Intelligent empennage control system and method based on multi-sensor fusion

Abstract

The invention discloses an intelligent empennage control system and method based on multi-sensor fusion, and the method comprises the steps: employing an online deduced self-induced aerodynamic load as a known feedforward input, stripping the unsteady interference caused by the action of an empennage from original data through a decoupling type extended Kalman filtering algorithm, thereby obtaining a real net vehicle body posture, and achieving the control of the empennage. On the basis, accurate and coordinated control of the downforce is realized through a multi-target optimization decision based on a net state. In this way, in a high-performance driving scene, it can be ensured that the vehicle obtains excellent high-speed stability, meanwhile, ideal front and rear axle load distribution and steering responsiveness are maintained, and the dynamic control limit and driving safety of the vehicle are comprehensively improved.

Description

Technical Field

[0001] This application relates to the field of intelligent control, and more specifically, to an intelligent tail wing control system and method based on multi-sensor fusion. Background Technology

[0002] With the continuous evolution of high-performance vehicle technology, active aerodynamic components have become crucial for improving vehicle stability and handling limits at high speeds. By constructing an intelligent rear wing control system, vehicles can optimize downforce in real time based on vehicle speed and body posture, thereby significantly improving tire grip utilization and enhancing dynamic performance.

[0003] However, existing intelligent rear wing control schemes have significant technical limitations in practical applications. Traditional control logic is mostly based on conventional vehicle dynamics data and basic filtering algorithms (such as standard Kalman filtering) for attitude estimation. In this framework, external aerodynamic loads are usually simplified as unknown disturbances or process noise. Under high-performance driving conditions, the rapid angle adjustment of the intelligent rear wing in a very short time will generate huge vertical loads of up to hundreds of kilograms. This sudden downforce will compress the rear suspension and cause the front of the car to lift relatively. Because existing algorithms cannot effectively distinguish between the suspension travel changes caused by the rear wing pressure and the attitude changes caused by the actual road surface slope undulations, the sensors will produce seriously ambiguous readings. This aerodynamic-mechanical coupling interference will cause data distortion acquired by the inertial measurement unit (IMU) and altitude sensor, causing the vehicle control system to misjudge the vehicle attitude, for example, mistaking the pitch angle change caused by the rear wing pressure as the vehicle going uphill. This misjudgment will lead to incorrect torque intervention, unreasonable braking force distribution, or incorrect rear wing angle compensation, thus forming a vicious cycle, destroying the dynamic balance of the entire vehicle, and making it difficult to meet the overall vehicle handling and stability requirements under harsh conditions.

[0004] Therefore, an optimized intelligent tail wing control method based on multi-sensor fusion is desired. Summary of the Invention

[0005] To address the aforementioned technical problems, this application provides an intelligent tail wing control system and method based on multi-sensor fusion.

[0006] According to one aspect of this application, a smart tail fin control method based on multi-sensor fusion is provided, comprising: Obtain the raw sensor dataset; The original sensor dataset is spatiotemporally aligned and preprocessed to obtain a synchronization state vector. The self-induced aerodynamic load is obtained by performing online simulation of the synchronization state vector and the current tail fin angle value. Decoupled extended Kalman filter estimation is performed on the self-induced aerodynamic load and the synchronization state vector to obtain the net vehicle body attitude state; The target tail wing angle value is obtained by performing downforce optimization decision based on the net vehicle body attitude state. The target tail fin angle value is converted into a PWM control signal for the motor actuator.

[0007] According to another aspect of this application, a smart tail fin control system based on multi-sensor fusion is provided, comprising: The data acquisition module is used to acquire the raw sensor dataset; The data spatiotemporal alignment and preprocessing module is used to perform spatiotemporal alignment and preprocessing on the original sensor dataset to obtain the synchronization state vector; The self-induced aerodynamic load online derivation module is used to perform self-induced aerodynamic load online derivation on the synchronization state vector and the current tail fin angle value to obtain the self-induced aerodynamic load; The decoupled extended Kalman filter estimation module is used to perform decoupled extended Kalman filter estimation on the self-induced aerodynamic load and the synchronization state vector to obtain the net vehicle body attitude state; The downforce optimization decision module is used to perform downforce optimization decisions based on the net vehicle body attitude state to obtain the target tail wing angle value. The PWM control signal generation module is used to convert the target tail fin angle value into a PWM control signal for the motor actuator.

[0008] Compared with existing technologies, this application provides an intelligent tail wing control system and method based on multi-sensor fusion. It utilizes online-derived self-induced aerodynamic loads as known feedforward inputs and employs a decoupled extended Kalman filter algorithm to remove unsteady-state disturbances caused by tail wing movements from the raw data, thereby obtaining the true net vehicle attitude. Based on this, it achieves precise and coordinated downforce control through multi-objective optimization decision-making based on the net state. In this way, under high-performance driving scenarios, it ensures that the vehicle achieves excellent high-speed stability while maintaining ideal front-rear axle load distribution and steering responsiveness, comprehensively improving the vehicle's dynamic handling limits and driving safety. Attached Figure Description

[0009] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 This is a flowchart of a smart tail wing control method based on multi-sensor fusion according to an embodiment of this application; Figure 2 This is a schematic diagram of the data flow of the intelligent tail wing control method based on multi-sensor fusion according to an embodiment of this application; Figure 3 This is a block diagram of a smart tail wing control system based on multi-sensor fusion according to an embodiment of this application. Detailed Implementation

[0011] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0012] As indicated in this application and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not specifically singular and may include plural forms. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0013] While this application makes various references to certain modules of the systems according to embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The modules described are merely illustrative, and different aspects of the systems and methods may use different modules.

[0014] Flowcharts are used in this application to illustrate the operations performed by the system according to embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.

[0015] In the technical solution of this application, an intelligent tail wing control method based on multi-sensor fusion is proposed. Figure 1 This is a flowchart of a smart tail wing control method based on multi-sensor fusion according to an embodiment of this application. Figure 2 This is a system architecture diagram of an intelligent tail fin control method based on multi-sensor fusion according to an embodiment of this application. Figure 1 and Figure 2As shown, the intelligent tail wing control method based on multi-sensor fusion according to an embodiment of this application includes the following steps: S1, acquiring the original sensor dataset; S2, performing spatiotemporal alignment and preprocessing on the original sensor dataset to obtain a synchronization state vector; S3, performing online derivation of self-induced aerodynamic load on the synchronization state vector and the current tail wing angle value to obtain the self-induced aerodynamic load; S4, performing decoupled extended Kalman filter estimation on the self-induced aerodynamic load and the synchronization state vector to obtain the net vehicle body attitude state; S5, performing downforce optimization decision based on the net vehicle body attitude state to obtain the target tail wing angle value; S6, converting the target tail wing angle value into a PWM control signal for the motor actuator.

[0016] Specifically, S1 involves acquiring the raw sensor dataset. This raw sensor dataset includes basic vehicle dynamics data, inertial measurement unit (IMU) data, chassis attitude data, and actuator feedback data. Specifically, the basic vehicle dynamics data mainly covers fundamental driving parameters such as longitudinal velocity scalars; the IMU data includes acceleration and angular velocity information reflecting changes in the vehicle's three-axis attitude; the chassis attitude data typically refers to real-time observations from suspension height or travel sensors; and the actuator feedback data specifically refers to the current rear wing angle value returned by the rear wing drive motor and displacement sensor. By acquiring multi-dimensional raw data, the system can provide essential underlying support for subsequent data preprocessing, self-induced aerodynamic load deduction, and decoupled state estimation, enabling the rear wing to automatically adjust according to real-time vehicle speed and attitude to achieve the technical objective of optimizing downforce.

[0017] In practice, firstly, the electrical signals output by each sensor are read in real time at a fixed sampling frequency through sensor interfaces deployed at different locations on the vehicle body. These signals are typically analog voltage values ​​or digital bus signals, directly reflecting the measurement results of physical quantities. Secondly, the collected signals are categorized according to their physical meaning and data source, and combined into a structured raw sensor dataset. This process ensures the consistency of the data format before it enters subsequent processing steps.

[0018] Specifically, S2 involves performing spatiotemporal alignment and preprocessing on the original sensor dataset to obtain a synchronization state vector. It should be understood that the original sensor dataset encompasses heterogeneous data from different physical locations on the vehicle, with varying sampling frequencies and transmission delays, including basic vehicle dynamics data, inertial measurement unit data, chassis attitude data, and actuator feedback data. Without unified spatiotemporal benchmarking of these data, subsequent online simulations of self-induced aerodynamic loads will face severe data misalignment, preventing the system from accurately distinguishing between aerodynamic loads caused by tail wing movement and attitude changes caused by actual road surface undulations, thus leading to erroneous torque intervention or tail wing angle compensation. Preprocessing to extract the synchronization state vector provides high-confidence observation input for subsequent decoupled extended Kalman filtering, ensuring accurate downforce optimization based on the net vehicle body attitude without increasing hardware costs.

[0019] In practice, the system first performs spatial alignment, transforming sensor data scattered across different mounting points on the vehicle body into the vehicle's center-of-gravity coordinate system (VCS) to eliminate geometric errors introduced by sensor mounting position deviations. The spatial transformation formula is as follows:

[0020] in, The position vector in the centroid coordinate system. The original vector in the sensor coordinate system. For rotation matrix, This is the translation offset vector. Next, an interpolation algorithm is used to align signals with different sampling frequencies (e.g., 100Hz IMU data and 20Hz suspension data) to the same control timestamp. Taking linear interpolation as an example, the calculation formula is:

[0021] in, The observation value at the target synchronization time. and For the known observations of adjacent sampling points, The target synchronization timestamp is then used. Subsequently, the aligned data undergoes preprocessing such as sliding window filtering or outlier removal to ultimately synthesize the aforementioned synchronization state vector, which includes kinematic reference state, vehicle dynamic response, and geometric attitude observations.

[0022] Specifically, in S3, the self-induced aerodynamic load is derived online from the synchronized state vector and the current tail wing angle value. It should be understood that the active aerodynamic components of high-performance vehicles generate enormous vertical loads of up to hundreds of kilograms when rapidly adjusting their angles (e.g., within less than 200 ms). This self-induced disturbance generated by the tail wing's own movement directly compresses the suspension and causes changes in vehicle attitude due to non-road factors. If the control system cannot deduce this aerodynamic load in real time, the inertial measurement unit (IMU) and altitude sensor will produce ambiguous readings, causing the system to misjudge whether the vehicle is going uphill or encountering road undulations, leading to incorrect torque intervention or control strategy failure. By deduce the self-induced aerodynamic load online, the system can provide accurate disturbance source input for subsequent decoupling filtering, thereby eliminating unsteady-state disturbances and ensuring the dynamic balance between the aerodynamic center and the center of mass.

[0023] In practice, the first step is to extract state parameters and perform aerodynamic mapping on the synchronized state vector and the current tail wing angle value to obtain the interpolated lift coefficient and longitudinal velocity scalar. Since the active aerodynamic components of high-performance vehicles generate significant vertical loads during rapid adjustments, the system cannot accurately deduce the self-induced disturbances generated by the tail wing if aerodynamic coefficients matching the current motion state and physical parameters cannot be obtained in real time. Existing algorithms often treat aerodynamic forces as unknown disturbances, leading to ambiguity in sensor readings under aerodynamic-mechanical coupling, causing the vehicle control system to misjudge road slope or undulations. Accurate extraction of the lift coefficient and velocity scalar provides the necessary parameter inputs for subsequent calculations of the net vehicle attitude, thereby eliminating attitude estimation distortion caused by tail wing downforce compressing the suspension and ensuring vehicle handling and stability under extreme conditions. It is worth noting that the aerodynamic map refers to a two-dimensional or multi-dimensional data matrix pre-stored in the controller, generated based on high-precision wind tunnel tests or fluid dynamics simulations. The lift coefficient is a dimensionless parameter characterizing the ability of a tail fin to generate lift (or downforce) at a specific angle of attack; the longitudinal velocity scalar is a physical quantity extracted from the kinematic reference state that characterizes the real-time rate of the vehicle in the direction of travel.

[0024] In this process, firstly, the kinematic reference state is resolved from the synchronization state vector, and the longitudinal velocity scalar is extracted from it. This scalar represents the physical quantity of the vehicle's real-time speed in the forward direction. Secondly, the current tail wing angle value fed back by the motor actuator is read in real time. Furthermore, As an index variable, the system accesses the expanded high-precision wind tunnel data map to locate the corresponding lift coefficient range. Since the physical tail fin angle may vary continuously, while the map data is typically discrete, the system needs to perform linear interpolation to obtain the accurate interpolated lift coefficient. The specific linear interpolation calculation formula is as follows:

[0025] in, This is the final interpolated lift coefficient. This is the current tail fin angle value. and These are adjacent nodes with known angles in the graph. and The corresponding known lift coefficient is given. Through the above calculations, the system can output high-precision aerodynamic characteristic parameters in real time.

[0026] Next, the dynamic pressure area factor is obtained by multiplying the longitudinal velocity scalar by the dynamic pressure reference area. It should be understood that the vertical load generated by the active aerodynamic components depends not only on the lift coefficient resulting from their physical angle, but also on the dynamic pressure intensity and the area of ​​force application during vehicle movement. By calculating the dynamic pressure area factor, the system can transform abstract vehicle speed information into a physically meaningful energy quantification index, providing a direct product basis for subsequent synthesis of self-induced vertical loads using the interpolated lift coefficient and dynamic pressure area factor. This step ensures that the derived self-induced aerodynamic load accurately reflects the aerodynamic force intensity under current driving conditions, effectively decoupling the aerodynamic-mechanical coupling interference caused by the tail wing's movement and preventing the vehicle control system from misjudging the vehicle's attitude due to aerodynamic downforce compressing the suspension. The dynamic pressure area factor is a composite parameter calculated based on aerodynamic principles, combined with ambient air density, vehicle instantaneous speed, and the geometric characteristics of the tail wing itself; it directly determines the magnitude of the aerodynamic coefficient converted into actual physical force.

[0027] In this process, the preset air density constant and tail fin reference area are first retrieved. Next, the longitudinal velocity scalar in the synchronization state vector is obtained and multiplied according to the dynamic pressure calculation formula to obtain the dynamic pressure area factor. During software execution, the algorithm first squares the longitudinal velocity scalar to obtain the velocity square term, then multiplies this term by the air density constant and takes half of it, and finally multiplies it by the tail fin reference area to produce the final dynamic pressure area factor.

[0028] Furthermore, the interpolated lift coefficient and dynamic pressure area factor are used to synthesize and quantify self-induced vertical loads to obtain self-induced aerodynamic loads. It is understandable that the rapid angle adjustments of active aerodynamic components in high-performance vehicles generate enormous vertical loads within a very short time. This self-induced disturbance directly compresses the suspension and causes changes in vehicle attitude due to non-road factors. If the system only treats aerodynamic forces as unknown external disturbances, ambiguous readings from sensors can cause the control system to misjudge road slope or undulations, leading to incorrect torque intervention commands. By synthesizing and quantifying the loads, the system can transform abstract aerodynamic coefficients into specific physical force values, thus providing accurate known feedforward inputs for subsequent decoupled extended Kalman filtering, eliminating unsteady-state disturbances, and ensuring the dynamic balance between the aerodynamic center and the center of mass.

[0029] In this process, the interpolated lift coefficient and the dynamic pressure area factor calculated from the longitudinal velocity scalar, air density, and tail fin reference area are obtained. Then, the interpolated lift coefficient and dynamic pressure area factor are used to synthesize and quantify the self-induced vertical load using the following formula:

[0030] in, This represents the interpolated lift coefficient. This is the dynamic pressure area factor. air density, The square of the longitudinal velocity scalar. This is the reference area for the tail fin.

[0031] Specifically, in S4, a decoupled extended Kalman filter is used to estimate the self-induced aerodynamic loads and the synchronized state vector to obtain the net vehicle attitude state. It should be understood that the active aerodynamic components of high-performance vehicles generate significant vertical loads when rapidly adjusting angles (e.g., from 0 degrees to 15 degrees instantaneously to provide braking assistance), resulting in suspension travel changes sufficient to interfere with the readings of the inertial measurement unit (IMU) and suspension height sensor. Existing Kalman filter or state observer designs typically treat external aerodynamic forces as slowly varying unknown disturbances or simple assimilation process noise, causing the algorithm to fail to distinguish the root causes of vehicle "nose-up" or "tail-down." This ambiguity in sensor readings can lead vehicle control systems (such as ESP) to misjudge road conditions and issue incorrect torque intervention commands. By performing decoupled extended Kalman filter estimation, the system can use the online derived self-induced aerodynamic load as a known feedforward input to remove the unsteady disturbances caused by the tail wing movement from the original motion signal, thereby obtaining the net body attitude that reflects the true motion characteristics of the vehicle, ensuring the dynamic balance between the aerodynamic center and the center of mass, and avoiding control strategy failure or deterioration of vehicle dynamic balance.

[0032] In practice, firstly, based on the self-induced aerodynamic load and the synchronization state vector, a priori state estimation vector and a priori covariance matrix are constructed. It should be understood that the filter cannot rely solely on current sensor measurements (i.e., the synchronization state vector) to estimate the state, because the measurements themselves contain noise and may not directly reflect all the state quantities that need to be estimated. Therefore, before incorporating new measurement data from the current moment, the best guess of the current state based on historical information and the system model is used to obtain the priori state estimation vector. Here, the priori state estimation vector is a preliminary prediction of the vehicle's attitude at the current moment, while the priori covariance matrix is ​​used to quantify the degree of uncertainty in this prediction process. Simultaneously, due to the inaccuracy of the model itself and the presence of process noise, the uncertainty of the prediction accumulates; this quantification of uncertainty is represented by the priori covariance matrix.

[0033] Next, based on the synchronized state vector, decoupling residual calculation and gain solving are performed on the prior state estimation vector and prior covariance matrix to obtain the decoupled measurement residual and Kalman gain matrix. The decoupled measurement residual refers to the deviation between sensor observations and model predictions, considering the known influence of self-induced aerodynamic loads on the suspension system; while the Kalman gain matrix is ​​an optimal weight matrix used to balance the confidence levels of the prior prediction model and the actual observation data in the posterior correction stage.

[0034] In this process, firstly, the geometric attitude observations in the synchronization state vector are extracted as the actual observation input at the current moment. These are then combined with the prior state estimation vector generated in the previous step and the online derived self-induced aerodynamic loads, and processed using an observation model function that includes aerodynamic compensation terms. By separating the expected compression of the chassis height by the self-induced aerodynamic loads from the observation terms, the system achieves decoupling at the measurement level. The decoupled measurement residual formula is expressed as:

[0035] in, The measurement residuals after decoupling. These are the geometric attitude observations in the synchronization state vector. For nonlinear observation model functions that include aerodynamic disturbance terms, This is the prior state estimation vector. This is for self-induced aerodynamic loads. Secondly, the Kalman gain is calculated based on the prior covariance matrix and the observation matrix to determine the contribution of the observation residuals to the state correction. The gain calculation formula is expressed as:

[0036] in, Here is the Kalman gain matrix. The prior covariance matrix, For the observation matrix, This represents the observation noise covariance matrix. In the specific implementation process, according to the operating logic of the decoupled extended Kalman filter estimation module, the system acquires a synchronous state vector containing vehicle dynamic response and geometric attitude observations in real time during each control cycle. This module utilizes the aforementioned decoupling formula to accurately identify sensor numerical drift caused by the rear wing downforce compressing the suspension, and separates it from attitude changes caused by actual road surface undulations. Through this refined residual decoupling and dynamic gain calculation, the system can eliminate the negative impact of aerodynamic-mechanical coupling on attitude estimation accuracy without increasing hardware sensor costs. This provides a high-confidence state benchmark for subsequent downforce optimization decisions based on the net vehicle attitude state, ultimately achieving the technical objective of improving overall vehicle handling stability and safety.

[0037] Furthermore, posterior state corrections are applied to the prior state estimation vector, the decoupled measurement residuals, and the Kalman gain matrix to obtain the net vehicle attitude state. It is understandable that in high-performance vehicles, under special conditions such as high-speed driving or large-angle rear wing adjustments, the aerodynamic pitch moment generated by the high angle of attack of the rear wing produces a lever-like effect, significantly raising the front axle while compressing the rear suspension. If the control algorithm only remains at the prior prediction stage without real-time correction and decoupling based on measured data, it will lead to a severe imbalance in the front and rear axle load distribution, potentially causing insufficient front wheel grip or steering failure. Through posterior state correction, the system can use the Kalman gain to optimally adjust the weights of the prediction model, thereby accurately constraining the loss of front axle load and eliminating unsteady-state interference without increasing hardware sensor costs. This ensures that the vehicle maintains the optimal aerodynamic center of gravity matching relationship under extreme conditions, significantly improving the vehicle's dynamic handling limits and driving safety. The final net vehicle attitude state refers to the final generated state vector that reflects the vehicle's true physical motion characteristics after eliminating the effects of aerodynamic-mechanical coupling.

[0038] In this process, a correction term for compensating for prior prediction bias is calculated by multiplying the Kalman gain matrix with the decoupled measurement residuals. This correction term is then applied to the prior state estimation vector, thereby achieving the optimal update of the system state. Through this linear weighted combination, the system can effectively suppress sensor noise and ensure the convergence of attitude estimates.

[0039] Specifically, in step S5, a downforce optimization decision is made based on the net vehicle body attitude state to obtain the target tail wing angle value. Since the net vehicle body attitude state has eliminated the influence of aerodynamic effects generated by the tail wing itself, it more accurately represents the vehicle's true attitude caused by road surface excitation, inertial forces, etc. Therefore, in the technical solution of this application, the aerodynamic downforce that the tail wing should generate is further determined based on this net state to optimize the vehicle's dynamic performance.

[0040] In practice, the system first performs multimodal weighting and dynamic parameter analysis on the net vehicle body attitude state and driving mode signal to obtain an optimized parameter tuple. Since high-performance vehicles exhibit significant differences in the priority definitions of downforce, driving stability, and energy consumption under different driving scenarios, the system can selectively adjust the objective function weights in subsequent optimization decisions by acquiring the decoupled net vehicle body attitude state, which reflects the vehicle's true motion characteristics, and combining it with the driving mode signal selected by the driver through the human-machine interface. This process ensures that the optimization decision is not only based on accurate physical states but also takes into account the driver's subjective intentions. This improves tire adhesion utilization and enhances handling limits while ensuring that the rear wing action conforms to the dynamic characteristics of the current driving mode, avoiding the limitations of a single control logic under complex conditions. The driving mode signal refers to the logical state bit representing the vehicle's current operating style (e.g., economy, sport, or track mode). The final optimized parameter tuple is a composite data structure encapsulating weight coefficients, dynamic constraints, and state reference benchmarks, serving as the direct input parameters for constructing the subsequent cost function.

[0041] In this process, firstly, the net vehicle attitude state from the filtering module is retrieved in real time, and the driving mode signal in the vehicle bus is monitored simultaneously. Secondly, the system executes weight configuration logic, retrieving the corresponding optimization target weight from the preset mapping matrix according to the current driving mode. For example, in high-performance mode, the system will allocate a higher weight to the downforce gain term. Then, combined with the current net vehicle attitude data (such as real-time vehicle speed and vehicle pitch state), the dynamic constraint boundary in the current physical environment is calculated. Subsequently, the weight coefficients and the resolved dynamic constraints are integrated into an optimization parameter tuple.

[0042] Next, the cost function is constructed and the extremum is solved based on the optimized parameter tuple to obtain the original optimal angle. It should be understood that traditional cost function construction only focuses on maximizing the aerodynamic downforce of the rear axle, i.e., minimizing... This stability term aims to improve the rear tire's grip utilization. However, this single-point-perspective optimization strategy has significant physical model flaws, neglecting the systematic impact of aerodynamic pitch moment coupling on the vehicle's attitude balance. Under special conditions such as high-speed driving or large-angle rear wing adjustments, the high angle-of-attack rear wing not only generates vertical downforce but also produces a huge aerodynamic pitch moment around the vehicle's lateral axis. This moment creates a lever-like effect, significantly lifting the front axle while compressing the rear suspension, leading to an unexpected decrease in the vertical load on the front wheels. If the control algorithm fails to capture this coupling relationship, solely pursuing rear axle grip will result in a severe imbalance in the front-rear axle load distribution, leading to insufficient front wheel grip. This causes the risk of rear axle instability to jump directly to the risk of front axle steering failure, disrupting the dynamic balance between the vehicle's aerodynamic center and center of gravity, making it difficult to meet the stringent requirements for overall vehicle handling stability in high-performance driving scenarios.

[0043] To address the aforementioned technical shortcomings, this improved scheme employs a global optimization mechanism based on aerodynamic center migration constraints. Specifically, firstly, multi-dimensional aerodynamic vector calculations are performed on the optimization parameter tuple and the currently traversed candidate tail fin angles to obtain the aerodynamic vectors. During this process, the system receives the optimization parameter tuple and the current candidate tail fin angles as input and accesses the expanded high-precision wind tunnel data map. Unlike traditional schemes that only query the lift-drag coefficient, this step simultaneously indexes the pitching moment coefficient, which characterizes the rotational effect. Then, based on the principles of fluid mechanics, the vertical aerodynamic load is calculated simultaneously. Aerodynamic drag And the aerodynamic pitching moment that can quantify the leverage effect. The calculation logic is as follows:

[0044] in, Vertical aerodynamic load (unit: N); Aerodynamic drag (unit: N); pneumatic pitching moment (unit: N·m); Air density (unit: kg / m³); The longitudinal speed of the vehicle (unit: m / s); For reference area (unit: m²); The reference chord length of the tail fin (unit: m); , and These are the lift, drag, and pitch moment coefficients at the corresponding angles. This step aims to extend the physical effect of the tail fin from a single vertical force dimension to a composite dimension of force and moment, providing accurate vector input for subsequent analysis of the vehicle's attitude deflection.

[0045] Secondly, axle load redistribution and balance calculations are performed on the aerodynamic vectors to obtain the axle load structure. That is, using the aerodynamic vectors generated in the previous step, combined with vehicle wheelbase and static axle load data, the dynamic axle load transfer caused by the aerodynamic pitching moment is calculated based on the principle of rigid body static equilibrium. This calculation quantifies the degree to which the front axle is lifted by decoupling the lever effect generated by the rear wing, and based on this, it calculates the total aerodynamic load on the vehicle's front axle at that moment. Total load on rear axle Then, the front and rear axle load ratio, which characterizes the current handling tendency, is obtained. Its operation logic is as follows:

[0046] in, The amount of axle load transfer due to aerodynamic pitching moment (unit: N); Vehicle wheelbase (unit: m); and Static front and rear axle loads (unit: N); and The dynamic total load on the front and rear axles after introducing aerodynamic effects (unit: N); The dynamic front-to-rear axle load distribution ratio (dimensionless) is used. This step reveals the potential weakening effect of tail wing movement on front wheel grip through mathematical modeling, preventing front axle instability caused by blindly increasing rear axle downforce.

[0047] Furthermore, a global cost optimization under multi-objective constraints is performed on the axle load structure to obtain the original optimal angle. That is, based on the calculated axle load structure, a comprehensive cost function is constructed that includes stability, energy consumption, and balance constraints. This function introduces a balance penalty term on top of the original grip and drag optimization. By adjusting the weights, while ensuring the vehicle receives sufficient vertical load, the front-to-rear axle load ratio is forcibly constrained to always remain near the ideal handling range (e.g., 40:60), thereby solving for the target rear wing angle that balances cornering limits and steering responsiveness. The optimization formula is as follows:

[0048] in, The final output target tail fin angle (unit: deg); , and These are the weighting coefficients for the grip, drag, and balance terms, respectively. The total lateral force requirement for the vehicle (unit: N); The road surface adhesion coefficient; The target value for the ideal axle load distribution ratio under the current driving mode (dimensionless).

[0049] This improved solution achieves a leap from local rear axle optimization to overall vehicle steady-state balance optimization, effectively solving the understeer risk caused by neglecting the aerodynamic pitch moment coupling effect in existing technologies. This allows the system to improve the high-speed rear axle stability of the vehicle while precisely constraining the loss of front axle load, ensuring that the front wheels still have sufficient steering grip. This enables the vehicle to maintain the optimal aerodynamic center and center of gravity matching relationship under extreme conditions, significantly improving the vehicle's dynamic handling limits and driving safety.

[0050] Furthermore, the original optimal angle is physically constrained and limited to obtain the target tail wing angle value. It should be understood that directly issuing the unfiltered theoretical angle to the execution layer could lead to mechanical overshoot, hardware interference, or excessive current damage to the drive system. Therefore, in the technical solution of this application, the original optimal angle is further physically constrained and limited to ensure that the issued target command always remains within the redundancy range of mechanical and electrical protection. This ensures the overall robustness and hardware lifespan of the tail wing control system while optimizing downforce in real time based on vehicle speed and body attitude. The final target tail wing angle value is an execution command that meets the physical implementation conditions and is used to directly convert it into physical actions.

[0051] In this process, firstly, the software algorithm retrieves physical extreme constants pre-stored in non-volatile memory. These constants include the minimum and maximum mechanical limit angles achievable by the tail fin. Secondly, the input angle is protected by executing conditional branch judgments or saturated function logic to ensure that the output target tail fin angle value no longer contains any invalid components that exceed physical dimensions or mechanical ranges.

[0052] Specifically, in step S6, the target tail wing angle value is converted into a PWM control signal for the motor actuator. It should be understood that after complex decoupling filtering and optimization decisions in the preceding steps, the obtained target tail wing angle value exists only in the controller's computational logic layer; essentially, it is a digital expected instruction. To drive the hardware-level motor actuator to generate physical displacement, the system must establish a mapping relationship between the logical angle and the driving voltage. By converting the angle value into a PWM control signal with specific duty cycle characteristics, the controller can instruct the actuator to adjust the active aerodynamic components to the ideal position, thereby achieving the ultimate technical objective of automatically adjusting the tail wing angle based on vehicle speed and body posture to optimize downforce. This step ensures that the software-level optimization strategy can accurately guide the hardware-level mechanical response, improving tire adhesion utilization while ensuring the dynamic balance of the vehicle's aerodynamic center and center of gravity.

[0053] Among them, the PWM control signal, or pulse width modulation signal, is the core means of motor control. It achieves precise control of motor speed or position by changing the width of the voltage pulse (i.e., duty cycle) at a fixed frequency.

[0054] In practice, firstly, the target tail fin angle value from the previous step is read in real time. At the same time, the current tail fin angle value is obtained in real time through feedback sensors such as the shaft angle encoder attached to the motor, and the angle deviation between the two is calculated. This deviation value is the direct basis for subsequent control decisions. Its sign represents the direction of motion (positive value means forward rotation is required, negative value means reverse rotation is required), and its absolute value represents the size of the difference from the target position.

[0055] Secondly, based on the calculated angle deviation, a specific control algorithm is applied to calculate the duty cycle (usually expressed as a percentage) of the PWM signal required to drive the motor. A common and effective method is to use a proportional-integral-derivative (PID) control law. The proportional term provides a fast response proportional to the deviation; the integral term eliminates steady-state error, ensuring accurate attainment of the target position; and the derivative term suppresses overshoot, increasing system stability. It is worth noting that the calculated duty cycle value needs to be limited to the range allowed by the motor driver.

[0056] Then, the microcontroller (MCU) generates a PWM waveform with a corresponding duty cycle based on the calculated duty cycle and the preset PWM frequency, using its internal timer and output comparator. Simultaneously, the system needs to determine the motor's rotation direction based on the sign (positive or negative) of the angle deviation. This is typically achieved by using an additional GPIO pin to output high and low levels to control the motor driver chip's steering logic (e.g., high level for forward rotation, low level for reverse rotation). Finally, this PWM signal, containing both direction and power (duty cycle) information, is output to the motor drive circuit, driving the motor to rotate and causing the tail fin to move towards the target angle.

[0057] In summary, the intelligent tail wing control method based on multi-sensor fusion according to the embodiments of this application is explained. It utilizes online-derived self-induced aerodynamic loads as known feedforward inputs and employs a decoupled extended Kalman filter algorithm to remove unsteady-state disturbances caused by tail wing movements from the raw data, thereby obtaining the true net vehicle attitude. Based on this, precise and coordinated downforce control is achieved through multi-objective optimization decision-making based on the net state. In this way, under high-performance driving scenarios, it ensures that the vehicle achieves excellent high-speed stability while maintaining ideal front-rear axle load distribution and steering responsiveness, comprehensively improving the vehicle's dynamic handling limits and driving safety.

[0058] Furthermore, a smart tail wing control system based on multi-sensor fusion is also provided.

[0059] Figure 3 This is a block diagram of a smart tail wing control system based on multi-sensor fusion according to an embodiment of this application. Figure 3 As shown, the intelligent tail wing control system 300 based on multi-sensor fusion according to an embodiment of this application includes: a data acquisition module 310 for acquiring a raw sensor dataset; a data spatiotemporal alignment and preprocessing module 320 for performing spatiotemporal alignment and preprocessing on the raw sensor dataset to obtain a synchronization state vector; a self-induced aerodynamic load online deduction module 330 for performing self-induced aerodynamic load online deduction on the synchronization state vector and the current tail wing angle value to obtain a self-induced aerodynamic load; a decoupled extended Kalman filter estimation module 340 for performing decoupled extended Kalman filter estimation on the self-induced aerodynamic load and the synchronization state vector to obtain a net vehicle body attitude state; a downforce optimization decision module 350 for performing downforce optimization decision based on the net vehicle body attitude state to obtain a target tail wing angle value; and a PWM control signal generation module 360 ​​for converting the target tail wing angle value into a PWM control signal for a motor actuator.

[0060] As described above, the intelligent tail wing control system 300 based on multi-sensor fusion according to the embodiments of this application can be implemented in various wireless terminals, such as servers with intelligent tail wing control algorithms based on multi-sensor fusion. In one possible implementation, the intelligent tail wing control system 300 based on multi-sensor fusion according to the embodiments of this application can be integrated into the wireless terminal as a software module and / or hardware module. For example, the intelligent tail wing control system 300 based on multi-sensor fusion can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the intelligent tail wing control system 300 based on multi-sensor fusion can also be one of many hardware modules of the wireless terminal.

[0061] Alternatively, in another example, the intelligent tail wing control system 300 based on multi-sensor fusion and the wireless terminal can also be separate devices, and the intelligent tail wing control system 300 based on multi-sensor fusion can be connected to the wireless terminal via wired and / or wireless networks, and transmit interactive information in accordance with an agreed data format.

[0062] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A multi-sensor fusion based intelligent tail wing control method, characterized in that, The method comprises the following steps: acquiring an original sensor data set; performing data space-time alignment and preprocessing on the original sensor data set to obtain a synchronous state vector; performing online deduction of self-induced aerodynamic load on the synchronous state vector and a current tail wing angle value to obtain self-induced aerodynamic load; performing decoupled extended Kalman filtering estimation on the self-induced aerodynamic load and the synchronous state vector to obtain a net vehicle body attitude state; performing net state-based downforce optimization decision on the net vehicle body attitude state to obtain a target tail wing angle value; converting the target tail wing angle value into a PWM control signal of a motor actuator. 2.The multi-sensor fusion based intelligent tail wing control method according to claim 1, wherein, The original sensor data set comprises vehicle dynamics basic data, inertial measurement unit data, chassis attitude data and actuator feedback data, and the synchronous state vector comprises kinematic reference states, vehicle body dynamic responses and geometric attitude observation values. 3.The multi-sensor fusion based intelligent tail wing control method according to claim 1, wherein, Performing online deduction of self-induced aerodynamic load on the synchronous state vector and a current tail wing angle value to obtain self-induced aerodynamic load comprises the following steps: extracting state parameters from the synchronous state vector and the current tail wing angle value and performing aerodynamic atlas addressing to obtain interpolated lift coefficients and a longitudinal velocity scalar; performing dynamic pressure reference region multiplication calculation on the longitudinal velocity scalar to obtain a dynamic pressure area factor; performing self-induced vertical load synthesis and quantization on the interpolated lift coefficients and the dynamic pressure area factor to obtain self-induced aerodynamic load. 4.The multi-sensor fusion based intelligent tail wing control method according to claim 3, wherein, Performing self-induced vertical load synthesis and quantization on the interpolated lift coefficients and the dynamic pressure area factor to obtain self-induced aerodynamic load comprises the following steps: performing self-induced vertical load synthesis and quantization on the interpolated lift coefficients and the dynamic pressure area factor by using the following formula: wherein, represents the interpolated lift coefficient, is the dynamic pressure area factor, is the air density, is the square of the longitudinal velocity vector, is the tail reference area. 5.The multi-sensor fusion based intelligent tail wing control method according to claim 1, wherein, Performing decoupled extended Kalman filtering estimation on the self-induced aerodynamic load and the synchronous state vector to obtain a net vehicle body attitude state comprises the following steps: constructing a prior state estimation vector and a prior covariance matrix based on the self-induced aerodynamic load and the synchronous state vector; performing decoupled residual calculation and gain solving on the prior state estimation vector and the prior covariance matrix based on the synchronous state vector to obtain a decoupled measurement residual and a Kalman gain matrix; performing posterior state correction on the prior state estimation vector, the decoupled measurement residual and the Kalman gain matrix to obtain the net vehicle body attitude state. 6.The multi-sensor fusion based intelligent tail wing control method according to claim 1, wherein, Performing net state-based downforce optimization decision on the net vehicle body attitude state to obtain a target tail wing angle value comprises the following steps: performing multi-modal weight configuration and dynamics parameter analysis on the net vehicle body attitude state and a driving mode signal to obtain an optimization parameter tuple; constructing a cost function and solving an extreme value based on the optimization parameter tuple to obtain an original optimal angle; performing physical constraint limiting on the original optimal angle to obtain the target tail wing angle value.

7. The multi-sensor fusion based intelligent tail wing control method according to claim 6, wherein, Constructing a cost function and solving an extreme value based on the optimization parameter tuple to obtain an original optimal angle comprises the following steps: performing multi-dimensional aerodynamic vector calculation on the optimization parameter tuple and a candidate tail wing angle currently traversed to obtain an aerodynamic vector; performing shaft load redistribution and balance degree calculation on the aerodynamic vector to obtain a shaft load structure; performing global cost optimization under multi-objective constraints on the shaft load structure to obtain the original optimal angle.

8. An intelligent tail wing control system based on multi-sensor fusion, characterized in that, The method comprises the following steps: a data acquisition module is configured to acquire an original sensor data set; The data space-time alignment and preprocessing module is configured to perform data space-time alignment and preprocessing on the original sensor data set to obtain a synchronization state vector; The self-induced aerodynamic load online deduction module is configured to perform self-induced aerodynamic load online deduction on the synchronization state vector and a current tail wing angle value to obtain a self-induced aerodynamic load; The decoupled extended Kalman filtering estimation module is configured to perform decoupled extended Kalman filtering estimation on the self-induced aerodynamic load and the synchronization state vector to obtain a net vehicle body attitude state; The downforce optimization decision module is configured to perform downforce optimization decision based on the net state on the net vehicle body attitude state to obtain a target tail wing angle value; The PWM control signal generation module is configured to convert the target tail wing angle value into a PWM control signal of a motor actuator.

Images