a lateral control method based on a vehicle dynamic prediction model. This method includes: inputting a first steering wheel angle within a pre-acquired target time window into the vehicle dynamic prediction model; convolving the first steering wheel angle within the input target time window with the unit impulse response of the vehicle dynamic prediction model to obtain a predicted yaw rate; and obtaining a predicted steering wheel angle based on the predicted yaw rate and a pre-established cost function, thereby controlling the vehicle according to the predicted steering wheel angle.

[0119] Furthermore, the logical instructions in the aforementioned memory 33 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0120] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the lateral control method based on the vehicle dynamic prediction model provided by the above methods. The method includes: inputting a first steering wheel angle within a pre-acquired target time window into the vehicle dynamic prediction model; convolving the first steering wheel angle within the input target time window with the unit impulse response of the vehicle dynamic prediction model to obtain a predicted yaw rate; and obtaining a predicted steering wheel angle based on the predicted yaw rate and a pre-established cost function, so as to control the vehicle according to the predicted steering wheel angle.

[0121] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the lateral control method based on a vehicle dynamic prediction model provided by the above methods. The method includes: inputting a first steering wheel angle within a pre-acquired target time window into a vehicle dynamic prediction model; convolving the first steering wheel angle within the input target time window with the unit impulse response of the vehicle dynamic prediction model to obtain a predicted yaw rate; and obtaining a predicted steering wheel angle based on the predicted yaw rate and a pre-established cost function, so as to control the vehicle according to the predicted steering wheel angle.

[0122] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0123] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0124] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A lateral control method based on a vehicle dynamic prediction model, characterized in that, include: The first steering wheel angle within the pre-acquired target time window is input into the vehicle dynamic prediction model. The vehicle dynamic prediction model convolves the first steering wheel angle within the input target time window with the unit impulse response of the vehicle dynamic prediction model to obtain the predicted yaw rate. Based on the predicted yaw rate and the pre-established cost function, the predicted steering wheel angle is obtained, and the vehicle is controlled according to the predicted steering wheel angle. The process of obtaining the predicted steering wheel angle based on the predicted yaw rate and the pre-established cost function includes: Based on the predicted yaw rate and combined with the pre-established cost function, multiple costs and the corresponding second steering wheel angle for each cost are obtained; Based on minimum cost and preset constraints, the corresponding second steering wheel angle is selected as the predicted steering wheel angle.

2. The lateral control method based on a vehicle dynamic prediction model according to claim 1, characterized in that, The step of obtaining multiple costs and the corresponding second steering wheel angle based on the predicted yaw rate and a pre-established cost function includes: Based on the predicted yaw rate and the pre-obtained inferred yaw rate, the yaw rate offset error is obtained; Based on the first steering wheel angle corresponding to the predicted yaw rate, obtain the heading angle, sideslip angle and lateral position of the same trajectory point corresponding to the first steering wheel angle within the target time window; The heading angle offset error is obtained based on the yaw rate offset error, the heading angle, and the sideslip angle; The lateral position offset error is obtained based on the yaw rate offset error, the heading angle offset error, and the lateral position. Based on the yaw rate offset error, the heading angle offset error, and the lateral position offset error, and in combination with... A pre-established cost function yields multiple costs and the corresponding second steering wheel angle for each cost.

3. The lateral control method based on a vehicle dynamic prediction model according to claim 2, characterized in that, Before obtaining the yaw rate offset error based on the predicted yaw rate and the pre-acquired inferred yaw rate, the process includes: Obtain the trajectory points within the target time window; Based on the trajectory points, the curvature corresponding to each trajectory point is obtained; Based on the curvature of each trajectory point and the vehicle speed corresponding to each trajectory point, the estimated yaw rate is obtained.

4. The lateral control method based on a vehicle dynamic prediction model according to claim 1, characterized in that, The step of selecting the corresponding second steering wheel angle based on minimum cost and preset constraints to obtain the predicted steering wheel angle includes: Based on the target driving area, determine the lateral position error constraint and the steering wheel speed constraint; Based on the lateral position error constraint and the steering wheel speed constraint, select the corresponding second steering wheel angle; Based on the cost of the selected second steering wheel angle, the second steering wheel angle with the lowest corresponding cost is selected as the predicted steering wheel angle.

5. The lateral control method based on a vehicle dynamic prediction model according to claim 1, characterized in that, Before inputting the first steering wheel angle within the pre-acquired target time window into the vehicle dynamic prediction model, the following steps are included: Obtain the first steering wheel angle and the first yaw rate within the target time window; The first steering wheel angle within the target time window is normalized. After obtaining the predicted yaw rate, the process includes: The vehicle dynamic prediction model is updated based on the predicted yaw rate, the first yaw rate within the target time window, and the normalized first steering wheel angle.

6. The lateral control method based on a vehicle dynamic prediction model according to claim 4, characterized in that, After acquiring the first steering wheel angle and the first yaw rate within the target time window, the method further includes: Based on the first steering wheel angle and the first yaw rate within the target time window, the corresponding average steering wheel angle and average yaw rate are obtained; Based on the average steering wheel angle, reduce each first steering wheel angle within the target time window; Based on the average yaw rate, the first yaw rate within each target time window is reduced respectively.

7. A lateral control device based on a vehicle dynamic prediction model, characterized in that, include: The yaw rate prediction module inputs the first steering wheel angle within the pre-acquired target time window into the vehicle dynamic prediction model. The vehicle dynamic prediction model convolves the first steering wheel angle within the input target time window with the unit impulse response of the vehicle dynamic prediction model to obtain the predicted yaw rate. An angle prediction module obtains a predicted steering wheel angle based on the predicted yaw rate and a pre-established cost function, and controls the vehicle according to the predicted steering wheel angle. The angle prediction module includes: The steering angle acquisition unit obtains multiple costs and the second steering wheel angle corresponding to each cost based on the predicted yaw rate and in combination with a pre-established cost function. The steering angle prediction unit selects the corresponding second steering wheel angle as the predicted steering wheel angle based on minimum cost and preset constraints.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the lateral control method based on the vehicle dynamic prediction model as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the lateral control method based on the vehicle dynamic prediction model as described in any one of claims 1 to 6.

Images