Intelligent vehicle trajectory tracking hybrid control method

Abstract

The invention discloses an intelligent vehicle trajectory tracking hybrid control method, comprising an NMPC controller and an LEMPC controller, calculating a transverse error and a transverse error change rate, and selecting a corresponding control mode according to the transverse error and the transverse error change rate; wherein in the switching control process of the NMPC controller and the LEMPC controller, a smooth switching function is introduced, weighted calculation is carried out on expected front wheel rotation angles output by the NMPC controller and the LEMPC controller, and the output amplitude of the expected front wheel rotation angles is restrained until the switching is finished, so that smooth switching is realized. The transverse control requirement of the intelligent vehicle under the complex working condition can be met, and the trajectory tracking control performance is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent driving technology, and specifically to a hybrid control method for intelligent vehicle trajectory tracking. Background Art

[0002] With the rapid development of intelligent driving technology, path tracking control, as one of the core tasks of intelligent vehicles, has received extensive attention. Its main goal is to ensure that the vehicle can accurately follow the predetermined trajectory in complex environments while ensuring driving stability and safety. However, due to complex factors such as road curvature changes, speed fluctuations, and external interference in actual working conditions, path tracking control faces great challenges.

[0003] Model Predictive Control (MPC), as an optimization control strategy, has been widely used in the field of trajectory tracking control due to its good ability to describe the dynamic characteristics of the system and effectively handle constraints. In particular, nonlinear model predictive control (NMPC) can provide high-precision control performance based on the nonlinear dynamic model of the vehicle. However, NMPC has a high computational complexity, which may lead to insufficient real-time performance and difficulty in adapting to rapidly changing working conditions. In contrast, linearized model-based predictive control (LEMPC) has significant advantages in computational efficiency, but its control accuracy may be limited under nonlinear working conditions. Summary of the invention

[0004] The purpose of the present invention is to provide a hybrid control method for intelligent vehicle trajectory tracking. The present invention can meet the lateral control requirements of intelligent vehicles under complex working conditions and improve trajectory tracking control performance.

[0005] The technical solution of the present invention is: a hybrid control method for intelligent vehicle trajectory tracking, including an NMPC controller and a LEMPC controller, calculating a lateral error and a lateral error change rate, and selecting a corresponding control mode according to the size of the lateral error and the lateral error change rate; when the lateral error is greater than or equal to a lateral error threshold and the lateral error change rate is less than or equal to a lateral error change rate threshold, the NMPC controller is selected, and the NMPC controller uses a vehicle dynamics model as a prediction model to output a first desired front wheel steering angle; when the lateral error is less than the lateral error threshold or the lateral error change rate is greater than the lateral error change rate threshold, the LEMPC controller is selected, and the LEMPC controller uses a vehicle dynamics error model as a prediction model to output a second desired front wheel steering angle;

[0006] In the process of switching control between the NMPC controller and the LEMPC controller, a smooth switching function is introduced to perform weighted calculation on the desired front wheel steering angles output by the NMPC controller and the LEMPC controller, and constrain the output amplitude of the desired front wheel steering angle until the switching is completed, thereby achieving smooth switching.

[0007] In the aforementioned intelligent vehicle trajectory tracking hybrid control method, the lateral error e d The specific calculation formula is:

[0008] e d =Y r -Y;

[0009] Among them, Y r is the reference trajectory, and Y is the actual trajectory;

[0010] The lateral error change rate is the lateral error e d Obtained by derivation.

[0011] In the aforementioned intelligent vehicle trajectory tracking hybrid control method, the lateral error threshold ρ pos The calculation formula is:

[0012]

[0013] Among them, e ymax is the maximum allowable lateral error threshold; K is the road curvature, and β is a constant used to control the influence of the square of the road curvature on the lateral error threshold.

[0014] In the aforementioned intelligent vehicle trajectory tracking hybrid control method, the lateral error change rate threshold τ max The calculation formula is:

[0015]

[0016] Where C is a constant, K is the road curvature, a is a parameter that controls the influence of road curvature on the threshold of the lateral error change rate, and e is the expected front wheel steering angle difference;

[0017] The calculation formula for the expected front wheel steering angle difference e is:

[0018] e = δ1-δ2;

[0019] Wherein, δ1 is the first desired front wheel steering angle obtained by the NMPC controller; δ2 is the second desired front wheel steering angle obtained by the LEMPC controller.

[0020] In the aforementioned intelligent vehicle trajectory tracking hybrid control method, the expression of the smooth switching function is as follows:

[0021]

[0022] Among them, α(e) is the smooth switching function, is the steepness of the control weight transition, mode indicates the switching control signal between the NMPC controller and the LEMPC controller, when the NMPC controller is selected, mode is 1, when the LEMPC controller is selected, mode is 2; e is the expected front wheel steering angle difference.

[0023] In the aforementioned intelligent vehicle trajectory tracking hybrid control method, the formula for weighted calculation of the first desired front wheel turning angle and the second desired front wheel turning angle is:

[0024] δ f1 =α(e)δ1+(1-α(e))δ2;

[0025] In the formula, δ f1 is the output front wheel steering angle, α(e) is the smooth switching function, δ1 is the first desired front wheel steering angle obtained by the NMPC controller; δ2 is the second desired front wheel steering angle obtained by the LEMPC controller.

[0026] In the aforementioned intelligent vehicle trajectory tracking hybrid control method, the smooth switch has a fuzzy controller, and the smooth switching function is controlled by the fuzzy controller. Dynamic adjustment is performed; the desired front wheel steering angle difference e is derived to obtain the change rate de, and the fuzzy controller takes the absolute value of the desired front wheel steering angle difference |e| of the NMPC controller and the LEMPC controller before and after the mode switching and the change rate de of the desired front wheel steering angle difference as input, as output.

[0027] In the aforementioned intelligent vehicle trajectory tracking hybrid control method, a sliding mode controller is introduced in the smooth switching process; the sliding mode controller drives the vehicle control system to slide along the sliding mode surface by defining a sliding mode surface control output.

[0028] In the aforementioned intelligent vehicle trajectory tracking hybrid control method, the expression for defining the sliding surface is:

[0029] s=||e||+ωe d ;

[0030] Among them, ω is a positive weight parameter and e is the expected front wheel steering angle difference.

[0031] In the aforementioned intelligent vehicle trajectory tracking hybrid control method, the output u of the sliding mode controller is smooth The calculation formula is:

[0032] u smooth = -k·tanh(s);

[0033] Where k is the sliding mode gain and tanh(s) is the hyperbolic tangent function of the sliding surface.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] The present invention combines the advantages of high precision of the NMPC controller and high computational efficiency of the LEMPC controller. When faced with complex working conditions, the NMPC controller can utilize the characteristics of the vehicle's nonlinear dynamics model to provide the vehicle with more accurate desired front wheel steering angle control instructions under complex road conditions, ensuring that the vehicle accurately tracks the target path; and under other working conditions, the LEMPC controller, with its efficient computing power, quickly responds to system requirements and ensures stable vehicle driving, effectively solving the problem that a single controller is difficult to balance accuracy and efficiency, and greatly improving the adaptability of the vehicle under various complex working conditions. The present invention avoids unstable switching of the control system under different working conditions by introducing a smooth switch in conjunction with a smooth switching function, significantly reduces transient oscillations and disturbances, and ensures a stable driving posture of the vehicle. In addition, the addition of the sliding mode controller of the present invention further enhances the robustness of the system during the switching process, suppresses the jitter phenomenon caused by frequent switching, and synergizes from multiple aspects to comprehensively improve the stability of the vehicle control system. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a control strategy diagram of the method of the present invention;

[0037] Figure 2 Selecting a flow chart for the controller of the present invention;

[0038] Figure 3 This is a comparison diagram of the double lane-shifting path tracking effect of the present invention;

[0039] Figure 4 This is a comparison diagram of the lateral acceleration of the double-lane shifting of the present invention;

[0040] Figure 5 This is a comparison diagram of the yaw angular velocity of the double-lane shifting method of the present invention;

[0041] Figure 6 A comparison diagram of the continuous lane change path tracking effect of the present invention;

[0042] Figure 7 This is a comparison diagram of lateral acceleration of continuous lane change according to the present invention;

[0043] Figure 8 This is a comparison diagram of the yaw angular velocity of the continuous lane change of the present invention. DETAILED DESCRIPTION

[0044] The present invention is further described below in conjunction with the accompanying drawings and embodiments, but they are not intended to limit the present invention.

[0045] Embodiment: A hybrid control method for intelligent vehicle trajectory tracking, such as Figure 1 As shown, it includes an NMPC controller and a LEMPC controller, calculates the lateral error and the lateral error change rate, and selects the corresponding control mode according to the size of the lateral error and the lateral error change rate, such as Figure 2 When the lateral error is greater than or equal to the lateral error threshold and the lateral error change rate is less than or equal to the lateral error change rate threshold, the NMPC controller is selected, and the NMPC controller uses the vehicle dynamics model as the prediction model to output the first desired front wheel steering angle; when the lateral error is less than the lateral error threshold or the lateral error change rate is greater than the lateral error change rate threshold, the LEMPC controller is selected, and the LEMPC controller uses the vehicle dynamics error model as the prediction model to output the second desired front wheel steering angle.

[0046] In this embodiment, the Carsim vehicle model is a simulation software specifically for vehicle dynamics. Based on the theory of multi-rigid body dynamics, it can accurately simulate various dynamic behaviors of the vehicle. The NMPC (nonlinear model predictive control) controller adopts an advanced control strategy, which is based on the nonlinear model of the controlled object and determines the control input at the current moment by solving a finite time domain optimization problem online. The LEMPC (linearized model predictive control) controller adopts an advanced control strategy that combines machine learning technology with economic model predictive control to better handle uncertainty, nonlinearity and complex dynamic characteristics in the system and improve control performance.

[0047] Specifically, the lateral error e d The specific calculation formula is:

[0048] e d =Y r -Y;

[0049] Among them, Y r is the reference trajectory, Y is the actual trajectory; the lateral error change rate is the lateral error e d Obtained by derivation.

[0050] In this embodiment, in a complex road environment, the curvature of the road is constantly changing, and the road curvature directly affects the difficulty of steering control. When the road curvature is large, the vehicle has more stringent requirements for steering control, and the vehicle needs to adjust the steering more accurately to avoid large lateral errors. Therefore, the tolerance for lateral errors should be reduced, and the lateral error threshold should be reduced accordingly. On the contrary, the vehicle can tolerate large lateral errors, and the lateral error threshold can be moderately increased. Since the curvature change has a significant impact on the vehicle's lateral acceleration and steering angle changes, in order to more accurately describe the relationship between the lateral error threshold and the road curvature, a quadratic function is used for modeling to achieve dynamic adjustment of the error tolerance under different curvature conditions.

[0051] Specifically, according to the road conditions, the lateral error threshold ρ is calculated pos and the lateral error change rate threshold τ max ; The lateral error threshold ρ pos The calculation formula is:

[0052]

[0053] Among them, e ymax is the maximum allowable lateral error threshold, K is the road curvature, and β is a constant used to control the influence of the square of the road curvature on the lateral error threshold;

[0054] The lateral error change rate threshold τ max The calculation formula is:

[0055]

[0056] Where C is a constant, K is the road curvature, a is a parameter that controls the influence of road curvature on the threshold of the lateral error change rate, and e is the expected front wheel steering angle difference;

[0057] The calculation formula for the expected front wheel steering angle difference e is:

[0058] e = δ1-δ2;

[0059] Wherein, δ1 is the first desired front wheel steering angle obtained by the NMPC controller; δ2 is the second desired front wheel steering angle obtained by the LEMPC controller.

[0060] In this embodiment, during the switching control between the NMPC controller and the LEMPC controller, a smooth switching function is introduced to perform weighted calculation on the desired front wheel steering angles output by the NMPC controller and the LEMPC controller to constrain the output amplitude of the desired front wheel steering angle until the switching is completed, thereby achieving smooth switching.

[0061] In this embodiment, when the control mode is switched, in order to avoid a jump in the output of the control quantity, which may cause instantaneous oscillation and disturbance, a smooth switching function is introduced. The expression of the smooth switching function is as follows:

[0062]

[0063] Among them, α(e) is the smooth switching function, is the steepness of the control weight transition, mode indicates the switching control signal between the NMPC controller and the LEMPC controller, when the NMPC controller is selected, mode is 1, when the LEMPC controller is selected, mode is 2; e is the expected front wheel steering angle difference.

[0064] The formula for weighted calculation of the first expected front wheel turning angle and the second expected front wheel turning angle is:

[0065] δ f1 =α(e)δ1+(1-α(e))δ2;

[0066] In the formula, δ f1 is the output front wheel steering angle, α(e) is the smooth switching function, δ1 is the first desired front wheel steering angle obtained by the NMPC controller; δ2 is the second desired front wheel steering angle obtained by the LEMPC controller.

[0067] In this embodiment, since the control weight of the smooth switching function needs to change in real time according to the change of curvature, a fuzzy controller is designed to The desired front wheel steering angle difference e is derived to obtain the rate of change de. The fuzzy controller takes the absolute value of the desired front wheel steering angle difference |e| of the NMPC controller and the LEMPC controller before and after the mode switching and the rate of change de of the desired front wheel steering angle difference as input. As the output. As shown in Table 1, the fuzzy domain of the input |e| of the fuzzy controller is {0,1,2,3}, the corresponding fuzzy subset is {Z0,PS,PM,PB}, the fuzzy domain of its change rate de is {-1,0,1}, the corresponding fuzzy subset is {N,ZO,P}, and the membership function adopts Gaussian type. The output of the fuzzy controller The fuzzy domain is {0,1,2,3}, the corresponding fuzzy subset is {Z0,PS,PM,PB}, and the membership function uses a triangle.

[0068] Table 1 Rules corresponding to the fuzzy controller

[0069]

[0070] According to the lateral error threshold ρ pos From the calculation formula, we can know that the lateral error threshold ρ posThere is a dependence on the road curvature K. As the road curvature K increases, the lateral error threshold ρ pos Should be reduced accordingly. This change may cause the control system to switch frequently when the road curvature K is large. Simply using a smooth switching function may not effectively avoid the cumulative error caused by switching, thereby affecting the control performance. Therefore, a sliding mode controller is introduced in the process of smooth switching to enhance the robustness during the switching process and effectively suppress the jitter caused by frequent switching, thereby improving the stability and control accuracy of the vehicle. The design steps of the sliding mode controller are as follows:

[0071] 1) Define the sliding surface

[0072] First, a sliding surface needs to be defined to represent the state of the vehicle. Since the switch between the LEMPC controller and the NMPC controller is usually based on the path tracking error of the vehicle, the lateral error e d and the rate of change of lateral error As the sliding surface. However, due to the inconsistency between the outputs of the NMPC controller and the LEMPC controller, a sudden change will occur at the switching point. For this reason, the sliding mode controller is directly used to smooth the transition between the NMPC controller and the LEMPC. Therefore, the sliding surface can be defined as:

[0073] s=||e||+ωe d ;

[0074] Among them, ω is a positive weight parameter and e is the expected front wheel steering angle difference.

[0075] 2) Design smooth control law

[0076] The goal of the sliding mode controller is to design the control output u smooth (t) is used to drive the vehicle's control system to slide along the sliding surface and ensure that it eventually remains on the sliding surface. In order to avoid oscillations caused by the discontinuity of the sign function, a smooth control law is usually used. The calculation formula of the smooth control law is:

[0077] u smooth = -k·tanh(s);

[0078] Where k is the sliding mode gain and tanh(s) is the hyperbolic tangent function of the sliding surface.

[0079] The present invention combines the advantages of high precision of the NMPC controller and high computational efficiency of the LEMPC controller. When faced with complex working conditions, the NMPC controller can utilize the characteristics of the vehicle's nonlinear dynamics model to provide the vehicle with more accurate desired front wheel steering angle control instructions under complex road conditions, ensuring that the vehicle accurately tracks the target path; and under other working conditions, the LEMPC controller, with its efficient computing power, quickly responds to system requirements and ensures stable vehicle driving, effectively solving the problem that a single controller is difficult to balance accuracy and efficiency, and greatly improving the adaptability of the vehicle under various complex working conditions. The present invention avoids unstable switching of the control system under different working conditions by introducing a smooth switch in conjunction with a smooth switching function, significantly reduces transient oscillations and disturbances, and ensures a stable driving posture of the vehicle. In addition, the addition of the sliding mode controller of the present invention further enhances the robustness of the system during the switching process, suppresses the jitter phenomenon caused by frequent switching, and synergizes from multiple aspects to comprehensively improve the stability of the vehicle control system.

[0080] Comparative example: In order to verify the control effect of the proposed hybrid path tracking controller in the path tracking process, the Carsim vehicle model and MATLAB / Simulink (imulink is a graphical software tool for modeling, simulation and analysis of dynamic systems. It is integrated in the MATLAB environment and provides engineers and researchers with an intuitive and efficient platform for designing, testing and verifying various complex systems) joint simulation method is adopted to verify the effectiveness of this method under double lane change conditions and continuous lane change conditions.

[0081] 1) Double lane shifting condition

[0082] The expression of double shift line is selected as:

[0083]

[0084] d x1 =25,d x2 =21.95,

[0085] d y1 =25,d y2 =21.95;

[0086] Among them, z1, z2, d x1 ,d y1 It has no real meaning, it is just a parameter name. Y is the reference horizontal position. is the reference yaw angle, and X is the reference longitudinal position.

[0087] Figure 3It shows that the path tracking hybrid control of the present invention has better path tracking effect than the single NMPC controller, with the maximum deviation of 0.083m. It can be seen that the actual path of the vehicle can track the target path well, and the speed at which the actual path converges to the target path at the curve is faster than that of the single LEMPC controller or NMPC controller. Figure 4 and Figure 5 It shows that when the vehicle is driving in a double lane change condition, the lateral acceleration and yaw angular velocity of the path tracking hybrid control without a switching controller change steeply and unstable, while the path tracking hybrid control with a smooth switch of the present invention has smaller fluctuations than a single LEMPC controller or NMPC controller, and the lateral acceleration and yaw angular velocity change relatively smoothly, indicating that the path tracking hybrid control with a smooth switch of the present invention can be controlled to be in a good stable state during the path tracking process.

[0088] 2) Continuous lane change conditions

[0089] The selected lane change curve expression is:

[0090]

[0091] Among them, d is the lateral displacement of the vehicle after the lane change is completed, and d=4m is selected; l is the longitudinal displacement of the vehicle after the lane change is completed, and l=100m is selected.

[0092] Figure 6 It shows that the vehicle can still track the path well under the continuous lane change curve. Figure 7 and Figure 8 It shows that when the vehicle is driving under continuous lane change conditions, the lateral acceleration and yaw rate of the path tracking hybrid control without a switching controller change steeply and vibrate severely, while the path tracking hybrid control with a smooth switch of the present invention has smaller fluctuations than a single LEMPC controller or NMPC controller, and the changes in lateral acceleration and yaw rate are relatively stable, indicating that the path tracking hybrid controller with a switching controller of the present invention can control the path tracking process to be in a good stable state.

[0093] In summary, the present invention can meet the lateral control requirements of intelligent vehicles under complex working conditions and improve trajectory tracking control performance.

Claims

1. A hybrid control method for intelligent vehicle trajectory tracking, comprising a NMPC controller and a LEMPC controller, characterized in that: Calculate the lateral error and the lateral error change rate, and select the corresponding control mode according to the size of the lateral error and the lateral error change rate; when the lateral error is greater than or equal to the lateral error threshold and the lateral error change rate is less than or equal to the lateral error change rate threshold, select the NMPC controller, and the NMPC controller uses the vehicle dynamics model as a prediction model to output the first desired front wheel steering angle; when the lateral error is less than the lateral error threshold or the lateral error change rate is greater than the lateral error change rate threshold, select the LEMPC controller, and the LEMPC controller uses the vehicle dynamics error model as a prediction model to output the second desired front wheel steering angle; In the process of switching control between the NMPC controller and the LEMPC controller, a smooth switching function is introduced to perform weighted calculation on the desired front wheel steering angles output by the NMPC controller and the LEMPC controller, and constrain the output amplitude of the desired front wheel steering angle until the switching is completed, thereby achieving smooth switching.

2. The intelligent vehicle trajectory tracking hybrid control method according to claim 1 is characterized in that: The lateral error e d The specific calculation formula is: and d =And r -AND; The lateral error change rate is the lateral error e d Obtained by derivation.

3. The intelligent vehicle trajectory tracking hybrid control method according to claim 1 is characterized in that: The lateral error threshold ρ pos The calculation formula is: Among them, e ymax is the maximum allowable lateral error threshold; K is the road curvature, and β is a constant used to control the influence of the square of the road curvature on the lateral error threshold.

4. The intelligent vehicle trajectory tracking hybrid control method according to claim 1 is characterized in that: The lateral error change rate threshold τ max The calculation formula is: Where C is a constant, K is the road curvature, a is a parameter that controls the influence of road curvature on the threshold of the lateral error change rate, and e is the expected front wheel steering angle difference; The calculation formula for the expected front wheel steering angle difference e is: e = δ1-δ2; Wherein, δ1 is the first desired front wheel steering angle obtained by the NMPC controller; δ2 is the second desired front wheel steering angle obtained by the LEMPC controller.

5. The intelligent vehicle trajectory tracking hybrid control method according to claim 3 is characterized by: The expression of the smooth switching function is as follows: Among them, α(e) is the smooth switching function, is the steepness of the control weight transition, mode indicates the switching control signal between the NMPC controller and the LEMPC controller. When the NMPC controller is selected, mode is 1, and when the LEMPC controller is selected, mode is 2; e is the expected front wheel steering angle difference.

6. The intelligent vehicle trajectory tracking hybrid control method according to claim 5 is characterized by: The formula for weighted calculation of the first expected front wheel turning angle and the second expected front wheel turning angle is: d f1 =α(e)δ1+(1-α(e))δ2; In the formula, δ f1 is the output front wheel steering angle, α(e) is the smooth switching function, δ1 is the first desired front wheel steering angle obtained by the NMPC controller; δ2 is the second desired front wheel steering angle obtained by the LEMPC controller.

7. The intelligent vehicle trajectory tracking hybrid control method according to claim 5 is characterized by: The smooth switch has a fuzzy controller, and the smooth switching function is controlled by the fuzzy controller. Dynamic adjustment is performed; the desired front wheel steering angle difference e is derived to obtain the change rate de, and the fuzzy controller takes the absolute value of the desired front wheel steering angle difference |e| of the NMPC controller and the LEMPC controller before and after the mode switching and the change rate de of the desired front wheel steering angle difference as input, as output.

8. The intelligent vehicle trajectory tracking hybrid control method according to claim 1 is characterized by: A sliding mode controller is introduced in the smooth switching process; the sliding mode controller controls the output by defining a sliding mode surface to drive the control system of the vehicle to slide along the sliding mode surface.

9. The intelligent vehicle trajectory tracking hybrid control method according to claim 8, characterized in that: The expression for defining the sliding surface is: s=||e||+ωe d ; Among them, ω is a positive weight parameter and e is the expected front wheel steering angle difference.

10. The intelligent vehicle trajectory tracking hybrid control method according to claim 9, characterized in that: The output u of the sliding mode controller smooth The calculation formula is: u smooth =-k·tanh(s); Where k is the sliding mode gain and tanh(s) is the hyperbolic tangent function of the sliding surface.

Images