An adaptive robust brake-by-wire method and system
By constructing a multi-disturbance observer that integrates an improved RBF neural network and a hierarchical adaptive robust control law, the problems of poor nonlinear adaptability and insufficient multi-coupled disturbance handling capability of the brake-by-wire system are solved, achieving fast, accurate, and stable control of braking pressure and improving the robustness and adaptability of the system.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HUNAN WEIFU AUTO PARTS CO LTD
- Filing Date
- 2026-03-25
- Publication Date
- 2026-06-26
AI Technical Summary
Existing brake-by-wire systems suffer from poor nonlinear adaptability, insufficient ability to handle multiple coupled disturbances, poor sliding mode chatter suppression, and difficulty in balancing response speed and control accuracy, thus failing to meet the high dynamic and high-precision control requirements of autonomous vehicles for braking systems.
A multi-disturbance observer integrating an improved RBF neural network is constructed. By combining a hierarchical adaptive robust control law and a fuzzy adaptive sliding mode gain regulator, the nonlinear characteristics of the system are accurately characterized through multi-coupled dynamic modeling. Virtual control law and actual control law are designed to achieve accurate, fast and stable control of braking pressure.
It significantly improves the robustness and adaptability of the brake-by-wire system under complex operating conditions, enables rapid tracking and precise control of braking pressure, solves the problem of balancing response speed and control accuracy, and enhances the system's stability and anti-interference capability.
Smart Images
Figure CN122275829A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of vehicle brake-by-wire technology, and in particular to an adaptive and robust brake-by-wire method and system. Background Technology
[0002] As a core technology of intelligent vehicle chassis electronic control, brake-by-wire systems abandon the traditional mechanical transmission structure of braking and achieve precise control of braking force through electrical signals, serving as a key support for realizing autonomous driving. Although significant progress has been made in the research of existing brake-by-wire control methods, there are still significant shortcomings due to the inherent characteristics of the system: brake-by-wire systems exhibit strong nonlinearity, time-varying parameters, and unmodeled dynamic characteristics. Factors such as hydraulic oil characteristic drift, changes in valve orifice flow field, and component wear can cause continuous changes in system parameters. Traditional PID and fuzzy PID control methods, due to fixed controller parameters, are difficult to adapt to the nonlinear characteristics under all operating conditions, and are prone to problems such as slow pressure build-up and large steady-state errors, failing to meet the requirements of high-precision braking control.
[0003] Sliding mode control is widely used in brake-by-wire systems due to its strong robustness. However, traditional sliding mode control is prone to chattering, which not only reduces the control accuracy of braking pressure but also exacerbates component wear and damages the control quality of the hydraulic system. Although some studies have mitigated chattering by improving sliding mode algorithms, these are mostly optimized for single operating conditions. In complex operating conditions with multiple coupled disturbances, chattering suppression is ineffective, and it is difficult to balance robustness and control smoothness. Furthermore, existing adaptive robust control methods mostly only observe and compensate for single disturbance sources, lacking the ability to handle multi-dimensional coupled disturbances such as hydraulic coupling, mechanical friction, and actuator hysteresis in brake-by-wire systems. When faced with external disturbances such as road surface disturbances and load changes, they are prone to control lag and weak anti-interference capabilities.
[0004] Meanwhile, existing linear adaptive braking methods that combine neural networks, while able to approximate unknown system disturbances through neural networks, mostly employ traditional RBF neural networks. These methods suffer from slow weight convergence speed and low disturbance approximation accuracy. Furthermore, the lack of coordinated design between the neural network weight adaptive law and controller parameter adjustment leads to a mismatch between disturbance observations and control law compensation, making it difficult to achieve real-time and accurate suppression of system disturbances. In addition, most control methods do not perform hierarchical optimization design for the virtual control law of backstepping control, which easily results in slow state quantity convergence and large overshoot during braking pressure tracking. They cannot balance the response speed and stability of braking control, making it difficult to meet the high dynamic and high-precision control requirements of autonomous vehicles for braking systems. Summary of the Invention
[0005] The purpose of this invention is to propose an adaptive and robust brake-by-wire method and system to solve the problems of poor nonlinear adaptability, insufficient ability to handle multiple coupled disturbances, poor effect of sliding mode chatter suppression, and difficulty in balancing response speed and control accuracy in existing brake-by-wire systems. This invention enables accurate, fast, and stable control of braking pressure under complex working conditions, thereby improving the robustness and adaptability of the brake-by-wire system.
[0006] To achieve the above objectives, this invention proposes an adaptive and robust linear braking method, the steps of which are as follows: Step S1: Establish a multi-coupled dynamic mathematical model of the brake-by-wire system, introduce multi-dimensional disturbance terms to characterize the system nonlinearity, parameter uncertainty and external disturbances, and determine the system state variables and control inputs; Step S2: Construct a multi-disturbance observer that integrates and improves the RBF neural network. Design the neural network weight adaptive law and parameter adaptive law based on the Lyapunov function to observe and approximate the hydraulic coupling disturbance, mechanical friction disturbance and external road surface disturbance of the system online. Step S3: Based on backstepping sliding mode control technology and combined with the observations of multiple disturbance observers, design a hierarchical adaptive robust control law, which includes a virtual control law and an actual control law. Step S4: Design a fuzzy adaptive sliding mode gain regulator and adjust the gain coefficient of the sliding mode control online; Step S5: Based on Lyapunov stability theory, verify the asymptotic stability of the entire control system until the state variables converge to the desired value; Step S6: Convert the designed control law into an electrical signal and input it to the actuator of the brake-by-wire system to achieve closed-loop adaptive robust control of the braking pressure.
[0007] Preferably, in step S1, the multi-coupled dynamic mathematical model of the brake-by-wire system uses the brake wheel cylinder pressure, valve core displacement, valve core movement speed, and brake wheel cylinder piston displacement as state variables, and the input voltage of the electromagnetic direct drive valve as the control input. Simultaneously, hydraulic coupling disturbance, mechanical friction disturbance, and external road surface disturbance are introduced as disturbance terms, as shown in the following formula: ; in, The cross-sectional area of the piston. P For the brake wheel cylinder pressure, Let be the piston mass, y be the piston displacement of the brake wheel cylinder, and t be the time parameter. This is the piston's equivalent damping coefficient. For the stiffness of the piston cylinder spring, The initial length of the spring. For hydraulic coupling disturbance, Brake fluid flow rate, For the volume of the brake wheel cylinder, The bulk modulus of elasticity of the liquid. The pressure-flow coefficient is... The total mass of the electromagnetic linear actuator. The rate of change of the valve core's movement speed. Input voltage, For motor constants, The valve core movement speed, For coil resistance, c The damping coefficient is... k For spring stiffness, x For valve core displacement, For friction, For steady-state hydraulic pressure, Transient hydraulic pressure, This is due to mechanical friction disturbance. This represents the rate of change of valve core displacement. This is due to external road surface disturbance.
[0008] Preferably, in step S2, the multi-perturbation observer that integrates the improved RBF neural network will measure the brake wheel cylinder pressure. P Brake wheel cylinder pressure change rate Valve core displacement x Valve core movement speed As the observer's state variable, an RBF neural network with an improved Gaussian function as the activation function is used to fit the total system perturbation. The observer calculation formula is as follows: ; in, For the estimated values of the state variables, for Time derivative, This is an estimate of the rate of change of brake wheel cylinder pressure. Brake wheel cylinder pressure P The estimation error, , , This is the disturbance estimate. for The second time derivative, , , This is an estimate of the nonlinear term. The estimation error is for the rate of change of brake wheel cylinder pressure. for Time derivative, for The estimated value, For position estimation error, for Time derivative, Valve core displacement x The estimated value, For the equivalent velocity estimation error, , , , The base gain of the observer, , To adjust the gain proportionally, , , , , , , , These are system parameters.
[0009] Preferably, the adaptive law for the neural network weights is calculated as follows: ; in, , These are the estimated weights for the neural network. , They are respectively , The first derivative, , For weighted gain coefficients, , To improve the estimation of the radial basis function set in an RBF neural network, , This is the correction coefficient matrix.
[0010] Preferably, in step S3, the design method of the hierarchical adaptive robust control law is as follows: Step S31: Define the braking pressure tracking error target as braking pressure, and design the first-level virtual control law, as shown in the following formula: ; in, For the target of braking pressure tracking error, For the target braking pressure, This is a first-level virtual control law. Target braking pressure The derivative, It is a positive real number; Step S32: Define the virtual control error and design the second-level virtual control law, as shown in the following formula: ; in, For virtual control error, This is a second-level virtual control law. , , , These are estimated values for the system parameters. The derivative of the first-order virtual control law. It is a positive real number. This is the hydraulic state correction coefficient for the second-level virtual control law; Step S33: Define the valve core displacement tracking error and design the third-level virtual control law, as shown in the following formula: ; in, For valve core displacement tracking error, This is a third-level virtual control law. The derivative of the second-order virtual control law. It is a positive real number; Step S34: Define the valve core speed tracking error, and design the actual control law based on the disturbance estimate from the multi-disturbance observer, as shown in the following formula: ; in, Valve core speed tracking error For actual control laws, , , , , These are estimated values for the system parameters. The derivative of the third-order virtual control law. It is a positive real number.
[0011] Preferably, in step S4, the fuzzy adaptive sliding mode gain regulator uses a sliding mode switching manifold. and its derivative As input variables, with sliding mode gain For output variables: ; Where T represents the transpose operation. They are respectively Sliding mode gain; Fuzzy inference employs a max-min synthesis method, defuzzification uses the centroid method, and the sliding mode gain adjustment value is output online. The corrected sliding mode control term is: ;in, It is a saturation function. Flutter suppression factor.
[0012] Preferably, in step S5, the stability verification method of the control system is as follows: Constructing Lyapunov functions: ; in, , For the weights of the neural network, , For the error in the weight estimation of the neural network, For parameter error, For parameter gain coefficient; By taking the time derivative of the Lyapunov function and combining the adaptive control laws of weights, parameters, and hierarchical adaptive robust control, we prove that: ; It satisfies the Lyapunov asymptotic stability condition; where, is the derivative of the Lyapunov function.
[0013] The present invention also provides an adaptive robust brake-by-wire system for implementing the above-mentioned adaptive robust brake-by-wire method, comprising a dual redundant execution module, a multi-sensor sensing module, a main control computing module, and a communication module; The dual-redundant execution module includes two sets of electromagnetic direct drive valves, brake wheel cylinders and hydraulic lines with the same structure. The two sets of electromagnetic direct drive valves are connected in parallel to the hydraulic source. The two sets of brake wheel cylinders are symmetrically arranged at the vehicle braking end. Each set of electromagnetic direct drive valves is driven by an independent dual pinion redundant motor. The multi-sensor sensing module includes a pressure sensor, a displacement sensor, a speed sensor, and a road condition sensor. The pressure sensor is installed at the brake wheel cylinder outlet, the displacement sensor and the speed sensor are installed at the valve core of the electromagnetic direct drive valve, and the road condition sensor is installed on the vehicle chassis. The main control computing module includes an embedded controller, which incorporates a multi-disturbance observer, a hierarchical adaptive robust control law, and a fuzzy adaptive sliding mode gain regulator. The communication module includes an on-board CAN bus and an Ethernet. The multi-sensor perception module, the main control computing module, and the dual-redundant execution module achieve real-time data interaction through the CAN bus, and the Ethernet is used to interact with the vehicle controller.
[0014] Preferably, the dual pinion redundant motor adopts dq-axis inversion adaptive control, with an inversion adaptive current controller designed for the d-axis to achieve accurate current tracking and an inversion adaptive position controller designed for the q-axis to achieve accurate control of the valve core displacement.
[0015] Therefore, this invention proposes an adaptive and robust linear braking method and system, the advantages of which are as follows: (1) By constructing a multi-disturbance observer that integrates and improves the RBF neural network, the present invention can accurately observe and approximate multi-dimensional coupled disturbances such as hydraulic coupling and mechanical friction online. Combined with the hierarchical adaptive robust control law, disturbance feedforward compensation is realized, which effectively solves the problem of insufficient multi-disturbance processing capability of traditional methods and greatly improves the robustness of the system under complex working conditions.
[0016] (2) This invention reduces the sliding mode chattering phenomenon at its source by adjusting the sliding mode gain online through fuzzy logic. At the same time, it optimizes the backstepping control virtual control law in layers, realizing the rapid tracking and precise control of braking pressure, and solving the problem of difficulty in balancing response speed and control accuracy.
[0017] (3) This invention accurately characterizes the nonlinear characteristics of the system through multi-coupled dynamic modeling. With the coordinated adaptive adjustment of neural network weights and controller parameters, it perfectly adapts to the time-varying parameters and unmodeled dynamics of the brake-by-wire system, completely solves the problem of poor nonlinear adaptability of traditional control methods, and ensures the stability of braking control under all working conditions.
[0018] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description
[0019] Figure 1 This is a flowchart of an adaptive and robust linear braking method according to the present invention. Detailed Implementation
[0020] To make the technical solutions, advantages, and objectives of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below. The described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the protection scope of the present invention.
[0021] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.
[0022] Example 1 like Figure 1 As shown, this invention provides an adaptive and robust linear braking method, as detailed below: Step S1: Establish a multi-coupled dynamic mathematical model of the brake-by-wire system, introduce multi-dimensional disturbance terms to characterize the system nonlinearity, parameter uncertainty and external disturbances, and determine the system state variables and control inputs; The multi-coupled dynamic mathematical model of the brake-by-wire system uses brake wheel cylinder pressure, valve spool displacement, valve spool speed, and brake wheel cylinder piston displacement as state variables, and electromagnetic direct drive valve input voltage as control input. It also introduces hydraulic coupling disturbance, mechanical friction disturbance, and external road surface disturbance as disturbance terms. The formula is as follows: ; in, The cross-sectional area of the piston. P For the brake wheel cylinder pressure, Let be the piston mass, y be the piston displacement of the brake wheel cylinder, and t be the time parameter. This is the piston's equivalent damping coefficient. For the stiffness of the piston cylinder spring, The initial length of the spring. For hydraulic coupling disturbance, Brake fluid flow rate, For the volume of the brake wheel cylinder, The bulk modulus of elasticity of the liquid. The pressure-flow coefficient is... The total mass of the electromagnetic linear actuator. The rate of change of the valve core's movement speed. Input voltage, For motor constants, The valve core movement speed, For coil resistance, c The damping coefficient is... k For spring stiffness, x For valve core displacement, For friction, For steady-state hydraulic pressure, Transient hydraulic pressure, This is due to mechanical friction disturbance. This represents the rate of change of valve core displacement. This is due to external road surface disturbance.
[0023] Step S2: Construct a multi-disturbance observer that integrates and improves the RBF neural network. Design the neural network weight adaptive law and parameter adaptive law based on the Lyapunov function to observe and approximate the hydraulic coupling disturbance, mechanical friction disturbance and external road surface disturbance of the system online. A multi-perturbation observer that integrates an improved RBF neural network will measure the brake wheel cylinder pressure. P Brake wheel cylinder pressure change rate Valve core displacement x Valve core movement speed As the observer's state variable, an RBF neural network with an improved Gaussian function as the activation function is used to fit the total system perturbation. The observer calculation formula is as follows: ; in, For the estimated values of the state variables, for Time derivative, This is an estimate of the rate of change of brake wheel cylinder pressure. Brake wheel cylinder pressure P The estimation error, , , This is the disturbance estimate. for The second time derivative, , , This is an estimate of the nonlinear term. The estimation error is for the rate of change of brake wheel cylinder pressure. for Time derivative, for The estimated value, For position estimation error, for Time derivative, Valve core displacement x The estimated value, For the equivalent velocity estimation error, , , , The base gain of the observer, , To adjust the gain proportionally, , , , , , , , These are system parameters.
[0024] The adaptive law for neural network weights is calculated as follows: ; in, , These are the estimated weights for the neural network. , They are respectively , The first derivative, , For weighted gain coefficients, , To improve the estimation of the radial basis function set in an RBF neural network, , This is the correction coefficient matrix.
[0025] Step S3: Based on backstepping sliding mode control technology and combined with the observations from multiple disturbance observers, design a hierarchical adaptive robust control law, including a virtual control law and an actual control law; the design method of the hierarchical adaptive robust control law is as follows: Step S31: Define the braking pressure tracking error target as braking pressure, and design the first-level virtual control law, as shown in the following formula: ; in, For the target of braking pressure tracking error, For the target braking pressure, This is a first-level virtual control law. Target braking pressure The derivative, It is a positive real number; Step S32: Define the virtual control error and design the second-level virtual control law, as shown in the following formula: ; in, For virtual control error, This is a second-level virtual control law. , , , These are estimated values for the system parameters. The derivative of the first-order virtual control law. It is a positive real number. This is the hydraulic state correction coefficient for the second-level virtual control law; Step S33: Define the valve core displacement tracking error and design the third-level virtual control law, as shown in the following formula: ; in, For valve core displacement tracking error, This is a third-level virtual control law. The derivative of the second-order virtual control law. It is a positive real number; Step S34: Define the valve core speed tracking error, and design the actual control law based on the disturbance estimate from the multi-disturbance observer, as shown in the following formula: ; in, Valve core speed tracking error For actual control laws, , , , , These are estimated values for the system parameters. The derivative of the third-order virtual control law. It is a positive real number.
[0026] Step S4: Design a fuzzy adaptive sliding mode gain regulator and adjust the gain coefficient of the sliding mode control online; Fuzzy adaptive sliding mode gain regulator to switch manifolds using sliding mode and its derivative As input variables, with sliding mode gain For output variables: ; Where T represents the transpose operation. They are respectively Sliding mode gain; Fuzzy inference is performed using the max-min synthesis method, and defuzzification is performed using the centroid method. The sliding mode gain adjustment value is output online, and the corrected sliding mode control term is... ;in, It is a saturation function. Flutter suppression factor.
[0027] Step S5: Based on Lyapunov stability theory, verify the asymptotic stability of the entire control system until the state variables converge to the desired value; The stability verification method for the control system is as follows: Constructing Lyapunov functions: ; in, , For the weights of the neural network, , For the error in the weight estimation of the neural network, For parameter error, For parameter gain coefficient; By taking the time derivative of the Lyapunov function and combining the adaptive control laws of weights, parameters, and hierarchical adaptive robust control, we prove that: ; It satisfies the Lyapunov asymptotic stability condition; where, is the derivative of the Lyapunov function.
[0028] Step S6: Convert the designed control law into an electrical signal and input it to the actuator of the brake-by-wire system to achieve closed-loop adaptive robust control of the braking pressure.
[0029] To verify the control performance of the adaptive robust brake-by-wire algorithm of this application, a hardware-in-the-loop simulation test platform for the brake-by-wire system was built. Traditional PID control, classical sliding mode control (SMC), and conventional adaptive robust control (ARC) were selected as comparison methods. Using the target braking pressure of 4MPa commonly used in autonomous driving as the test benchmark, the actual working conditions of conventional braking on dry asphalt roads of vehicles were simulated, and comparative tests on braking pressure tracking performance were carried out. The two core indicators of each algorithm, pressure response speed and steady-state control accuracy, were verified in particular. The test results are shown in Table 1.
[0030] Table 1 Test results of pressure response speed and steady-state control accuracy of each algorithm
[0031] As shown in Table 1, traditional PID control, due to its fixed parameters, struggles to adapt to the nonlinear characteristics of linear control braking systems, resulting in a pressure build-up time of 125ms and the slowest response. While classical sliding mode control exhibits strong robustness with a build-up time of 70ms, it suffers from significant chattering. The conventional ARC algorithm optimizes nonlinear adaptability, achieving a build-up time of 65ms, a significant improvement over the former two. The algorithm of this invention achieves real-time feedforward compensation for multi-dimensional coupled disturbances by integrating a multi-disturbance observer with an improved RBF neural network. Combined with the fast tracking characteristics of the hierarchical adaptive robust control law, the pressure build-up time is reduced to only 40ms, a reduction of 68% compared to traditional PID control, 42.9% compared to classical SMC, and 38.5% compared to the conventional ARC algorithm. This significantly improves the braking pressure build-up speed, enabling rapid response to braking commands from autonomous driving systems.
[0032] In terms of steady-state accuracy, traditional PID control has a steady-state error of 0.15 MPa and an 8% overshoot, resulting in the worst control accuracy and failing to meet the requirements of high-precision braking. Classical sliding mode control is affected by chattering, with a steady-state error of 0.05 MPa and continuous small fluctuations in pressure. Although the conventional ARC algorithm suppresses chattering and reduces the steady-state error to 0.035 MPa, it still lags in compensating for multi-dimensional coupled disturbances. The algorithm of this invention completely eliminates sliding mode chattering through precise gain adjustment of the fuzzy adaptive sliding mode gain regulator. At the same time, through the coordinated adaptive adjustment of neural network weights and controller parameters, it achieves precise disturbance compensation, with a steady-state error of only 0.008 MPa. This represents a 94.7% improvement over traditional PID control, an 84% improvement over classic SMC, and a 77.1% improvement over the conventional ARC algorithm. The braking pressure can stably track the target value without overshoot or fluctuations, and the control accuracy is significantly better than the comparative methods.
[0033] Example 2 The present invention also provides an adaptive robust brake-by-wire system for implementing the above-mentioned adaptive robust brake-by-wire method, comprising a dual redundant execution module, a multi-sensor sensing module, a main control computing module, and a communication module; The dual-redundant actuator module includes two sets of identical electromagnetic direct-drive valves, brake wheel cylinders, and hydraulic lines. The two sets of electromagnetic direct-drive valves are connected in parallel to the hydraulic power source, and the two sets of brake wheel cylinders are symmetrically arranged at the vehicle's braking end. Each set of electromagnetic direct-drive valves is driven by an independent dual-pinion redundant motor. The dual-pinion redundant motor adopts dq-axis inversion adaptive control. The d-axis is designed with an inversion adaptive current controller to achieve precise current tracking, and the q-axis is designed with an inversion adaptive position controller to achieve precise control of the valve core displacement.
[0034] The multi-sensor sensing module includes a pressure sensor, a displacement sensor, a speed sensor, and a road condition sensor. The pressure sensor is installed at the brake wheel cylinder outlet, the displacement sensor and the speed sensor are installed at the valve core of the electromagnetic direct drive valve, and the road condition sensor is installed on the vehicle chassis. The main control computing module includes an embedded controller with a built-in multi-disturbance observer, a hierarchical adaptive robust control law, and a fuzzy adaptive sliding mode gain regulator; The communication module includes an onboard CAN bus and an Ethernet. The multi-sensor perception module, the main control computing module, and the dual-redundant execution module achieve real-time data interaction through the CAN bus, while the Ethernet is used for information interaction with the vehicle controller.
[0035] It is worth noting that all contents not described in detail in this invention are existing technologies and are well known to those skilled in the art.
[0036] Therefore, this invention provides an adaptive and robust brake-by-wire method and system. By integrating an improved RBF neural network with a multi-disturbance observer, it achieves accurate observation and feedforward compensation of multi-dimensional coupled disturbances. Combined with a hierarchical adaptive robust control law, it improves the response speed and control accuracy of braking pressure tracking. At the same time, a fuzzy adaptive sliding mode gain regulator is designed and a chattering suppression factor ε is introduced to optimize the sliding mode control term, effectively reducing the sliding mode chattering phenomenon at its source. This comprehensively improves the robustness, adaptability, and control stability of the brake-by-wire system under complex operating conditions.
[0037] 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 preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.
Claims
1. An adaptive and robust linear braking method, characterized in that, The specific steps are as follows: Step S1: Establish a multi-coupled dynamic mathematical model of the brake-by-wire system, introduce multi-dimensional disturbance terms to characterize the system nonlinearity, parameter uncertainty and external disturbances, and determine the system state variables and control inputs; Step S2: Construct a multi-disturbance observer that integrates and improves the RBF neural network. Design the neural network weight adaptive law and parameter adaptive law based on the Lyapunov function to observe and approximate the hydraulic coupling disturbance, mechanical friction disturbance and external road surface disturbance of the system online. Step S3: Based on backstepping sliding mode control technology and combined with the observations of multiple disturbance observers, design a hierarchical adaptive robust control law, which includes a virtual control law and an actual control law. Step S4: Design a fuzzy adaptive sliding mode gain regulator and adjust the gain coefficient of the sliding mode control online; Step S5: Based on Lyapunov stability theory, verify the asymptotic stability of the entire control system until the state variables converge to the desired value; Step S6: Convert the designed control law into an electrical signal and input it to the actuator of the brake-by-wire system to achieve closed-loop adaptive robust control of the braking pressure.
2. The adaptive robust linear braking method according to claim 1, characterized in that, In step S1, the multi-coupled dynamic mathematical model of the brake-by-wire system uses the brake wheel cylinder pressure, valve core displacement, valve core movement speed, and brake wheel cylinder piston displacement as state variables, and the electromagnetic direct drive valve input voltage as the control input. It also introduces hydraulic coupling disturbance, mechanical friction disturbance, and external road surface disturbance as disturbance terms. The formula is as follows: ； in, The cross-sectional area of the piston. P For the brake wheel cylinder pressure, Let y be the piston mass, y be the piston displacement of the brake wheel cylinder, and t be the time parameter. This is the piston's equivalent damping coefficient. For the stiffness of the piston cylinder spring, The initial length of the spring. For hydraulic coupling disturbance, Brake fluid flow rate, For the volume of the brake wheel cylinder, The bulk modulus of elasticity of the liquid. The pressure-flow coefficient is... The total mass of the electromagnetic linear actuator. The rate of change of the valve core's movement speed. Input voltage, For motor constants, The valve core movement speed, For coil resistance, c The damping coefficient is... k For spring stiffness, x For valve core displacement, For friction, For steady-state hydraulic pressure, Transient hydraulic pressure, This is due to mechanical friction disturbance. This represents the rate of change of valve core displacement. This is due to external road surface disturbance.
3. The adaptive robust linear braking method according to claim 2, characterized in that, In step S2, the multi-perturbation observer that integrates the improved RBF neural network will measure the brake wheel cylinder pressure. P Brake wheel cylinder pressure change rate Valve core displacement x Valve core movement speed As the observer's state variable, an RBF neural network with an improved Gaussian function as the activation function is used to fit the total system perturbation. The observer calculation formula is as follows: ； in, For the estimated values of the state variables, for Time derivative, This is an estimate of the rate of change of brake wheel cylinder pressure. Brake wheel cylinder pressure P The estimation error, , , This is the disturbance estimate. for The second time derivative, , , This is an estimate of the nonlinear term. The estimation error is for the rate of change of brake wheel cylinder pressure. for Time derivative, for The estimated value, For position estimation error, for Time derivative, Valve core displacement x The estimated value, For the equivalent velocity estimation error, , , , The base gain of the observer, , To adjust the gain proportionally, , , , , , , , These are system parameters.
4. The adaptive robust linear braking method according to claim 3, characterized in that, The adaptive law for the weights of the neural network is calculated as follows: ； in, , These are the estimated weights for the neural network. , They are respectively , The first derivative, , For weighted gain coefficients, , To improve the estimation of the radial basis function set in an RBF neural network, , This is the correction coefficient matrix.
5. The adaptive robust linear braking method according to claim 4, characterized in that, In step S3, the design method of the hierarchical adaptive robust control law is as follows: Step S31: Define the braking pressure tracking error target as braking pressure, and design the first-level virtual control law, as shown in the following formula: ； in, For the target of braking pressure tracking error, For the target braking pressure, This is a first-level virtual control law. Target braking pressure The derivative, It is a positive real number; Step S32: Define the virtual control error and design the second-level virtual control law, as shown in the following formula: ； in, For virtual control error, This is a second-level virtual control law. , , , These are estimated values for the system parameters. The derivative of the first-order virtual control law. It is a positive real number. This is the hydraulic state correction coefficient for the second-level virtual control law; Step S33: Define the valve core displacement tracking error and design the third-level virtual control law, as shown in the following formula: ； in, For valve core displacement tracking error, This is a third-level virtual control law. The derivative of the second-order virtual control law. It is a positive real number; Step S34: Define the valve core speed tracking error, and design the actual control law based on the disturbance estimate from the multi-disturbance observer, as shown in the following formula: ； in, Valve core speed tracking error For actual control laws, , , , , These are estimated values for the system parameters. The derivative of the third-order virtual control law. It is a positive real number.
6. The adaptive robust linear braking method according to claim 5, characterized in that, In step S4, the fuzzy adaptive sliding mode gain regulator switches manifolds using sliding mode. and its derivative As input variables, with sliding mode gain For output variables: ； Where T represents the transpose operation. They are respectively Sliding mode gain; Fuzzy inference employs a max-min synthesis method, defuzzification uses the centroid method, and the sliding mode gain adjustment value is output online. The corrected sliding mode control term is: ;in, It is a saturation function. Flutter suppression factor.
7. The adaptive robust linear braking method according to claim 6, characterized in that, In step S5, the stability verification method of the control system is as follows: Constructing Lyapunov functions: ； in, , For the weights of the neural network, , For the error in the weight estimation of the neural network, For parameter error, For parameter gain coefficient; By taking the time derivative of the Lyapunov function and combining the adaptive control laws of weights, parameters, and hierarchical adaptive robust control, we prove that: ； It satisfies the Lyapunov asymptotic stability condition; where, is the derivative of the Lyapunov function.
8. An adaptive robust brake-by-wire system for implementing the adaptive robust brake-by-wire method according to any one of claims 1-7, characterized in that, It includes a dual-redundant execution module, a multi-sensor perception module, a main control computing module, and a communication module; The dual-redundant execution module includes two sets of electromagnetic direct drive valves, brake wheel cylinders and hydraulic lines with the same structure. The two sets of electromagnetic direct drive valves are connected in parallel to the hydraulic source. The two sets of brake wheel cylinders are symmetrically arranged at the vehicle braking end. Each set of electromagnetic direct drive valves is driven by an independent dual pinion redundant motor. The multi-sensor sensing module includes a pressure sensor, a displacement sensor, a speed sensor, and a road condition sensor. The pressure sensor is installed at the brake wheel cylinder outlet, the displacement sensor and the speed sensor are installed at the valve core of the electromagnetic direct drive valve, and the road condition sensor is installed on the vehicle chassis. The main control computing module includes an embedded controller, which incorporates a multi-disturbance observer, a hierarchical adaptive robust control law, and a fuzzy adaptive sliding mode gain regulator. The communication module includes an on-board CAN bus and an Ethernet. The multi-sensor perception module, the main control computing module, and the dual-redundant execution module achieve real-time data interaction through the CAN bus, and the Ethernet is used to interact with the vehicle controller.
9. An adaptive robust brake-by-wire system according to claim 8, characterized in that, The dual pinion redundant motor adopts dq-axis inversion adaptive control. The d-axis is designed with an inversion adaptive current controller to achieve accurate current tracking, and the q-axis is designed with an inversion adaptive position controller to achieve precise control of valve core displacement.