A method and system for target dynamic optimization of an axle front preview active suspension control

Abstract

The application relates to a kind of axle front preview active suspension control target dynamic optimization method and system, the axle front preview active suspension control target dynamic optimization method, with driving style quantitative parameter as core control variable, reconstructs axle front preview active suspension multi-objective control logic, including steps S1: coupling information acquisition and preprocessing;S2: driving style feature extraction and quantification;S3: the mapping relationship between driving style quantification value, vehicle dynamics work interval, axle front preview space mismatch deviation is established, and the deviation of axle front preview path and timing is corrected;S4: dynamic weight evolution and normalized solution;S5: dynamic weight coefficient is integrated into axle front preview model predictive control algorithm, and the corrected preview road information is coupled, and the optimal suspension control instruction is solved and the actuator is driven to act;S6: online iterative optimization, realizes full-condition adaptive closed-loop control.The application significantly improves the preview compensation precision, and greatly improves the full-scene adaptability.

Description

Technical Field

[0001] This invention relates to a vehicle control technology, specifically to a dynamic optimization method and system for active suspension control targets that integrates driver handling style and axle front aiming correction, applicable to intelligent chassis of passenger cars and commercial vehicles, and particularly for active suspension systems with axle front aiming function. Background Technology

[0002] Active suspension with front axle anti-collision capability can significantly improve vehicle ride quality by proactively sensing road disturbances and compensating for them in advance. However, existing technology has three major drawbacks:

[0003] (1) Fixed control targets and poor adaptability to all scenarios: Traditional control schemes adopt fixed target weight allocation strategies without dynamic adjustment based on driver's handling characteristics, resulting in serious lack of scenario adaptability. For example, under stable driving conditions, the handling stability weight is redundant and the comfort control is weak, resulting in obvious driving and riding bumps; under aggressive driving conditions, the comfort weight is too high and the tire grip is insufficient, which can easily cause handling stability hazards such as body roll and understeer. It cannot match the driver's real handling needs and driving scenario requirements throughout the entire process.

[0004] (2) Pre-aiming accuracy is severely affected by driving behavior: Aggressive driving is prone to cause wheel track deviation, resulting in mismatch between the pre-aiming path and the actual wheel track, compensation lag, or even control failure. Existing technologies have not established a correlation correction mechanism between driving behavior and pre-aiming accuracy.

[0005] (3) The project is complex, costly, and not practical enough: some solutions rely on high-end peripherals and have high modification costs, making it difficult to achieve mass implementation and affecting the implementation and promotion of the technology.

[0006] Based on the above shortcomings, the existing axle front anti-sight suspension can only perform under ideal and stable conditions. Its advantages are greatly reduced in real and complex driving scenarios. Therefore, there is an urgent need for an optimization solution that fits the driver's control, adapts to anti-sight deviations, and is low-cost and easy to implement. Summary of the Invention

[0007] This invention addresses the technical shortcomings of existing axle-front anti-aiming active suspension systems, such as fixed control target weights, neglecting the influence of driver handling characteristics on anti-aiming accuracy, poor adaptability to all scenarios, and insufficient engineering practicality. It proposes a dynamic optimization method and system for the control target of axle-front anti-aiming active suspension.

[0008] This invention breaks through the traditional fixed-weight control framework, transforming driving style from a disturbance source into a core control variable. With driving style as the core regulation variable, it constructs a triple nonlinear mapping system of driving style-dynamics-preview deviation, designs a non-jumping dynamic weight evolution mechanism, and realizes that the control target weight is adaptively adjusted according to driving style, synchronously correcting preview timing and trajectory deviation. The dynamic weight is integrated into the preview control algorithm to complete precise advance compensation, achieving deep synergy between preview compensation and driving control.

[0009] Invention technical solution:

[0010] A dynamic optimization method for front axle anti-axle active suspension control targets, using driving style quantification parameters as core control variables, reconstructs the multi-objective control logic of front axle anti-axle active suspension. Specific steps include:

[0011] S1: Coupled Information Acquisition and Data Preprocessing

[0012] Simultaneously collect driver control behavior data, vehicle dynamic response data, and axle-front anti-road elevation data to complete spatiotemporal synchronization of multi-source data, noise reduction filtering, and normalization preprocessing.

[0013] S2: Driving Style Feature Extraction and Continuous Quantization

[0014] Extract multidimensional features of driving style, construct a continuous quantitative representation model, and output normalized quantitative values ​​of driving style to realize the transformation of driving style from qualitative classification to quantitative parameters;

[0015] S3: Triple Mapping Relationship Modeling

[0016] Establish a nonlinear mapping relationship among driving style quantification value, vehicle dynamics working range, and axle front aiming space mismatch deviation, and correct the axle front aiming path and timing deviation based on the mapping relationship;

[0017] S4: Dynamic Weight Evolution and Normalization Solution

[0018] Based on the quantitative value of driving style, a dynamic weight evolution mechanism for control objectives is constructed to calculate the dynamic weight coefficients of three major objectives: vehicle comfort, suspension travel constraint, and tire adhesion in real time. The weight coefficients satisfy the normalization constraint and have no step jump.

[0019] S5: Integrated Pre-aiming Closed-Loop Control

[0020] The dynamic weighting coefficients are incorporated into the axle-front aiming model predictive control algorithm, coupled with the corrected aiming road surface information, and the optimal suspension control command is solved in a rolling manner to drive the actuator action.

[0021] S6: Online Iterative Optimization

[0022] By collecting real-time feedback data from vehicles and iteratively optimizing mapping parameters and weight evolution parameters online, adaptive closed-loop control under all operating conditions can be achieved.

[0023] In step S2, a multi-dimensional driving style feature vector is constructed based on the driver's control behavior and vehicle dynamic response. The driving style is continuously and quantitatively represented by a neural network model, and a normalized driving style quantification value is output to achieve real-time driving style recognition without manual threshold judgment or classification delay.

[0024] The multidimensional feature vector includes at least handling intensity features, handling frequency features, anti-aiming and following characteristics, and dynamic response hysteresis features. The continuous quantitative representation output range is λ∈[0,1], which can adaptively match all types of driving behaviors, including smooth driving, mild driving, and aggressive driving: λ∈[0,0.3] represents smooth driving, with gentle handling and high anti-aiming and following; λ∈(0.3,0.7] represents mild driving, with balanced handling, balancing comfort and stability; λ∈(0.7,1] represents aggressive driving, with intense handling, large wheel track deviation, and high risk of anti-aiming mismatch. This quantification method accurately tracks the handling changes under transient driving conditions, adapts to transient conditions such as rapid acceleration, sharp steering, and continuous lane changes, and provides a unified quantitative input for the dynamic control of the axle front anti-aiming active suspension.

[0025] In step S3, a nonlinear explicit mapping relationship is established between the driving style quantification value, vehicle dynamics parameters, and axle front aiming deviation.

[0026] The mapping relationship uses normalized driving style quantification as the independent variable and vehicle comfort parameters, suspension travel parameters, anti-track deviation parameters, and tire adhesion parameters as dependent variables, where: a z-rms : The effective value of the vehicle's vertical acceleration, representing an indicator of ride comfort; a z0 : Reference value of vertical acceleration of vehicle body under stable driving conditions; S max S0: Suspension dynamic travel limit value, representing the safety constraint threshold of the suspension structure; Δe: Suspension dynamic travel reference value under stable driving conditions; max ∆e0: Maximum lateral deviation between the pre-aiming trajectory and the wheel track, representing the degree of pre-aiming mismatch; ∆e0: Reference value of lateral deviation of pre-aiming under stable driving conditions; : Tire vertical dynamic load change rate, a characterizing tire adhesion stability index; : Baseline value of tire vertical dynamic load change rate under stable driving conditions; λ: Normalized driving style quantification value, ranging from [0,1]; k a k s k e k f ξ: First-order linear calibration coefficient, characterizing the linear rate of change of the parameter with driving style; aξ f Second-order nonlinear calibration coefficients characterize the nonlinear correction magnitude of parameters with driving style; the included first-order linear calibration coefficients k a k s k e k f With the second-order nonlinear calibration coefficient ξ a ξ f The solution is obtained by fitting multi-scenario real vehicle test data using the nonlinear least squares method, and can be directly solved in the vehicle controller.

[0027] Based on the mapping relationship, the pre-aiming path and timing deviation of the axle are corrected. The core function of the above model is to predict the vehicle dynamic offset and the pre-aiming lateral deviation according to the current λ value, and correct the pre-aiming trajectory in advance to avoid compensation lag and suspension over-adjustment problems caused by sudden changes in driving style.

[0028] In step S4, the dynamic weight evolution mechanism is an exponential smooth evolution function, specifically satisfying:

[0029] In the formula, w1(λ): the dynamic weighting coefficient of vehicle comfort that adapts to driving style, representing the proportion of adjustment for the driving comfort target; w2(λ): the dynamic weighting coefficient of suspension travel constraint that adapts to driving style, representing the proportion of adjustment for the suspension structure safety target; w3(λ): the dynamic weighting coefficient of tire adhesion that adapts to driving style, representing the proportion of adjustment for the vehicle handling and grip target; w 10 : Initial weight calibration values ​​for the comfort target under stable driving baseline conditions; w 20 : Initial weight calibration values ​​of the suspension dynamic travel constraint target under stable driving baseline conditions; w 30 : Initial weight calibration value of tire adhesion target under stable driving benchmark conditions; λ: Normalized driving style quantification value, ranging from [0,1], the larger the value, the more aggressive the driving behavior; k1, k3: Weight exponent evolution rate calibration coefficients, which are positive real numbers used to adjust the speed of weight change with driving style; η: Suspension dynamic travel weight smoothing correction coefficient, which is a positive real number to ensure that the weight transition is without sudden changes or shocks; eThe natural constant is approximated at 2.71828. A non-jumping weighting mechanism is constructed to adapt to driving style, focusing on three control objectives: vehicle comfort, suspension travel constraint, and tire grip. This mechanism strictly adheres to two hard constraints: the sum of weights must be 1, and the suspension travel weight w2(λ) ≥ 0.15. The weight evolution logic is clear: during smooth driving, comfort accounts for approximately 60% of the weight, prioritizing the filtering of road vibrations; during gentle driving, the three weights are evenly distributed, balancing comfort and handling; during aggressive driving, tire grip accounts for approximately 75% of the weight, suppressing roll and pitch and ensuring grip performance. The exponential weighting function achieves a smooth transition, completely eliminating suspension impact and improving ride comfort.

[0030] In step S5, the model predicts the control period with a time domain of ≤10ms and a preview time domain of ≤200ms. Dynamic weights are substituted into the cost function, coupled with the corrected preview road information, and the optimal control quantity is solved through rolling optimization. The cost function is:

[0031] In the formula, J : Model predictive control cost function, which is the target value for control optimization. The minimum value is obtained to acquire the optimal suspension control command; Np—model prediction step size, a positive integer representing the number of prediction steps for future road disturbances by the control algorithm; w1(λ): Dynamic weighting coefficient for vehicle comfort, adaptively adjusted according to the driving style quantization value λ; a z (i): The vertical acceleration of the vehicle body predicted in the i-th step, representing a quantitative indicator of ride comfort; w2(λ): The dynamic weight coefficient of the suspension travel constraint, which is adaptively adjusted according to the quantitative value λ of the driving style; S(i): The real-time value of the suspension travel predicted in the i-th step; S0: The reference calibration value of the suspension travel, which is a preset threshold for structural safety; w3(λ): The dynamic weight coefficient of tire adhesion, which is adaptively adjusted according to the quantitative value λ of the driving style. (i): The i-th step predicts the rate of change of the vertical dynamic load on the tire, which characterizes the tire grip stability index; ρ : Control increment penalty coefficient, a positive real number, used to suppress abrupt changes in control commands, ensure smooth execution, and extend actuator life; ∆u: Suspension control command increment, representing the magnitude of change in control quantity between adjacent cycles.

[0032] In step S6, the online iterative optimization uses the gradient descent method to update the mapping coefficients and weight parameters every 100 control cycles to adapt to long-term use scenarios with different drivers, different road surfaces, and vehicle component degradation.

[0033] In the aforementioned method for dynamic optimization of the axle-front anti-suspension control target, step S1 involves collecting data including driver longitudinal and lateral control parameters, vehicle dynamic parameters, and axle-front anti-suspension road surface elevation parameters. The driver longitudinal and lateral control parameters include steering wheel angle, steering wheel angle rate, throttle opening rate of change, brake master cylinder pressure, and longitudinal acceleration / deceleration. The vehicle dynamic parameters include vertical acceleration, roll / pitch angle, suspension travel, and tire dynamic load.

[0034] A multi-threaded synchronous acquisition mode is adopted, with a sampling frequency of 100Hz for control parameters and pre-aiming data, and a sampling frequency of no less than 50Hz for dynamic parameters. The pre-aiming parameters of the axle are collected by LiDAR to obtain the road surface elevation of 0.5-5m in front of the axle, with a resolution of no less than 1cm. Spatiotemporal alignment of multi-source data is achieved through CAN bus timestamps. A combination of sliding mean filtering and Kalman filtering is used for noise reduction to remove outliers and high-frequency noise interference. All data are subjected to Min-Max normalization to eliminate dimensional differences and uniformly map them to the [0,1] interval, laying the foundation for subsequent feature extraction and model calculation. For the pre-aiming data of the axle, the initial correction of spatial mismatch is completed simultaneously, and invalid pre-aiming points caused by changes in vehicle posture and speed are removed, while valid road surface elevation data are retained to reduce the subsequent computational load.

[0035] A dynamic optimization system for axle-front anti-alignment active suspension control target, implementing the above method, adopts an integrated on-board ECU architecture, is fully integrated into the original vehicle controller without any additional hardware increments, and includes:

[0036] The multi-source information sensing unit reuses existing on-board sensors to collect and preprocess three types of data: driver control, vehicle dynamics, and axle front aiming.

[0037] The driving style quantization unit, integrated into the vehicle ECU, is used to extract style features and output continuous style quantization values ​​λ, with a recognition latency of no more than 5ms.

[0038] The mapping and aiming correction unit has a built-in nonlinear mapping model, which is used to solve the dynamic range and aiming deviation and complete the aiming trajectory correction;

[0039] The dynamic weight optimization unit is used to generate dynamic weight coefficients for the three major control objectives in real time without any jumps.

[0040] The fusion-based anti-aiming control unit combines dynamic weights with corrected anti-aiming information to output optimal suspension control commands.

[0041] The suspension actuator is compatible with mainstream active suspension actuators, with a response time of ≤50ms, and is used to execute control commands to complete suspension adjustment.

[0042] The communication unit uses the vehicle-mounted CAN bus to realize data interaction between units and ensure real-time control.

[0043] The closed-loop iterative unit is used to collect vehicle feedback data and optimize model parameters online.

[0044] The aforementioned axle-front anti-aiming active suspension control target dynamic optimization system has online self-learning capabilities. It collects vehicle dynamic feedback data every 100 control cycles and uses the gradient descent method to fine-tune the mapping coefficients and weight evolution parameters. It automatically adapts to long-term usage scenarios such as different driver driving habits, different road conditions, and aging vehicle parts, without the need for manual secondary calibration.

[0045] Beneficial effects of the invention:

[0046] 1. This invention breaks through the barrier of fixed weight control, realizing dynamic adaptation of the control target weight according to driving style, achieving deep coordination between pre-aiming compensation and driving operation, and greatly improving the adaptability of all scenarios: the control performance of all scenarios is improved by more than 25%, the dynamic weight matches various driving styles, the comfort of smooth driving is improved by 32%, the handling stability of aggressive driving is improved by 40%, and the gentle driving achieves a balanced optimization of comfort and handling stability; taking into account both driving comfort and driving stability.

[0047] This invention proposes and constructs a global nonlinear mapping system (driving style-dynamics-pre-aiming deviation triple coupling mapping) that consists of "driving style quantification value → vehicle dynamics working range → axle pre-aiming spatial mismatch". It upgrades driving style from a traditional disturbance source to a core input variable for pre-aiming correction and control decision-making, rather than an additional correction term. This mapping mechanism is an explicit real-time solvable structure that not only adapts to steady-state conditions but also covers transient aggressive driving scenarios such as sharp turns and rapid acceleration, completely solving the common technical problems of pre-aiming trajectory deviation and compensation lag failure under aggressive driving in the industry.

[0048] 2. This invention eliminates the pre-aiming mismatch problem caused by aggressive driving by constructing a correlation correction mechanism between driving style and pre-aiming deviation, thereby improving the accuracy of advance compensation; the pre-aiming compensation accuracy is significantly improved: relying on triple mapping to correct the pre-aiming mismatch problem caused by driving style, the compensation lag time is shortened by 60%, and the control accuracy in complex working conditions is greatly improved;

[0049] This invention addresses the shortcomings of traditional fixed weights and step-weight adjustments by designing an exponential smooth weight evolution mechanism (without abrupt dynamic weights) that satisfies normalization constraints and has structural hard protection. The weights for comfort and handling stability change continuously and gradually with driving style, without abrupt changes or control shocks. At the same time, a lower limit for the suspension travel weight is set to balance ride quality and suspension structural safety. The weight evolution rules, smooth adjustment methods, and constraint logic are separately included in the subordinate rights protection, fully covering equivalent infringement schemes with similar control logic.

[0050] 3. This invention achieves human-machine collaborative anti-aiming control combined with low-cost engineering innovation. The hardware relies on existing vehicle components without modification, reusing vehicle hardware and simplifying the calibration process to create a low-cost axle front anti-aiming active suspension optimization control scheme. It covers the core architecture, control logic, and all-scenario applications, demonstrating strong engineering feasibility and extremely low implementation costs: it completely reuses the original vehicle sensors and ECU, requiring no additional hardware investment, simplifying the calibration process, and resulting in low incremental costs for mass production. Furthermore, it exhibits strong control robustness and durability: possessing online self-learning optimization capabilities, it can adapt to different drivers, varying road surfaces, and aging vehicle components, ensuring stable performance over long-term use. Attached Figure Description

[0051] Figure 1 The diagram shows the overall system architecture of this invention;

[0052] Figure 2 The diagram shown is a flowchart of the method of the present invention;

[0053] Figure 3 The figure shown is a curve illustrating the evolution of the driving style-control target weights according to the present invention. Detailed Implementation

[0054] To make the technical concept and advantages of the invention clearer, the technical solution of the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the following embodiments are merely preferred embodiments for explaining and illustrating the present invention, and should not be considered as, nor constitute a limitation on, the scope of patent protection claimed by the present invention.

[0055] Example 1: See Figure 2 This invention discloses a dynamic optimization method for front axle anti-axle active suspension control targets. Using driving style quantification parameters as the core control variables, it reconstructs the multi-objective control logic of the front axle anti-axle active suspension. Specifically, it includes the following steps:

[0056] S1: Coupled Information Acquisition and Data Preprocessing

[0057] Simultaneously collect driver control behavior data, vehicle dynamic response data, and axle-front anti-road elevation data to complete spatiotemporal synchronization of multi-source data, noise reduction filtering, and normalization preprocessing.

[0058] S2: Driving Style Feature Extraction and Continuous Quantization

[0059] Extract multidimensional features of driving style, construct a continuous quantitative representation model, and output normalized quantitative values ​​of driving style to realize the transformation of driving style from qualitative classification to quantitative parameters;

[0060] S3: Triple Mapping Relationship Modeling

[0061] Establish a nonlinear mapping relationship among driving style quantification value, vehicle dynamics working range, and axle front aiming space mismatch deviation, and correct the axle front aiming path and timing deviation based on the mapping relationship;

[0062] S4: Dynamic Weight Evolution and Normalization Solution

[0063] Based on the quantitative value of driving style, a dynamic weight evolution mechanism for control objectives is constructed to calculate the dynamic weight coefficients of three major objectives: vehicle comfort, suspension travel constraint, and tire adhesion in real time. The weight coefficients satisfy the normalization constraint and have no step jump.

[0064] S5: Integrated Pre-aiming Closed-Loop Control

[0065] The dynamic weighting coefficients are incorporated into the axle-front aiming model predictive control algorithm, coupled with the corrected aiming road surface information, and the optimal suspension control command is solved in a rolling manner to drive the actuator action.

[0066] S6: Online Iterative Optimization

[0067] By collecting real-time feedback data from vehicles and iteratively optimizing mapping parameters and weight evolution parameters online, adaptive closed-loop control under all operating conditions can be achieved.

[0068] Example 2: The specific technical process of each step in the dynamic optimization method for the axle-front anti-aiming active suspension control target in this example is as follows:

[0069] S1 Coupled Information Acquisition and Preprocessing

[0070] A multi-threaded synchronous acquisition mode was adopted, with a sampling frequency of 100Hz for control parameters, 50Hz for dynamic parameters, and 100Hz for pre-aiming data. Spatiotemporal alignment of multi-source data was achieved through CAN bus timestamps. A combination of moving mean filtering and Kalman filtering was used for noise reduction to remove outliers and high-frequency noise interference. All data were subjected to Min-Max normalization to eliminate dimensional differences and uniformly map them to the [0,1] interval, laying the foundation for subsequent feature extraction and model calculation. For the axle-front pre-aiming data, preliminary spatial mismatch correction was performed simultaneously, invalid pre-aiming points caused by changes in vehicle posture and speed were removed, and valid road elevation data were retained to reduce the computational load in the future.

[0071] S2 Driving Style Feature Extraction and Continuous Quantization

[0072] Abandoning traditional qualitative classification methods of stable / mild / aggressive approaches, this paper adopts a continuous quantitative representation model, constructing a four-dimensional feature vector encompassing manipulation intensity, manipulation frequency, anticipation and following accuracy, and dynamic hysteresis. This feature vector is then applied through an 8-node hidden layer and a training error E. errAn improved BP neural network with a value ≤0.01 outputs style quantization values ​​λ∈[0,1]: λ∈[0,0.3] represents smooth driving, characterized by gentle handling and high anti-aiming accuracy; λ∈(0.3,0.7] represents mild driving, characterized by balanced handling, balancing comfort and stability; and λ∈(0.7,1] represents aggressive driving, characterized by intense handling, large wheel track deviation, and high risk of anti-aiming mismatch. This quantization method has no classification delay and can track changes in driver control in real time, adapting to transient conditions such as rapid acceleration, sharp turns, and continuous lane changes.

[0073] S3 Triple Mapping Relationship Modeling

[0074] Through real-world road tests across various scenarios, including straight roads, curves, and rough surfaces, over 1000 sets of sample data were collected. A nonlinear least squares fitting method was used to establish a triple explicit mapping model of driving style, dynamics, and anti-misalignment. This model can be solved quickly and directly within the vehicle's ECU without iterative fitting. The core function of the model is to predict the vehicle's dynamic offset and anti-misalignment lateral deviation based on the current λ value, correcting the anti-misalignment trajectory in advance and avoiding compensation lag and suspension over-adjustment issues caused by sudden changes in driving style.

[0075] S4 Dynamic Weight Evolution and Normalization

[0076] Around the three control objectives of vehicle comfort, suspension travel constraint, and tire grip, a non-jumping weighting mechanism is constructed that adapts to driving style, strictly adhering to two hard constraints: the sum of weights must be 1, and the suspension travel weight w2(λ) ≥ 0.15. The weight evolution logic is clear: during smooth driving, the comfort weight accounts for about 60%, prioritizing the filtering of road vibrations; during gentle driving, the three weights are evenly distributed, balancing comfort and handling stability; during aggressive driving, the tire grip weight accounts for about 75%, suppressing body roll and pitch and ensuring grip performance. The exponential weighting function achieves a smooth transition, completely eliminating suspension impact and improving ride comfort.

[0077] S5 Fusion Pre-aiming Closed-Loop Control

[0078] Model predictive control (MPC) with a control cycle of 10ms and a prediction time of 200ms is employed. Dynamic weights are substituted into the cost function, coupled with corrected road surface information, and the optimal control quantity is solved through rolling optimization. Control commands are transmitted to the suspension actuators via the CAN bus, and the adjustable damping shock absorbers and controllable stiffness air springs respond synchronously, achieving both advance compensation and vehicle attitude control, while balancing control accuracy and actuator lifespan.

[0079] S6 Online Iterative Optimization (Practicality Enhancement)

[0080] The system has online self-learning capabilities, collecting vehicle dynamic feedback data every 100 control cycles (1 second), and using the gradient descent method to fine-tune the mapping coefficients and weight evolution parameters. It automatically adapts to long-term use scenarios such as different drivers' driving habits, different road conditions, and aging vehicle parts, without the need for manual secondary calibration, thus greatly improving the stability of control throughout the entire life cycle.

[0081] Example 3: The dynamic optimization method for the axle front-aiming active suspension control target in this example differs from Example 1 or Example 2 in that: in step S2, the continuous quantitative representation of driving style adopts a four-dimensional feature vector X=[x1, x2, x3, x4] containing x1 control intensity, x2 control frequency, x3 anti-aiming following degree, and x4 dynamic response lag degree. The continuous output of λ∈[0,1] is achieved by improving the BP neural network, without the need for manual classification threshold determination;

[0082] Where x1 = 0.4| |+0.3|∆α|+0.3|∆P b | Handling strength: ∆α is the average steering wheel angle, representing the steering amplitude; ∆α is the change in longitudinal acceleration, representing the degree of acceleration or deceleration; ∆P b x1 represents the change in brake master cylinder pressure, characterizing braking intensity; x2 represents the steering frequency, extracted using Fast Fourier Transform to capture the center-of-gravity frequency of the steering wheel angle signal, used to characterize the frequency of steering actions; x3 = 1 - ∆e / L p Pre-aiming tracking accuracy: ∆e is the lateral deviation between the pre-aiming trajectory and the actual wheel track, L p The forward aiming reference distance is 1, and the closer the value is to 1, the better the aiming follow-up. The x4 dynamic hysteresis is the delay difference between the control input and the vehicle attitude response, which is used to characterize the degree of sluggishness of the vehicle's dynamic response.

[0083] This invention abandons the traditional qualitative classification methods of stable / mild / aggressive approaches, and adopts a continuous quantitative representation mode to construct a four-dimensional feature vector of manipulation intensity, manipulation frequency, anticipation and following degree, and dynamic hysteresis. This is achieved through an 8-node hidden layer and a training error E. err An improved BP neural network with a value ≤0.01 outputs style quantization values ​​λ∈[0,1]: λ∈[0,0.3] represents smooth driving, characterized by gentle handling and high anti-aiming accuracy; λ∈(0.3,0.7] represents mild driving, characterized by balanced handling, balancing comfort and stability; and λ∈(0.7,1] represents aggressive driving, characterized by intense handling, large wheel track deviation, and high risk of anti-aiming mismatch. This quantization method has no classification delay and can track changes in driver control in real time, adapting to transient conditions such as rapid acceleration, sharp turns, and continuous lane changes.

[0084] Example 4: The dynamic optimization method for the axle-front anti-aiming active suspension control target in this example differs from the aforementioned examples in that:

[0085] In step S3, the nonlinear mapping relationship is an explicitly solvable function, and its expression after calibration is:

[0086] In the formula, a z-rms : The effective value of the vehicle's vertical acceleration, representing an indicator of ride comfort; a z0 : Reference value of vertical acceleration of vehicle body under stable driving conditions; S max S0: Suspension dynamic travel limit value, representing the safety constraint threshold of the suspension structure; Δe: Suspension dynamic travel reference value under stable driving conditions; max ∆e0: Maximum lateral deviation between the pre-aiming trajectory and the wheel track, representing the degree of pre-aiming mismatch; ∆e0: Reference value of lateral deviation of pre-aiming under stable driving conditions; : Tire vertical dynamic load change rate, a characterizing tire adhesion stability index; : Baseline value of tire vertical dynamic load change rate under stable driving conditions; λ: Normalized driving style quantification value, ranging from [0,1]; k a k s k e k f ξ: First-order linear calibration coefficient, characterizing the linear rate of change of the parameter with driving style; a ξ f : Second-order nonlinear calibration coefficients, characterizing the nonlinear correction magnitude of parameters with driving style; the above parameters are used to correct the pre-aiming wheel track deviation and timing lag deviation in real time.

[0087] Through real-world road tests across various scenarios, including straight roads, curves, and rough surfaces, over 1000 sets of sample data were collected. A nonlinear least squares fitting method was used to establish a triple explicit mapping model of driving style, dynamics, and anti-misalignment. This model can be solved quickly and directly within the vehicle's ECU without iterative fitting. The core function of the model is to predict the vehicle's dynamic offset and anti-misalignment lateral deviation based on the current λ value, correcting the anti-misalignment trajectory in advance and avoiding compensation lag and suspension over-adjustment issues caused by sudden changes in driving style.

[0088] Example 5: The dynamic optimization method for the axle-front anti-aiming active suspension control target in this example differs from the aforementioned examples in that:

[0089] In step S4, the dynamic weight evolution mechanism is an exponential smooth evolution function, specifically satisfying:

[0090] In the formula, w1(λ): the dynamic weighting coefficient of vehicle comfort that adapts to driving style, representing the proportion of adjustment for the driving comfort target; w2(λ): the dynamic weighting coefficient of suspension travel constraint that adapts to driving style, representing the proportion of adjustment for the suspension structure safety target; w3(λ): the dynamic weighting coefficient of tire adhesion that adapts to driving style, representing the proportion of adjustment for the vehicle handling and grip target; w 10 : Initial weight calibration values ​​for the comfort target under stable driving baseline conditions; w 20 : Initial weight calibration values ​​of the suspension dynamic travel constraint target under stable driving baseline conditions; w 30 : Initial weight calibration value of tire adhesion target under stable driving benchmark conditions; λ: Normalized driving style quantification value, ranging from [0,1], the larger the value, the more aggressive the driving behavior; k1, k3: Weight exponent evolution rate calibration coefficient, positive real number, used to adjust the speed of weight change with driving style; η: Suspension dynamic travel weight smoothing correction coefficient, positive real number, to ensure that the weight transition is without sudden changes or shocks; e: Natural constant, taken as an approximation of 2.71828; and satisfying the weight normalization constraint w1(λ)+w2(λ)+w3(λ)=1, and the suspension dynamic travel weight hard constraint w2(λ)≥0.15; the comfort weight decreases monotonically with increasing λ, the tire adhesion weight increases monotonically with increasing λ, and the weight transition is without step shocks.

[0091] Example 6: The dynamic optimization method for the axle-front anti-aiming active suspension control target in this example differs from the aforementioned examples in that:

[0092] In step S5, the model predicts the control period with a time domain of ≤10ms and a preview time domain of ≤200ms. Dynamic weights are substituted into the cost function, coupled with the corrected preview road information, and the optimal control quantity is solved through rolling optimization. The cost function is:

[0093] In the formula, J Model predictive control cost function: This function is the objective value for control optimization. The minimum value is calculated to obtain the optimal suspension control command. N p —Model prediction step size, a positive integer, representing the number of prediction steps the control algorithm takes for future road disturbances; w1(λ): Dynamic weighting coefficient for vehicle comfort, adaptively adjusted according to the driving style quantization value λ; a z (i): The vertical acceleration of the vehicle body predicted in the i-th step, representing a quantitative indicator of ride comfort; w2(λ): The dynamic weight coefficient of the suspension travel constraint, which is adaptively adjusted according to the quantitative value λ of the driving style; S(i): The real-time value of the suspension travel predicted in the i-th step; S0: The reference calibration value of the suspension travel, which is a preset threshold for structural safety; w3(λ): The dynamic weight coefficient of tire adhesion, which is adaptively adjusted according to the quantitative value λ of the driving style. (i): The i-th step predicts the rate of change of the vertical dynamic load on the tire, which characterizes the tire grip stability index; ρ : Control increment penalty coefficient, a positive real number, used to suppress abrupt changes in control commands, ensuring smooth execution and extending actuator life; ∆u: Suspension control command increment, representing the magnitude of change in control quantity between adjacent cycles. This cost function balances control accuracy, ride comfort, and suspension structural safety, achieving multi-objective collaborative optimization.

[0094] The present invention provides a dynamic optimization method for the front-axle anti-suspension control target, as described above. In step S1, the collected data includes driver longitudinal and lateral control parameters, vehicle dynamic parameters, and front-axle anti-suspension road surface elevation parameters. Among them, the driver longitudinal and lateral control parameters include steering wheel angle, steering wheel angle rate, throttle opening rate of change, brake master cylinder pressure, and longitudinal acceleration / deceleration; the vehicle dynamic parameters include vehicle vertical acceleration, roll angle / pitch angle, suspension dynamic travel, and tire dynamic load. A multi-threaded synchronous acquisition mode is adopted, with a sampling frequency of 100Hz for control parameters and pre-aiming data, and a sampling frequency of no less than 50Hz for dynamic parameters. The pre-aiming parameters of the axle are collected by LiDAR to obtain the road surface elevation of 0.5-5m in front of the axle, with a resolution of no less than 1cm. Spatiotemporal alignment of multi-source data is achieved through CAN bus timestamps. A combination of sliding mean filtering and Kalman filtering is used for noise reduction to remove outliers and high-frequency noise interference. All data are subjected to Min-Max normalization to eliminate dimensional differences and uniformly map them to the [0,1] interval, laying the foundation for subsequent feature extraction and model calculation. For the pre-aiming data of the axle, the initial correction of spatial mismatch is completed simultaneously, and invalid pre-aiming points caused by changes in vehicle posture and speed are removed, while valid road surface elevation data are retained to reduce the subsequent computational load.

[0095] In step S6, the online iterative optimization uses the gradient descent method to update the mapping coefficients and weight parameters every 100 control cycles to adapt to different drivers' driving habits, different road surfaces, and long-term use scenarios with vehicle component degradation.

[0096] Figure 3 The graph shows the evolution of the driving style-control target weights of this invention, illustrating the smooth evolution of the weights and highlighting the innovative design of non-jump and adaptive operation.

[0097] Example 7: See Figure 1 This embodiment describes a dynamic optimization system for axle-front anti-aiming active suspension control target, as described above. The system employs an integrated modular architecture with an onboard ECU, integrating all functional units into the original vehicle controller without requiring additional hardware; only a low-cost data synchronization unit is added. Specifically, it includes:

[0098] The multi-source information sensing unit reuses existing on-board sensors, including on-board steering wheel angle sensor, IMU, suspension displacement sensor, and front axle lidar, to collect and preprocess three types of data: driver control, vehicle dynamics, and front axle aiming.

[0099] The driving style quantization unit, integrated into the vehicle ECU, is used to extract style features and output continuous style quantization values ​​λ. It is developed using embedded C language, with concise code, low computing power consumption, and recognition latency of no more than 5ms.

[0100] The mapping and pre-aiming correction / weight optimization unit has a built-in nonlinear mapping model to solve the dynamic range and pre-aiming deviation and complete the pre-aiming trajectory correction. It stores calibration data through built-in Flash, which has fast reading and operation speed and no computing power overflow.

[0101] The dynamic weight optimization unit is used to generate dynamic weight coefficients for the three major control objectives in real time without any jumps.

[0102] The fusion-based anti-aiming control unit combines dynamic weights with corrected anti-aiming information to output optimal suspension control commands.

[0103] The suspension actuator is compatible with mainstream active suspension actuators, with a response time of ≤50ms, and is used to execute control commands to complete suspension adjustment.

[0104] The communication unit uses the vehicle-mounted CAN bus to realize data interaction between units and ensure real-time control.

[0105] The closed-loop iterative unit is used to collect vehicle feedback data and optimize model parameters online. The fusion-based anti-targeting control unit optimizes the MPC algorithm to adapt to the computing power of mass-produced vehicle ECUs, resulting in strong real-time control.

[0106] Suspension actuator: compatible with mainstream electromagnetic dampers and air springs on the market, with rapid response and full adjustment range;

[0107] Communication and Iteration Unit: Relying on the original vehicle's CAN bus to achieve data interaction, and equipped with online optimization logic, it ensures long-term reliable operation.

[0108] This invention relates to a dynamic optimization system for active suspension control based on driving style front axle anti-aiming. It has online self-learning capabilities, collects vehicle dynamic feedback data every 100 control cycles (1 second), and fine-tunes the mapping coefficients and weight evolution parameters using the gradient descent method. It automatically adapts to long-term usage scenarios such as different drivers' driving habits, different road conditions, and aging vehicle parts, without the need for manual secondary calibration, which can significantly improve the stability of control throughout the entire life cycle.

[0109] This invention focuses on three major control objectives: vehicle comfort, suspension travel constraint, and tire grip. It constructs a non-jumping weighting mechanism that adapts to driving style, strictly adhering to two hard constraints: the sum of weights must be 1, and the suspension travel weight w2(λ) ≥ 0.15. The weight evolution logic is clear: during smooth driving, the comfort weight accounts for approximately 60%, prioritizing the filtering of road vibrations; during gentle driving, the three weights are evenly distributed, balancing comfort and handling stability; during aggressive driving, the tire grip weight accounts for approximately 75%, suppressing roll and pitch and ensuring grip performance. The exponential weighting function achieves a smooth transition, completely eliminating suspension impact and improving ride comfort. Practical application examples are as follows:

[0110] Application Example 1: Smooth Driving (Straight Urban Road, Speed ​​40km / h)

[0111] 1. Implementation conditions and basic parameter calibration

[0112] The test vehicle was a compact pure electric family sedan with a wheelbase of 2.7m, equipped with a front-axle lidar pre-aiming active suspension system, and a curb weight of 1.5t; the basic calibration parameters of the control algorithm (formulas in claims 4 and 5): initial weight calibration value w 10 =0.6、w 20 =0.2、w 30 =0.2; weight evolution rate coefficients k1=1.5, k3=2.0, smoothing correction coefficient η=0.2; triple mapping calibration coefficient k a =0.3、k s =0.1、k e =0.2、k f =0.25, second-order coefficient ξ a =0.1、ξ f =0.15.

[0113] Driving conditions: The vehicle is on a straight asphalt road in the city, without potholes or bumps. The speed is constant at 40km / h. The driver does not accelerate or decelerate suddenly or make large-angle turns. The driving style is controlled smoothly. The normalized driving style quantification value λ=0.12 is output by the style quantification module (the smooth driving range λ∈[0,0.3]).

[0114] 2. Implementation Steps

[0115] Step S1: Multi-source data acquisition and preprocessing

[0116] Synchronous data collection: Mean steering wheel angle =2.1°, longitudinal acceleration change ∆α=0.05g, brake master cylinder pressure change ∆P b =0.1MPa; Reference value of vertical acceleration of vehicle body a z0 =0.21m / s 2The reference value for suspension dynamic travel is S0 = 35mm, the reference value for anti-sight lateral deviation is ∆e0 = 1.2cm, and the reference value for tire vertical dynamic load change rate is... =120N / s; data sampling frequency 100Hz, normalized after Kalman filtering and timestamp alignment.

[0117] Step S2: Quantitative Calculation of Driving Style

[0118] Substituting into the control strength formula, x1 = 0.4 | |+0.3|∆α|+0.3|∆P b The calculated control intensity x1=0.84; combined with the four-dimensional features of control frequency, anti-aiming and following degree x3=1-∆e / Lp=0.97, and dynamic lag degree, the output λ=0.12 of the BP neural network is determined to be a smooth driving type.

[0119] Step S3: Triple mapping pre-aiming correction

[0120] Substitute the parameters into the nonlinear mapping formula to calculate the correction parameters:

[0121] =0.21×(1+0.3×0.12+0.1×0.122) =0.218m / s 2 ;

[0122] =35×(1+0.1×0.12)=35.42mm;

[0123] =1.2×(1+0.2×0.12)=1.2288cm;

[0124] Fine-tuning the pre-aiming trajectory based on the correction value eliminates minor pre-aiming mismatches.

[0125] Step S4: Dynamic weight calculation

[0126] Substituting into the exponential weight evolution formula:

[0127] =0.6×e^{-1.5×0.12}=0.53 (Dynamic comfort weight);

[0128] =0.2×(1-0.2×0.38)≈0.2 (suspension dynamic travel weight);

[0129] =0.2×e^{2.0×0.12}=0.27 (Tire adhesion weight);

[0130] Verification constraints: w1 + w2 + w3 = 1, w2 ≥ 0.15, which satisfies the normalization and hard constraint requirements, and the weights do not change abruptly.

[0131] Step S5: Fusion Pre-aiming Control Execution

[0132] Control period 10ms, prediction step size Np=20, substitute into the cost function

[0133] (Penalty coefficient) ρ =0.8)

[0134] The optimal control command is solved by rolling; the suspension actuator response is as follows: the damping is adjusted to 0.6 kN·s / m and the stiffness is adjusted to 8 kN / m to compensate for minor road vibrations in advance.

[0135] Step S6: Online Iterative Optimization

[0136] Feedback data is collected every 100 control cycles. The current operating parameters are stable, and there is no need to fine-tune the mapping coefficients and weight parameters, thus maintaining closed-loop steady-state control.

[0137] 3. Implementation Results and Data Comparison

[0138] Based on actual vehicle testing, the effective value of the vertical acceleration a of the vehicle body under this scheme is... z-rms The suspension travel is approximately 0.218 m / s², a 32% reduction compared to the traditional fixed-weight scheme; the suspension travel is stable at 32–35 mm with no overtravel; the anti-sight deviation is controlled within 1.2 cm, and the compensation lag time is <10 ms; the ride is smooth and shock-free, with excellent comfort performance, fully meeting the requirements for smooth driving.

[0139] Application Example 2: Aggressive Driving (Changing Lanes on an Expressway at a Speed ​​of 80 km / h)

[0140] 1. Implementation conditions

[0141] The test vehicle was the same as in Example 1, with driving conditions including continuous lane changes and sharp turns, a peak steering wheel angle of 25°, longitudinal deceleration of 0.6g, and a style quantification value λ=0.93.

[0142] 2. Implementation steps and effects

[0143] High-frequency control data and anti-aim information are collected to identify aggressive driving. A mapping model is used to significantly correct anti-aim trajectory deviations, with weights w1=0.1, w2=0.15, and w3=0.75. The outer suspension damping is increased to 1.8 kN·s / m and stiffness to 18 kN / m, suppressing body roll. Ultimately, the body roll angle is ≤1.8°, tire dynamic load fluctuation is reduced by 36%, lane changes are stable with no understeer, and handling stability is significantly improved.

[0144] Application Example 3: Gentle Driving (Mixed Urban and Suburban Roads, Speed ​​60km / h)

[0145] The style quantification value λ=0.52, and the three weights are evenly distributed as w1=0.36, w2=0.2, and w3=0.44. The suspension damping and stiffness are adjusted to the middle level, taking into account both comfort and handling stability. Compared with the traditional solution, the comfort is improved by 21% and the handling stability is improved by 26%, which is suitable for complex mixed road conditions in urban and suburban areas.

[0146] This invention addresses the shortcomings of existing axle-mounted anti-aim active suspension systems, such as fixed control target weights, neglecting the impact of driver control characteristics on anti-aim accuracy, poor adaptability across all scenarios, and insufficient engineering practicality. It transforms driving style from a disturbance source into a core control variable, constructs a triple nonlinear mapping system of driving style, dynamics, and anti-aim deviation, and designs a non-jumping dynamic weight evolution mechanism. This enables the control target weights to adaptively adjust with driving style, simultaneously correcting anti-aim timing and trajectory deviations. The dynamic weights are integrated into the anti-aim control algorithm to achieve precise advance compensation, breaking through the traditional fixed-weight control framework and achieving deep synergy between anti-aim compensation and driving control, improving control performance by over 25% across all scenarios. The hardware relies on existing vehicle components, allowing for low-cost engineering implementation.

Claims

1. A method for dynamic optimization of a target in axle-front anti-suspension active suspension control, characterized in that: Using driving style quantification parameters as the core control variables, the multi-objective control logic of the axle-front anti-suspension active suspension is reconstructed. Specific steps include: S1: Coupled Information Acquisition and Data Preprocessing Simultaneously collect driver control behavior data, vehicle dynamic response data, and axle-front anti-road elevation data to complete spatiotemporal synchronization of multi-source data, noise reduction filtering, and normalization preprocessing. S2: Driving Style Feature Extraction and Continuous Quantization Extract multidimensional features of driving style, construct a continuous quantitative representation model, and output normalized quantitative values ​​of driving style to realize the transformation of driving style from qualitative classification to quantitative parameters; S3: Mapping Relationship Modeling Establish a nonlinear mapping relationship among driving style quantification, vehicle dynamics parameters, and axle front aiming space mismatch deviation: The mapping relationship uses normalized driving style quantification as the independent variable and vehicle comfort parameters, suspension travel parameters, anti-track deviation parameters, and tire adhesion parameters as dependent variables, where: a z-rms : The effective value of the vehicle's vertical acceleration, representing an indicator of ride comfort; a z0 : Reference value of vertical acceleration of vehicle body under stable driving conditions; S max S0: Suspension dynamic travel limit value, representing the safety constraint threshold of the suspension structure; Δe: Suspension dynamic travel reference value under stable driving conditions; max ∆e0: Maximum lateral deviation between the pre-aiming trajectory and the wheel track, representing the degree of pre-aiming mismatch; ∆e0: Reference value of lateral deviation of pre-aiming under stable driving conditions; : Tire vertical dynamic load change rate, a characterizing tire adhesion stability index; : Baseline value of tire vertical dynamic load change rate under stable driving conditions; λ: Normalized driving style quantification value, ranging from [0,1]; k a k s k e k f ξ: First-order linear calibration coefficient, characterizing the linear rate of change of the parameter with driving style; a ξ f Second-order nonlinear calibration coefficients characterize the nonlinear correction magnitude of parameters with driving style; obtained by fitting multi-scenario real vehicle test data using nonlinear least squares method, and can be directly solved in the vehicle controller; Based on the mapping relationship, the axle front aiming path and timing deviation are corrected. According to the current λ value, the vehicle dynamic offset and aiming lateral deviation are predicted, and the aiming trajectory is corrected in advance to avoid compensation lag and suspension over-adjustment problems caused by sudden changes in driving style. S4: Dynamic Weight Evolution and Normalization Solution Based on the quantitative value of driving style, a dynamic weight evolution mechanism for control objectives is constructed to calculate the dynamic weight coefficients of three major objectives: vehicle comfort, suspension travel constraint, and tire adhesion in real time. The weight coefficients satisfy the normalization constraint and have no step jump. S5: Integrated Pre-aiming Closed-Loop Control The dynamic weighting coefficients are incorporated into the axle-front aiming model predictive control algorithm, coupled with the corrected aiming road surface information, and the optimal suspension control command is solved in a rolling manner to drive the actuator action. S6: Online Iterative Optimization By collecting real-time feedback data from vehicles and iteratively optimizing mapping parameters and weight evolution parameters online, adaptive closed-loop control under all operating conditions can be achieved.

2. The dynamic optimization method for axle-front anti-aiming active suspension control target according to claim 1, characterized in that: In step S2, a multi-dimensional driving style feature vector is constructed based on the driver's control behavior and vehicle dynamic response. The driving style is continuously and quantitatively represented by a neural network model, and a normalized driving style quantification value is output to achieve real-time driving style recognition without manual threshold judgment or classification delay. The multidimensional driving style feature vector includes at least the control intensity feature, control frequency feature, anti-aiming and following feature, and dynamic response hysteresis feature; the continuous quantitative representation output range is λ∈[0,1], which can adaptively match all types of driving behaviors such as smooth driving, mild driving, and aggressive driving: λ∈[0,0.3] is smooth driving, with gentle control and high anti-aiming and following. λ∈(0.3,0.7] represents mild driving, with balanced handling, comfort, and stability; λ∈(0.7,1] represents aggressive driving, with intense handling, large wheel track deviation, and high risk of anti-aiming mismatch.

3. The dynamic optimization method for axle-front anti-aiming active suspension control target according to claim 1, characterized in that: In step S4, the dynamic weight evolution mechanism is an exponential smooth evolution function, specifically satisfying: In the formula, w1(λ): the dynamic weighting coefficient of vehicle comfort that adapts to driving style, representing the proportion of adjustment of the driving comfort target; w2(λ): The dynamic weighting coefficient of the suspension travel constraint that adapts to driving style, representing the proportion of adjustment of the suspension structure safety target; w3(λ): The dynamic weighting coefficient of tire adhesion that adapts to driving style, representing the proportion of the control of the vehicle's handling and grip target; w 10 : Initial weight calibration values ​​for the comfort target under stable driving baseline conditions; w 20 : Initial weight calibration values ​​of the suspension dynamic travel constraint target under stable driving baseline conditions; w 30 : Initial weight calibration value of tire adhesion target under stable driving benchmark conditions; λ: Normalized driving style quantification value, ranging from [0,1], the larger the value, the more aggressive the driving behavior; k1, k3: Weight exponent evolution rate calibration coefficients, which are positive real numbers used to adjust the speed at which the weight changes with driving style; η The suspension dynamic travel weight smoothing correction coefficient is a positive real number, ensuring that the weight transition is smooth and without sudden changes or shocks. e : Natural constant, approximated at 2.71828; Around the three major control objectives of vehicle comfort, suspension travel constraint, and tire adhesion, a non-jump weighting mechanism that adapts to driving style is constructed, strictly adhering to two hard constraints: the sum of weights is 1 and the dynamic weighting coefficient of suspension travel constraint w2(λ) ≥ 0.

15.

4. The dynamic optimization method for axle-front anti-aiming active suspension control target according to claim 1, characterized in that: In step S5, the model predicts the control period with a time domain of ≤10ms and a preview time domain of ≤200ms. The dynamic weight coefficients are substituted into the cost function, coupled with the corrected preview road information, and the optimal control quantity is solved through rolling optimization. The cost function is: In the formula, J Model predictive control cost function: This function is the objective value for control optimization. The minimum value is calculated to obtain the optimal suspension control command. N p —Model prediction step size, a positive integer, representing the number of prediction steps the control algorithm takes for future road disturbances; w1(λ): Dynamic weighting coefficient for vehicle comfort, adaptively adjusted according to the driving style quantization value λ; a z (i): The vertical acceleration of the vehicle body predicted in the i-th step, which represents the quantitative index of driving comfort; w2(λ): The dynamic weight coefficient of the suspension travel constraint, which is adaptively adjusted according to the quantitative value λ of the driving style; S(i): The real-time value of the suspension travel predicted in the i-th step. S0: The reference value of suspension dynamic travel under stable driving conditions, which is a preset threshold for structural safety; w3(λ): Dynamic weighting coefficient for tire adhesion, which is adaptively adjusted according to the driving style quantification value λ; (i): The i-th step predicts the rate of change of the vertical dynamic load on the tire, which characterizes the tire grip stability index; ρ : Control increment penalty coefficient, a positive real number, used to suppress abrupt changes in control commands, ensure smooth execution, and extend actuator life; ∆u: Suspension control command increment, representing the magnitude of change in control quantity between adjacent cycles.

5. The dynamic optimization method for axle-front anti-aiming active suspension control target according to claim 1, characterized in that: In step S6, the online iterative optimization uses the gradient descent method to update the mapping coefficients and weight parameters every 100 control cycles to adapt to different drivers' driving habits, different road surfaces, and long-term use scenarios with vehicle component degradation.

6. The method for dynamic optimization of the axle-front anti-aiming active suspension control target according to any one of claims 1-5, characterized in that: In step S1, the collected data includes driver longitudinal and lateral control parameters, vehicle dynamic parameters, and axle-front pre-aiming road surface elevation parameters. Among them, the driver longitudinal and lateral control parameters include steering wheel angle, steering wheel angle rate, throttle opening rate of change, brake master cylinder pressure, and longitudinal acceleration / deceleration. The vehicle dynamic parameters include body vertical acceleration, roll angle / pitch angle, suspension dynamic travel, and tire dynamic load.

7. A dynamic optimization system for axle-front anti-aiming active suspension control target that implements the method of any one of claims 1-6, characterized in that: It adopts an integrated vehicle ECU architecture, which is fully integrated into the original vehicle controller without any additional hardware additions, including: The multi-source information sensing unit reuses existing on-board sensors to collect and preprocess three types of data: driver control behavior data, vehicle dynamic response data, and axle-front pre-aiming road elevation data. The driving style quantization unit, integrated into the vehicle ECU, is used to extract multi-dimensional features of driving style and output a continuously normalized driving style quantization value λ, with a recognition delay of no more than 5ms. The mapping and pre-aiming correction unit has a built-in nonlinear mapping model, which is used to solve the mismatch deviation between vehicle dynamic parameters and pre-aiming space and complete the pre-aiming trajectory correction. The dynamic weight optimization unit is used to generate dynamic weight coefficients for the three control objectives without step jumps in real time. The fusion-based anti-aiming control unit combines dynamic weighting coefficients with corrected anti-aiming information to output optimal suspension control commands. The suspension actuator unit, adapted to the active suspension actuator, has a response time of ≤50ms and is used to execute control commands to complete suspension adjustment. The communication unit uses the vehicle-mounted CAN bus to realize data interaction between units and ensure real-time control. The closed-loop iterative unit is used to collect vehicle feedback data and optimize model parameters online.

Images