Vehicle active suspension dual-adaptive control method, system, equipment and medium

Abstract

The invention provides a vehicle active suspension dual-adaptive control method, system and device and a medium, and belongs to the technical field of vehicle active suspension control, and the method comprises the steps: building a linear parameter change dynamic model of a vehicle active suspension system; establishing an augmented linear parameter change dynamical model with an exogenous system, and obtaining preview state output based on the augmented linear parameter change dynamical model; constructing a preview feedforward controller, and decomposing the control input into a first control component based on preview state output; constructing an LMI optimization solver to obtain a vertex gain, and obtaining a second control component through an HPPD gain scheduling interpolation controller; a dynamic adaptive robust gain is calculated by adopting a dual adaptive law with a rigorous physical projection operator and is input into a smooth robust compensator based on a hyperbolic tangent function to obtain a third control component; and the vehicle active suspension system is driven to move. The method can adaptively track the sudden change of the mass of the system and expand the feasible stable area of the suspension system.

Description

Technical Field

[0001] This invention belongs to the field of active vehicle suspension control technology, and particularly relates to a dual adaptive control method, system, device and medium for active vehicle suspension. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] In modern automotive engineering, active suspension systems utilize real-time sensor feedback to actively inject dynamic compensation forces, thereby fundamentally breaking the unavoidable physical trade-off between shock absorption (comfort) and tire contact (handling) in traditional passive suspensions.

[0004] However, real-world vehicles face significant parameter uncertainties, particularly large fluctuations in sprung mass, leading to a substantial deterioration in the performance of traditional fixed-gain controllers. While robust strategies such as sliding mode control can handle matching uncertainties, their hard-switching mechanisms inevitably induce high-frequency actuator chattering, severely reducing comfort and accelerating mechanical wear. Existing linear parameter variation frameworks employ passive algebraic boundary techniques to address parameter mismatches, forcing an unrealistic increase in the decision variables of the linear matrix inequality. Furthermore, traditional feedback control is inherently a causal system, requiring a passive response only after a road impact occurs and attitude deviations are generated, resulting in an inherent physical delay and making it difficult to completely eliminate severe mismatch disturbances.

[0005] Therefore, existing control strategies often suffer from one disadvantage at the expense of another, making it difficult to simultaneously achieve high-fidelity damping performance, eliminate actuator chatter, and ensure the absolute feasibility of the controller's theoretical solution in complex real-world driving environments. Summary of the Invention

[0006] To overcome the shortcomings of the prior art, this invention proposes a dual adaptive control method, system, device, and medium for vehicle active suspension, in order to solve the problems of deteriorated control performance, high-frequency chattering of actuators, and infeasibility of theoretical solutions for controllers and inherent physical delays in the mechanism when the vehicle active suspension system faces unknown load mutations (severe fluctuations in sprung mass) and continuous nonlinear disturbances.

[0007] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: In a first aspect, the present invention discloses a dual adaptive control method for active suspension of a vehicle, comprising: Establish a dynamic model of linear parameter variation for the vehicle's active suspension system; Obtain the reference speed signal of the road ahead and establish an augmented linear parameter variation dynamic model with an exogenous system. Based on the augmented linear parameter variation dynamic model, obtain the aiming state output. A preview feedforward controller is constructed based on the separation control principle, and the control input is decomposed into a first control component based on the preview state output. Based on Schur's complement theorem, an LMI optimization solver is constructed to obtain the vertex gain, and the vertex gain is used by the HPPD gain scheduling interpolation controller to obtain the second control component; Based on vertex gain, optimization results, and overall control commands, a dynamic adaptive robust gain is calculated using a dual adaptive law with a rigorous physical projection operator; the dynamic adaptive robust gain is then input into a smooth robust compensator based on the hyperbolic tangent function to obtain the third control component. The controlled object output from the linear parameter variation dynamic model of the vehicle active suspension system is combined with the first control component, the second control component and the third control component to obtain the overall control command, which drives the vehicle active suspension system to move.

[0008] Secondly, this invention discloses a dual adaptive control system for a vehicle's active suspension, comprising: The first model building module is used to establish a linear parameter variation dynamic model of the vehicle active suspension system. The second model construction module is used to acquire the reference speed signal of the road ahead and establish an augmented linear parameter change dynamic model with an exogenous system, and obtain the aiming state output based on the augmented linear parameter change dynamic model. The aiming module is used to construct an aiming feedforward controller based on the separation control principle, and decompose the control input into a first control component based on the aiming state output; The interpolation module is used to construct an LMI optimization solver based on Schur's complement theorem to obtain the vertex gain, and the vertex gain is used to obtain the second control component by the HPPD gain scheduling interpolation controller; The adaptive and robust compensation module is used to calculate the dynamic adaptive robust gain based on the vertex gain, optimization results and total control command using a dual adaptive law with a rigorous physical projection operator; the dynamic adaptive robust gain is then input into a smooth robust compensator based on the hyperbolic tangent function to obtain the third control component; The control module is used to combine the controlled object output by the linear parameter change dynamic model of the vehicle active suspension system with the first control component, the second control component and the third control component to obtain the overall control command, and drive the vehicle active suspension system to move.

[0009] Thirdly, the present invention discloses an electronic device, including a memory and a processor, and computer instructions stored in the memory and running on the processor, wherein the computer instructions, when run by the processor, complete the steps of the above-described vehicle active suspension dual adaptive control method.

[0010] Fourthly, the present invention discloses a computer-readable storage medium for storing computer instructions, which, when executed by a processor, complete the steps of the above-described vehicle active suspension dual adaptive control method.

[0011] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention constructs a composite control law with pre-aiming feedforward, integrates an adaptive nominal feedback mechanism with homogeneous polynomial parameter-dependent gain scheduling, designs a smooth and robust compensator based on hyperbolic tangent function, and develops an energy-driven dual adaptive law for dynamically estimating time-varying mass and unknown disturbance upper limits, enabling the active suspension system to achieve high-fidelity vibration isolation while ensuring excellent transient ride comfort.

[0012] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0013] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0014] Figure 1 This is a logic block diagram of the vehicle active suspension dual adaptive control method described in Embodiment 1 of the present invention.

[0015] Figure 2 This is a comparison diagram of the suspension dynamic deflection in scenario one as described in embodiment one of the present invention.

[0016] Figure 3 This is a comparison diagram of the vertical speed of the vehicle body in scenario one as described in embodiment one of the present invention.

[0017] Figure 4 This is a comparison diagram of tire dynamic deflection in scenario one as described in embodiment one of the present invention.

[0018] Figure 5 This is a comparison diagram of the vertical speed of the wheel in scenario one as described in embodiment one of the present invention.

[0019] Figure 6 This is the control force diagram output in scenario one as described in embodiment one of the present invention.

[0020] Figure 7 This is a graph showing the estimation error of the mass parameter in the dual adaptive law scenario described in Embodiment 1 of the present invention.

[0021] Figure 8 This is the dynamic robust gain diagram in the dual adaptive law scenario described in Embodiment 1 of the present invention.

[0022] Figure 9 This is a comparison diagram of the suspension dynamic deflection in scenario two as described in Embodiment 1 of the present invention.

[0023] Figure 10 This is a comparison diagram of the vertical speed of the vehicle body in scenario two as described in embodiment one of the present invention.

[0024] Figure 11 This is the control force diagram output in scenario two as described in embodiment one of the present invention.

[0025] Figure 12 This is the dynamic robust gain diagram in the dual adaptive law of scenario two described in Embodiment 1 of the present invention. Detailed Implementation

[0026] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0027] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0028] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0029] Terminology Explanation: Linear Parameter Variation (LPV) Framework: A control theory framework for handling systems with time-varying or nonlinear parameters, allowing the system matrix to change in real time with the parameters.

[0030] Homogeneous polynomial parameter dependence (HPPD): A gain scheduling mechanism used to provide adaptive smooth gain interpolation as parameters change, thereby extending the stable region of the system.

[0031] Sliding mode control (SMC): A traditional robust control strategy often used to handle matching uncertainties, but its hard switching mechanism is prone to causing high-frequency chattering.

[0032] Linear matrix inequalities (LMI): A mathematical tool commonly used for controller synthesis and solution; traditional passive algebraic boundary methods often lead to excessively large LMI decision variables, causing infeasibility problems.

[0033] Pre-aiming control: an advanced control strategy that uses sensors to obtain information about future system disturbances (such as the elevation of the road ahead) in advance, and outputs control commands in advance through a feedforward mechanism to counteract the disturbances.

[0034] Performance-preserving control: A comprehensive control method that, under the premise of satisfying the robust stability of the system, ensures that a certain performance index (such as a quadratic cost function) of the system has an upper bound through mathematical inequality constraints.

[0035] Schur's complement lemma: a matrix theory tool that can equivalently reduce the dimensionality of high-dimensional matrix inequalities containing nonlinear quadratic terms to linear LMI form, and is a key mathematical means to solve nonlinear matrix inequalities (BMI).

[0036] Uniformly Eventually Bounded (UUB): A stability criterion that states all signals in a closed-loop system will eventually converge and remain within a compact set of residuals.

[0037] Example 1 Active suspension systems, by injecting dynamic compensation forces in real time, theoretically break the physical trade-off between ride comfort and tire contact in traditional passive suspensions. However, in practical applications, vehicles often face significant parameter uncertainties, particularly drastic fluctuations in sprung mass (typically up to ±30%). Traditional fixed-gain controllers exhibit significantly deteriorated performance under these uncertainties, potentially leading to severe system oscillations and performance degradation under extreme overcompensated load mutations. Therefore, this invention aims to provide a scheduling mechanism that adaptively tracks system mass mutations to effectively extend the feasible stability region of the suspension system under extreme load fluctuations, maintaining excellent dynamic control performance.

[0038] While traditional robust strategies (such as sliding mode control, SMC) are often used in practical applications to handle matching uncertainties in systems, they typically rely on rigid hard-switching mechanisms. This mechanism inevitably induces high-frequency actuator chattering, which not only severely reduces ride comfort but also accelerates the mechanical wear of physical actuators. Therefore, this invention aims to design a smooth, continuous robust compensation mechanism that actively suppresses external matching uncertainties while completely eliminating high-frequency actuator chattering, thereby protecting the mechanical structure and optimizing passenger ride comfort.

[0039] From the perspective of theoretical control synthesis, existing robust linear parameter variation (LPV) frameworks typically handle parameter mismatch problems through passive algebraic boundary techniques. This passive approach forces an unrealistic increase in the decision variables of the linear matrix inequality (LMI), leading to severe mathematical conservatism and often causing the fatal problem of LMI unsolvability (i.e., LMI infeasibility) during controller synthesis. Therefore, this invention aims to propose a constructive method to dynamically estimate unknown disturbance boundaries to actively neutralize disturbance terms, thereby systematically separating the feasibility constraints of LMI from rigid disturbance boundaries, achieving controller synthesis with extremely low conservatism and high flexibility.

[0040] While existing active suspension control strategies have made some progress in suppressing disturbances, traditional feedback control is essentially a "causal system," meaning that the controller can only passively generate compensating forces after the vehicle experiences a physical impact from the road surface and undergoes a posture deviation. This "reactive" mechanism has an inherent physical delay, especially when facing severe mismatched disturbances such as speed bumps, making it difficult to completely eliminate vertical impacts. Furthermore, existing LPV robustness frameworks not only suffer from the problem of often being unable to solve LMI (Least Conformity Index), but conventional LMI feasibility solutions often result in excessively small calculated control gains (i.e., "conservative softness"), making the suspension susceptible to breakdown under large impacts.

[0041] Based on the above analysis, although active suspension technology has been extensively explored in theory, existing control strategies often suffer from trade-offs, making it difficult to simultaneously achieve high-fidelity damping performance, eliminate actuator chatter, and ensure the absolute feasibility of the controller's theoretical solution in complex real-world driving environments. Therefore, the engineering community urgently needs a novel comprehensive control scheme that can overcome both theoretical and practical challenges. This patent decouples system stability constraints from external unknown disturbance boundaries, achieving highly flexible and low-conservatism active adaptive control while ensuring excellent transient ride comfort. It breaks causal limitations, proactively reshapes the vehicle's posture in advance, and provides a novel control scheme that balances absolute stability and optimal support stiffness at the mathematical level.

[0042] In one or more embodiments, a dual adaptive control method for vehicle active suspension is disclosed, such as... Figure 1 As shown, in order to fundamentally overcome the physical delay caused by the strict causality of traditional active suspension systems and completely solve the inherent conservative 'soft leg' problem in LMI solution of conventional robust control methods, this invention proposes a composite control method with active predictive capability. When the vehicle's active suspension system faces extreme parameter uncertainties (sprung mass changes) and unknown mismatch nonlinear disturbances (such as the massive impact of speed bumps), this invention first actively reshapes the vehicle's attitude to offset the impact momentum before the physical impact arrives through an optimal preview feedforward mechanism based on the exogenous system. Simultaneously, it constructs a comprehensive control law by combining optimal HPPD gain scheduling based on Schur's performance-based LMI design, a smooth robust compensator, and a dual adaptive law with a projection operator to prevent integral saturation. This allows the vehicle chassis to maintain absolute stability and excellent physical ride comfort even under the combined severe conditions of extreme overcompensation traps and massive impacts. The method specifically includes the following steps: Step S1: Establish a dynamic model of the linear parameter changes of the vehicle's active suspension system.

[0043] This embodiment takes into account the large range of time-varying sprung mass characteristics and continuous nonlinear external disturbances in the active suspension system of a quarter-vehicle during driving. Based on Newton's second law, a dynamic model of the active suspension with decoupling between matched and unmatched disturbances is expressed by dynamic state-space equations.

[0044] Specifically, according to Newton's second law, the equation of motion is expressed as:

[0045]

[0046] in, For time-varying spring load mass, Unsprung mass; The vertical velocity of the vehicle body; The vertical velocity of the wheel; and These represent the vertical displacement of the vehicle body, the vertical displacement of the wheels, and the vertical displacement of the road surface profile, respectively. for; and These represent the stiffness and damping coefficients of the suspension, respectively. This indicates the tire stiffness. The control force is the real-time control input applied to the active suspension.

[0047] The dynamic model of the linear parameter variation (LPV) of a vehicle's active suspension system is represented by the following equation:

[0048] in, Let be the state vector of the system, specifically defined as Each component represents the suspension dynamic deflection, the vehicle vertical velocity, the tire dynamic deflection, and the wheel vertical velocity, respectively. The real and unknown time-varying mass reciprocal parameter is defined as follows: ; This is the real-time control input force acting on the active suspension; The passive spring damping force of the suspension, and External road surface mismatch disturbance input is defined as... ; Nonlinear interference for unknown matching channels; The fundamental state matrix of a quarter-vehicle model. For the input matrix, This is the road surface interference matrix.

[0049] Step S2: Obtain the reference speed signal of the road ahead, and establish an augmented linear parameter change dynamic model with an exogenous system. Based on the augmented linear parameter change dynamic model, obtain the aiming state output.

[0050] This embodiment takes into account that the active suspension system not only faces drastic fluctuations in sprung mass during actual operation, but also needs to handle complex road excitations. In order to achieve proactive defense with "foresight," this embodiment introduces a sensor pre-aiming exogenous system based on the traditional quarter-vehicle model.

[0051] Step S201: Use onboard sensors to obtain information about the road surface ahead.

[0052] Specifically, using onboard sensors (such as LiDAR or stereo cameras) in advance It can obtain the reference speed signal of the road ahead in seconds.

[0053] Step S202: Construct a first-order low-pass pre-aiming exogenous system to simulate the dynamic response characteristics of the sensor and filter out high-frequency noise. The first-order low-pass pre-aiming exogenous system is as follows:

[0054] in, and These are the exogenous system state matrix and input matrix, which determine the cutoff frequency (bandwidth) and time constant of the low-pass filter, respectively (in specific implementations, for example, setting...). This means that a bandwidth of (first-order smoothing filter); The output is the pre-aiming state after smoothing and filtering; It is a first-order low-pass pre-aiming exogenous system; This is the reference speed signal.

[0055] This realistically simulates the hardware signal processing delay of real physical sensors and effectively filters out high-frequency electronic noise and fine gravel interference from asphalt roads.

[0056] Step S203: Combine the linear parameter variation dynamic model of the pre-aiming exogenous system with the one-quarter vehicle active suspension system to construct an augmented LPV dynamic model that includes time-varying sprung mass, nonlinear matching disturbance, and mismatched road surface disturbance.

[0057] The first-order low-pass preview exogenous system and the vehicle's original 4-dimensional state variables are compared. (Representing suspension dynamic deflection, vehicle vertical velocity, tire dynamic deflection, and wheel vertical velocity, respectively) are combined to construct an augmented LPV dynamics system that includes road surface prediction capabilities:

[0058] in, The fundamental state matrix of a quarter-vehicle model. For the input matrix and The road surface interference matrix is ​​specifically expanded as follows:

[0059] In the formula, , These are the stiffness and damping coefficients of the passive suspension. , These represent tire stiffness and unsprung mass, respectively. In the formula... This represents unmeasurable nonlinear lumped disturbances in the matching channels of the suspension system, such as nonlinear mechanical friction within the actuator, hydraulic hysteresis, or unmodeled high-frequency dynamic disturbances. This invention assumes that this disturbance has an unknown physical upper limit. And it satisfies the spatial boundedness assumption. .in, For the unknown nonlinear matching disturbance, the physical upper limit constant is given. The unique symmetric positive definite Lyapunov matrix obtained offline by the Schur-complement LMI optimization solver represents the energy decay characteristics of the closed-loop system. That is, the current 4-dimensional state variable vector of the system. This augmented model legally incorporates external road surface signals, which were originally outside the system, into the system's internal state variables, laying the mathematical foundation for subsequent comprehensive optimal feedforward gain.

[0060] This invention introduces a pre-aiming exogenous system and optimal feedforward gain, breaking the causal limitation of traditional feedback control that must wait for the vehicle body to deviate before exerting force. The controller can pre-command the actuators to produce a small reverse displacement (i.e., active flexion to absorb force) before the wheels actually hit a raised surface. At the moment of impact, the downward momentum of the vehicle body perfectly counteracts the upward impact force of the road surface, reducing the original passive amplitude of tens of millimeters by more than 60%, achieving a leap from passive disturbance rejection to active dynamic neutralization.

[0061] Step S3: Construct a preview feedforward controller based on the separation control principle, and decompose the control input into a first control component based on the preview state output.

[0062] The active control input is decoupled into three functionally independent modules: a nonlinear cancellation term that directly cancels the known passive forces of the suspension, a nominal anticipation state feedback component responsible for global stability and early unloading, and a smooth and robust compensation component dedicated to suppressing unknown lumped nonlinear disturbances.

[0063] Specifically, this invention adopts the concept of separate control, which separates the final control input of the active actuator. The decoupled design consists of three functionally independent components:

[0064] in, To counteract the nonlinear cancellation term of the known passive force of the suspension; The reciprocal of the time-varying quality parameter estimated by active dynamic estimation; The nominal advance state feedback component is responsible for the system's closed-loop stability and early unloading. This is a smooth and robust compensation component specifically designed to suppress unknown lumped disturbances, thus completely overcoming the conservatism of traditional algebraic boundary techniques.

[0065] The first control component is the aiming feedforward control component. It consists of the product of the optimal pre-aiming feedforward gain and the smoothed pre-aiming state output of the first-order low-pass pre-aiming exogenous system, i.e.:

[0066] in, For optimal aiming feedforward gain, This represents the smoothed pre-aiming state output by the first-order low-pass pre-aiming exogenous system.

[0067] This control component is independent of the vehicle body status feedback. Its core function is to instruct the actuator to generate a reverse downward displacement (active knee flexion to unload force) before the wheel actually presses on the road bump, so as to counteract the upcoming road impact momentum.

[0068] Step S4: Based on Schur's complement theorem, construct an LMI optimization solver to obtain the vertex gain. The vertex gain is then processed by an HPPD gain scheduling interpolation controller to obtain the second control component. It is obtained by interpolating the unloaded vertex gain and the fully loaded vertex gain using the barycentric coordinate system weights, and then multiplying by the current system state variable, i.e.:

[0069] in, For unloaded vertex gain, To achieve full vertex gain, The weights are for the centroid coordinate system.

[0070] This control component is specifically responsible for providing the underlying support force to stabilize the vehicle body posture under different time-varying loads.

[0071] It should be noted that the aforementioned first control component With the second control component Together, they constitute the aforementioned nominal aiming state feedback component. ,Right now:

[0072] This embodiment introduces an optimal control quadratic cost function as the performance boundary and uses Schur's complement lemma to reduce the nonlinear algebraic Riccati inequality to a linear polytopic LMI. By setting the optimization objective of maximizing the matrix trace in the solver, the robust HPPD vertex feedback gain, which overcomes the "conservative weakness," and the optimal pre-aiming feedforward gain for the exogenous pavement system are solved offline. During online execution, smooth gain interpolation is performed between polyhedral vertices using the weights of the barycentric coordinate system. By deeply integrating optimal control theory with the LMI method, the nominal preview state feedback control law is designed as follows:

[0073] in, ( The vertex gain represents the locally optimal state feedback obtained through offline solution. It is the underlying support force responsible for "firmly stabilizing the vehicle's attitude." Specifically, The feedback gain is specifically calculated for the absolute no-load limit condition of the vehicle. This is a feedback gain specifically calculated for the vehicle's absolute full-load limit condition. The specific calculation method for both is as follows: [The text abruptly shifts to a different topic] ...within a preset state deviation penalty matrix... and control input penalty matrix Under performance constraints, to maximize matrix traces ( For a convex optimization objective, the linear matrix inequality (LMI) of the multicellular body after dimensionality reduction based on Schur's complement theorem is solved offline by a solver to obtain the global positive definite decision matrix. and the auxiliary solution matrices corresponding to the no-load and full-load boundaries, respectively. and And then through analytical expression and The calculation yielded the result.

[0074] This represents the optimal pre-aiming feedforward gain obtained for the exogenous pre-aiming channel. The specific calculation method is as follows: the vehicle's basic dynamics state space is concatenated with the aforementioned first-order low-pass pre-aiming exogenous system to construct an augmented state space model; then, under a set augmented quadratic cost function, the global optimal feedback gain matrix of the augmented system is solved offline using optimal control theory (such as the linear quadratic regulator LQR algorithm), and the pre-aiming state variables are extracted from it. The gain coefficient is It is independent of the status feedback and determines the magnitude and timing of the "active knee flexion preload unloading" action when the active suspension "sees" an impact on the road ahead.

[0075] The first term of the nominal anti-sighting state feedback control law formula is the state feedback component interpolated based on real-time mass HPPD, and the second term is the optimal anti-sighting feedforward component that triggers active pre-compression unloading of the suspension. Interpolation weights. It is a parameter-dependent convex weight function constructed based on the principle of the centroid coordinate system, and its specific calculation formula is as follows:

[0076]

[0077] in, These represent the physical boundary values ​​of the vehicle's mass reciprocals under absolute full load and absolute no-load conditions, respectively. This construction mathematically guarantees... and This allows the system to smoothly and seamlessly switch and adjust between peak gains when faced with drastic fluctuations in sprung mass.

[0078] The performance-preserving LMI optimization solution process is as follows: A stringent quadratic performance weighting matrix is ​​manually set. and control consumption penalty By introducing the Schur Complement Lemma, the high-dimensional nonlinear matrix inequalities that were originally impossible to solve directly are ingeniously reduced in dimension and equivalently transformed into the following purely linear polytopic LMI constraints:

[0079] in, Let f(x) be a matrix symmetric operator, representing the sum of the matrix and its transpose, i.e. ; Let be the symmetric positive definite decision matrix to be solved by the LMI solver, which is the inverse of the Lyapunov matrix. ; ( ) is an auxiliary variable matrix introduced to transform nonlinear Riccati inequalities into linear matrix inequalities (LMI).

[0080] Set the optimization objective in the solver as follows: (i.e., minimizing the Lyapunov matrix) (norm of). This optimization objective forces the solver to accurately extract the hard core vertex gain that provides the maximum control bandwidth and the strongest physical support on the boundary of the LMI feasible region. Simultaneously, the optimal feedforward gain is independently solved for the exogenous channel. .

[0081] This embodiment introduces Schur's complement lemma and performance-preserving optimization through the aforementioned scheme, completely overcoming the "conservative weakness" phenomenon easily caused by traditional LMI solutions. It not only decouples the nonlinear disturbance bounds of LMI through adaptive laws, but also embeds optimal control (LQR) costs into the LMI constraints, using Schur's complement lemma to transform high-dimensional nonlinear matrix inequalities into a solvable convex optimization problem. Through the optimization mechanism of maximizing matrix traces, the solver is forced to output a hard-core feedback gain with optimal damping characteristics, completely solving the industry pain point of soft suspension and susceptibility to breakdown in conventional robust control.

[0082] Based on this, the attached diagram The parameter is the HPPD gain scheduling interpolation control component. The specific implementation process is as follows: During the online operation phase, the system uses the reciprocal of the currently dynamically estimated quality parameter... Real-time calculation of the weights of the centroid coordinate system and Subsequently, the offline-obtained idle vertex gain was calculated. and full-load vertex gain Perform smooth interpolation according to this weight, and multiply by the current state variable vector of the system. Finally, the second control component is obtained by synthesis. This control component is specifically responsible for providing smooth and adaptive optimal underlying support force for the vehicle chassis when facing a wide range of time-varying loads.

[0083] Step S5: Based on the vertex gain, optimization results and overall control command, the dynamic adaptive robust gain is calculated using a dual adaptive law with a rigorous physical projection operator; the dynamic adaptive robust gain is input into a smooth robust compensator based on the hyperbolic tangent function to obtain the third control component.

[0084] Step S501: Design an anti-windup dual adaptive law with a rigorous physical projection operator.

[0085] Under extreme conditions (such as massive road impacts coupled with sudden mass shifts), a surge in the system's energy gradient can easily lead to catastrophic "integrator windup" in parameter estimation. To address this, the energy-driven adaptive law of this invention forcibly incorporates a rigorous physical projection operator. proj ):

[0086]

[0087] in, The adaptive update rate of the inverse estimate of the time-varying spring load mass (i.e., the derivative of the estimated parameter with respect to time) is the corresponding... The inverse of the unknown time-varying mass (true parameter) Online dynamic estimates of ) The update rate for the dynamically adaptive robust gain (i.e., the estimated derivative of the gain with respect to time). For dynamic adaptive robust gain, i.e., the physical upper limit of unknown nonlinear matching interference ( The estimated value; For physical projection operators; The learning rate is adapted to the quality parameter; Leakage gain is a quality parameter. For dynamic robust gain learning rate; This is for dynamic robust leakage gain. The anti-integral saturation mechanism works as follows: when the system encounters a massive impact, causing the estimated value to hit the physical limit boundary (such as reaching the full load limit),... Furthermore, when the derivative still attempts to drive the parameter to integrate beyond its limits, the projection operator will instantly truncate the derivative to 0. This safety valve mechanism ensures that the online estimated parameters never exceed the actual physical bearing limit, greatly improving the system's extreme survivability.

[0088] This embodiment relies on the system's Lyapunov energy gradient to design an adaptive update law that drives the dynamic approximation of unknown mass parameters and perturbation upper limits. The adaptive law innovatively introduces a physical projection operator and a leakage term (…). σ (-modification) While actively neutralizing disturbances and decoupling LMI constraints, it strictly cuts off the integral wind saturation phenomenon under extreme impacts, ensuring that the parameters estimated online never exceed the physical bearing capacity boundary.

[0089] Based on an improved dual-adaptive mechanism with a rigorous projection operator and HPPD interpolation, this invention provides extremely strong fault tolerance and protection against integral saturation under extreme load mutations. It not only perfectly absorbs mass mutations of up to ±30%, but also eliminates parameter integral windup caused by massive shocks at the algorithm's underlying level. Simulations confirm that even under extreme overcompensation traps, the parameter estimation error is locked at 10. Within the Level 3 safety boundary, ensure the absolute stability of the vehicle chassis system.

[0090] Step S502: Design a smooth and robust compensator to eliminate high-frequency wear of the actuator.

[0091] To suppress the uncertainty of lumped matching Furthermore, to strictly avoid the "chattering" defect of sliding mode control, this invention utilizes a continuously differentiable hyperbolic tangent function (tanh) to reconstruct a smooth compensator:

[0092] in, It is the thickness of the smooth boundary layer; For the dynamic adaptive robust gain, its current value is calculated from the update rate obtained from the aforementioned double adaptive law formula. The result is obtained through real-time time integration, thus achieving on-demand dynamic compensation based on the degree of system disturbance. This design ensures absolute smoothness of control force and significantly extends the service life of mechanical actuators.

[0093] This embodiment utilizes a continuously differentiable hyperbolic tangent function to replace the hard-switching surface of the sign function in traditional sliding mode control. Combined with dynamic robust gain, it actively suppresses the uncertainty of the matching channel and completely eliminates high-frequency actuator chatter. Specifically, this invention addresses the shortcomings of traditional robust control strategies that rely on hard-switching mechanisms, leading to severe high-frequency actuator chatter, accelerated mechanical wear, and deteriorated ride comfort. It innovatively designs a smooth robust compensator based on the hyperbolic tangent function (tanh). This compensator not only actively and efficiently suppresses nonlinear matching uncertainties in the system but also outputs continuous, smooth control compensation force, eliminating high-frequency actuator chatter and greatly optimizing the vehicle's transient ride comfort.

[0094] Step S6: Combine the controlled object output by the linear parameter change dynamic model of the vehicle active suspension system with the first control component, the second control component and the third control component to obtain the total control command, and drive the vehicle active suspension system to move.

[0095] The master control command is:

[0096] Meanwhile, the overall control command is fed back to the dual adaptive law with physical projection operator for cyclic optimization.

[0097] Step S7: Verify the uniformly eventually bounded (UUB) stability of the system by constructing a comprehensive set of Lyapunov candidate functions and perform simulation verification.

[0098] Preferably, a rigorous mathematical proof is provided for the uniformly eventually bounded (UUB) stability of the system.

[0099] The core technical advantage of this invention lies in its complete decoupling of the perturbation boundary from the LMI solution through an energy-driven dual adaptive law, while ensuring global system stability. A rigorous Lyapunov stability proof is presented below: Define the parameter estimation error as The dynamic robust gain estimation error is .

[0100] Step S701: Construct synthetic Lyapunov candidate functions: Choose a positive definite comprehensive energy function that includes system state energy, parameter estimation error, and gain estimation error. :

[0101] Step S702: Derivative decomposition and algebraic summation of system energy: right Differentiate the equations and substitute them into the controlled dynamic equations. Use algebraic identities. Passive suspension force With precise cancellation, the system's energy derivative can be fully expanded as follows:

[0102] in, For external road surface mismatched disturbance input (i.e., road surface vertical velocity excitation, satisfying) ); the first term of the formula Defined as the current nominal state feedback gain of the system. It is the dynamic lumped gain obtained by mixing and interpolating real-time quality estimates using the HPPD mechanism when the controller is running online.

[0103] Step S703, Proof of component scaling and error cancellation: (1) Nominal pre-aiming stabilization: Based on the strict constraint of performance-preserving LMI in step S3, the vertex gain calculated offline ensures that regardless of How to interpolate between vertices, the closed-loop nominal system must satisfy This represents the system's powerful ability to dissipate and stabilize energy. It also provides protection against residual pre-targeting exogenous interference. Applying Young's inequality for scaling:

[0104] in, Represents the performance weighting matrix The minimum eigenvalue is an important mathematical parameter for measuring the lower bound of the system's energy dissipation; due to the optimal aiming gain The active compensation effect, the interference term after physical cancellation It is greatly simplified and compressed.

[0105] (2) Perfect neutralization of parameter errors: Substitute the adaptive update law from step S5. Based on the core mathematical properties of the projection operator... That is, the projection operator only accelerates convergence and never diverges, safely removing... After the operator, the cross-coupling terms are perfectly neutralized, leaving only the leakage terms. :

[0106] (3) Smoothing suppression of matched interference: Introducing a known upper limit for interference. Substitute And applying the Smooth Sliding Mode Lemma, that is, there exists a constant... Make We can obtain:

[0107] Substitution Adaptive law The cross terms are canceled out again, and the leakage term is obtained after formula processing:

[0108] Step S704, Dissipative Inequality Synthesis and Uniform Final Bounded UUB Conclusion: Integrating all the scaling steps above, the final energy dissipation inequality of the system converges strictly as follows:

[0109] Among them, the bounded set residual .

[0110] Define global decay rate Then the inequality can be simplified to its standard form:

[0111] in, The aforementioned positive definite synthetic Lyapunov candidate energy function The derivative with respect to time represents the rate of decay (or rate of change) of the total energy of the closed-loop system. The validity of the above standard form dissipation inequality rigorously proves mathematically that the composite control law of this invention can guarantee that all state errors and parameter estimation errors of the closed-loop system are uniformly and ultimately bounded (UUB).

[0112] Furthermore, this example uses the computer software MATLAB to rigorously simulate and verify the invented control method. The specific simulation implementation steps are as follows: 1. Simulation system parameters and solver settings: For the quarter vehicle active suspension system, the basic physical parameters of the model are selected as shown in Table 1.

[0113] Table 1 Selection of Basic Physical Parameters

[0114] Because this invention relates to large-scale mass jumps and smooth, robust compensation (tanh function), the system exhibits pronounced stiff dynamic characteristics. The simulation employs the high-precision solver ode15s, specifically designed for stiff differential equation systems, with the relative error tolerance (RelTol) strictly set to [value missing]. The maximum integration step size (MaxStep) is limited to 1. Second.

[0115] The basic physical parameters of the one-quarter of the vehicles under investigation are as follows: nominal sprung mass Unsprung mass Suspension passive stiffness Damping coefficient Tire stiffness

[0116] The physical boundary of mass is set as the nominal mass. That is, no-load limit Full load limit The physical projection boundary of the adaptive parameter estimation is strictly locked as follows: The bandwidth matrix of the pre-aiming exogenous filter is set to Pre-aiming time .

[0117] 2. Controller offline integration and adaptive parameter calibration.

[0118] Offline performance-preserving LMI solution: Define the state performance weighting matrix. Control and punishment Using the YALMIP toolbox To maximize the objective, solve the Schur-complemented multicellular LMI to obtain the vertex feedback gain with optimal support stiffness. and aiming feedforward gain

[0119] Online adaptive parameter setting: quality parameter adaptive learning rate Leakage gain Dynamic robust gain learning rate Set according to the intensity of the working conditions. Between, leakage gain To ensure rapid recovery after impact; to smooth the boundary layer thickness. Meanwhile, a conventional linear quadratic regulator (LQR) based on full-load harsh operating conditions was designed as a comparison baseline.

[0120] 3. Simulation experiments in different scenarios.

[0121] (1)Scenario 1 is the road surface protrusion and quality change test.

[0122] Combined working condition of single speed bump impact and sudden mass change, working condition setting: simulation duration 6 seconds. During this period, an angular frequency of 10 ... The integral amplitude reached Typical speed bump impact; in At any given moment, the sprung mass changes from the nominal value. Sudden increase Full load; nonlinear matching friction disturbance injected throughout the process. .

[0123] Simulation Results Analysis: Active Pre-stressing and Rapid Convergence: In Before the actual physical impact of the speed bump occurs (approximately The suspension controlled by this invention utilizes optimal pre-aiming gain. It initiated an extremely smooth downward "knee-bending pre-compression" maneuver. At the moment of impact, the downward momentum of the vehicle body counteracted the upward impact force, raising the LQR baseline to nearly [missing information]. The passive amplitude was flattened The above, and in There were no aftershocks and the earthquake subsided perfectly.

[0124] Verification of the anti-integral saturation safety valve: Faced with severe impact, the physical projection operator (Proj) in the underlying code acts like an iron wall, preventing the integrator from accumulating beyond its limits. This prevents the quality estimation error from diverging; instead, it exhibits a controlled "small square wave" rebound, with the error peak locked within a certain range. The extremely small safety level.

[0125] Smooth Control Force: Dynamic Robust Gain It instantly awakens and suppresses nonlinear disturbances, and the output control force is extremely smooth throughout the process, completely eliminating the high-frequency chattering of traditional sliding mode control.

[0126] (2) Scenario 2 is an extremely random road surface and overcompensated trap test.

[0127] Simulation duration: Continuous random road surface and extreme mass trap combined working condition setting The road surface excitation was extremely severe: superimposed with and Two consecutive random sinusoidal road surfaces of different frequencies, and in High-intensity sinusoidal turbulence was superimposed; the matching disturbance intensity was increased to

[0128] Creating an "overcompensation trap": vehicle front At the absolute no-load limit ( ),exist At that moment, the mass undergoes a sudden jump across the entire range, abruptly increasing to the full load limit. ).

[0129] Simulation effect analysis: A stability limit akin to a dimensional reduction attack: Under continuous and severe turbulence, the LQR baseline, due to its "soft legs" and fixed gain defects, experienced large-amplitude and prolonged violent oscillations (suspension deflection approaching...). The vehicle's speed approached This invention, however, smoothly reduces the oscillation amplitude by nearly [amount missing]. .

[0130] Perfectly resolving quality pitfalls: In During the ultimate mass transition, the HPPD centroid coordinates are utilized. With its underlying real-time interpolation mechanism and smooth transition of the tanh function, the system responds seamlessly to load mutations without generating any mathematical singularities or mutation spikes, demonstrating the algorithm's extreme survivability and excellent robustness in extremely harsh environments.

[0131] Results Explanation: like Figure 1 The diagram shows the structural interaction of an active suspension dual adaptive smooth robust controller based on performance-preserving HPPD and anti-seepage feedback. Each box in the diagram represents a core control module within the system, and the arrows clearly indicate the calculation and transmission path of sensor states (such as anti-seepage road surface elevation), parameter estimates (such as mass estimates and disturbance limits), and control commands between the controlled object and the control module at a specific moment.

[0132] like Figure 2 , Figure 3 , Figure 4 and Figure 5 As shown, the diagram illustrates the response comparison of the active suspension in four core states (suspension dynamic deflection, vehicle vertical velocity, tire dynamic deflection, and wheel vertical velocity) under Scenario 1 (single speed bump impact and mass mutation test). The black dashed line in the diagram represents the traditional LQR baseline, and the black solid line represents the control method of this invention. It can be clearly observed that, before the actual physical impact of the speed bump occurs at 0.5s (approximately 0.35s), this invention, with its optimal anticipation gain, preemptively executes a smooth reverse dive "pre-compression" action. At the moment of impact, compared to the extremely large passive amplitude of the LQR baseline, this invention reduces the peak values ​​of indicators such as dynamic deflection by more than 60%, and achieves perfect convergence with no aftershocks within 1.5s. It achieves the optimal balance between suspension deflection and vehicle velocity within a very small predetermined range, effectively preventing attitude loss of control.

[0133] like Figure 6 As shown, this demonstrates the control force output in Scene 1. Curve. Under the same nonlinear matching friction and environmental parameters, the smooth and robust control method of this invention (solid black line) outputs an extremely continuous and smooth compensation force, remaining stable within safe physical boundaries throughout the entire process. It eliminates the high-frequency chattering problem caused by the hard switching mechanism relied upon by traditional sliding mode control (SMC), actively isolates external disturbances, protects the physical actuator from high-frequency fatigue wear, and greatly optimizes transient smoothness.

[0134] like Figure 7 and Figure 8 As shown, the estimation error of the mass parameter of the dual adaptive law in scenario one ( and dynamic robust gain The evolution curve. For example... Figure 7 As shown, when faced with a mass shift and combined impact of up to 30% at time 1.0s, the underlying physical projection operator (Proj) acts as a "safety valve" against integral saturation, forcibly truncating out-of-bounds integration. The mass estimation error not only does not diverge, but instead exhibits a controlled, slight rebound, with its peak value locked at an extremely small level. Within safe limits. Figure 8 This demonstrates that the dynamic gain can instantly recover and provide strong compensation when a disturbance occurs, and smoothly fall back after the disturbance, confirming the efficiency and reliability of the dual adaptive mechanism.

[0135] like Figure 9 and Figure 10 The figure shows the comparison curves of suspension dynamic deflection and vehicle vertical speed, the core comfort indicators, under scenario two (the extreme condition of superimposed continuous random road surface and a 2.5s full-load passage through a compensation trap). During continuous severe bumps, the LQR baseline (black dashed line) experienced large-amplitude and prolonged violent oscillations due to its conservatism and fixed gain limitations; while the method of this invention successfully compressed the oscillation amplitude by nearly 70%. Furthermore, when the extreme mass jump occurs at 2.5s, thanks to the real-time interpolation mechanism of HPPD scheduling, the system responded seamlessly without generating any mathematical singularities or abrupt spikes, demonstrating significantly stronger extreme survivability compared to the LQR method.

[0136] like Figure 11 and Figure 12 As shown, the control force under extremely harsh combined working conditions in Scenario 2 With dynamic robust gain The results further confirm that even when encountering continuous high-intensity random sinusoidal turbulence and unknown nonlinear disturbances, the present invention can still maintain a smooth control output without chattering, and the robust gain can be adaptively and smoothly scheduled at extremely high frequencies according to the strength of the disturbance energy, proving the absolute robustness and high-fidelity vibration reduction performance of the control system in complex and harsh environments.

[0137] Existing active suspension feedback control strategies are all strictly causal systems, meaning the actuators can only passively generate compensating force after the wheel experiences a physical impact and displacement deviation, resulting in an unavoidable physical response delay. This invention introduces a first-order low-pass exogenous system and an optimal anticipation feedforward control link. By anticipating... The system acquires the road surface elevation ahead in seconds and independently solves for the feedforward gain under the augmented LPV model, enabling the controller to proactively output compensation commands before the impact occurs. Simulation tests show that, when facing amplitude... When encountering a typical speed bump impact, the method of this invention can drive the suspension to generate a pre-compression relief action with early reverse dive. Compared with the traditional LQR controller, this pre-compression mechanism significantly reduces the passive amplitude of the suspension dynamic deflection peak by more than 60%, and shortens the system's oscillation convergence time to less than 1.5 seconds. It breaks the causal physical delay of traditional feedback control and significantly improves the transient vibration reduction performance when facing sudden road changes.

[0138] To address the problems of existing robust LPV control frameworks often employing passive algebraic boundary conditions to handle parameter mismatch, leading to excessively large LMI (linear matrix inequality) decision variables, frequent infeasibility (no solution), or excessively small feedback gain (i.e., soft suspension support), this invention addresses these issues. Firstly, it actively neutralizes nonlinear disturbances in the Lyapunov derivative through an energy-driven dual adaptive law, completely decoupling the disturbance boundary from the LMI decision matrix. Secondly, it introduces Schur's complement lemma to equivalently reduce the nonlinear Riccati inequality of optimal control to a polycellular LMI, and maximizes the matrix trace (… The objective function is optimized and solved. Theoretically, this method strictly guarantees the uniformly eventually bounded (UUB) stability of the closed-loop system. Simulation and quantization results show that, in superposition... and Under harsh conditions of compound continuous random bumps and extreme nonlinear disturbances, compared with traditional controllers that experience severe oscillations, this invention reduces the oscillation amplitude of vehicle vertical velocity and suspension deflection by nearly 70% thanks to its optimal damping characteristics. It eliminates the conservative problem of LMI solution from the mathematical level, greatly expanding the optimal stability margin of the system.

[0139] To address the problem that traditional robust sliding mode control (SMC) strategies rely on hard-switching mechanisms using a sign function (sgn) to suppress uncertainties, leading to high-frequency oscillations (chattering) in the control force output and accelerated actuator wear, this invention constructs a method based on a hyperbolic tangent function combined with smooth boundary layer parameters (…). The smooth, robust compensator replaces the discontinuous hard switching surface with a continuously differentiable smooth transition layer, and incorporates dynamic robust gain. Compensation is performed as needed. When continuously injecting nonlinear matched friction disturbances of equal intensity, compared to the extremely high-frequency jumps of hundreds of Newtons in traditional SMC, the compensation control force output by this invention is continuously smooth (stable throughout the entire process). Within the physical safety boundaries, this eliminates high-frequency fatigue wear of electromagnetic or hydraulic actuators at its source, while effectively reducing the vehicle's vertical acceleration. This eliminates high-frequency vibration of the actuators, extends the lifespan of the mechanical structure, and improves ride smoothness.

[0140] When encountering massive road impacts or large-scale mass shifts, the energy gradient of traditional adaptive controllers surges, easily leading to divergent "integral wind saturation" in parameter estimates, thus causing system collapse and instability. This invention rigorously embeds the physical projection operator (Proj) and leakage terms into the energy-driven mass adaptive update law. Simultaneously, it combines the HPPD mechanism, using barycentric coordinate weights to interpolate between the unloaded and fully loaded vertex gains. Therefore, when encountering massive road impacts or large-scale mass shifts, the energy gradient of traditional adaptive controllers surges, easily causing divergent "integral wind saturation" in parameter estimates, leading to system collapse and instability. Instantaneous change in sprung mass (e.g.) No-load jump to Under extreme overcompensation traps involving full load and combined massive road impacts, the projection operator successfully forcibly truncated the divergent integral derivative. Simulation results show that the peak value of the mass parameter estimation error is safely and rigorously locked within a certain range. Within an extremely small range, algorithm divergence is prevented, ensuring the vehicle chassis's extreme survivability under harsh conditions. The integral wind saturation path under extreme impact is completely severed, providing extremely strong fault tolerance against load mutations.

[0141] Example 2 In one or more embodiments, a dual adaptive control system for active vehicle suspension is disclosed, specifically including: The first model building module is used to establish a linear parameter variation dynamic model of the vehicle active suspension system. The second model construction module is used to acquire the reference speed signal of the road ahead and establish an augmented linear parameter change dynamic model with an exogenous system, and obtain the aiming state output based on the augmented linear parameter change dynamic model. The aiming module is used to construct an aiming feedforward controller based on the separation control principle, and decompose the control input into a first control component based on the aiming state output; The interpolation module is used to construct an LMI optimization solver based on Schur's complement theorem to obtain the vertex gain, and the vertex gain is used to obtain the second control component by the HPPD gain scheduling interpolation controller; The adaptive and robust compensation module is used to calculate the dynamic adaptive robust gain based on the vertex gain, optimization results and total control command using a dual adaptive law with a rigorous physical projection operator; the dynamic adaptive robust gain is then input into a smooth robust compensator based on the hyperbolic tangent function to obtain the third control component; The control module is used to combine the controlled object output by the linear parameter change dynamic model of the vehicle active suspension system with the first control component, the second control component and the third control component to obtain the overall control command, and drive the vehicle active suspension system to move.

[0142] Example 3 This embodiment provides an electronic device, including a memory and a processor, as well as computer instructions stored in the memory and running on the processor. When the computer instructions are executed by the processor, they complete the steps of the above-described vehicle active suspension dual adaptive control method.

[0143] Example 4 This embodiment provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, complete the steps of the above-described vehicle active suspension dual adaptive control method.

[0144] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0145] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0146] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment, whereby a series of operational steps are performed to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0147] The descriptions of each embodiment in the above embodiments have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0148] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A dual adaptive control method for active vehicle suspension, characterized in that, include: Establish a dynamic model of linear parameter variation for the vehicle's active suspension system; Obtain the reference speed signal of the road ahead and establish an augmented linear parameter variation dynamic model with an exogenous system. Based on the augmented linear parameter variation dynamic model, obtain the aiming state output. A preview feedforward controller is constructed based on the separation control principle, and the control input is decomposed into a first control component based on the preview state output. Based on Schur's complement theorem, an LMI optimization solver is constructed to obtain the vertex gain, and the vertex gain is used by the HPPD gain scheduling interpolation controller to obtain the second control component; Based on vertex gain, optimization results, and overall control commands, a dynamic adaptive robust gain is calculated using a dual adaptive law with a rigorous physical projection operator; the dynamic adaptive robust gain is then input into a smooth robust compensator based on the hyperbolic tangent function to obtain the third control component. The controlled object output from the linear parameter variation dynamic model of the vehicle active suspension system is combined with the first control component, the second control component and the third control component to obtain the overall control command, which drives the vehicle active suspension system to move.

2. The vehicle active suspension dual adaptive control method as described in claim 1, characterized in that, The augmented linear parameter variation dynamic model with exogenous system is specifically as follows: in, The basic state matrix of a quarter-vehicle model; The input matrix; This is the road surface interference matrix; and These are the exogenous system state matrix and input matrix, which determine the cutoff frequency and time constant of the low-pass filter, respectively. The output is the pre-aiming state after smoothing and filtering; It is a first-order low-pass pre-aiming exogenous system; For reference speed signal; Let this be the system's state vector; The time-varying reciprocal of the mass parameter; This refers to the passive spring damping force of the suspension. This is the real-time control input force acting on the active suspension; This is nonlinear interference for an unknown matching channel.

3. The vehicle active suspension dual adaptive control method as described in claim 1, characterized in that, The physical projection operator is: in, The adaptive update rate is the inverse estimate of the time-varying spring load mass. The update rate for the dynamically adaptive robust gain; For dynamic adaptive robust gain; For physical projection operators; These represent the physical boundaries of the reciprocal mass of the vehicle under absolute full load and absolute no load limits, respectively. This is the current 4-dimensional state variable vector of the system; It is a Lyapunov matrix; The learning rate is adapted to the quality parameter; Leakage gain is a quality parameter. For dynamic robust gain learning rate; For dynamic robust leakage gain; The input matrix; This is the nominal aiming state feedback component; For smooth and robust compensation components.

4. The vehicle active suspension dual adaptive control method as described in claim 1, characterized in that, The smooth robust compensator based on the hyperbolic tangent function obtains the third control component as follows: in, It is the thickness of the smooth boundary layer; For dynamic adaptive robust gain; The input matrix; It is a Lyapunov matrix; This is the current 4-dimensional state variable vector of the system.

5. The vehicle active suspension dual adaptive control method as described in claim 1, characterized in that, The solution method for constructing the LMI optimization solver based on Schur's complement theorem is as follows: in, It is a matrix symmetric operator. Let be the symmetric positive definite decision matrix to be solved by the LMI solver. For auxiliary variable matrix; This is the state deviation penalty matrix; To control the input penalty matrix; The basic state matrix of a quarter-vehicle model; The input matrix; The optimization objective in the solver is set to minimize the Lyapunov matrix. The norm is used to obtain the vertex gain.

6. The vehicle active suspension dual adaptive control method as described in claim 5, characterized in that, The vertex gain is used by the HPPD gain scheduling interpolation controller to obtain the second control component. This component is then interpolated between the unloaded and fully loaded vertex gains using the centroid coordinate system weights, and multiplied by the current system state variable to obtain: in, For unloaded vertex gain, To achieve full vertex gain, The weights are for the centroid coordinate system.

7. The vehicle active suspension dual adaptive control method as described in claim 6, characterized in that, The interpolation weights are parameter-dependent convex weight functions constructed based on the principle of the centroid coordinate system, specifically: in, These represent the physical boundaries of the reciprocal mass of the vehicle under absolute full load and absolute no load limits, respectively. This is an online dynamic estimate of the inverse of the unknown time-varying mass.

8. A dual adaptive control system for vehicle active suspension, characterized in that, include: The first model building module is used to establish a linear parameter variation dynamic model of the vehicle active suspension system. The second model construction module is used to acquire the reference speed signal of the road ahead and establish an augmented linear parameter change dynamic model with an exogenous system, and obtain the aiming state output based on the augmented linear parameter change dynamic model. The aiming module is used to construct an aiming feedforward controller based on the separation control principle, and decompose the control input into a first control component based on the aiming state output; The interpolation module is used to construct an LMI optimization solver based on Schur's complement theorem to obtain the vertex gain, and the vertex gain is used to obtain the second control component by the HPPD gain scheduling interpolation controller; The adaptive and robust compensation module is used to calculate the dynamic adaptive robust gain based on the vertex gain, optimization results and total control command using a dual adaptive law with a rigorous physical projection operator; the dynamic adaptive robust gain is then input into a smooth robust compensator based on the hyperbolic tangent function to obtain the third control component; The control module is used to combine the controlled object output by the linear parameter change dynamic model of the vehicle active suspension system with the first control component, the second control component and the third control component to obtain the overall control command, and drive the vehicle active suspension system to move.

9. An electronic device, characterized in that, It includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, which, when executed by the processor, perform the vehicle active suspension dual adaptive control method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, Used to store computer instructions, which, when executed by a processor, complete the vehicle active suspension dual adaptive control method according to any one of claims 1-7.

Images