Semi-active suspension preview control method fusing long and short time domain prediction

Abstract

The invention discloses a semi-active suspension preview control method fusing long and short time domain prediction, which comprises the following steps: designing a disturbance invariant set to perform feedback correction on uncertainty existing in long time domain preview information so as to ensure the robustness of a Tube-MPC controller, and outputting an optimal suspension active force as a lower layer actuator target force; obtaining respective target forces of the double-chamber air spring and the CDC damper; and designing a double-chamber air spring-CDC damper collaborative distribution and compensation control strategy based on the short-time domain preview information to control the opening and closing of a double-chamber switching solenoid valve in the double-chamber air spring and the current of the CDC damper. According to the method, the long-time-domain preview information is introduced into the vehicle body attitude control, and the uncertainty is fed back and corrected by introducing the disturbance invariant set of the Tube-MPC aiming at the problem of uncertainty of the long-time-domain preview information, so that the robustness of the Tube-MPC in the face of the uncertainty is ensured.

Description

Technical Field

[0001] The invention belongs to the technical field of automobile suspension control, and in particular relates to a semi-active suspension preview control method integrating long- and short-time domain prediction. Background Art

[0002] At present, semi-active suspension control systems are usually based on a hierarchical control architecture, in which the upper-level controller calculates the global control target based on the dynamic model of the whole vehicle and generates the target force of each suspension to optimize the pitch, roll and vertical movement of the vehicle; the lower-level controller directly controls the actuator to achieve the execution of the target force. Among them, the solution represented by the dual-actuator suspension (such as the dual-chamber air spring and continuous damping control (CDC damper) combines the force adjustment capability of the air spring and the advantages of the damping control of the damper, which can better cope with dynamic loads and road disturbances. In addition, in recent years, intelligent control algorithms that integrate road preview information are considered to be an effective way to further optimize the performance of the suspension system. In the existing research on the preview control method of the suspension system, for example: China's invention patent application number is CN202110623048 .8, entitled “A preview control method for an automotive electronic suspension” realizes effective identification and detection of different road types and optimal suspension parameter matching control for different road types; the Chinese invention patent application number is CN202010677973.4, entitled “Active suspension preview control method based on road information identification by camera sensor”, which uses camera sensors to obtain road preview information, and combines H∞ optimal control to improve the time lag phenomenon in the adjustment process of active suspension damping parameters, and realizes the balanced control of multiple performance indicators of active suspension; the Chinese invention patent application number is CN202410102300.4, entitled “MPC preview suspension control method and system based on neural dynamics model”, which uses neural networks to establish a seven-degree-of-freedom suspension motion mathematical model, and integrates preview MPC to achieve control.

[0003] However, the above existing control strategies have the following problems:

[0004] First, in terms of the preview time domain, existing studies generally use a single preview time domain information, and do not consider integrating the long-time domain and short-time domain preview information into the controller design respectively. Therefore, it is difficult to take into account both the global body posture control performance and the local actuator response performance at the same time, which restricts the control effect of the semi-active suspension under complex road conditions.

[0005] Second, in terms of suspension hierarchical control architecture, existing research mainly focuses on the upper-level vehicle body posture preview control, and lacks research on the lower-level actuator preview control. In addition, although both upper-level and lower-level controllers can benefit from preview information, due to their different functional positioning in the hierarchical control architecture, their information application methods are also significantly different.

[0006] Third, in terms of the integration of long- and short-time preview information with upper and lower-level control, existing research often ignores the fact that the upper-level long-time preview information usually has uncertainty that affects the control accuracy, while the lower-level actuators, such as the dual-chamber air spring, have a charging and deflating time lag during the switching of adjustment modes. In theory, this lag can be compensated by short-time preview to achieve more efficient coordinated control between the air spring and the CDC damper.

[0007] Therefore, in-depth research on incorporating long-time and short-time preview information into the controller design respectively, exploring how to effectively reduce uncertainty in long-time preview and improve the hysteresis characteristics of the actuator through short-time preview technology, has important theoretical significance and application value for further improving the performance of the suspension system. Summary of the invention

[0008] In view of the above-mentioned deficiencies of the prior art, the purpose of the present invention is to provide a semi-active suspension preview control method integrating long-term and short-term predictions, so as to solve the problems of uncertainty in long-term preview information and serious time lag between lower-level actuators in the prior art;

[0009] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0010] A semi-active suspension preview control method integrating long- and short-time domain predictions of the present invention is based on a semi-active suspension system, the system comprising: a dual-chamber air spring, a CDC damper, a sensor assembly and an electronic control unit;

[0011] The dual-chamber air spring and CDC damper are both installed between the vehicle body and the wheel, one end of each is connected to the vehicle body through the tower top, and the other end is connected to the wheel through the lower arm; the dual-chamber air spring is used to provide support force and variable stiffness; the CDC damper is used for dynamic damping adjustment;

[0012] The sensor assembly includes: a vehicle body acceleration sensor, a suspension travel sensor, an inertial measurement unit and a camera;

[0013] The vehicle body acceleration sensor is used to collect the vibration acceleration of the vehicle body in the vertical direction;

[0014] Suspension travel sensor, used to collect the compression or extension of the suspension;

[0015] Inertial measurement unit, used to collect vehicle roll and pitch angular velocity;

[0016] A camera is used to collect road interference data in front of the vehicle to provide long-term preview information and short-term preview information;

[0017] The electronic control unit is electrically connected to the solenoid valve in the CDC damper, the dual-chamber switching solenoid valve in the dual-chamber air spring, and the sensor assembly, and performs calculations based on the data collected by the sensor assembly to obtain output control instructions, thereby achieving real-time adjustment of the stiffness of the dual-chamber air spring and the damping force of the CDC damper;

[0018] The steps are as follows:

[0019] 1) Based on the vehicle's seven-degree-of-freedom suspension model, the long-term preview information is used as the road surface input to build a Tube-MPC controller. The vehicle body vibration acceleration, vehicle roll angular velocity, and vehicle pitch angular velocity are combined to achieve control of the vehicle body posture. A disturbance invariant set is designed to feedback correct the uncertainty of the long-term preview information to ensure the robustness of the Tube-MPC controller, and the optimal suspension active force is output as the target force of the lower actuator.

[0020] 2) Based on the target force of the lower actuator in step 1), the target force of the lower actuator is separated in the frequency domain to obtain the target forces of the dual-chamber air spring and the CDC damper respectively;

[0021] 3) Based on the target forces of the dual-chamber air spring and CDC damper obtained in step 2), the mapping relationship between different time delays and short preview time domains between the lower-level actuators is analyzed, and a dual-chamber air spring-CDC damper collaborative allocation and compensation control strategy based on short-time domain preview information is designed to control the switch of the dual-chamber switching solenoid valve in the dual-chamber air spring and the current of the CDC damper.

[0022] Furthermore, the step 1) specifically includes:

[0023] 11) Set the long-term preview distance to l p , then preview time domain The preview distance in front of the vehicle l p The road disturbance with uncertainty at is w k , w k =w(t)+Δw, then the road disturbance with uncertainty that the vehicle contacts is w k (t) = w(tt p )+Δw, where Δw is the uncertainty contained in the preview information, v is the vehicle speed, t is the time, w(t) is the road disturbance without uncertainty at time t, w(tt p ) is tt p No uncertain road disturbance at any time;

[0024] w k (t) Discretization to obtain discrete road disturbance w k * (t), the expression is:

[0025]

[0026] In the formula, w k+1 (t) represents the road disturbance at the k+1th step at time t, w k (t) represents the road disturbance at the kth step at time t, w k+1 (t p ) indicates t p The road disturbance at the k+1th step, w k+1 (t p -t s ) indicates t p -t s The road disturbance at the k+1th step;

[0027] 12) Based on the discrete road disturbance w obtained in step 11) k * (t) Build a Tube-MPC controller and design a disturbance invariant set; the Tube-MPC controller contains a nominal system and an error system. The minimum disturbance invariant set Z is designed to tighten the constraints on the state and control quantities of the nominal system. Its constraint set for Where X and U are the set of real system state quantities and the set of control quantities respectively;

[0028] The body and wheel status values ​​are obtained based on the data collected by the body vibration acceleration sensor and the suspension travel sensor. The constraint range is (z b -z w ) min ≤z b -z w ≤(z b -z w ) max , to ensure that the semi-active suspension system is within the working range, represents the body vibration acceleration, is the wheel vibration acceleration, z b represents the vertical displacement of the vehicle body, z w represents the vertical displacement of the wheel; the nominal system and error system of the designed Tube-MPC controller are as follows:

[0029]

[0030]

[0031] In the formula, is the nominal system state quantity at time k+1; are the nominal system state quantity and control quantity at time k respectively; are the nominal system state quantity set and control quantity set respectively; e k+1 is the error between the real system and the nominal system at time k+1, e k is the error between the real system and the nominal system at time k, that is, Defined as the error system control quantity; is the control quantity of the error system at time k; A d is the state transfer matrix of the discrete system; B d is the control quantity matrix of the discrete system; W is a convex subset covering all possible values ​​of Δw and including the origin;

[0032] The control variable u of the Tube-MPC controller is k It is divided into the nominal system control quantity and the error system state feedback control quantity, and the expression is as follows:

[0033]

[0034] Where K is the state feedback matrix, Ke k is the error system state feedback control quantity;

[0035] Get e k+1 =A K e k +Δw, where the closed-loop system matrix A with state feedback K =A d +B d K; assume that K exists to ensure A K is stable, then the error set is bounded, proving that there exists a set Z that satisfies Z is the disturbance invariant set of the Tube-MPC controller; if the disturbance sequence Δw∈W, then at time k, the nominal system state quantity of the system deviates from the actual system state quantity, if it satisfies Then the error at the next moment is in the perturbation invariant set,

[0036] Constraint spectrum radius ρ(A k ) to ensure the stability of Tube-MPC controller, the design disturbance invariant set formula is:

[0037]

[0038] In the formula, Z(t p ) is the Tube-MPC controller and the preview time domain t p the associated perturbation-invariant set;

[0039] 13) Based on the constructed disturbance invariant set, the Tube-MPC controller is further solved, the problem of solving the nominal system is transformed into the problem of solving the optimal control, and the optimal suspension force is output as the target force of the lower actuator; at the same time, considering the Lyapunov stability of the system, the design objective function and constraints are as follows:

[0040]

[0041] Where J is the objective function; is the value of the nominal system state quantity predicted at time k at the future time k+i; ref|k+i|k is the expected value of the real system state at the future k+i time; is the value of the nominal system control quantity change at time k at the future k+i time; R is the weight matrix of the nominal system control quantity change; P is the weight matrix of the terminal state; Np is the prediction time domain; X f is the terminal constraint; To predict the nominal system state quantity of the terminal in the time domain; Q 1 , Q 2 , Q 3 They correspond to the outputs of the Tube-MPC controller, including the weights of the vehicle body vibration acceleration, roll angular velocity, and pitch angular velocity;

[0042] According to the vehicle historical data, real-time body vibration acceleration data, suspension travel data, roll angular velocity data, pitch angular velocity data and future road disturbance data collected by the sensor component, the future output response of the Tube-MPC controller is predicted, and the real system in the Tube-MPC controller is discretized. The discrete real system is expressed as:

[0043] x k+1 =A d x k +B d u k

[0044] In the formula, x k+1 is the real system state at time k+1; x k is the real system state at time k, which includes the body vibration acceleration Wheel vibration acceleration Suspension travel z b -z w ;u k is the real system control quantity at time k, and the control quantity is the optimal suspension force;

[0045] Combining the real system state quantity and the real system control quantity, the discrete real system after dimension expansion is expressed as:

[0046]

[0047] In the formula, represents the state error vector of the real system after dimension expansion; u k-1 is the real system control quantity at time k-1; the expanded dimension matrix Expanded Matrix Expanded Matrix η k is the state vector of the real system at time k after dimension expansion; I Nu is a unit matrix with dimension Nu, which is related to the control quantity; I Nx is the unit matrix with dimension Nx, which is related to the state quantity; Δu k is the change of the real system control quantity at time k and subsequent times;

[0048] Obtain the predicted output y of the real system after dimension expansion in the prediction time domain k+1∣k :

[0049] y k+1∣k =S x ξ k +S u ΔU k

[0050] In the formula,

[0051]

[0052] N c To control the time domain, It is the sequence of control variable changes, and is the optimal suspension active force as the target force of the lower actuator.

[0053] Furthermore, the step 2) specifically includes:

[0054] The target force of the lower actuator obtained in step 1) is distributed, and the time domain signal is converted into a frequency domain signal using Fourier transform, and the frequency domain separation is performed to divide the target force into a low-frequency active force and a high-frequency active force; given a time domain signal x(t), its frequency domain X(f) is expressed as:

[0055]

[0056] In the formula, f is the frequency in the frequency domain, and t is the time in the time domain;

[0057] Use a low-pass filter to extract the low-frequency active force F low (t) is used as the target force of the dual-chamber air spring and the transfer function H of the first-order low-pass filter is low (jω) and low-frequency active force F low (t) is expressed as:

[0058]

[0059] Where, X c is the reactance of capacitor C, R is the resistance value in the filter, ω is the target force angular frequency of the lower actuator, C is the capacitance value in the filter, F Tube-MPC (t) represents the actuator target force output by the Tube-MPC controller, j is an imaginary unit;

[0060] Use low-pass filter to extract high-frequency active force F high (t) is used as the target force of the CDC damper, and the transfer function H of the first-order high-pass filter is high (jω) and high-frequency active force F high (t) is expressed as:

[0061]

[0062] Furthermore, the step 3) specifically includes:

[0063] 31) Based on the target forces of the dual-chamber air spring and CDC damper obtained in step 2), analyze the specific required short preview time domain T preview The mapping relationship between different time delays between the lower-level actuators; design and run a series of simulations, and collect system response data under different time delays by means of parameter scanning. Specifically, the simulation scenarios include various road disturbance characteristics at different vehicle speeds. Different time delays between the lower-level actuators are simulated by adjusting the inflation and deflation time delay of the dual-chamber air spring and the response time of the CDC damper. The preview time domain is dynamically adjusted in each simulation, and the accuracy of the system response is recorded.

[0064] The generalized regression neural network is used to analyze the specific short preview time domain T preview The mapping relationship between different time delays between the lower-level actuators is as follows:

[0065] The radial basis function is used to measure the similarity between the input data and the training sample, and the activation function is calculated as follows:

[0066]

[0067] In the formula, φ i is the output of the i-th neuron; D i is the Euclidean distance between the input data and the training sample, where the input data includes the time lag t delay , vehicle speed v and road disturbance characteristics r; σ is the smoothing factor of the radial basis function;

[0068] Calculate normalized weights Where W iis the normalized weight of the i-th training sample; N is the number of training samples; the final output is calculated as the short preview time domain Y i is the short preview time domain corresponding to the i-th training sample;

[0069] 32) Based on the short preview time domain T obtained in step 31) preview , the collaborative allocation and compensation control strategy of the dual-chamber air spring-CDC damper based on short-time preview information is designed as follows:

[0070] According to the low-frequency active force F low (t) controls the mode switching of the dual-chamber air spring to adapt to low-frequency vibrations of different amplitudes. The mode switching logic is as follows:

[0071]

[0072] In the formula, u valve F is the switch state of the solenoid valve in the dual-chamber air spring, 1 represents open, 0 represents closed; thresh It is the active power threshold of the solenoid valve switch in the dual-chamber air spring. In order to improve the stability of the dual-chamber air spring and avoid the adverse effects caused by frequent mode switching, the threshold needs to be controlled within a reasonable range during the design process;

[0073] Output force F according to the switch state of the solenoid valve in the double-chamber air spring AS (t), the active force output by the dual-chamber air spring in a certain stiffness mode, is expressed as follows:

[0074]

[0075] In the formula, F open (t) is the main force when opening the valve, F closed (t) is the main force when closing the valve;

[0076] Control the CDC damper in the high frequency range by tracking the target damping force F high (t) Absorb vibrations and use the CDC damper in the low-frequency range to compensate for the insufficient dual-chamber air spring active force F in the low-frequency range of the next time period. low (t+T preview )-F AS (t+T preview ), while maintaining the high-frequency damping control at the current time t; the final target damping force F of the CDC damper is calculated CDC (t) is the superposition of the damping force in the high-frequency range and the compensation force in the low-frequency range, and the expression is as follows:

[0077] F CDC (t) = F high (t)+Flow (t+T preview )-F AS (t+T preview )

[0078] Sliding mode control is used to realize the CDC damper to the final target damping force T CDC (t), the sliding surface s is defined as the combination of the error between the final target damping force and the actual damping force and its derivative as follows:

[0079]

[0080] Where, γ is the sliding surface weight coefficient, which is used to adjust the response speed of the system; F error =e high (t)+e low (t+T preview ), the error in the high frequency range e high (t) represents the actual damping force of the CDC damper in the high frequency range and F high (t); the error in the low frequency range e low (t+T preview )=F low (t+T preview )-F AS (t+T preview );

[0081] The designed sliding mode control law is:

[0082]

[0083] Where ε is the constant approach rate, κ is the exponential approach coefficient, and δ is the boundary layer thickness.

[0084] Beneficial effects of the present invention:

[0085] The present invention introduces long-term preview information into vehicle body posture control, and aims at the uncertainty problem of long-term preview information. By introducing the disturbance invariant set of Tube-MPC controller, the uncertainty is feedback-corrected, so as to ensure the robustness of Tube-MPC controller in the face of uncertainty.

[0086] Aiming at the serious time lag problem between the dual-chamber air spring and the CDC damper in the lower actuator, the present invention proposes a dual-chamber air spring-CDC damper collaborative allocation and compensation control strategy based on short-time domain preview information. The mapping relationship between the short preview time domain and the different time lags between the lower actuators is fitted based on a generalized regression neural network model, and the short-time domain preview information is used to compensate for the time lag between the lower actuators, thereby improving the accuracy of system response. BRIEF DESCRIPTION OF THE DRAWINGS

[0087] Figure 1 The schematic diagram of the method of the present invention is shown in FIG. DETAILED DESCRIPTION

[0088] In order to facilitate the understanding of those skilled in the art, the present invention is further described below in conjunction with embodiments and drawings. The contents mentioned in the implementation modes are not intended to limit the present invention.

[0089] A semi-active suspension preview control method integrating long- and short-time domain predictions of the present invention is based on a semi-active suspension system, the system comprising: a dual-chamber air spring, a CDC damper, a sensor assembly and an electronic control unit;

[0090] The dual-chamber air spring and CDC damper are both installed between the vehicle body and the wheel, one end of each is connected to the vehicle body through the tower top, and the other end is connected to the wheel through the lower arm; the dual-chamber air spring is used to provide support force and variable stiffness; the CDC damper is used for dynamic damping adjustment;

[0091] The sensor assembly includes: a vehicle body acceleration sensor, a suspension travel sensor, an inertial measurement unit and a camera;

[0092] The vehicle body acceleration sensor is used to collect the vibration acceleration of the vehicle body in the vertical direction;

[0093] Suspension travel sensor, used to collect the compression or extension of the suspension;

[0094] Inertial measurement unit, used to collect vehicle roll and pitch angular velocity;

[0095] A camera is used to collect road interference data in front of the vehicle to provide long-term preview information and short-term preview information;

[0096] The electronic control unit is electrically connected to the solenoid valve in the CDC damper, the dual-chamber switching solenoid valve in the dual-chamber air spring, and the sensor assembly, and performs calculations based on the data collected by the sensor assembly to obtain output control instructions, thereby achieving real-time adjustment of the stiffness of the dual-chamber air spring and the damping force of the CDC damper;

[0097] Reference Figure 1 As shown, the method steps are as follows:

[0098] 1) Based on the vehicle's seven-degree-of-freedom suspension model, the long-term preview information is used as the road input to build a Tube-MPC controller, and the vehicle body vibration acceleration, vehicle roll angular velocity, and vehicle pitch angular velocity are combined to achieve control of the vehicle body posture. The disturbance invariant set is designed to feedback correct the uncertainty of the long-term preview information to ensure the robustness of the Tube-MPC controller, and the optimal suspension active force is output as the target force of the lower-level actuator; specifically, it includes:

[0099] 11) Set the long-term preview distance to l p , then preview time domain The preview distance in front of the vehicle l p The road disturbance with uncertainty at is w k , w k =w(t)+Δw, then the road disturbance with uncertainty that the vehicle contacts is w k (t) = w(tt p )+Δw, where Δw is the uncertainty contained in the preview information, v is the vehicle speed, t is the time, w(t) is the road disturbance without uncertainty at time t, w(tt p ) is tt p No uncertain road disturbance at any time;

[0100] w k (t) Discretization to obtain discrete road disturbance w k * (t), the expression is:

[0101]

[0102] Where w k+1 (t) represents the road disturbance at the k+1th step at time t, w k (t) represents the road disturbance at the kth step at time t, w k+1 (t p ) indicates t p The road disturbance at the k+1th step, w k+1 (t p -t s ) indicates t p -t s The road disturbance at time k+1;

[0103] 12) Based on the discrete road disturbance w obtained in step 11) k * (t) Build a Tube-MPC controller and design a disturbance invariant set; the Tube-MPC controller contains a nominal system and an error system. The minimum disturbance invariant set Z is designed to tighten the constraints on the state and control quantities of the nominal system. Its constraint set for Where X and U are the set of real system state quantities and the set of control quantities respectively;

[0104] The body and wheel status values ​​are obtained based on the data collected by the body vibration acceleration sensor and the suspension travel sensor. The constraint range (set X) is (z b -z w )min ≤z b -z w ≤(z b -z w ) max , to ensure that the semi-active suspension system is within the working range, represents the body vibration acceleration, is the wheel vibration acceleration, z b represents the vertical displacement of the vehicle body, z w represents the vertical displacement of the wheel; the nominal system and error system of the designed Tube-MPC controller are as follows:

[0105]

[0106] In the formula, is the nominal system state quantity at time k+1; are the nominal system state quantity and control quantity at time k respectively; are the nominal system state quantity set and control quantity set respectively; e k+1 is the error between the real system and the nominal system at time k+1, e k is the error between the real system and the nominal system at time k, that is, Defined as the error system control quantity; is the control quantity of the error system at time k; A d is the state transfer matrix of the discrete system; B d is the control quantity matrix of the discrete system; W is a convex subset covering all possible values ​​of Δw and including the origin;

[0107] The control variable u of the Tube-MPC controller is k It is divided into the nominal system control quantity and the error system state feedback control quantity, and the expression is as follows:

[0108]

[0109] Where K is the state feedback matrix, Ke k is the error system state feedback control quantity;

[0110] Get e k+1 =A K e k +Δw, where the closed-loop system matrix A with state feedback K =A d +B d K; assume that K exists to ensure A K is stable, then the error set is bounded, proving that there exists a set Z that satisfies Z is the disturbance invariant set of the Tube-MPC controller; if the disturbance sequence Δw∈W, then at time k, the nominal system state quantity of the system deviates from the actual system state quantity, if it satisfies Then the error at the next moment is in the perturbation invariant set,

[0111] Constraint spectrum radius ρ(A k ) to ensure the stability of Tube-MPC controller, the design disturbance invariant set formula is:

[0112]

[0113] In the formula, Z(t p ) is the Tube-MPC controller and the preview time domain t p the associated perturbation-invariant set;

[0114] 13) Based on the constructed disturbance invariant set, the Tube-MPC controller is further solved, the problem of solving the nominal system is transformed into the problem of solving the optimal control, and the optimal suspension force is output as the target force of the lower actuator; at the same time, considering the Lyapunov stability of the system, the design objective function and constraints are as follows:

[0115]

[0116] Where J is the objective function; is the value of the nominal system state quantity predicted at time k at the future time k+i; ref|k+i|k is the expected value of the real system state at the future k+i time; is the value of the nominal system control quantity change at time k at the future k+i time; R is the weight matrix of the nominal system control quantity change; P is the weight matrix of the terminal state; N p For the prediction time domain X f is the terminal constraint; To predict the nominal system state quantity of the terminal in the time domain; Q 1 , Q 2 , Q 3 They correspond to the outputs of the Tube-MPC controller, including the weights of the vehicle body vibration acceleration, roll angular velocity, and pitch angular velocity;

[0117] According to the vehicle historical data, real-time body vibration acceleration data, suspension travel data, roll angular velocity data, pitch angular velocity data and future road disturbance data collected by the sensor component, the future output response of the Tube-MPC controller is predicted, and the real system in the Tube-MPC controller is discretized. The discrete real system is expressed as:

[0118] x k+1 =A d x k +B d u k

[0119] In the formula, x k+1 is the real system state at time k+1; x k is the real system state at time k, which includes the body vibration acceleration Wheel vibration acceleration Suspension travel z b -z w ;u k is the real system control quantity at time k, and the control quantity is the optimal suspension force;

[0120] Combining the real system state quantity and the real system control quantity, the discrete real system after dimension expansion is expressed as:

[0121]

[0122] In the formula, represents the state error vector of the real system after dimension expansion; u k-1 is the real system control quantity at time k-1; the expanded dimension matrix Expanded Matrix Expanded Matrix η k is the state vector of the real system at time k after dimension expansion; I Nu is a unit matrix with dimension Nu, which is related to the control quantity; I Nx is the unit matrix with dimension Nx, which is related to the state quantity; Δu k is the change of the real system control quantity at time k and subsequent times;

[0123] Obtain the predicted output y of the real system after dimension expansion in the prediction time domain k+1∣k :

[0124] y k+1∣k =S x ξ k +S u ΔU k

[0125] In the formula,

[0126]

[0127] N c To control the time domain, It is the sequence of control variable changes, and is the optimal suspension active force as the target force of the lower actuator.

[0128] 2) Based on the target force of the lower actuator in step 1), the target force of the lower actuator is separated in the frequency domain to obtain the target forces of the dual-chamber air spring and the CDC damper; specifically, the following steps are included:

[0129] The target force of the lower actuator obtained in step 1) is distributed, and the time domain signal is converted into a frequency domain signal using Fourier transform, and the frequency domain separation is performed to divide the target force into a low-frequency active force and a high-frequency active force; given a time domain signal x(t), its frequency domain X(f) is expressed as:

[0130]

[0131] In the formula, f is the frequency in the frequency domain, and t is the time in the time domain;

[0132] Use a low-pass filter to extract the low-frequency active force F low (t) is used as the target force of the dual-chamber air spring and the transfer function H of the first-order low-pass filter is low (jω) and low-frequency active force F low (t) is expressed as:

[0133]

[0134] Where, X c is the reactance of capacitor C, R is the resistance value in the filter, ω is the target force angular frequency of the lower actuator, C is the capacitance value in the filter, F Tube-MPC (t) represents the actuator target force output by the Tube-MPC controller, j is an imaginary unit;

[0135] Use low-pass filter to extract high-frequency active force F high (t) is used as the target force of the CDC damper, and the transfer function H of the first-order high-pass filter is high (jω) and high-frequency active force F high (t) is expressed as:

[0136]

[0137] 3) Based on the target forces of the dual-chamber air spring and CDC damper obtained in step 2), the mapping relationship between different time delays and short preview time domains between the lower-level actuators is analyzed, and a dual-chamber air spring-CDC damper collaborative allocation and compensation control strategy based on short-time preview information is designed to control the switch of the dual-chamber switching solenoid valve in the dual-chamber air spring and the current of the CDC damper; specifically including:

[0138] 31) Based on the target forces of the dual-chamber air spring and CDC damper obtained in step 2), analyze the specific required short preview time domain T previewThe mapping relationship between different time delays between the lower-level actuators; design and run a series of simulations, and collect system response data under different time delays by means of parameter scanning. Specifically, the simulation scenarios include various road disturbance characteristics at different vehicle speeds, and different time delays between the lower-level actuators (such as 100ms, 300ms, and 500ms) are simulated by adjusting the inflation and deflation time delay of the dual-chamber air spring and the response time of the CDC damper. In each simulation, the preview time domain is dynamically adjusted to record the accuracy of the system response;

[0139] The generalized regression neural network (GRNN) is used to analyze the specific short preview time domain T preview The mapping relationship between different time delays between the lower-level executors is as follows:

[0140] Use the radial basis function (Gaussian kernel) to measure the similarity between the input data and the training sample, and calculate the activation function as follows:

[0141]

[0142] In the formula, φ i is the output of the i-th neuron; D i is the Euclidean distance between the input data and the training sample, where the input data includes the time lag t delay , vehicle speed v and road disturbance characteristics r; σ is the smoothing factor of the radial basis function;

[0143] Calculate normalized weights Where W i is the normalized weight of the i-th training sample; N is the number of training samples; the final output is calculated as the short preview time domain Y i is the short preview time domain corresponding to the i-th training sample;

[0144] 32) Based on the short preview time domain T obtained in step 31) preview , the collaborative allocation and compensation control strategy of the dual-chamber air spring-CDC damper based on short-time preview information is designed as follows:

[0145] According to the low-frequency active force F low (t) controls the mode switching of the dual-chamber air spring (the switch of the solenoid valve in the dual-chamber air spring) to adapt to low-frequency vibrations of different amplitudes. The mode switching logic is as follows:

[0146]

[0147] In the formula, u valve F is the switch state of the solenoid valve in the dual-chamber air spring, 1 represents open, 0 represents closed;thresh It is the active power threshold of the solenoid valve switch in the dual-chamber air spring. In order to improve the stability of the dual-chamber air spring and avoid the adverse effects caused by frequent mode switching, the threshold needs to be controlled within a reasonable range during the design process;

[0148] Output force F according to the switch state of the solenoid valve in the double-chamber air spring AS (t), the active force output by the dual-chamber air spring in a certain stiffness mode, is expressed as follows:

[0149]

[0150] In the formula, F open (t) is the main force when opening the valve, F closed (t) is the main force when closing the valve;

[0151] Control the CDC damper in the high frequency range by tracking the target damping force F high (t) Absorb vibrations and use the CDC damper to compensate for the insufficient dual-chamber air spring active force F in the low-frequency range of the next time period (short-time preview time domain) low (t+T preview )-F AS (t+T preview ), while maintaining the high-frequency damping control at the current time t; the final target damping force F of the CDC damper is calculated CDC (t) is the superposition of the damping force in the high-frequency range and the compensation force in the low-frequency range, and the expression is as follows:

[0152] F CDC (t) = F high (t)+F low (t+T preview )-F AS (t+T preview )

[0153] Sliding mode control (SMC) is used to achieve the CDC damper to the final target damping force F CDC (t), the sliding surface s is defined as the combination of the error between the final target damping force and the actual damping force and its derivative as follows:

[0154]

[0155] Where, γ is the sliding surface weight coefficient, which is used to adjust the response speed of the system; F error =e high (t)+e low (t+T preview ), the error in the high frequency range e high (t) represents the actual damping force of the CDC damper in the high frequency range and Fhigh (t); the error in the low frequency range e low (t+T preview )=F low (t+T preview )-F AS (t+T preview );

[0156] The designed sliding mode control law is:

[0157]

[0158] Where ε is the constant approach rate, κ is the exponential approach coefficient, and δ is the boundary layer thickness.

[0159] The present invention has many specific application paths. The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements can be made without departing from the principle of the present invention. These improvements should also be regarded as the protection scope of the present invention.

Claims

1. A semi-active suspension preview control method integrating long-term and short-term predictions, based on a semi-active suspension system, comprising: Dual-chamber air springs, CDC dampers, sensor components and electronic control units; The dual-chamber air spring and the CDC damper are both installed between the vehicle body and the wheel, one end of each of which is connected to the vehicle body through the tower top, and the other end is connected to the wheel through the lower arm; the dual-chamber air spring is used to provide support force and variable stiffness; The CDC damper is used for dynamic damping adjustment; The sensor assembly includes: a vehicle body acceleration sensor, a suspension travel sensor, an inertial measurement unit and a camera; The vehicle body acceleration sensor is used to collect the vibration acceleration of the vehicle body in the vertical direction; Suspension travel sensor, used to collect the compression or extension of the suspension; Inertial measurement unit, used to collect vehicle roll and pitch angular velocity; A camera is used to collect road interference data in front of the vehicle to provide long-term preview information and short-term preview information; The electronic control unit is electrically connected to the solenoid valve in the CDC damper, the dual-chamber switching solenoid valve in the dual-chamber air spring, and the sensor assembly, and performs calculations based on the data collected by the sensor assembly to obtain output control instructions, thereby achieving real-time adjustment of the stiffness of the dual-chamber air spring and the damping force of the CDC damper; The method is characterized in that the steps are as follows: 1) Based on the vehicle's seven-degree-of-freedom suspension model, the long-term preview information is used as the road surface input to build a Tube-MPC controller. The vehicle body vibration acceleration, vehicle roll angular velocity, and vehicle pitch angular velocity are combined to achieve control of the vehicle body posture. A disturbance invariant set is designed to feedback correct the uncertainty of the long-term preview information to ensure the robustness of the Tube-MPC controller, and the optimal suspension active force is output as the target force of the lower actuator. 2) Based on the target force of the lower actuator in step 1), the target force of the lower actuator is separated in the frequency domain to obtain the target forces of the dual-chamber air spring and the CDC damper respectively; 3) Based on the target forces of the dual-chamber air spring and CDC damper obtained in step 2), the mapping relationship between different time delays and short preview time domains between the lower-level actuators is analyzed, and a dual-chamber air spring-CDC damper collaborative allocation and compensation control strategy based on short-time domain preview information is designed to control the switch of the dual-chamber switching solenoid valve in the dual-chamber air spring and the current of the CDC damper.

2. The semi-active suspension preview control method integrating long- and short-term prediction according to claim 1 is characterized in that: The step 1) specifically includes: 11) Set the long-term preview distance to l p , then preview time domain The preview distance in front of the vehicle l p The road disturbance with uncertainty at is w k , w k =w(t)+Δw, then the road disturbance with uncertainty that the vehicle contacts is w k (t) = w(tt p )+Δw, where Δw is the uncertainty contained in the preview information, v is the vehicle speed, t is the time, w(t) is the road disturbance without uncertainty at time t, w(tt p ) is tt p No uncertain road disturbance at any time; w k (t) Discretization to obtain discrete road disturbance w k * (t), the expression is: In the formula, w k+1 (t) represents the road disturbance at the k+1th step at time t, w k (t) represents the road disturbance at the kth step at time t, w k+1 (t p ) indicates t p The road disturbance at the k+1th step, w k+1 (t p -t s ) indicates t p -t s The road disturbance at the k+1th step; 12) Based on the discrete road disturbance w obtained in step 11) k * (t) Build a Tube-MPC controller and design a disturbance invariant set; the Tube-MPC controller contains a nominal system and an error system. The minimum disturbance invariant set Z is designed to tighten the constraints on the state and control quantities of the nominal system. Its constraint set for Where X and U are the set of real system state quantities and the set of control quantities respectively; The body and wheel status values ​​are obtained based on the data collected by the body vibration acceleration sensor and the suspension travel sensor. The constraint range is (z b -z w ) min ≤z b -z w ≤(z b -z w ) max , to ensure that the semi-active suspension system is within the working range, represents the body vibration acceleration, is the wheel vibration acceleration, z b represents the vertical displacement of the vehicle body, z w represents the vertical displacement of the wheel; the nominal system and error system of the designed Tube-MPC controller are as follows: In the formula, is the nominal system state quantity at time k+1; are the nominal system state quantity and control quantity at time k respectively; are the nominal system state quantity set and control quantity set respectively; e k+1 is the error between the real system and the nominal system at time k+1, e k is the error between the real system and the nominal system at time k, that is, Defined as the error system control quantity; is the control quantity of the error system at time k; A d is the state transfer matrix of the discrete system; B d is the control quantity matrix of the discrete system; W is a convex subset covering all possible values ​​of Δw and including the origin; The control variable u of the Tube-MPC controller is k It is divided into the nominal system control quantity and the error system state feedback control quantity, and the expression is as follows: Where K is the state feedback matrix, Ke k is the error system state feedback control quantity; Get e k+1 =A K e k +Δw, where the closed-loop system matrix A with state feedback K =A d +B d K; assume that K exists to ensure A K is stable, then the error set is bounded, proving that there exists a set Z that satisfies Z is the disturbance invariant set of the Tube-MPC controller; if the disturbance sequence Δw∈W, then at time k, the nominal system state quantity of the system deviates from the actual system state quantity, if it satisfies Then the error at the next moment is in the perturbation invariant set, Constraint spectrum radius ρ(A k ) to ensure the stability of Tube-MPC controller, the design disturbance invariant set formula is: In the formula, Z(t p ) is the Tube-MPC controller and the preview time domain t p the associated perturbation-invariant set; 13) Based on the constructed disturbance invariant set, the Tube-MPC controller is further solved, the problem of solving the nominal system is transformed into the problem of solving the optimal control, and the optimal suspension force is output as the target force of the lower actuator; at the same time, considering the Lyapunov stability of the system, the design objective function and constraints are as follows: Where J is the objective function; is the value of the nominal system state quantity predicted at time k at the future time k+i; ref|k+i|k is the expected value of the real system state at the future k+i time; is the value of the nominal system control quantity change at time k at the future k+i time; R is the weight matrix of the nominal system control quantity change; P is the weight matrix of the terminal state; N p is the prediction time domain; X f is the terminal constraint; To predict the nominal system state quantity of the terminal in the time domain; Q1, Q2, and Q3 correspond to the outputs of the Tube-MPC controller, including the weights of the vehicle body vibration acceleration, roll angular velocity, and pitch angular velocity; According to the vehicle historical data, real-time body vibration acceleration data, suspension travel data, roll angular velocity data, pitch angular velocity data and future road disturbance data collected by the sensor component, the future output response of the Tube-MPC controller is predicted, and the real system in the Tube-MPC controller is discretized. The discrete real system is expressed as: x k+1 =A d x k +B d u k In the formula, x k+1 is the real system state at time k+1; x k is the real system state at time k, which includes the body vibration acceleration Wheel vibration acceleration Suspension travel z b -z w ;u k is the real system control quantity at time k, and the control quantity is the optimal suspension force; Combining the real system state quantity and the real system control quantity, the discrete real system after dimension expansion is expressed as: In the formula, represents the state error vector of the real system after dimension expansion; u k-1 is the real system control quantity at time k-1; the expanded dimension matrix Expanded Matrix Expanded Matrix η k is the state vector of the real system at time k after dimension expansion; I Nu is a unit matrix with dimension Nu, which is related to the control quantity; I Nx is the unit matrix with dimension Nx, which is related to the state quantity; Δu k is the change of the real system control quantity at time k and subsequent times; Obtain the predicted output y of the real system after dimension expansion in the prediction time domain k+1|k : y k+1|k =S x x k +S u D.U. k In the formula, Nc is the control time domain, It is the sequence of control variable changes, and is the optimal suspension active force as the target force of the lower actuator.

3. The semi-active suspension preview control method integrating long- and short-time domain prediction according to claim 2 is characterized in that: The step 2) specifically includes: The target force of the lower actuator obtained in step 1) is distributed, and the time domain signal is converted into a frequency domain signal using Fourier transform, and the frequency domain separation is performed to divide the target force into a low-frequency active force and a high-frequency active force; given a time domain signal x(t), its frequency domain X(f) is expressed as: In the formula, f is the frequency in the frequency domain, and t is the time in the time domain; Use a low-pass filter to extract the low-frequency active force F low (t) is used as the target force of the dual-chamber air spring and the transfer function H of the first-order low-pass filter is low (jω) and low-frequency active force F low (t) is expressed as: Where, X c is the reactance of capacitor C, R is the resistance value in the filter, ω is the target force angular frequency of the lower actuator, C is the capacitance value in the filter, F Tube-MPC (t) represents the actuator target force output by the Tube-MPC controller, j is an imaginary unit; Use low-pass filter to extract high-frequency active force F high (t) is used as the target force of the CDC damper, and the transfer function H of the first-order high-pass filter is high (jω) and high-frequency active force F high (t) is expressed as:

4. The semi-active suspension preview control method integrating long- and short-term prediction according to claim 1 is characterized in that: The step 3) specifically includes: 31) Based on the target forces of the dual-chamber air spring and CDC damper obtained in step 2), analyze the specific required short preview time domain T preview The mapping relationship between different time delays between the lower-level actuators; design and run a series of simulations, and collect system response data under different time delays by means of parameter scanning. Specifically, the simulation scenarios include various road disturbance characteristics at different vehicle speeds. Different time delays between the lower-level actuators are simulated by adjusting the inflation and deflation time delay of the dual-chamber air spring and the response time of the CDC damper. The preview time domain is dynamically adjusted in each simulation, and the accuracy of the system response is recorded. The generalized regression neural network is used to analyze the specific short preview time domain T preview The mapping relationship between different time delays between the lower-level executors is as follows: The radial basis function is used to measure the similarity between the input data and the training sample, and the activation function is calculated as follows: In the formula, φ i is the output of the i-th neuron; D i is the Euclidean distance between the input data and the training sample, where the input data includes the time lag t delay , vehicle speed v and road disturbance characteristics r; σ is the smoothing factor of the radial basis function; Calculate normalized weights Where W i is the normalized weight of the i-th training sample; N is the number of training samples; the final output is calculated as the short preview time domain Y i is the short preview time domain corresponding to the i-th training sample; 32) Based on the short preview time domain T obtained in step 31) preview , the collaborative allocation and compensation control strategy of the dual-chamber air spring-CDC damper based on short-time preview information is designed as follows: According to the low-frequency active force F low (t) controls the mode switching of the dual-chamber air spring to adapt to low-frequency vibrations of different amplitudes. The mode switching logic is as follows: In the formula, u valve F is the switch state of the solenoid valve in the dual-chamber air spring, 1 represents open, 0 represents closed; thresh It is the active power threshold of the solenoid valve switch in the dual-chamber air spring. In order to improve the stability of the dual-chamber air spring and avoid the adverse effects caused by frequent mode switching, the threshold needs to be controlled within a reasonable range during the design process; Output force F according to the switch state of the solenoid valve in the double-chamber air spring AS (t), the active force output by the dual-chamber air spring in a certain stiffness mode, is expressed as follows: In the formula, F open (t) is the main force when opening the valve, F closed (t) is the main force when closing the valve; Control the CDC damper in the high frequency range by tracking the target damping force F high (t) Absorb vibrations and use the CDC damper in the low-frequency range to compensate for the insufficient dual-chamber air spring active force F in the low-frequency range of the next time period. low (t+T preview )-F AS (t+T preview ), while maintaining the high-frequency damping control at the current time t; the final target damping force F of the CDC damper is calculated CDC (t) is the superposition of the damping force in the high-frequency range and the compensation force in the low-frequency range, and the expression is as follows: F CDC (t)=F high (t)+F low (t+T preview )-F AS (t+T preview ) Sliding mode control is used to realize the CDC damper to the final target damping force F CDC (t), the sliding surface s is defined as the combination of the error between the final target damping force and the actual damping force and its derivative as follows: Where, γ is the sliding surface weight coefficient, which is used to adjust the response speed of the system; F error =e high (t)+e low (t+T preview ), the error in the high frequency range e high (t) represents the actual damping force of the CDC damper in the high frequency range and F high (t); the error in the low frequency range e low (t+T preview )=F low (t+T preview )-F AS (t+T preview ); The designed sliding mode control law is: Where ε is the constant approach rate, κ is the exponential approach coefficient, and δ is the boundary layer thickness.

Images