Air suspension and active stabilizer bar cross-domain coupling control method

Abstract

The invention discloses a cross-domain coupling control method for an air suspension and an active stabilizer bar. The cross-domain coupling control method comprises the steps that a signal input layer collects and fuses data and transmits the data to a decision-making layer; the decision-making layer forms a control instruction based on a model predictive control algorithm, and comprises the steps of constructing an automobile three-degree-of-freedom coupling dynamic model, and iteratively solving a cost function minimum value in a rolling time domain to obtain the control instruction; the three-degree-of-freedom coupling dynamic model quantifies multi-dimensional motion association through a stiffness matrix off-diagonal term, can accurately predict the cross influence of air pressure adjustment of the air suspension on the roll motion, and provides a nonlinear prediction basis for model prediction control. According to the invention, by cooperating with the multi-dimensional dynamic signal and the dynamic arbitration control logic, the complex composite working condition is adapted, and the cooperative optimization of the vertical comfort, the roll stability and the pitching dynamic characteristic of the vehicle is realized.

Description

Technical Field

[0001] This invention relates to the field of vehicle control system technology, and in particular to a method for coupling control of air suspension and active stabilizer bar. Background Technology

[0002] During vehicle operation, vertical comfort (such as vibration suppression when going over speed bumps or bumpy roads), roll stability (such as body roll control during cornering), and pitch dynamics (such as nose-diving and nose-up phenomena during acceleration / braking) are key performance indicators affecting driving experience and driving safety. Air suspension optimizes vertical stiffness and height by adjusting air chamber pressure, while active stabilizer bars adjust torsional stiffness in real time via a motor / hydraulic system. Both act on the vertical and roll / pitch motion dimensions, respectively, and are the core actuators for improving vehicle dynamics.

[0003] In existing technologies, air suspension and active stabilizer bars mostly employ independent control strategies. Air suspension focuses on vertical vibration attenuation, adjusting based on signals such as vehicle vertical acceleration and suspension travel; active stabilizer bars focus on suppressing roll and pitch moments. However, actual vehicle motion is a multi-dimensional coupled dynamic process involving vertical, roll, and pitch, and independent control has the following drawbacks:

[0004] 1) Coupling characteristics are not coordinated: When cornering, if the vertical adjustment of the air suspension is not linked with the roll control of the active stabilizer bar, the vertical load transfer caused by the body roll will interfere with the optimization of suspension comfort, and vice versa; under braking pitch conditions, pitch motion and vertical vibration are superimposed on each other, and independent control is prone to causing attitude loss and comfort conflict.

[0005] 2) Isolated control logic: The lack of cross-system information interaction and dynamic arbitration may weaken the roll suppression effect of the active stabilizer bar due to the stiffness adjustment of the air suspension, or cause the actuator to move frequently due to target conflicts, increasing energy consumption and hardware wear.

[0006] 3) Performance boundary limitations: When faced with complex working conditions (such as asymmetric road surface + steering + braking combined scenarios), independent control is difficult to balance vertical comfort, roll stability and pitch dynamic characteristics. It either sacrifices comfort to ensure stability or causes rollover risk due to excessive softening of the suspension, making it impossible to achieve optimal synergy of multiple performances. Summary of the Invention

[0007] To address the shortcomings of the existing technology, this invention provides a cross-domain coupling control method for air suspension and active stabilizer bar. This method solves the problems caused by neglecting the cross-domain coupling characteristics of vertical, roll, and pitch movements when the air suspension and active stabilizer bar are controlled independently. These problems include conflicts between comfort and stability under various operating conditions, internal friction caused by isolated control logic, and limited performance boundaries in complex scenarios.

[0008] The technical solution of the present invention is as follows:

[0009] A cross-domain coupling control method for air suspension and active stabilizer bar includes the following steps:

[0010] The signal input layer collects and fuses data and transmits it to the decision layer. The data transmitted to the decision layer includes: road curvature, road surface unevenness, vehicle longitudinal acceleration, lateral acceleration, vertical acceleration, roll rate and pitch rate.

[0011] The decision-making layer generates control commands based on model predictive control algorithms, including constructing a three-degree-of-freedom coupled dynamics model of the vehicle and iteratively solving for the minimum cost function in the rolling time domain to obtain control commands; the three-degree-of-freedom coupled dynamics model of the vehicle is as follows:

[0012] ,

[0013] ,

[0014] , , ,

[0015] in, This represents the vertical displacement of the vehicle's center of gravity. Vertical velocity, It is the vertical acceleration of the vehicle body. The body roll angle, The angular velocity is the roll rate. This is the roll acceleration. The vehicle body pitch angle, The pitch angular velocity, For pitch acceleration, Here is the mass-moment-of-inertia matrix. Here is the damping matrix. Here is the stiffness matrix. To control the allocation matrix, It is an external excitation term. The damping matrix and stiffness matrix are asymmetric matrices. The off-diagonal terms of the damping matrix include vertical and roll coupling damping, vertical and pitch coupling damping, vertical force coupling damping caused by roll angular velocity, and vertical force coupling damping caused by pitch angular velocity. The off-diagonal terms of the stiffness matrix include vertical-pitch coupling stiffness, vertical-roll coupling stiffness, vertical force coupling stiffness caused by roll angle, and vertical force coupling stiffness caused by pitch angle. For control commands, This represents the change in air spring pressure. For the output torque of the active stabilizer bar, For pitch actuator control torque;

[0016] The cost function is a combination of vehicle state terms and control command terms. The vehicle state terms are a dynamically weighted sum of vehicle vertical acceleration, roll angle, pitch angle, vertical velocity, roll angular velocity, and pitch angular velocity, with the dynamic weights determined by the vehicle's operating conditions. By dynamically adjusting the priorities of objectives such as vertical comfort, roll stability, and pitch suppression, the system can achieve an optimal balance between comfort and stability under different operating conditions, adapting to all scenarios and alleviating the problems of isolated control logic and limited performance boundaries.

[0017] Furthermore, the cost function is

[0018] ,

[0019] in, To predict the time domain, To control the time domain, This represents the maximum vertical acceleration. This represents the maximum roll angle. This represents the maximum pitch angle. This represents the maximum vertical velocity. This represents the maximum roll rate. This represents the maximum pitch angular velocity. To dynamically adjust the weighting coefficients of each objective priority, This is the actuator energy consumption weight matrix. Representing the current moment, Represents an integer index variable. Represents the prediction of the future starting from the current moment. A specific point in time.

[0020] Furthermore, , , , , , , , , , , ,

[0021] in For the lateral acceleration of the vehicle, For vehicle speed, For the longitudinal acceleration of the vehicle, This represents the rate of change of air suspension pressure. This represents the maximum pressure change rate of the air suspension. This is the maximum output torque of the stabilizer bar. For the pitch coordinator torque variation rate, The maximum torque change rate of the pitch coordinator. , , , , All of these are calibration coefficients.

[0022] Furthermore, when iteratively solving for the minimum cost function in the rolling time domain, the following constraints are set: vehicle roll angle constraint, vehicle center of gravity vertical acceleration constraint, air spring pressure change constraint, air spring pressure change rate constraint, active stabilizer bar output torque constraint, and active stabilizer bar output torque change rate constraint. This ensures that the output control commands meet the physical limits and engineering safety constraints of all actuators, and can be directly and safely applied to the actuators. It prevents actuator overshoot or vehicle instability, ensures the safe and reliable operation of the system, avoids problems such as decreased comfort and increased stability risks caused by control commands exceeding constraints, and further addresses the performance boundary limitation problem.

[0023] Furthermore, feedforward compensation is set for the control commands, including road surface preview compensation. = Curveside tilt feed And pitch feedforward for braking / acceleration ,in, For calibration coefficients, To anticipate the future The road surface unevenness at all times, t Indicates the current moment. The moment of inertia of the vehicle body during roll. For vehicle speed, To predict the road curvature, r is the tire radius. For the sprung mass, For longitudinal acceleration, It is the height of the center of mass.

[0024] Furthermore, when the signal input layer collects and fuses data for transmission to the decision layer, it collects the original road curvature and vehicle speed data to construct road curvature observations, and performs Kalman filtering estimation to obtain the road curvature transmitted to the decision layer. The road curvature observations are: , The original curvature of the road, It is lateral acceleration. For vehicle speed, This involves dynamic weighting based on visibility. The scheme achieves accurate curvature estimation across all operating conditions through dynamic weight allocation and signal decoupling, overcoming the inherent limitations of single-sensor vision and IMU methods in curvature detection.

[0025] Furthermore, when the signal input layer collects and fuses data to the decision layer, it constructs roll angle observations from the collected roll angular velocity and the predicted roll angle output by the three-degree-of-freedom coupled dynamics model of the vehicle, and performs Kalman filtering estimation to obtain the roll angle transmitted to the decision layer. The roll angle observations are as follows: , The roll angle is calculated from the collected roll angular velocity. It predicts the roll angle. These are fixed weighting coefficients.

[0026] Furthermore, the actuator layer converts the control commands into physical quantities that can be directly controlled by the underlying actuators through an optimized objective function, and then converts them into drive signals for each actuator for control.

[0027] The physical quantities that can be directly controlled at the underlying level of each actuator are:

[0028] ,

[0029] in, The values ​​represent the air pressure changes of the four air springs in the suspension. FL indicates the front left, FR indicates the front right, RL indicates the rear left, and RR indicates the rear right. For controlling the current of the active stabilizer bar motor, For the pitch hydraulic actuator pressure difference,

[0030] The objective function is:

[0031] ,

[0032] in, Let B be the adaptive weight matrix for the operating conditions, and let B be the linear mapping matrix constructed based on the actuator dynamics model. This is a matrix used to punish excessive actions by the actuator. For smoothing matrices, Control commands output by the decision-making level The previous cycle instruction for a directly controllable physical quantity.

[0033] Furthermore, , , , , The weights for vertical stiffness control. The vertical impact frequency, For vertical basic weights, For frequency sensitivity coefficient, It is the vehicle's vertical resonant frequency. As the basic weight for lateral tilt, This is the roll sensitivity coefficient. For the safe roll strength threshold, This is the tilt strength coefficient. The deceleration coefficient, For the safe deceleration threshold, For deceleration sensitivity coefficient, For pitch foundation weights;

[0034] , These are the weighting coefficients for the air pressure control quantities of the front and rear axles, respectively, satisfying... , This is the weighting coefficient for the current control quantity of the active stabilizer bar motor. Weighting coefficients for pressure control quantities of hydraulic actuators;

[0035] , The smoothing weighting coefficient for air suspension air pressure changes is given by vehicle speed. A monotonically increasing function. The smoothing weighting coefficient for the change in current of the active stabilizer bar motor is given by the vehicle speed. A monotonically increasing function. It is the smoothing weighting coefficient for the pressure change of the hydraulic actuator, and is the vehicle speed. It is a monotonically increasing function.

[0036] Compared with the prior art, the present invention has the following advantages:

[0037] 1. Cross-domain collaborative control was achieved: By using a three-degree-of-freedom coupled model and MPC algorithm, the target conflict problem when the air suspension and stabilizer bar are controlled independently was fundamentally solved, and the collaborative optimization of vertical, roll and pitch motions was achieved.

[0038] 2. Improved system adaptability: The system adopts a dynamic arbitration mechanism, which can intelligently adjust the weight of control targets based on real-time operating conditions such as vehicle speed and acceleration, so that the system can achieve the best balance between comfort and stability and adapt to all operating conditions.

[0039] 3. Enhanced anti-interference and forward-looking capabilities: By utilizing visual preview information to provide feedforward compensation and combining it with feedback control, the system's ability to suppress predictable disturbances such as road surface undulations, curves, braking / acceleration is significantly improved, and response delay is reduced.

[0040] 4. Ensuring the robustness and reliability of the control system: A detailed dynamic model, real-time constraint processing, and fault tolerance strategy were designed at the actuator level to ensure the physical realizability of control commands and maintain the basic stability of the vehicle through redundant control when some components fail.

[0041] In summary, this invention achieves optimal synergy between vehicle vertical comfort, roll stability, and pitch dynamic characteristics by coordinating multi-dimensional dynamic signals, dynamic arbitration control logic, and adapting to complex composite working conditions. This breaks through the performance bottleneck of traditional independent control and improves the vehicle's driving quality and system energy efficiency in all scenarios. Attached Figure Description

[0042] Figure 1 The figure shows the simulation results of the vertical displacement response of the cross-domain coupling control method of air suspension and active stabilizer bar in the embodiment.

[0043] Figure 2 The figure shows the simulation results of the roll angle response of the cross-domain coupling control method of air suspension and active stabilizer bar in the example.

[0044] Figure 3 The figure shows the simulation results of the pitch angle response of the cross-domain coupling control method of air suspension and active stabilizer bar in the example.

[0045] Figure 4 The figure shows the simulation results of the vertical acceleration response of the cross-domain coupling control method of air suspension and active stabilizer bar in the embodiment. Detailed Implementation

[0046] The present invention will be further described below with reference to embodiments, but these are not intended to limit the scope of the invention.

[0047] The cross-domain coupling control method for air suspension and active stabilizer bars in this embodiment mainly includes: a signal input layer acquiring and fusing data and transmitting it to a decision layer; the decision layer generating control commands based on a model predictive control algorithm; and an actuator layer converting the control commands into physical quantities that can be directly controlled by the underlying actuators through an optimized objective function, and then converting them into drive signals for each actuator for control. The processing procedures of each layer are described in detail below.

[0048] The signal input layer includes a sensor layer and a data fusion layer.

[0049] The sensor layer includes a binocular vision system, a 6-axis IMU (Inertial Measurement Unit), a stabilizer bar torque sensor, an air pressure sensor, wheel speed sensors, and a steering wheel angle sensor. The data measured by the sensor layer is transmitted as input signals to the decision layer.

[0050] Binocular vision systems are primarily used to measure the curvature of the road in front of a vehicle. and road surface unevenness This vision system adjusts its preview distance and preview time according to changes in vehicle speed. Therefore, its main function is to provide preview information to the system, enabling proactive control and optimization of the vehicle's dynamic state.

[0051] The 6-axis IMU primarily measures the vehicle's longitudinal, lateral, and vertical acceleration, as well as its roll and pitch angular velocities. The data obtained from the 6-axis IMU is transmitted to the decision-making level as the vehicle's driving status.

[0052] Stabilizer bar torque sensor is mainly used to measure the actual output torque of active stabilizer bars.

[0053] An air pressure sensor is mainly used to measure the pressure in the air chamber of an air spring.

[0054] Wheel speed sensors monitor the rotational speed of each wheel in real time.

[0055] The data fusion layer, implemented using the Kalman filter algorithm, outputs the roll angle. Road curvature Decision parameters, etc.

[0056] 1) A method for fusing road curvature κ.

[0057] To address the inherent limitations of single-sensor systems like vision and IMU in curvature detection, this system employs a hierarchical confidence fusion strategy. Through dynamic weight allocation and signal decoupling, it achieves accurate curvature estimation across all operating conditions. Specifically, it includes the following steps:

[0058] Step 1: Receive the original curvature of the binocular vision system Compared with the raw lateral acceleration signal from the IMU;

[0059] Step 2: Design a dynamic weight allocation mechanism:

[0060] ,

[0061] This represents the weights of the original curvature signal in a binocular vision system. When visibility is < 50m, the vision sensor is highly susceptible to environmental interference. Pick Under other operating conditions, the vision sensor is reliable. Pick The above range was calibrated through multi-condition testing, covering a reasonable parameter range for similar scenarios.

[0062] Step 3: Constructing a composite observation set:

[0063] ,

[0064] In the formula, The measured value is for curvature blending. Lateral acceleration signal (m / s) output by the IMU 2 ), This is the vehicle speed value (m / s) resulting from the wheel speed combination. The formula for calculating vehicle speed is shown below:

[0065]

[0066] In the formula, R represents the effective diameter of the tire (m). This represents the rotational speed (rad / s) of each tire.

[0067] Step 4: Optimization of the Kalman filter. This includes the following steps:

[0068] (1) State prediction: ( =1, which means a first-order uniform velocity model is used.

[0069] (2) Covariance prediction and Kalman gain calculation:

[0070] , ,

[0071] In the formula, Let be the error covariance matrix. Kalman gain matrix, For the observation matrix ( ), The covariance matrix of the process noise represents the abruptness of changes in road curvature and can be calibrated experimentally. The value range is [0.01, 0.05]. To observe the noise covariance matrix and characterize the reliability difference between visual curvature and IMU dynamic curvature, it can be jointly calibrated through dual lane-change simulation and real-vehicle step input testing, with the value range being []. , ].

[0072] (3) State update and covariance matrix update.

[0073] , .

[0074] (4) Boundary constraints: ,

[0075] In the formula, The road surface adhesion coefficient is the preset or real-time estimated value, which is taken here. It can guarantee more than 95% of the working conditions of the road and has reserved a safety boundary (the good road adhesion coefficient can reach 0.7-0.8, take the lower limit). This is the acceleration due to gravity.

[0076] 2) Roll angle The observation fusion method.

[0077] To address the issues of IMU angular velocity integral drift and model prediction lag, a pre-fused Kalman filter architecture based on fixed weights is proposed. Specifically, it includes the following steps:

[0078] Step 1: Observation construction.

[0079] ,

[0080] In the formula, It is the roll angle obtained by integrating the IMU roll rate, i.e. , It is the roll rate measured by the IMU (in practical applications, discrete equations can be used). The roll angle is predicted based on a three-degree-of-freedom vehicle model (which will be established in the next section). These are the weighting coefficients. Value range: 0.7 0.9 is an empirically calibrated weight, which tends to favor the IMU, which has good real-time performance but is prone to drift, and supplements it with the predictions of the model, which have strong physical constraints but rely on them. The weight can be adjusted according to the actual working conditions.

[0081] Step 2: Optimize the state estimation using a Kalman filter.

[0082] (1) State prediction: ( ),

[0083] in, for Prior roll angle at all times The roll angle is predicted based on the vehicle dynamics model mentioned above.

[0084] (2) Covariance prediction and Kalman gain calculation:

[0085] ,

[0086] ,

[0087] In the formula, The process noise covariance, which characterizes the unpredictable disturbances in vehicle body roll motion, needs to be calibrated experimentally. The value range is [0.02, 0.06]. It is the observation matrix, also . To observe the noise covariance matrix and characterize the difference in reliability between the IMU integral drift and the vehicle model prediction, calibration can be performed experimentally, with values ​​ranging from [...]. , ].

[0088] (3) Posterior state update and posterior covariance update:

[0089] , ,

[0090] The decision-making layer includes the vehicle's three-degree-of-freedom coupled model, dynamic arbitration mechanism, MPC controller, and forward-looking collaborative decision-making.

[0091] This three-degree-of-freedom coupled vehicle model serves as the prediction model for the MPC algorithm, accurately predicting the vehicle's state in the rolling time domain and providing a foundation for optimization solutions. This model quantifies multi-dimensional motion relationships through off-diagonal terms of the stiffness matrix, providing a nonlinear prediction basis for subsequent MPC (Model Predictive Control). Unlike traditional linearized models, this model can accurately predict the cross-effects of air suspension pressure adjustments on roll motion, making the MPC optimization sequence more closely reflect the vehicle's actual dynamic characteristics. The specific steps are as follows:

[0092] 1) Definition of system state vector.

[0093] Construct a three-degree-of-freedom coupled model and select the vertical displacement of the vehicle's center of gravity. Body roll angle Vehicle pitch angle As core motion parameters, the system state vector and its derivative are defined as follows:

[0094] , , ,

[0095] in, It reflects the compression or extension state of the vehicle's suspension; Characterizes the degree of roll of a vehicle about its longitudinal axis when it is turning; It reflects the pitch changes around the lateral axis when the vehicle accelerates or brakes.

[0096] 2) Establish a three-degree-of-freedom coupled dynamic model of the vehicle.

[0097] Traditional solutions decouple vertical (air suspension) and roll / pitch (stabilizer bar) control, neglecting the mechanical relationship between the three (e.g., vertical load transfer during roll can react on suspension stiffness). This equation uses state vectors... A unified description of three-degree-of-freedom motion is provided, and the "vertical-roll coupling" (the relationship between vertical force and roll angle caused by the difference in left and right suspension stiffness) is quantified using off-diagonal terms of the stiffness matrix, revealing the root cause of cross-domain dynamic conflicts at the model level. The three-degree-of-freedom coupled dynamics model of the vehicle is as follows:

[0098] ,

[0099] ,

[0100] In the formula, Representing the change in air spring pressure (kPa), it is the dynamic adjustment of the gas pressure inside the air spring cavity, used to adjust the vertical stiffness of the suspension and the vehicle height (m). It is the output torque (N·m) of the active stabilizer bar that directly suppresses the body roll angle and reduces the risk of body roll in corners; It is the pitch actuator control torque (N·m), which suppresses pitch motion during acceleration / braking. Here is the mass-moment-of-inertia matrix. Here is the damping matrix. Here is the stiffness matrix. To control the allocation matrix, It is an external incentive.

[0101] , , ,

[0102] , ,

[0103] In the formula, The load is the sprung mass (kg). The moment of inertia is the tilting motion (kg·m²). The pitch moment of inertia is (kg·m²).

[0104] Stiffness matrix The off-diagonal terms in the equation demonstrate the asymmetry of coupling. Among them, Represents the total vertical stiffness (N / m). It is the vertical-pitch coupling stiffness (N / rad); It is the vertical-tilt coupling stiffness (N / rad); It is the vertical force coupling stiffness (N / rad) caused by the roll angle. It is the vertical force coupling stiffness (N / rad) caused by the pitch angle. Equivalent roll stiffness (N·m / rad). It is the equivalent pitch stiffness (N·m / rad). and This characterizes the core capability of the suspension system to generate anti-roll and anti-pitch moments, which is crucial for coupled control; therefore, it is retained and accurately modeled. The item... and This characterizes the secondary feedback effect of vehicle body attitude on vertical force. Its magnitude is much smaller than that of the main coupling term and the control target, so it can be approximately ignored while ensuring accuracy. , This simplifies controller design.

[0105] Similarly, the damping matrix The off-diagonal terms also follow asymmetric coupling. Among them, It is vertical damping (N·s / m), It is the roll damping (N·m·s / rad), which describes the ratio of damping torque to angular velocity during roll motion. It is pitch damping (N·m·s / rad), which describes the ratio of damping torque to angular velocity during pitch motion. It is the coupling damping coefficient between vertical and lateral directions; It is the coupling damping coefficient between vertical and pitch directions. It is the vertical force coupled damping (N·s / rad) caused by the roll angular velocity. This is the vertical force coupled damping (N·s / rad) caused by the pitch angular velocity. Based on the same simplification principle as the stiffness matrix, the coupled damping term... and It can also be ignored (i.e.) , ), while retaining the core items and .

[0106] To stabilize the lever arm (m), The lever arm (m) of the pitch actuator.

[0107] In the expression, Represents the vertical excitation force (N) for road surface unevenness. Represents the vehicle's center of gravity height (m). This represents the excitation torque for road surface tilting (N·m). The pitching excitation torque is (N·m).

[0108] ,

[0109] ,

[0110] ,

[0111] It is the vertical unevenness input of the road surface. It is lateral acceleration (m / s²) 2 ), It is longitudinal acceleration (m / s²) 2 ), It is the pitch control efficiency coefficient. It is the vertical force (N) of the front axle suspension. It is the vertical force (N) of the rear axle suspension. It is the effective area (m²) of the air spring. It is the length of the stabilizer lever arm (m). For the pitch lever ratio, and .

[0112] ,

[0113] ,

[0114] ,

[0115] ,

[0116] ,

[0117] ,

[0118] ,

[0119] ,

[0120] ,

[0121] It is the vertical critical damping ratio (dimensionless), which needs to be calibrated experimentally. In engineering, it is taken as [0.2, 0.4]. It is the critical roll damping ratio (dimensionless), which needs to be calibrated experimentally. In engineering, it is taken as [0.1, 0.3]. This is the critical pitch damping ratio, which needs to be calibrated experimentally. Generally... Take [0.15, 0.35]. It is the additional roll stiffness coefficient (N·m / rad) provided by the active stabilizer bar. It is the vertical stiffness (N·m) of the left front axle suspension. It is the vertical stiffness (N·m) of the right front axle suspension. It is the vertical stiffness (N·m) of the left rear axle suspension. It is the vertical stiffness (N·m) of the right rear axle suspension. It is the front axle track (between the centers of the left and right wheels) (m). It is the rear axle track (m). It is the horizontal distance (m) from the front axle to the center of gravity. It is the horizontal distance (m) from the rear axle to the center of mass. It is the moment of inertia of the sprung mass of the vehicle about the X-axis (the longitudinal axis of the vehicle) (kg / m). 2 ), It is the moment of inertia of the sprung mass of the vehicle about the Y-axis (the vehicle's transverse axis) (kg / m). 2 ), These are the damping coefficients (N·s / m) of the left and right front axle wheel suspensions, respectively. These are the damping coefficients (N·s / m) of the left and right rear axle wheel suspensions, respectively.

[0122] The dynamic arbitration mechanism, by establishing a unified mathematical model, transforms multiple objectives such as vertical comfort, lateral handling, and tire load stability into quantifiable cost functions, and dynamically adjusts the priority of objectives based on real-time operating conditions to achieve a dynamic and coordinated control paradigm.

[0123] Under different operating conditions, vehicles exhibit varying degrees of focus on vertical comfort, roll stability, and pitch suppression. Traditional control methods struggle to dynamically adjust target priorities, leading to inconsistent control performance or wasted resources. This invention designs a dynamic arbitration mechanism to drive the controller to adjust target weights in real time, thereby improving control adaptability.

[0124] The dynamic arbitration module receives vehicle status (vehicle speed, lateral acceleration, longitudinal acceleration) in real time, generates a state weight vector, and a cost function. The following discrete form is adopted:

[0125] ,

[0126] In the formula, To predict the time domain, To control the time domain. It is the vertical acceleration of the vehicle body. The maximum vertical acceleration is taken as 1.5 m / s². 2 ; The roll angle reflects the degree of body roll in a curve. To determine the maximum roll angle, take... =5°; The pitch angle represents the degree to which a car pitches up during acceleration or down during deceleration. To determine the maximum pitch angle, take... =4°; Vertical velocity, To determine the maximum vertical velocity, take... =0.5m / s; The angular velocity is the roll rate. This represents the maximum roll angular velocity. Pick ; The pitch angular velocity, This represents the maximum pitch angular velocity. Pick ; The aforementioned control inputs correspond to changes in air suspension pressure, stabilizer bar torque, and pitch control torque, respectively. To dynamically adjust the weighting coefficients of each objective priority, This is the actuator energy consumption weight matrix, used to suppress invalid actions. Representing the current moment, Represents an integer index variable. Represents the prediction of the future starting from the current moment. A specific point in time.

[0127] The weight values ​​adopt an adaptive mechanism:

[0128] This formula represents With lateral acceleration Increase and decrease (temporarily reduce comfort priority during aggressive driving).

[0129] This formula represents With vehicle speed Increased size leads to increased body roll (high-speed curves require enhanced roll suppression).

[0130] In the formula, For the longitudinal acceleration of the vehicle, The calibration coefficient, verified through simulation, has a value range of 3 to 5. This formula represents... The larger, The closer to 1 (pitch priority); when driving smoothly, Release the weight to other targets.

[0131] Characterized by weighting the square of the vertical velocity. The vertical velocity weighting coefficient is used to optimize suspension travel and tire contact with the ground.

[0132] It is a weighted average of the squares of the roll velocity. This is a weighting coefficient for rollover speed, reducing the risk of rollover.

[0133] It is a squared weighted average of pitch speeds. This is a pitch velocity weighting coefficient to improve attitude stability.

[0134] The value of can be determined by using existing position weights and defining the velocity weight through a proportional coefficient.

[0135] , , Here The design incorporates an adaptive coefficient that dynamically adjusts according to operating conditions, namely: This indicates that the speed should be increased when the speed is high. To strengthen speed constraints.

[0136] Energy consumption weight The design must be tailored to the physical characteristics of the actuators (air suspension, stabilizer bar motors, and pitch coordinators have significantly different energy consumption modes). The core design principle is to prioritize low-power control, penalize ineffective actions, and achieve a balance between control performance and energy consumption. Specific rules include:

[0137] 1) Energy consumption weight of air suspension (Pressure regulation).

[0138] Air compressor power consumption and pressure change rate Strong correlation (high-frequency / high-pressure adjustments will drastically increase energy consumption), therefore the design is as follows:

[0139] ,

[0140] In the formula, The maximum pressure change rate of the air suspension is determined by the air suspension hardware and can be calibrated experimentally. , The calibration coefficients, verified through simulation, have a value range of [value range missing]. Take 0.8, Take 0.2.

[0141] 2) Stabilizer bar motor energy consumption weight (Torque output).

[0142] Stabilizer motor power consumption and the square of torque Positive correlation (the greater the torque, the higher the heat generation / energy consumption), therefore the design... :

[0143] ,

[0144] In the formula, The maximum output torque of the stabilizer bar can be determined experimentally from the stabilizer bar's technical parameters. This is a calibration coefficient, which can be adjusted based on the energy consumption-performance balance. =1.

[0145] 3) Pitch coordinator energy consumption weight (Torque output).

[0146] Pitch coordinators (such as active suspension cylinders and torque motors) have high energy consumption and torque variation rate. Related (high-frequency torque adjustment will increase hydraulic or motor losses), therefore the design... :

[0147] ,

[0148] In the formula, The maximum torque change rate of the pitch coordinator is determined by the system hardware and can be calibrated experimentally. These are calibration coefficients, primarily balancing energy consumption and response speed. They can be optimized through simulation or experimentation. Take 0.6.

[0149] The MPC-based rolling optimization strategy relies on a three-degree-of-freedom coupled model to accurately predict the dynamic coupling relationship of the vehicle's vertical-roll-pitch motion. It integrates the adaptive weights of the operating conditions output by the dynamic arbitration mechanism and iteratively solves the minimum cost function in the rolling time domain to achieve coordinated optimization of air suspension air pressure regulation and active stabilizer rod torque distribution. Through closed-loop correction of feedforward preview information and feedback state error, the control strategy can both anticipate road disturbances and adapt to changes in vehicle dynamic characteristics in real time, breaking through the limitations of traditional control in adapting to multi-actuator coupling conflicts and dynamic switching of operating conditions.

[0150] The execution flow of the MPC-based rolling optimization strategy is as follows:

[0151] Step 1: Discretization and prediction initialization of the three-degree-of-freedom coupled model.

[0152] 1) Establish a continuous-time model.

[0153] First, define the extension variables. The three-degree-of-freedom coupled dynamics equations of the vehicle Establish a continuous-time state-space model:

[0154] ,

[0155] in, For the system matrix, To control the input matrix, This is the interference vector.

[0156] , ,

[0157] , ,

[0158] ,

[0159] , .

[0160] 2) The continuous-time model is converted into a discrete state-space form using the Euler discretization method.

[0161] ,

[0162] It is a discretized state vector (containing displacement and angular velocity information). Let be the system state transition matrix, and , The sampling period is taken here. . To predict noise for the model.

[0163] ,

[0164] .

[0165] 3) Substitute the above cost function again.

[0166] ,

[0167] The vertical acceleration in the cost function is processed using velocity difference:

[0168] .

[0169] 4) Rolling time domain optimization solution (generation of optimal control quantity).

[0170] (1) Setting constraints.

[0171] To prevent actuator overshoot or vehicle instability, the following constraints are set:

[0172] State constraint: roll angle (Preventing rollover); Vertical acceleration (Comfort threshold);

[0173] Control constraints: Air suspension air pressure changes (Actuator physical limits); stabilizer bar torque (Motor power limitation);

[0174] Rate constraint: Rate of change of air pressure (Avoid suspension impact); Torque variation rate (Motor response limitation).

[0175] (2) Solve for the optimal control sequence using quadratic programming (QP).

[0176] make To predict state variables, construct the state sequence and control sequence in the prediction time domain.

[0177] Predicted state sequence: .

[0178] Control sequence: .

[0179] in, As before, .

[0180] Prediction equations are developed recursively:

[0181] ,

[0182] ,

[0183] ,

[0184] The cost function can then be rearranged as follows:

[0185] ,

[0186] In the formula, the weight matrix =blkdiag( () is the state weight matrix in the prediction time domain, enabling real-time switching of multi-objective priorities, where, Includes dynamic weights (updated in real time by the arbitration mechanism). =blkdiag( ) is the diagonal matrix of the control weight blocks. Same as above.

[0187] Feedforward compensation integration.

[0188] Pre-aiming collaborative decision-making utilizes pre-aiming information (road curvature, unevenness) to adjust control parameters in advance. The output of the pre-aiming collaborative decision-making is used as a known feedforward term to counteract disturbances in the prediction equation.

[0189] Traditional technical solutions rely on real-time feedback (e.g., adjusting stabilizer rod torque only after the roll angle exceeds a threshold), failing to anticipate predictable disturbances (e.g., curves ahead, bumpy roads), resulting in response delays (control actions lag behind the arrival of the disturbance). In complex operating conditions, independent control struggles to balance multiple objectives (comfort, stability, safety), often leading to frequent actuator movements or performance trade-offs (e.g., sacrificing comfort for stability) due to priority conflicts. Therefore, a pre-aiming cooperative technology was designed.

[0190] By sensing future road disturbances through binocular vision and vehicle dynamics, feedforward control parameters are generated to proactively counteract the dynamic effects of road surface, curves, and braking. Specifically, this includes:

[0191] 1) Road surface anticipation compensation (vertical comfort).

[0192] = ,

[0193] In the formula, This is the feedforward compensation term for the change in air spring pressure. The calibration coefficients need to be calibrated on a real vehicle, and the values ​​are (50~100 kPa / m). Predicting the future with binocular vision The road surface unevenness (in meters) at a given time is processed using a low-pass filter. The significance of this formula is to adjust the air suspension pressure in advance to counteract the impending road impact and reduce vertical acceleration. .

[0194] 2) Cornering roll feedforward (roll stability).

[0195] ,

[0196] In the formula, It is the feedforward compensation term for the output torque of the active stabilizer bar; The moment of inertia of vehicle body roll ( (This is determined by the three-degree-of-freedom model in the following section). For vehicle speed, denoted as the pre-aimed road curvature; r is the tire radius.

[0197] 3) Pitch feedforward for braking / acceleration (pitch dynamic characteristics).

[0198] ,

[0199] In the formula, For pitch feedforward torque compensation term during braking / acceleration; For the sprung mass, This is longitudinal acceleration (sensed in real time by the IMU); It is the height of the center of mass (m).

[0200] The actuator layer is the key link that transforms the control commands output by the decision layer into physical actions. It includes the following core modules: actuator dynamic characteristic modeling, control quantity allocation algorithm, real-time constraint processing, actuator interface conversion module, and fault tolerance strategy.

[0201] The dynamic characteristic modeling of actuators mainly involves clarifying the response delay, physical constraints, and nonlinear characteristics of each actuator. Through the aforementioned three-degree-of-freedom coupled model, the control requirements of the vehicle's roll, vertical, and pitch movements on the suspension / stabilizer actuators are clarified (e.g., vertical stiffness needs to be dynamically adjusted with the roll angle). The following section models the air suspension actuators, active stabilizer bar actuators, and pitch control actuators, establishing their physical characteristic models to provide a basis for the physical implementation of control commands.

[0202] 1. Modeling the dynamic characteristics of the actuator.

[0203] 1) Modeling of air suspension actuators.

[0204] (1) Vertical stiffness model.

[0205] Air springs are typical nonlinear spring structures with a total vertical stiffness of [missing information]. It can be calculated using the following formula:

[0206] ,

[0207] This represents the i-th air spring at the initial air pressure The underlying gas stiffness, The basic mechanical stiffness of the rubber base and sealing structure representing the i-th air spring. This represents the change in air pressure of the i-th air spring. Represents real-time air pressure. For the gas polytropic index, take [value], here we take [value]. =1.3. These parameters of the air spring can be calibrated in the laboratory.

[0208] (2) Dynamic response model of air pressure.

[0209] The rate of change of air pressure is limited by the power of the air compressor and exhibits a first-order hysteresis characteristic:

[0210] ,

[0211] In the formula, The change in air pressure of the i-th air spring (kPa) is the difference between the target air pressure output by the decision layer and the initial air pressure. The response time constant is typically set to 0.5s to 1.5s (larger values ​​for small-flow compressors and smaller values ​​for large-flow compressors). This is the gain coefficient, which is experimentally calibrated to be 200~500kPa / V. The control voltage (V) for the actuator drive circuit, ranging from 0 to 5V, is derived from the target air pressure change obtained through optimization by the decision-making layer MPC.

[0212] (3) Physical constraints

[0213] Pressure range: [-0.5 bar, 0.5 bar], to prevent overpressure rupture or negative pressure collapse of the air chamber;

[0214] Rate of change of air pressure: [-0.3bar / s, 0.5bar / s]) to prevent suspension impact or compressor overload.

[0215] 2) Modeling of active stabilizer bar actuator.

[0216] The active stabilizer bar is driven by a permanent magnet synchronous motor and a reduction gear mechanism. It suppresses body roll by outputting torsional torque, and its core is torque transmission and dynamic response modeling.

[0217] (1) Torque output model.

[0218] The relationship between the actual output torque of the stabilizer bar and the motor control current is as follows, taking into account transmission efficiency and friction torque:

[0219] ,

[0220] In the formula, To ensure the transmission efficiency of the reduction mechanism, considering the friction between gear meshing and bearings, a value of 0.85~0.95 is adopted; The reduction ratio is determined by the mechanical structure and is usually taken as 15~30 (to convert the high speed and low torque of the motor into the low speed and high torque of the stabilizer bar). The torque constant of the motor (N·m / A) is determined by the motor parameters and is typically taken as 2.5~4.0 N·m / A. The motor control current command (A) is determined by the target tilt torque from the decision-making level. Reverse push ( ); The torsional friction torque (N·m) of the stabilizer bar is generated by the resistance of the bushing and bearing, and is taken as 5~15N·m; The sign for the roll rate direction (measured by a 6-axis IMU) is used to ensure that the frictional torque is opposite to the torsional direction.

[0221] (2) Dynamic response model.

[0222] The inductance of the motor and the inertia of the reduction gear cause a first-order lag in torque output.

[0223] ,

[0224] In the formula, The response time constant (s) is affected by the motor speed and reduction ratio, and is taken as 0.2~0.8s; The target roll suppression torque (N·m) output by the decision-making level MPC must meet the following requirements. [−800N m, 800N m] (motor power limit).

[0225] (3) Nonlinear correction.

[0226] Motor current saturation: hour, ( The maximum allowable current for the motor is set to 11A to ensure... 800 N·m);

[0227] Torsion angle limitation: When the stabilizer bar torsional angle At that time, the torque output saturates, that is... , , which is the limit of the mechanical structure.

[0228] 3) Modeling of pitch control actuators.

[0229] The pitch control actuator uses a hydraulic actuator, which generates pitch suppression torque through telescopic movement, adapting to acceleration pitching / braking pitching conditions.

[0230] (1) Torque output model.

[0231] Pitching moment is directly related to the pressure difference in the hydraulic system and needs to be considered in conjunction with the lever arm at the installation position.

[0232] ,

[0233] In the formula, The actuator lever arm (m) is the horizontal distance from the actuator mounting point to the vehicle's center of gravity, geometrically measured to be 0.8~1.2m (in the three-degree-of-freedom coupled model). Consistent); The effective area of ​​the actuator piston (m²) is determined by the actuator specifications and is typically 0.01~0.02 m². The pressure difference (Pa) between the rodless and rod-side chambers of the actuator is regulated by a hydraulic valve control command. Maximum pressure difference. Pa (hydraulic system rated value).

[0234] (2) Dynamic response model.

[0235] The compressibility of the hydraulic fluid and the flow delay at the valve orifice cause a lag in the pressure differential response, which conforms to first-order dynamic characteristics.

[0236] ,

[0237] In the formula, The response time constant (s) is affected by the oil viscosity and pipeline length, and is taken as 0.3~1.0s (viscosity is higher at low temperatures). Increased size can compensate for oil temperature: , C, This is the value measured by the oil temperature sensor. For the hydraulic system gain (Pa / V), take 5 × 10⁻⁶. 6 ~1×10 7 Pa / V; This refers to the hydraulic valve control voltage command (V), ranging from 0 to 5V, corresponding to a pressure difference of 0~ .

[0238] (3) Physical constraints.

[0239] Maximum pitch moment: [-1000 N·m, 1000 N·m]);

[0240] Flow limit: Maximum flow rate of hydraulic pump The extension / retraction speed of the actuator is determined by (L / min), and must meet the following requirements. , The actuator extension / retraction speed (m / s) is used to avoid response delays caused by insufficient flow.

[0241] 2. Control quantity allocation algorithm.

[0242] The control quantity allocation algorithm decomposes the generalized virtual control quantity (total demand) calculated by the decision layer into specific physical instructions for each actuator through real-time optimization of the allocation framework, taking into full account the dynamic characteristics, physical limits and multi-objective priorities of each actuator.

[0243] 1) Optimize variable definitions.

[0244] The optimization variables are the physical quantities that can be directly controlled at the underlying level of each actuator:

[0245] ,

[0246] In the formula, The air pressure change (kPa) of the four air springs in the suspension is given by the constraint. [-0.5bar, 0.5bar], FL indicates left front, FR indicates right front, RL indicates left rear, RR indicates right rear; The control current (A) for the active stabilizer bar motor is constrained as follows: [0, 11A]; The pressure difference (Pa) of the pitch hydraulic actuator is constrained as follows: Pa.

[0247] 2) Control efficiency matrix (B).

[0248] Based on the actuator dynamics model, construct the linear mapping matrix B:

[0249] ,

[0250] In the formula, The initial vertical stiffness (N / m) of the front left air spring. The initial vertical stiffness (N / m) of the front right air spring. The initial vertical stiffness of the left rear air spring is (N / m), and the initial vertical stiffness of the right rear air spring is (N / m). The meaning is the same as before; it is the gas polyvariance index. This is the reference air pressure for the left front air spring. This is the reference air pressure for the right front air spring. This is the reference air pressure for the left rear air spring. This is the reference air pressure for the right rear air spring; all four parameters are set to [specific values]. (Initial air pressure value). , These are the front and rear wheel track widths (m). , These are the distances (m) from the front and rear axles to the vehicle's center of gravity, respectively. For the effective area of ​​the piston, As the lever arm, This is the motor torque constant.

[0251] 2) Objective function design.

[0252] Design a weighted quadratic objective function to achieve multi-objective optimization:

[0253] ,

[0254] The formula contains three sub-terms:

[0255] (1) Tracking error term middle, The total demand vector for the decision-making level (vertical stiffness increment, roll torque, pitch moment); This is the adaptive weight matrix for operating conditions (diagonal matrix). .in, , , . This is the weight (priority) for vertical stiffness control. The larger the value, the more preferentially the algorithm adjusts the vertical stiffness. This refers to the vertical impact frequency (such as the frequency of vertical vibration of the vehicle body when going over a speed bump). The vertical base weight (an empirical value that determines the basic priority of vertical control). This is the frequency sensitivity coefficient (which determines the strength of the weight's response to frequency changes). The larger the value, the faster the weight decreases when the frequency deviates from the resonance point. It is the vehicle's vertical resonant frequency (usually 1.5~2.5Hz), an inherent characteristic of the vehicle, determined experimentally. The basic weight of roll (an empirical value that determines the basic priority of roll control). The roll sensitivity coefficient (calibrated experimentally). The safe roll strength threshold (an empirical value, which can be taken as 0.2). This is the tilt strength coefficient. ( Vehicle speed (m / s) The turning radius is (m). The acceleration due to gravity is m / s² 2 ), The deceleration coefficient, For the safe deceleration threshold, This is the deceleration sensitivity coefficient (the value needs to be experimentally calibrated). The pitch base weight (an empirical value that determines the base priority of pitch control).

[0256] (2) Control cost item It is used to punish excessive actuator movement, reduce energy consumption and wear. It is a diagonal matrix, and the element values ​​are based on the energy consumption characteristics of the actuator; In the formula, These are the weighting coefficients for the air pressure control quantities of the front and rear axles, representing the unit air pressure change cost of the front and rear axle air springs, and satisfying the following conditions: This is to reflect the energy consumption difference between the front and rear axle actuators. This is the weighting coefficient for the current control quantity of the active stabilizer bar motor. These are the weighting coefficients for the pressure control parameters of the hydraulic actuator. These four parameters need to be calibrated through bench tests or road tests.

[0257] (3) Smoothing term It is mainly used to suppress instruction mutations and improve comfort. The previous cycle instruction for a directly controllable physical quantity. This is the weight matrix, and it is also a diagonal matrix:

[0258] ,

[0259] The smoothing weighting coefficient for air suspension air pressure changes is given by vehicle speed. A monotonically increasing function. The smoothing weighting coefficient for the change in current of the active stabilizer bar motor is also for vehicle speed. It is a monotonically increasing function. It is the smoothing weighting coefficient for the pressure change of the hydraulic actuator, and is the vehicle speed. It is a monotonically increasing function. These three parameters also need to be calibrated through bench tests or road tests.

[0260] 3) Constraints.

[0261] Transform the physical limit of the actuator into an inequality constraint of a QP problem:

[0262] Amplitude constraints: , =[−50,−50,−50,−50,0,0] T , =[50,50,50,50,11,2×10 7 ] T .

[0263] 3. Real-time constraint processing strategy.

[0264] Real-time constraint processing is crucial for ensuring the safe and reliable operation of the system by guaranteeing that all actuators' physical limits and engineering safety constraints are met before control commands are issued. This module receives the raw control sequence from the MPC controller and compares, corrects, or uses it as boundary conditions for the optimization problem, specifically handling the following three types of constraints:

[0265] 1) Actuator amplitude and range constraints: Ensure that the control commands of each actuator do not exceed the physical limits allowed by its hardware. This includes: air suspension air pressure changes. [−50, 50] kPa, active stabilizer bar output torque [−800,800] N·m and pitch control torque [−1000, 1000] N·m.

[0266] 2) Rate of Change Constraint: Limits the rate of change of control commands to prevent shocks and protect the actuator. This includes: air pressure rate of change. [−30, 50] kPa / s, Torque variation rate [−200, 200] N m / s, [−3.5,3.5]×10 6 Pa / s (used to limit pitch moment) (generation rate).

[0267] 3) Vehicle state safety constraints: Real-time monitoring and ensuring the predicted state sequence remains within safe boundaries, prioritizing vehicle stability. This mainly includes: vehicle roll angle. [−2°, 2°] (to prevent rollover), vertical acceleration [−0.3g, 0.3g] (Standard for ensuring comfort).

[0268] The above constraints are all embedded in the quadratic programming (QP) solver of MPC in the form of inequalities as boundary conditions for the optimization problem, ensuring that the optimal control sequence can be directly and safely applied to the actuator.

[0269] 4. Actuator interface conversion module.

[0270] The actuator interface conversion module is a bridge connecting the control quantity allocation algorithm and the physical actuator. Its core function is to convert the physical quantity commands output by the optimization algorithm into electrical control signals that drive the actuator, ensuring that the control commands are ultimately implemented.

[0271] The specific signal conversion relationship is as follows:

[0272] The optimization variables output by the control quantity allocation algorithm As a physical quantity, it needs to be converted into an actuator drive signal through the following relationship:

[0273] (1) Air suspension drive signal.

[0274] Air pressure commands for each air spring Converted into solenoid valve drive voltage The conversion relationship is as follows:

[0275] ,

[0276] in, The pressure-to-voltage conversion coefficient (kPa / V) is determined through actuator calibration experiments. As a preferred embodiment, it can be taken as... =25 kPa / V.

[0277] (2) Active stabilizer bar drive signal.

[0278] stabilizer bar torque command Motor current in the optimization variables The transformation is directly achieved by the motor in the optimization variables, and its conversion relationship is based on the motor model:

[0279] ,

[0280] in, is the motor torque constant (N·m / A). For transmission efficiency, This is the reduction ratio. Therefore, the drive circuit directly receives the current command. This will generate the required torque.

[0281] (3) Pitch hydraulic actuator drive signal.

[0282] Hydraulic actuator pressure difference command Converted into servo valve drive voltage The conversion relationship is as follows:

[0283] ,

[0284] in, The value is the hydraulic-to-voltage conversion factor (Pa / V), which is determined through hydraulic system calibration experiments.

[0285] 5. Fault tolerance strategy.

[0286] The fault-tolerance strategy aims to maintain the vehicle's basic stability and safety, achieving a gradual performance degradation, by controlling redundancy and algorithmic reconfiguration when some system components fail. This strategy comprises two parts: fault diagnosis and redundancy control.

[0287] Fault Diagnosis and Monitoring: Real-time monitoring based on the consistency between actuator commands and sensor feedback. Fault determination thresholds and delays are set; for example, if the deviation between the air spring pressure command and the measured value consistently exceeds the threshold... (Take 10 kPa) Exceeds If the current surges (take 100 ms), the air circuit is considered to be faulty; if the current of the stabilizer bar motor surges but the torque sensor feedback is zero, the motor or transmission mechanism is considered to be stuck.

[0288] Redundant control logic:

[0289] (1) Single-circuit failure of air suspension: If a certain air spring cannot build up pressure (e.g. leakage), the pressure of the other three air chambers is adjusted by the control quantity allocation algorithm, the coupling relationship of the vehicle body posture is used for compensation, the vehicle speed is appropriately limited, and the driver is alerted at the same time.

[0290] (2) Active stabilizer bar failure: If the stabilizer bar motor fails, the weight of the differential pressure control term of the air suspension left and right sides will be dynamically increased in the cost function of MPC. The anti-roll moment will be generated by the asymmetric stiffness of the suspension, which will partially replace the function of the stabilizer bar.

[0291] (3) Failure of key sensors: If the data of key sensors such as IMU are abnormal, switch to a model-based state observer (such as Kalman filter) to reconstruct and estimate the vehicle state (such as roll angle and pitch angle), and continue to run the degraded control mode based on the estimated value.

[0292] All fault information is reported to the instrument panel via the vehicle network, prompting the driver to perform maintenance and ensuring driving safety.

[0293] The control method of this invention was simulated in the MATLAB / Simulink environment. The simulation results are as follows. Figures 1-4 As shown. The vehicle parameters used in this example are: sprung mass of 1450 kg; roll moment of inertia of 480 kg·m.

[0294] The pitch moment of inertia is 2200 kg·m²; the center of gravity height is 0.45 m. Suspension system parameters: vertical stiffness: 28000 N / m; vertical damping: 1800 N·s / m; roll stiffness: 22000 N·m / rad; roll damping: 1300 N·m·s / rad; pitch stiffness: 24000 N·m / rad; pitch damping: 1500 N·m·s / rad. Excitation conditions: road surface excitation is a composite sine wave (1.8Hz + 3.5Hz + 6Hz), amplitude 0.025m; lateral excitation is 0.24g; steering input (exponential growth); longitudinal excitation is -0.21g braking input (exponential growth); simulation time starts from 0 seconds and lasts for 10 seconds; sampling time is 0.005 seconds.

[0295] Simulation Result Analysis:

[0296] 1. Vertical displacement response ( Figure 1 ).

[0297] Response characteristics: It exhibits multi-frequency oscillations under the excitation of composite pavement.

[0298] Key data: Initial fluctuation range: ±40 mm; Stabilization time: approximately 4 seconds; Final fluctuation range: ±20 mm.

[0299] Depend on Figure 1 It can be seen that the suspension system effectively absorbed most of the road impacts, with only continuous small oscillations, indicating that the suspension stiffness and damping were well matched.

[0300] 2. Roll angle response ( Figure 2 ).

[0301] Response characteristics: The steering input causes the angular velocity to rise rapidly and then tend to stabilize.

[0302] Key data: Peak angle: 4.7°; Overshoot: approximately 5%; Steady-state value: 4°.

[0303] Depend on Figure 2 It can be seen that the roll control effect is good, within the safe range (usually <5°), and the steady-state error indicates the presence of a continuous lateral force.

[0304] 3. Pitch angle response ( Figure 3 ).

[0305] Response characteristics: Braking input causes negative changes (nodding phenomenon).

[0306] Key data: Maximum pitch angle: -4.0°; Steady-state value: -2.5° to -3.0°; Oscillation amplitude: ±0.5°.

[0307] Depend on Figure 3It can be seen that the pitch motion control is reasonable, with only slight oscillations.

[0308] 4. Vertical acceleration response ( Figure 4 ).

[0309] Response characteristics: It exhibits high-frequency oscillation characteristics.

[0310] Key data: Maximum acceleration: ±1.0 m / s²; RMS value: approximately 0.5 m / s² (0.05g); Main frequencies: 1.8Hz / 3.5Hz / 6Hz.

[0311] Depend on Figure 4 It can be seen that the ride comfort is good (ISO 2631 rating is "comfortable"), and the high-frequency oscillations are caused by road surface excitation.

Claims

1. A cross-domain coupling control method for air suspension and active stabilizer bar, characterized in that, Includes the following steps: The signal input layer collects and fuses data and transmits it to the decision layer. The data transmitted to the decision layer includes: road curvature, road surface unevenness, vehicle longitudinal acceleration, lateral acceleration, vertical acceleration, roll rate and pitch rate. The decision-making layer generates control commands based on model predictive control algorithms, including constructing a three-degree-of-freedom coupled dynamics model of the vehicle and iteratively solving for the minimum cost function in the rolling time domain to obtain control commands; the three-degree-of-freedom coupled dynamics model of the vehicle is as follows: ， ， ， ， ， in, This represents the vertical displacement of the vehicle's center of gravity. Vertical velocity, It is the vertical acceleration of the vehicle body. The body roll angle, The angular velocity is the roll rate. This is the roll acceleration. The vehicle body pitch angle, The pitch angular velocity, For pitch acceleration, Here is the mass-moment-of-inertia matrix. Here is the damping matrix. Here is the stiffness matrix. To control the allocation matrix, It is an external excitation term. The damping matrix and stiffness matrix are asymmetric matrices. The off-diagonal terms of the damping matrix include vertical and roll coupling damping, vertical and pitch coupling damping, vertical force coupling damping caused by roll angular velocity, and vertical force coupling damping caused by pitch angular velocity. The off-diagonal terms of the stiffness matrix include vertical-pitch coupling stiffness, vertical-roll coupling stiffness, vertical force coupling stiffness caused by roll angle, and vertical force coupling stiffness caused by pitch angle. For control commands, This represents the change in air spring pressure. For the output torque of the active stabilizer bar, For pitch actuator control torque; The cost function is a combination of vehicle state terms and control command terms. The vehicle state terms are a dynamic weighted sum of vehicle vertical acceleration, roll angle, pitch angle, vertical velocity, roll rate, and pitch rate. The dynamic weights are determined by the vehicle operating conditions.

2. The cross-domain coupling control method for air suspension and active stabilizer bar according to claim 1, characterized in that, The cost function is: ， in, To predict the time domain, To control the time domain, This represents the maximum vertical acceleration. This represents the maximum roll angle. This represents the maximum pitch angle. This represents the maximum vertical velocity. This represents the maximum roll rate. This represents the maximum pitch angular velocity. To dynamically adjust the weighting coefficients of each objective priority, This is the actuator energy consumption weight matrix. Representing the current moment, Represents an integer index variable. Represents the prediction of the future starting from the current moment. A specific point in time.

3. The cross-domain coupling control method for air suspension and active stabilizer bar according to claim 2, characterized in that, ， ， ， ， ， ， ， ， ， ， ， in For the lateral acceleration of the vehicle, For vehicle speed, For the longitudinal acceleration of the vehicle, This represents the rate of change of air suspension pressure. This represents the maximum pressure change rate of the air suspension. This is the maximum output torque of the stabilizer bar. For the pitch coordinator torque variation rate, The maximum torque change rate of the pitch coordinator. , , , , All of these are calibration coefficients.

4. The cross-domain coupling control method for air suspension and active stabilizer bar according to claim 1, characterized in that, When iteratively solving for the minimum cost function in the rolling time domain, the following constraints are set: vehicle roll angle constraint, vehicle center of gravity vertical acceleration constraint, air spring pressure change constraint, air spring pressure change rate constraint, active stabilizer output torque constraint, and active stabilizer output torque change rate constraint.

5. The cross-domain coupling control method for air suspension and active stabilizer bar according to claim 1, characterized in that, Feedforward compensation is set for control commands, and the feedforward compensation includes road surface preview compensation. = Curveside tilt feed And pitch feedforward for braking / acceleration ,in, For calibration coefficients, To anticipate the future The road surface unevenness at all times, t Indicates the current moment. The moment of inertia of the vehicle body during roll. For vehicle speed, To predict the road curvature, r is the tire radius. For the sprung mass, For longitudinal acceleration, It is the height of the center of mass.

6. The cross-domain coupling control method for air suspension and active stabilizer bar according to claim 1, characterized in that, When the signal input layer collects and fuses data for transmission to the decision layer, it collects the original road curvature and vehicle speed data to construct road curvature observations, and performs Kalman filtering estimation to obtain the road curvature transmitted to the decision layer. The road curvature observations are: , The original curvature of the road, It is lateral acceleration. For vehicle speed, This is a dynamic weight adjusted based on visibility.

7. The cross-domain coupling control method for air suspension and active stabilizer bar according to claim 1, characterized in that, When the signal input layer collects and fuses data and transmits it to the decision layer, the roll angle observation value is constructed from the collected roll angular velocity and the predicted roll angle output by the three-degree-of-freedom coupled dynamics model of the vehicle, and Kalman filtering is performed to estimate the roll angle transmitted to the decision layer. The roll angle observation value is: , The roll angle is calculated from the collected roll angular velocity. It predicts the roll angle. These are fixed weighting coefficients.

8. The cross-domain coupling control method for air suspension and active stabilizer bar according to claim 1, characterized in that, The actuator layer converts the control commands into physical quantities that can be directly controlled by the underlying actuators through an optimized objective function, and then converts them into drive signals for each actuator for control. The physical quantities that can be directly controlled at the underlying level of each actuator are: ， in, The values ​​represent the air pressure changes of the four air springs in the suspension. FL indicates the front left, FR indicates the front right, RL indicates the rear left, and RR indicates the rear right. For controlling the current of the active stabilizer bar motor, For the pitch hydraulic actuator pressure difference, The objective function is: ， in, Let B be the adaptive weight matrix for the operating conditions, and let B be the linear mapping matrix constructed based on the actuator dynamics model. This is a matrix used to punish excessive actions by the actuator. For smoothing matrices, Control commands output by the decision-making level The previous cycle instruction for a directly controllable physical quantity.

9. The cross-domain coupling control method for air suspension and active stabilizer bar according to claim 8, characterized in that, , , , , The weights for vertical stiffness control. The vertical impact frequency, As the roll torque weight, For pitch torque weighting, For vertical basic weights, For frequency sensitivity coefficient, It is the vehicle's vertical resonant frequency. As the basic weight for lateral tilt, This is the roll sensitivity coefficient. For the safe roll strength threshold, This is the tilt strength coefficient. The deceleration coefficient, For the safe deceleration threshold, For deceleration sensitivity coefficient, For pitch foundation weights; , These are the weighting coefficients for the air pressure control quantities of the front and rear axles, respectively, satisfying... , This is the weighting coefficient for the current control quantity of the active stabilizer bar motor. Weighting coefficients for pressure control quantities of hydraulic actuators; , The smoothing weighting coefficient for air suspension air pressure changes is given by vehicle speed. A monotonically increasing function. The smoothing weighting coefficient for the change in current of the active stabilizer bar motor is given by the vehicle speed. A monotonically increasing function. It is the smoothing weighting coefficient for the pressure change of the hydraulic actuator, and is the vehicle speed. It is a monotonically increasing function.

Images