Longitudinal control method, system and vehicle for autonomous vehicle

Abstract

The application belongs to the technical field of automatic driving control, and particularly relates to a longitudinal control method and system of an automatic driving vehicle and the vehicle, the method comprising the following steps: establishing MPC feedback control, calculating feedback acceleration by using the MPC feedback control; 2) establishing feedforward control, calculating feedforward acceleration by using the feedforward control; establishing distance and / or speed integral compensation, obtaining distance speed compensation acceleration; setting a vehicle target acceleration according to the feedback acceleration, the feedforward acceleration and the distance speed compensation acceleration, and controlling the vehicle according to the vehicle target acceleration. Thus, the application solves the problems of poor control effect and low precision of the longitudinal control of the automatic driving vehicle in the prior art when the dynamic model is nonlinear.

Description

Technical Field

[0001] This invention belongs to the field of autonomous driving control technology, specifically relating to a longitudinal control method, system, and vehicle for autonomous vehicles. Background Technology

[0002] Currently, the mainstream longitudinal control methods for autonomous vehicles include position-velocity dual-loop PID (Proportion-Integral-Differential) and MPC (Model Predictive Control). Position-velocity dual-loop PID control, however, does not consider the vehicle's dynamics model. In practice, calibrating control parameters based on experience is cumbersome, and the calibrated parameters are difficult to adapt to all driving conditions. Furthermore, when vehicle parameters (such as mass) change, the settling time for the same KP parameter will change, resulting in an inability to adapt quickly. Position-velocity dual-loop PID longitudinal control is often used for autonomous driving scenarios that do not require fast and accurate responses or operate under relatively simple conditions.

[0003] However, since the calibration parameters of the position-velocity dual closed-loop PID are cumbersome and difficult to adapt to all operating conditions, the dynamic error model used by MPC control is linearized and time-invariant.

[0004] Another mainstream vertical control method is MPC control, such as... Figure 1 As shown, the target acceleration equals the reference acceleration of the target trajectory plus the compensation acceleration calculated based on the slope and the feedback acceleration calculated by MPC. The feedback acceleration calculated by MPC is mainly based on the vehicle's dynamic error model and the current state variables, which can predict the system output over a period of time (i.e., prediction time and control time). By solving optimization problems that satisfy the objective function and various constraints, the control sequence in the control time domain is obtained. Then, the first value in the control time domain is used as the current output value and applied to the controlled object. The above process is repeated in the next cycle, continuously completing each constrained optimization problem, thereby achieving continuous longitudinal control of the vehicle. However, MPC control uses a linear time-invariant (LTI) dynamic model to predict future behavior. If the object is strongly nonlinear, or its characteristics change over time, the LTI prediction accuracy may be severely reduced, making MPC performance unacceptable. If the dynamic model of the autonomous vehicle is nonlinear, traditional MPC control methods will linearize it during actual control, leading to model errors. Furthermore, when the parameters of the autonomous vehicle (such as mass) change over time, it will cause inaccurate prediction accuracy, thus affecting the control effect, and this steady-state error cannot accumulate over time.

[0005] In summary, due to the linearization of the dynamic error model and the inaccuracy of vehicle parameters, the existing longitudinal control of autonomous vehicles suffers from poor control performance and low prediction accuracy when the dynamic model is highly nonlinear or its characteristics change over time. Summary of the Invention

[0006] The purpose of this invention is to provide a longitudinal control method, system, and vehicle for autonomous vehicles, in order to solve the problems of poor control effect and low accuracy in the longitudinal control of autonomous vehicles when the dynamic model is nonlinear in the prior art.

[0007] To solve the above-mentioned technical problems, the technical solution provided by this invention and the corresponding beneficial effects of the technical solution are as follows:

[0008] The present invention provides a longitudinal control method for an autonomous vehicle, comprising the following steps:

[0009] 1) Establish MPC feedback control and use MPC feedback control to calculate the feedback acceleration;

[0010] 2) Establish feedforward control and use it to calculate the feedforward acceleration;

[0011] 3) Establish integral compensation for distance and / or velocity to obtain distance-velocity compensated acceleration;

[0012] 4) Set the vehicle target acceleration based on feedback acceleration, feedforward acceleration and distance-velocity compensation acceleration, and control the vehicle according to the vehicle target acceleration.

[0013] The beneficial effects of the above technical solution are as follows: This invention introduces distance integral compensation and velocity integral compensation to obtain distance-velocity compensated acceleration; the feedforward acceleration, feedback acceleration, and distance-velocity compensated acceleration are set as the vehicle target acceleration, thereby controlling the vehicle according to the vehicle target acceleration. Because this invention uses distance integral compensation and velocity integral compensation to obtain distance-velocity compensated acceleration, it can solve the problems of inaccurate dynamic error models and low control accuracy caused by linearization of the dynamic error model and inaccurate vehicle parameters, thus achieving an adaptive effect and enhancing the robustness of the control system.

[0014] Furthermore, step 3) includes:

[0015] The speed deviation is integrated to obtain the speed integral compensation acceleration. The speed deviation is the difference between the reference speed of the target trajectory and the actual speed of the vehicle.

[0016] The distance deviation is integrated to obtain the distance integral compensation acceleration. The distance deviation is the difference between the reference position of the target trajectory and the actual position of the vehicle.

[0017] The distance-velocity compensation acceleration is determined based on the velocity integral compensation acceleration and the distance integral compensation acceleration.

[0018] Furthermore, to improve speed following accuracy during road travel and the precision of reaching the target point, step 3) determines the distance-velocity compensation acceleration in the following manner:

[0019] If the distance deviation is greater than the set threshold, the value of the distance velocity compensation acceleration is equal to the value of the velocity integral compensation acceleration;

[0020] If the distance deviation is less than or equal to the set threshold, the value of the distance velocity compensation acceleration is equal to the value of the distance integral compensation acceleration.

[0021] Furthermore, to improve accuracy, integration limiting is included when integrating the velocity deviation and when integrating the distance deviation, so as to limit the maximum and minimum values ​​of the integral compensation acceleration and the distance integral compensation acceleration respectively; the maximum and minimum values ​​are determined by calibration.

[0022] Furthermore, the speed integration parameters used when integrating the speed deviation are... The distance integration parameters are determined based on the speed deviation and are used when integrating the distance deviation. It is determined based on the distance deviation.

[0023] Furthermore, in step 2), the feedforward acceleration is obtained based on the trajectory planning target acceleration and the slope compensation acceleration.

[0024] Furthermore, in step 1), the MPC feedback control includes a vehicle dynamics error model, the formula of which is as follows:

[0025]

[0026]

[0027]

[0028]

[0029] A=

[0030] B=

[0031] C=

[0032] in, To control the cycle, Let k be the state variables at time k, which include lateral deviation, rate of change of lateral deviation, orientation angle deviation, rate of change of orientation angle deviation, distance error, and velocity error. = Let 'a' be the predicted steering wheel angle, 'a' be the predicted acceleration, 'I' denote the identity matrix, and 'C' be the predicted steering wheel angle. f C r These represent the lateral stiffness of the front and rear axles on a single side, respectively, where m is the vehicle mass and l is the weight of the vehicle. f l is the distance from the front axle to the center of gravity of the vehicle. r V is the distance from the rear axle to the center of gravity of the vehicle. x The speed of the vehicle's center of gravity. This represents the vehicle's moment of inertia. This refers to the vehicle's angular velocity.

[0033] Furthermore, the MPC feedback control model in step 1) includes an optimization objective function. And constraint functions, optimization objective function The formulas for the constraint functions are as follows:

[0034]

[0035]

[0036] Where N represents the prediction and control time domains, Q represents the state weighting matrix, and R represents the control weighting matrix. These represent the minimum and maximum constraints of the state matrix, respectively. These represent the minimum and maximum constraints of control, respectively. This represents the initial state. This represents the state matrix at time k. Represents the reference state matrix. The control variables in the state-space equations consist of the steering wheel angle and acceleration. This represents the transpose of a matrix.

[0037] To address the aforementioned problems, the present invention provides a longitudinal control system for an autonomous vehicle, comprising a processor for executing computer instructions to implement a longitudinal control method for an autonomous vehicle according to the present invention, and achieving the same beneficial effects as the method.

[0038] To address the aforementioned problems, the present invention provides an autonomous driving vehicle, including a vehicle body and an autonomous driving vehicle longitudinal control system of the present invention, achieving the same beneficial effects as the system. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the longitudinal control principle for autonomous driving in existing technologies;

[0040] Figure 2 This is a schematic diagram of a longitudinal control system for an autonomous vehicle according to the present invention. Detailed Implementation

[0041] This invention adds distance and speed integral (KI) control to the MPC feedback control. A switch is used to select between distance integral and speed integral control. When the distance error is less than a set threshold, distance integral control is selected for precise approach control; when the distance error is greater than the set threshold, speed integral control is selected for adaptive compensation of model errors, thereby improving control accuracy and enhancing control robustness. This invention is applicable to any vehicle that supports longitudinal drive-by-wire control.

[0042] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0043] Method Implementation Examples:

[0044] An embodiment of the longitudinal control method for an autonomous vehicle according to the present invention, such as... Figure 2 As shown, the longitudinal control method of this invention includes three parts: feedforward control calculation, MPC feedback control calculation, and distance-velocity KI compensation calculation. The target acceleration of the vehicle is equal to the slope compensation acceleration calculated by the feedforward and the reference acceleration of the target trajectory, plus the feedback acceleration calculated by the MPC and the distance-velocity KI (proportional-integral, K is the coefficient, I is the integral) compensation acceleration. The distance-velocity KI compensation acceleration is switched by a switch.

[0045] The calculations for the three parts of the longitudinal control method are explained below:

[0046] Step 1: Calculate the longitudinal feedforward to obtain the feedforward acceleration.

[0047] The acceleration of feedforward compensation equals the reference acceleration of the target trajectory plus the acceleration of slope compensation (the acceleration of feedforward compensation is also called feedforward acceleration, and the reference acceleration of the target trajectory is also called the target acceleration of trajectory planning), where the acceleration of slope compensation... The calculation formula is as follows:

[0048] (1)

[0049] Where g is the acceleration due to gravity, 9.8 m / s²; Slope, in radians.

[0050] Step 2: MPC feedback control calculation to obtain feedback acceleration.

[0051] The formula for the vehicle's dynamic error model is as follows:

[0052] (2)

[0053] in, This represents the lateral deviation, with the unit being meters (m). Indicates the rate of change of lateral deviation; This indicates the deviation in orientation angle, expressed in radians. This indicates the rate of change of the orientation angle deviation; This represents the distance error, with the unit being meters (m). This indicates the speed error, expressed in m / s. The moment of inertia of a vehicle is expressed in kg·m². f l is the distance from the front axle to the center of gravity of the vehicle. r C is the distance from the rear axle to the center of gravity of the vehicle. f C r These are the lateral stiffness of the front and rear axles of the vehicle, respectively, for each wheel on one side. x l is the speed of the vehicle's center of gravity. z For the vehicle's moment of inertia, Let 'a' be the predicted steering wheel angle, 'a' be the predicted acceleration, and 'm' be the vehicle mass. This refers to the vehicle's angular velocity.

[0054] The above equation (2) can be written as a state equation as follows:

[0055] (3)

[0056] Where, in the formula A= X= B= u= C= A and B are coefficient matrices, X is the vehicle state variable, and u is the predicted steering wheel angle / acceleration.

[0057] Bilinear discretization is as follows:

[0058] (4)

[0059] in, The control period is represented in seconds (s), and I represents the identity matrix.

[0060] Therefore, the dynamic error model equation can be written as:

[0061] (5)

[0062] in, Let k be the state variables at time k, which include lateral deviation, rate of change of lateral deviation, orientation angle deviation, rate of change of orientation angle deviation, distance error, and velocity error. = Let 'a' be the predicted steering wheel angle and 'a' be the predicted acceleration.

[0063] Set the optimization objective function The constraint functions are as follows:

[0064] (6)

[0065] (7)

[0066] Where N represents the prediction and control time domain, in seconds; Q represents the state weighting matrix; and R represents the control weighting matrix. Indicates the reference state, which defaults to 0; These represent the minimum and maximum constraints of the state matrix, respectively. These represent the minimum and maximum constraints of control, respectively. Indicates the initial state; x k Denotes the state matrix at time k; x r Represents the reference state matrix; , denoted by , where 'a' represents the predicted steering wheel angle, 'a' represents the predicted acceleration, and 'T' represents the transpose of the matrix.

[0067] MPC feedback control includes the following steps:

[0068] At time k, combining historical information and the current state and prediction models Predict the system output for N steps;

[0069] Based on the constraints of formula (7), design the objective function formula (6) and calculate the optimal control solution. The input is given to the controlled vehicle, causing it to move under the current control parameters;

[0070] Then, repeat the above steps at time k+1. In this way, the constrained optimization problem is solved in a rolling manner, thereby achieving continuous control of the controlled object.

[0071] Step 3: Distance-velocity KI compensation control, perform integral control calculations to obtain distance-velocity compensation acceleration.

[0072] When distance error Greater than a certain threshold (e.g., 1.5m, threshold) When the distance is obtained through calibration, speed integral control is activated. The distance error is the distance deviation obtained by subtracting the actual position from the reference position of the target trajectory; this distance error is also called distance deviation. The speed deviation obtained by subtracting the actual vehicle speed from the reference speed of the target trajectory is used to obtain the speed integral compensation acceleration through speed integral control. The speed integral parameters... The parameters are based on a one-dimensional table of speed errors (the speed integral and distance integral tables are empirical tables, obtained by collecting data from human drivers and analyzing data under similar conditions to derive speed and position compensation tables). When the error distance... Less than or equal to the threshold (e.g., 1.5m, which can be calibrated) When the distance is calibrated, the distance integral control is activated. The distance deviation obtained by subtracting the actual position from the reference position of the target trajectory is used for integral control to obtain the distance integral compensated acceleration, where the distance integral parameter... The parameters are a one-dimensional table based on the distance error. In this invention, the velocity integral compensation acceleration and the distance integral compensation acceleration are also subject to integral limiting, which restricts the maximum and minimum values ​​of the velocity integral compensation acceleration and the distance integral compensation acceleration, such as limiting them to [-0.5, 1.0] m / s2. The maximum and minimum values ​​can be calibrated, and finally the final distance velocity compensation acceleration is obtained.

[0073] Step 4: Use the sum of feedforward acceleration, feedback acceleration, and distance-velocity compensation acceleration as the target acceleration for the vehicle, and control the vehicle according to the target acceleration.

[0074] This invention introduces an integral model to eliminate the poor control accuracy problem caused by the inaccuracy of the MPC dynamic error model. It solves the problem of poor control accuracy caused by the linearization of the dynamic error model and the inaccuracy of the vehicle parameters, and can achieve an adaptive effect, thereby enhancing the robustness of the control system.

[0075] This invention switches between distance integral compensation and speed integral compensation based on the magnitude of the distance error and the threshold value, which can solve the problems of speed following accuracy and precise arrival distance control when driving on the road.

[0076] System Implementation Example:

[0077] An embodiment of the longitudinal control system for an autonomous vehicle according to the present invention includes a memory, a processor, and an internal bus. The processor and the memory communicate and interact with each other via the internal bus. The memory includes at least one software function module stored in the memory. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby implementing the longitudinal control method for an autonomous vehicle described in the method embodiment of the present invention.

[0078] The processor can be a microprocessor (MCU), a programmable logic device (FPGA), or other processing devices. The memory can be any type of memory that stores information using electrical energy, such as RAM or ROM. In other embodiments, the longitudinal control system for an autonomous vehicle according to the present invention can also be as follows: Figure 2 As shown, the system includes feedforward control, MPC feedback control, and distance-velocity KI compensation. Feedforward control uses the sum of the target acceleration from trajectory planning and the gradient compensation acceleration as the feedforward acceleration; MPC feedback control obtains the feedback acceleration; and distance-velocity KI compensation obtains the distance-velocity compensation acceleration. Finally, the sum of the feedforward acceleration, feedback acceleration, and distance-velocity compensation acceleration is used as the target acceleration. This system offers high control accuracy, achieves adaptive performance, and enhances the robustness of the control system.

[0079] Vehicle Example:

[0080] An embodiment of the autonomous vehicle of the present invention includes a vehicle body, a slope detection sensor, and an autonomous vehicle longitudinal control system described in the system embodiment, which can accurately realize the longitudinal control of the vehicle, improve the vehicle's adaptive effect, and accurately control the vehicle's stopping distance.

Claims

1. A longitudinal control method for an autonomous vehicle, characterized in that: Includes the following steps: 1) Establish MPC feedback control and use MPC feedback control to calculate the feedback acceleration; 2) Establish feedforward control and use it to calculate the feedforward acceleration; 3) Integrate the speed deviation to obtain the speed integral compensation acceleration. The speed deviation is the difference between the reference speed of the target trajectory and the actual speed of the vehicle. The distance deviation is integrated to obtain the distance integral compensation acceleration. The distance deviation is the difference between the reference position of the target trajectory and the actual position of the vehicle. If the distance deviation is greater than the set threshold, the value of the distance velocity compensation acceleration is equal to the value of the velocity integral compensation acceleration; If the distance deviation is less than or equal to the set threshold, the value of the distance velocity compensation acceleration is equal to the value of the distance integral compensation acceleration. 4) Set the vehicle target acceleration based on feedback acceleration, feedforward acceleration and distance-velocity compensation acceleration, and control the vehicle according to the vehicle target acceleration.

2. The longitudinal control method for autonomous vehicles according to claim 1, characterized in that: Integral limiting is included when integrating velocity deviation and when integrating distance deviation, to limit the maximum and minimum values ​​of integral compensation acceleration and distance integral compensation acceleration respectively. The maximum and minimum values ​​are determined by calibration.

3. The longitudinal control method for autonomous vehicles according to claim 1, characterized in that: The speed integration parameter used when integrating the speed deviation. The distance integration parameters are determined based on the speed deviation and are used when integrating the distance deviation. It is determined based on the distance deviation.

4. The longitudinal control method for autonomous vehicles according to claim 1, characterized in that: In step 2), the feedforward acceleration is obtained based on the trajectory planning target acceleration and the slope compensation acceleration.

5. The longitudinal control method for autonomous vehicles according to claim 1, characterized in that: Step 1) includes a vehicle dynamics error model in the MPC feedback control. The formula for the dynamics error model is as follows: A= B= C= in, To control the cycle, Let k be the state variables at time k, which include lateral deviation, rate of change of lateral deviation, orientation angle deviation, rate of change of orientation angle deviation, distance error, and velocity error. = Let 'a' be the predicted steering wheel angle, 'a' be the predicted acceleration, 'I' denote the identity matrix, and 'C' be the predicted steering wheel angle. f C r These represent the lateral stiffness of the front and rear axles on a single side, respectively, where m is the vehicle mass and l is the weight of the vehicle. f l is the distance from the front axle to the center of gravity of the vehicle. r V is the distance from the rear axle to the center of gravity of the vehicle. x The speed of the vehicle's center of gravity. This represents the vehicle's moment of inertia. This refers to the vehicle's angular velocity.

6. The longitudinal control method for an automated vehicle according to any one of claims 1 to 5, characterized in that: Step 1) The MPC feedback control model includes an optimization objective function. And constraint functions, optimization objective function The formulas for the constraint functions are as follows: Where N represents the prediction and control time domains, Q represents the state weighting matrix, and R represents the control weighting matrix. These represent the minimum and maximum constraints of the state matrix, respectively. These represent the minimum and maximum constraints of control, respectively. This represents the initial state. This represents the state matrix at time k. Represents the reference state matrix. The control variables in the state-space equations consist of the steering wheel angle and acceleration. This represents the transpose of a matrix.

7. A longitudinal control system for an autonomous vehicle, characterized in that: The system includes a processor for executing computer instructions to implement the longitudinal control method for an autonomous vehicle as described in any one of claims 1 to 5.

8. An autonomous driving vehicle, comprising a vehicle body, characterized in that: It also includes the longitudinal control system for autonomous vehicles as described in claim 7.

Images