Personalized steering correction target curve construction method based on state expansion observer

Abstract

This invention relates to the field of automotive steering control technology and provides a method for constructing a personalized steering return-to-center target curve based on a state extension observer (ESO). The method includes: real-time driving preference identification, which compares the driver's actual operating speed with the expected speed of the baseline model to extract steering speed correction; self-evolutionary construction of the personalized return-to-center target curve, which uses an ESO to transform a discrete personalized control target preference dataset into a personalized return-to-center target curve; and adaptive sliding mode return-to-center tracking control, which designs an adaptive sliding mode control law combined with ESO feedforward compensation for the personalized return-to-center target curve to drive the actuator motor to accurately track the target trajectory. This method establishes a closed loop of "identification - ESO evolution target - ASM tracking," enabling online growth of the target trajectory and solving the problems of existing strategies being unable to identify driver preferences online and lacking robustness under complex conditions.

Description

Technical Field

[0001] This invention belongs to the field of automotive steering control technology, and particularly relates to a method for constructing a personalized steering return target curve based on a state extension observer. Background Technology

[0002] In recent years, with the increasing demand for intelligent and personalized vehicles, consumers' expectations for steering return-to-center functionality have moved beyond simple center alignment, focusing instead on the feel and quality during the return-to-center process. Traditional methods for constructing steering return-to-center target curves mostly rely on a simple classification of the return-to-center target based on preset driving styles (such as sport, comfort, etc.), and then use this as the basis for calibration and weighted control. These methods struggle to adapt to the diverse preferences of individual drivers regarding the linearity of the return-to-center feel, convergence speed, and damping, and also fail to meet drivers' true expectations under complex and varied operating conditions.

[0003] To optimize steering return-to-center control, existing technologies have proposed several improvement schemes. For example, Chinese invention patent CN119283958A, "A Control Method for Switching Between EPS Assist Mode and Return-to-Center Mode," designs a switching time length module to determine a return-to-center time that better meets the driver's expectations based on the current steering wheel angle and vehicle speed, making the mode switching process smoother. However, this method mainly focuses on the smoothness control of the transient mode switching, and its core return-to-center target curve is still generated based on fixed parameters, making it unable to adaptively adjust according to individual driver preferences. Therefore, it is difficult to resolve the contradiction between standardized control and personalized needs.

[0004] Another Chinese invention patent, CN121316964A, entitled "An Active Return-to-Center Control Method and System for Electric Power Steering," improves control robustness by estimating the steering wheel's inertial torque to correct the requested torque. While this method suppresses interference to some extent through inertial torque compensation, its core architecture still relies on a deterministic feedback adjustment mechanism. This method fails to incorporate online identification of driver operating habits and dynamic updating of the target curve, resulting in a fixed return-to-center characteristic. It cannot continuously learn and adapt to the driver's personalized style throughout the vehicle's lifecycle, and its tracking accuracy and robustness under complex conditions still need improvement. Summary of the Invention

[0005] The purpose of this invention is to provide a method for constructing a personalized steering return target curve based on a state expansion observer, aiming to solve the problems mentioned in the background art.

[0006] The present invention is implemented as follows: a method for constructing a personalized steering return target curve based on a state expansion observer includes the following steps:

[0007] Step 1: Real-time driving preference identification;

[0008] Based on the real-time operating status of the vehicle, the driver's actual operating speed is compared with the expected speed of the baseline model to determine the consistency between the driver's needs and the baseline model; if the driver's current needs for the return-to-center quality are inconsistent with the baseline model, the steering speed correction amount reflecting the driver's personalized intention is extracted to construct a personalized control target preference dataset.

[0009] Step 2: Self-evolutionary construction of personalized recovery target curve;

[0010] An extended state observer (ESO) is used to establish a self-evolutionary mechanism from the baseline model to the personalized target, transforming the discrete personalized control target preference dataset into a continuous, smooth, and dynamically adjustable personalized homing target curve.

[0011] Step 3: Adaptive sliding mode return-to-center tracking control;

[0012] For the personalized return-to-center target curve, a control law based on adaptive sliding mode (ASM) combined with ESO feedforward compensation is designed to compensate for system uncertainties and model errors online, and drive the actuator motor to accurately track the target return-to-center trajectory.

[0013] A further technical solution, in step 1, the specific process of real-time driving preference identification is as follows:

[0014] During the steering wheel return phase, the steering wheel angle is collected simultaneously. Rotation speed Driver input torque and vehicle speed ;

[0015] The driver's actual operating speed is compared with the expected speed of the baseline model to determine the consistency between the driver's needs and the baseline model.

[0016] Define a real-time deviation observation. :

[0017]

[0018] in, As a preset baseline model, This refers to the driver's actual operating speed. The expected speed of the baseline model;

[0019] Preset For personalized consistency activation threshold, if If the driver's current requirement for the return-to-center quality is consistent with the baseline model, the system maintains baseline control; otherwise... If the system determines that the driver has unique driving preferences or that the baseline model cannot meet the requirements under the current operating conditions, it will trigger an activation signal and extract the steering speed observation correction amount that reflects the driver's personalized intentions. :

[0020]

[0021] The recursive least squares (RLS) method is used to perform online preference learning on the correction amount, eliminating high-frequency noise and system transient response errors, and obtaining a stable and convergent personalized return-to-normal target velocity. :

[0022]

[0023] in, This is the learning correction amount for steering speed.

[0024] A further technical solution, in step 2, is to implement the self-evolution mechanism based on ESO as follows:

[0025] The process of personalized goal evolution is abstracted into a second-order subsystem, and driver preference bias and nonlinear friction are regarded as an expansion state. ;

[0026] Constructing the discretized form of the ESO equations:

[0027]

[0028]

[0029]

[0030] in, The target rotation angle after fitting. The target recovery speed of self-evolution output , The sampling step size, , , For ESO observer gain, For observation error, , For nonlinear function parameters, For nonlinear error threshold, Given the system's known control gain, Input for target curve control. This is the linearization switching function.

[0031] A further technical solution, in step 3, specifically includes the control law based on ASM and combined with ESO feedforward compensation, which includes:

[0032] Establish the error dynamic equation: Define the tracking error Establish a system that includes control inputs The second-order dynamic equation:

[0033]

[0034] in, The second derivative of the tracking error, To correct the angle for personalized goals, The coefficients of the second-order error equation, This is a comprehensive disturbance term that includes friction and parameter perturbations. To achieve a personalized target positive angular acceleration;

[0035] Design the sliding surface: Select a first-order linear sliding surface. ,in To adjust the positive definite coefficients of the convergence rate, This is the first derivative of the tracking error;

[0036] Constructing the control law: The motor control torque is determined by the nominal equivalent term. ESO feedforward compensation term and adaptive approach term composition:

[0037]

[0038] in, The real-time estimate of the upper bound of the perturbation is obtained through an adaptive law. Update For adaptive gain updates; The gain coefficient for sliding mode control; It is a saturation function; The thickness of the saturation layer; This is the proportionality coefficient.

[0039] In a further technical solution, in step 2, ESO ensures the smoothness and approximation accuracy of the target curve by dynamically adjusting the observer gain.

[0040] The present invention provides a method for constructing a personalized steering return target curve based on a state extension observer. This method overcomes the limitations of traditional fixed calibration. By combining RLS and ESO, it achieves online learning and self-evolution of driver operating preferences, enabling the return characteristics to dynamically adjust according to driver habits and providing personalized adaptation capabilities. ESO is used to smooth and filter the target trajectory, ensuring no sudden changes in motor torque during the return process. Furthermore, sliding mode control achieves high-precision steady-state tracking of the personalized target. The use of an adaptive sliding mode algorithm combined with ESO feedforward compensation can offset the adverse effects of nonlinear friction, parameter perturbations, and road surface disturbances in real time. By combining personalized preference learning with underlying robust control, the system exhibits a high degree of personalization and robust steering quality under all operating conditions. Attached Figure Description

[0041] Figure 1 A schematic diagram illustrating the implementation scheme of the personalized steering and homing target curve construction method based on state expansion observer provided in an embodiment of the present invention;

[0042] Figure 2 A schematic diagram illustrating the implementation scheme of step 1 in the personalized steering and homing target curve construction method based on state expansion observer provided in the embodiments of the present invention;

[0043] Figure 3 A schematic diagram illustrating the implementation scheme of step 2 in the personalized steering and homing target curve construction method based on state expansion observer provided in the embodiments of the present invention;

[0044] Figure 4 A schematic diagram illustrating the implementation scheme of step 3 in the personalized steering and homing target curve construction method based on state expansion observer provided in the embodiments of the present invention;

[0045] Figure 5 This is a comparison diagram of the positive phase trajectory before and after self-evolution. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0047] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0048] like Figures 1-4 As shown, a method for constructing a personalized steering return target curve based on a state expansion observer, according to an embodiment of the present invention, includes the following steps:

[0049] Step 1: Real-time driving preference identification;

[0050] Based on the vehicle's real-time operating status, the driver's actual operating speed is compared with the expected speed of the baseline model to determine the consistency between the driver's needs and the baseline model. If the driver's current demand for return-to-center quality is inconsistent with the baseline model, the steering speed correction amount reflecting the driver's personalized intention is extracted to construct a personalized control target preference dataset.

[0051] Step 2: Self-evolutionary construction of personalized recovery target curve;

[0052] By utilizing ESO, a self-evolutionary mechanism from baseline model to personalized target is established, transforming discrete personalized control target preference datasets into continuous, smooth, and dynamically adjustable personalized positive target curves.

[0053] Step 3: Adaptive sliding mode return-to-center tracking control;

[0054] For the personalized return-to-center target curve, a control law based on ASM and combined with ESO feedforward compensation is designed to compensate for system uncertainties and model errors online, and drive the actuator motor to accurately track the target return-to-center trajectory.

[0055] like Figure 2 As shown, in a preferred embodiment of the present invention, the specific process of real-time driving preference identification in step 1 is as follows:

[0056] During the steering wheel return phase, the steering wheel angle is collected simultaneously. Rotation speed Driver input torque and vehicle speed ;

[0057] The driver's actual operating speed is compared with the expected speed of the baseline model to determine the consistency between the driver's needs and the baseline model.

[0058] Define a real-time deviation observation. :

[0059]

[0060] in, As a preset baseline model, This refers to the driver's actual operating speed. This represents the expected speed of the baseline model.

[0061] Preset This is the activation threshold for personalized consistency. If... If the driver's current requirement for the centering quality matches the baseline model, the system maintains baseline control and does not extract any correction values. If the system determines that the driver has unique driving preferences or that the baseline model cannot meet the requirements under the current operating conditions, it will trigger an activation signal and extract the steering speed correction amount reflecting the driver's personalized intentions. The steering speed observation correction amount is defined as follows: :

[0062]

[0063] The recursive least squares method is used to perform online preference learning on the correction amount, eliminating high-frequency noise and system transient response errors, and obtaining a stable and convergent personalized homing target velocity. :

[0064]

[0065] in, This is the learning correction amount for steering speed.

[0066] like Figure 3 As shown, in a preferred embodiment of the present invention, in step 2, the self-evolution mechanism based on ESO is implemented as follows:

[0067] The process of personalized goal evolution is abstracted into a second-order subsystem, and driver preference bias and nonlinear friction are regarded as an expansion state. ;

[0068] Constructing the discretized form of the ESO equations:

[0069]

[0070]

[0071]

[0072] in, The target rotation angle after fitting. The target recovery speed of self-evolution output , The sampling step size, , , For ESO observer gain, For observation error, , For nonlinear function parameters, For nonlinear error threshold, Given the system's known control gain, Input for target curve control. The linearization switching function ensures the smoothness and approximation accuracy of the target curve by dynamically adjusting the observer gain.

[0073] like Figure 4 As shown, in a preferred embodiment of the present invention, in step 3, the control law based on ASM and combined with ESO feedforward compensation specifically includes:

[0074] Establish the error dynamic equation: Define the tracking error Establish a system that includes control inputs The second-order dynamic equation:

[0075]

[0076] in, The second derivative of the tracking error, To correct the angle for personalized goals, The coefficients of the second-order error equation, This is a comprehensive disturbance term that includes unmodeled dynamics such as friction and parameter perturbations. To achieve a personalized target positive angular acceleration;

[0077] Design the sliding surface: Select a first-order linear sliding surface. ,in To adjust the positive definite coefficients of the convergence rate, This is the first derivative of the tracking error;

[0078] Constructing the control law: The motor control torque is determined by the nominal equivalent term. ESO feedforward compensation term and adaptive approach term composition:

[0079]

[0080] in, The real-time estimate of the upper bound of the perturbation is obtained through an adaptive law. Update For adaptive gain updates; The gain coefficient for sliding mode control; It is a saturation function; The thickness of the saturation layer; This is the proportionality coefficient.

[0081] like Figure 5The image shows a comparison of the self-centering trajectory before and after self-evolution, contrasting the baseline before self-evolution with the more personalized target curve after self-evolution that better matches the driver's expectations. The dashed line, representing the baseline trajectory, exhibits a standard linear or simple non-linear relationship based on preset parameters. While it can complete the basic centering task, it cannot capture the subtle preferences of a particular driver for changes in centering force and speed. The solid line, representing the target curve after ESO self-evolution, shows significant differences in characteristics during the initial and near-midpoint stages of centering. The evolved trajectory is smoother and conforms to the driver's operating logic under specific conditions, demonstrating that the system can adjust according to the driver's actual correction amount. The target trajectory is dynamically reconstructed, making the turning angle and speed during the return-to-center process more closely match the driver's psychological expectations.

[0082] In summary, the personalized steering return control method based on the State Extended Observer (ESO) proposed in this invention breaks through the limitations of traditional "fixed parameter" control by learning driver operating preferences in real time online. Experimental and simulation results show that the personalized return curve generated by this method can more accurately match the unique expectations of different drivers for return feel and convergence speed compared to the baseline model. Meanwhile, the introduced adaptive sliding mode control and ESO feedforward compensation mechanism effectively enhance the system's robustness under nonlinear disturbances and complex operating conditions, achieving an organic unity of personalized adaptation and high-precision steady-state tracking during the steering return process.

[0083] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for constructing a personalized steering and homing target curve based on a state-extended observer, characterized in that, Includes the following steps: Step 1: Real-time driving preference identification; Based on the real-time operating status of the vehicle, the driver's actual operating speed is compared with the expected speed of the baseline model to determine the consistency between the driver's needs and the baseline model. If the driver's current demand for return-to-center quality is inconsistent with the baseline model, the steering speed correction amount reflecting the driver's personalized intention is extracted to construct a personalized control target preference dataset. Step 2: Self-evolutionary construction of personalized recovery target curve; By using ESO to establish a self-evolutionary mechanism from baseline model to personalized target, discrete personalized control target preference datasets are transformed into continuous, smooth and dynamically adjustable personalized positive target curves. Step 3: Adaptive sliding mode return-to-center tracking control; For the personalized return-to-center target curve, a control law based on ASM and combined with ESO feedforward compensation is designed to compensate for system uncertainties and model errors online, and drive the actuator motor to accurately track the target return-to-center trajectory.

2. The method for constructing a personalized steering and homing target curve based on a state expansion observer according to claim 1, characterized in that, In step 1, the specific process of real-time driving preference identification is as follows: During the steering wheel return phase, the steering wheel angle is collected simultaneously. Rotation speed Driver input torque and vehicle speed ; The driver's actual operating speed is compared with the expected speed of the baseline model to determine the consistency between the driver's needs and the baseline model. Define a real-time deviation observation. : in, As a preset baseline model, This refers to the driver's actual operating speed. The expected speed of the baseline model; Preset For personalized consistency activation threshold, if If the driver's current requirement for the return-to-center quality is consistent with the baseline model, the system maintains baseline control; otherwise... If the system determines that the driver has unique driving preferences or that the baseline model cannot meet the requirements under the current operating conditions, it will trigger an activation signal and extract the steering speed observation correction amount that reflects the driver's personalized intentions. : The recursive least squares method is used to perform online preference learning on the correction amount, eliminating high-frequency noise and system transient response errors, and obtaining a stable and convergent personalized homing target velocity. : in, This is the learning correction amount for steering speed.

3. The method for constructing a personalized steering and homing target curve based on a state-extended observer according to claim 2, characterized in that, In step 2, the self-evolution mechanism based on ESO is implemented as follows: The process of personalized goal evolution is abstracted into a second-order subsystem, and driver preference bias and nonlinear friction are regarded as an expansion state. ; Constructing the discretized form of the ESO equations: in, The target rotation angle after fitting. The target recovery speed of self-evolution output , The sampling step size, , , For ESO observer gain, For observation error, , For nonlinear function parameters, For nonlinear error threshold, Given the system's known control gain, Input for target curve control. This is the linearization switching function.

4. The method for constructing a personalized steering and homing target curve based on a state expansion observer according to claim 3, characterized in that, In step 2, ESO ensures the smoothness and approximation accuracy of the target curve by dynamically adjusting the observer gain.

5. The method for constructing a personalized steering and homing target curve based on a state expansion observer according to claim 3, characterized in that, In step 3, the control law based on ASM and combined with ESO feedforward compensation specifically includes: Establish the error dynamic equation: Define the tracking error Establish a system that includes control inputs The second-order dynamic equation: in, The second derivative of the tracking error, To correct the angle for personalized goals, The coefficients of the second-order error equation, This is a comprehensive disturbance term that includes friction and parameter perturbations. To achieve a personalized target positive angular acceleration; Design the sliding surface: Select a first-order linear sliding surface. ,in To adjust the positive definite coefficients of the convergence rate, This is the first derivative of the tracking error; Constructing the control law: The motor control torque is determined by the nominal equivalent term. ESO feedforward compensation term and adaptive approach term composition: in, The real-time estimate of the upper bound of the perturbation is obtained through an adaptive law. Update For adaptive gain updates; The gain coefficient for sliding mode control; It is a saturation function; The thickness of the saturation layer; This is the proportionality coefficient.

Images