Anti-motion sickness intelligent shock absorption control system

Abstract

The present application relates to the technical field of vehicle damping control, and discloses an anti-car-sickness intelligent damping control system. A multi-dimensional sensing array of the system collects vehicle driving state, passenger physiological indexes and environmental disturbance data in real time, and generates multi-dimensional driving data set in time and space alignment through time stamp alignment and space coordinate synchronization. A data separation engine carries out feature decoupling reconstruction on the data set, separates the main feature matrix and the abnormal disturbance identification set, and a feature screening unit screens the core feature sequence in combination with the vehicle dynamics parameter library. A damping strategy generator generates a multi-level collaborative damping control strategy accordingly. An abnormal feedback regulator analyzes the abnormal disturbance identification set and triggers strategy dynamic tuning. A control instruction distributor converts the tuned strategy into a driving signal and issues it to the suspension actuator cluster. The system can comprehensively perceive multi-dimensional information, dynamically adjust the damping strategy, effectively relieve car sickness and improve passenger comfort.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vehicle shock absorption control, in particular to an anti-car-sickness intelligent shock absorption control system. BACKGROUND

[0002] During vehicle driving, car sickness is a common problem affecting the riding experience of passengers, which is closely related to the change of vehicle driving state, environmental disturbance and passenger physiological state. The current vehicle shock absorption system on the market mostly adopts a passive shock absorption structure, which can only buffer the road bumps within a fixed frequency range through a pre-set mechanical damping characteristic, and cannot adjust the shock absorption strategy according to the real-time changes of vehicle driving state and passenger physiological needs.

[0003] When the vehicle drives on an unpaved road and encounters irregular bumps, or frequently starts and stops and turns on urban roads, the passive shock absorption system is difficult to quickly respond to these dynamic driving scenarios, resulting in large amplitude vibration and attitude change of the vehicle body. This unstable driving state will continuously and irregularly stimulate the vestibular system of the passenger's inner ear, and further cause car sickness symptoms such as dizziness and nausea, especially for the elderly, children and other people with sensitive vestibular function, and such discomfort reactions are more obvious.

[0004] Although some existing vehicles are equipped with active shock absorption systems, most of them only use a single or a few vehicle state parameters such as vehicle driving speed and suspension stroke as control basis, and lack real-time monitoring and feedback of passenger physiological indicators. These systems cannot accurately identify whether the passenger has shown signs of car sickness or discomfort, and can only adjust the shock absorption parameters according to the fixed control logic, making it difficult to achieve personalized shock absorption control for different passenger individual differences. In addition, the existing active shock absorption system has a lag in adjusting the control strategy when facing sudden road disturbances, and cannot timely suppress the instantaneous severe vibration of the vehicle body, which still causes strong physiological stimulation to the passengers, and cannot fundamentally solve the problem of car sickness.

[0005] With the continuous improvement of consumers' requirements for vehicle riding comfort, the shortcomings of traditional shock absorption systems in adaptability, personalization and response speed are increasingly prominent, and an intelligent control system that can comprehensively perceive the vehicle driving state, passenger physiological indicators and environmental disturbance, and dynamically adjust the shock absorption strategy is urgently needed to effectively alleviate or even eliminate car sickness and improve the riding experience of passengers. SUMMARY

[0006] The purpose of the present application is to provide an anti-car-sickness intelligent shock absorption control system to solve the problems raised in the background.

[0007] To achieve the above purpose, the present application provides an anti-car-sickness intelligent shock absorption control system, which comprises:

[0008] A multi-dimensional sensor array is used to collect vehicle driving state data, passenger physiological index data and environmental disturbance data in real time, and the vehicle driving state data, passenger physiological index data and environmental disturbance data are subjected to timestamp alignment and spatial coordinate synchronization to generate a multi-dimensional driving data set aligned in time and space.

[0009] A data separation engine is used to perform feature decoupling reconstruction processing on the multi-dimensional driving data set aligned in time and space, and separate a main feature matrix for shock absorption control and an abnormal disturbance identification set for feedback tuning.

[0010] A feature screening unit is used to perform multi-dimensional feature correlation analysis based on the main feature matrix and in combination with a preconfigured vehicle dynamics parameter library to screen a core feature sequence for shock absorption decision-making.

[0011] A shock absorption strategy generator is used to generate a multi-level coordinated shock absorption control strategy according to the core feature sequence.

[0012] An abnormal feedback regulator is used to analyze the abnormal disturbance identification set, identify a sudden disturbance event in the vehicle driving state data or an abnormal fluctuation pattern in the passenger physiological index data, and trigger dynamic tuning of the shock absorption control strategy.

[0013] A control instruction distributor is used to convert the tuned shock absorption control strategy into a driving signal and distribute it to a suspension actuator cluster.

[0014] Preferably, the data separation engine comprises:

[0015] A matrix conversion module is used to convert the multi-dimensional driving data set aligned in time and space into a standardized feature tensor.

[0016] A tensor decomposition module is used to perform rank-constrained decomposition on the standardized feature tensor to obtain a steady-state feature basis and a transient disturbance component.

[0017] The steady-state feature basis constitutes the main feature matrix, and the transient disturbance component constitutes the abnormal disturbance identification set.

[0018] Preferably, the feature screening unit comprises:

[0019] An association graph construction module is used to extract vehicle longitudinal acceleration features, lateral swing angle features, vertical vibration frequency spectrum features, passenger heart rate variation features and eye movement trajectory features from the main feature matrix to construct a multi-modal feature association graph.

[0020] An entropy calculation module is used to calculate the dynamic correlation entropy of each feature based on the multi-modal feature association graph.

[0021] A sequence generation module configured to generate a core feature sequence according to the top K features in the dynamic correlation entropy descending order.

[0022] Preferably, the shock absorption strategy generator comprises:

[0023] A strategy mapping module configured to map the core feature sequence to a preset vehicle dynamics parameter library to obtain a basic shock absorption control curve.

[0024] A multi-level optimization module configured to perform hierarchical optimization decomposition on the basic shock absorption control curve according to a vehicle suspension topology to generate the multi-level cooperative shock absorption control strategy comprising a main suspension level control strategy and an auxiliary suspension level control strategy.

[0025] Preferably, the abnormal feedback regulator comprises:

[0026] A disturbance mode recognition module configured to analyze time-frequency features of the transient disturbance component to recognize a sudden road impact mode, a continuous resonance mode or a passenger physiological instability mode.

[0027] A tuning parameter generation module configured to retrieve a preset tuning rule library according to the recognized disturbance mode to generate a suspension stiffness correction coefficient or a damping compensation gradient.

[0028] A strategy recalculation module configured to inject the suspension stiffness correction coefficient or the damping compensation gradient into the multi-level cooperative shock absorption control strategy for real-time recalculation.

[0029] Preferably, the control instruction distributor comprises:

[0030] A signal conversion module configured to convert the recalculated multi-level cooperative shock absorption control strategy into a pulse width modulation signal.

[0031] A partitioned distribution module configured to distribute the pulse width modulation signal to a corresponding main suspension controller or auxiliary suspension controller according to a position code of a suspension actuator.

[0032] Preferably, the system further comprises:

[0033] A cloud cooperative processor configured to receive a cross-vehicle cooperative identifier in the transient disturbance component and dispatch a shock absorption feature matrix of a neighboring vehicle from the cloud.

[0034] A coupling strategy optimizer configured to generate a coupling constraint term based on the shock absorption feature matrix of the neighboring vehicle to implement cooperative optimization on the multi-level cooperative shock absorption control strategy.

[0035] Preferably, the system further comprises:

[0036] a feature tracing module configured to trace associated feature vectors in historical driving data sets when the transient disturbance component exceeds a preset threshold;

[0037] a disturbance clustering module configured to cluster and group the associated feature vectors according to vehicle working condition labels to generate device-level disturbance clusters and environment-level disturbance clusters;

[0038] the device-level disturbance clusters are input into a suspension health diagnosis unit, and the environment-level disturbance clusters are input into an environment adaptation learning unit.

[0039] Preferably, the system further comprises:

[0040] an edge computing partition comprising a real-time processing area and a non-volatile storage area that are isolated from each other;

[0041] the real-time processing area runs the data separation engine and the feature screening unit, and the non-volatile storage area solidifies the preset vehicle dynamics parameter library and the tuning rule library.

[0042] Preferably, the system further comprises:

[0043] a secure isolation repeater configured to establish an encrypted channel between the edge computing partition and a cloud-side co-processor and implement runtime verification on the coupled constraint items issued.

[0044] Compared with the prior art, the present application has the following beneficial effects:

[0045] The anti-motion sickness intelligent shock absorption control system realizes real-time collection of vehicle driving state data, passenger physiological index data and environmental disturbance data through a multi-dimensional sensing array, and completes timestamp alignment and spatial coordinate synchronization to generate a time-space aligned multi-dimensional driving data set. Compared with the traditional shock absorption system which relies only on the perception of a single or a small number of vehicle parameters, the anti-motion sickness intelligent shock absorption control system can more comprehensively and accurately capture various key information that affects the passenger's riding experience and motion sickness phenomenon, providing a richer and more reliable data foundation for the subsequent development of shock absorption control strategies.

[0046] The data separation engine performs feature decoupling and reconstruction processing on the time-space aligned multi-dimensional driving data set to separate a main feature matrix for shock absorption control and an abnormal disturbance identification set for feedback tuning. This processing method can effectively screen out core data features directly related to shock absorption control, accurately identify abnormal disturbance information in the data set, avoid interference of irrelevant data on shock absorption control decisions, improve the pertinence and efficiency of data processing, and enable the subsequent feature screening and strategy generation links to focus more on key control elements, reduce redundant calculations, and ensure the efficiency of system operation.

[0047] The feature screening unit screens out a core feature sequence used for shock absorption decision based on the main feature matrix in combination with a preset vehicle dynamics parameter library, the process fully integrates professional knowledge of vehicle dynamics, can accurately extract key features that play a decisive role in shock absorption effect from complex multi-dimensional data features, ensures that the generated shock absorption decision basis conforms to the mechanical law of vehicle driving, makes the subsequently generated shock absorption control strategy more scientific and reasonable, and avoids the problem of poor shock absorption effect caused by improper feature selection.

[0048] The shock absorption strategy generator generates a multi-level coordinated shock absorption control strategy according to the core feature sequence, compared with the fixed single control mode of the traditional shock absorption system, the multi-level coordinated control strategy can formulate differentiated shock absorption schemes for different driving scenes and different passenger physiological states, realize fine control of the body vibration and posture, effectively suppress the body movement that may cause car sickness in different driving scenes, and better adapt to diversified driving needs and passenger states.

[0049] The abnormal feedback regulator identifies sudden disturbance events in the vehicle driving state data or abnormal fluctuation patterns in the passenger physiological index data by analyzing the abnormal disturbance identification set, and triggers dynamic tuning of the shock absorption control strategy, the real-time feedback tuning mechanism can quickly respond to sudden road disturbances or changes in passenger physiological state, timely adjust the shock absorption control strategy, avoid the aggravation of the body vibration or passenger discomfort caused by the lag of the control strategy, and ensure that the system still maintains good shock absorption effect under various sudden conditions, effectively alleviating the car sickness symptoms of passengers.

[0050] The control instruction distributor converts the tuned shock absorption control strategy into a driving signal and issues it to the suspension actuator cluster, ensuring accurate transmission and rapid execution of the control instruction, so that the suspension actuator can respond in time according to the adjusted strategy, quickly adjust the damping characteristics and support force of the suspension, thereby quickly suppressing the body vibration, stabilizing the body posture, reducing the stimulation to the passenger's vestibular system, achieving effective relief of car sickness phenomenon through multi-link cooperation, significantly improving the passenger's ride comfort, and also adapting to the physiological differences of different passengers, meeting the high-quality demand of consumers for vehicle ride experience. BRIEF DESCRIPTION OF DRAWINGS

[0051] Figure 1 The timing diagram of the anti-car sickness intelligent shock absorption control system described in the present application;

[0052] Figure 2 The working principle flowchart of the data separation engine;

[0053] Figure 3 The working principle flowchart of the shock absorption strategy generator;

[0054] Figure 4 Flow chart of the working principle of the system containing the feature tracing and disturbance clustering module. DETAILED DESCRIPTION

[0055] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0056] Please refer to Figure 1 The present application provides an anti-motion sickness intelligent shock absorption control system, which comprises a multi-dimensional sensing array, a data separation engine, a feature screening unit, a shock absorption strategy generator, an abnormal feedback regulator and a control instruction distributor.

[0057] The multi-dimensional sensing array is arranged in the vehicle chassis, seat and passenger wearing equipment, and is used for collecting vehicle driving state data of vehicle longitudinal acceleration, transverse angular velocity and vertical vibration frequency, physiological index data of passenger heart rate, skin electric response and eye tremor frequency, and environmental disturbance data of road slope and air turbulence intensity in real time. All data are time-stamped by a high-precision clock source and are spatially coordinated based on the vehicle coordinate system, to generate a multi-dimensional driving data set with time and space alignment. The data separation engine receives the data set, separates a main feature matrix and an abnormal disturbance identification set through matrix conversion and tensor decomposition processing. The feature screening unit performs correlation analysis based on the main feature matrix and a preset vehicle dynamics parameter library, and screens a core feature sequence. The shock absorption strategy generator maps the core feature sequence to a multi-level coordinated shock absorption control strategy, including main suspension level and auxiliary suspension level control parameters. The abnormal feedback regulator analyzes the abnormal disturbance identification set, identifies a sudden disturbance event or physiological abnormal fluctuation, and triggers dynamic tuning of the shock absorption control strategy. The control instruction distributor converts the tuned strategy into a driving signal and distributes it to the suspension actuator cluster.

[0058] Embodiment 1: Please refer to Figure 2, multidimensional driving data set is input to the matrix conversion module of the data separation engine. The module receives a heterogeneous data stream from the multidimensional sensor array, including a time series signal of vehicle longitudinal acceleration, angular velocity sampling values of lateral roll angle, three-axis frequency spectrum data of vertical vibration, RR interval sequence of occupant heart rate, admittance change curve of skin conductance, angular displacement trajectory of nystagmus, inclination measurement value of road surface slope, and pressure fluctuation reading of air turbulence. The matrix conversion module first implements Z-score standardization processing on all numerical features to eliminate dimensional differences: for each dimensional feature data, the mean and standard deviation within the sliding time window are calculated, and the original data is subtracted from the mean and divided by the standard deviation. For classification features (such as road type encoding), one-hot encoding is used to convert to a binary vector. The standardized data is arranged in chronological order to form a three-dimensional standardized feature tensor, with dimensions of time step, feature channel number, and spatial sampling point number.

[0059] The tensor decomposition module implements rank constraint processing based on Tucker decomposition on the standardized feature tensor, setting the rank constraint value of the core tensor as R, where R is dynamically adjusted according to the correlation analysis of the feature dimensions of historical driving data. The decomposition process first initializes the factor matrix, and iteratively optimizes it by the least squares method to minimize the Frobenius norm error between the original tensor and the reconstructed tensor. After decomposition, three factor matrices and a core tensor are obtained. The steady-state feature basis is generated by the product of the core tensor and the factor matrix, and its data characteristics are represented as low-frequency trend components: the smooth curve of vehicle longitudinal acceleration reflects uniform driving state, the baseline fluctuation of occupant heart rate variability represents resting physiological state, and the road slope in environmental disturbance data presents a slow change trend. The output of this basis is a two-dimensional main feature matrix, with row vectors corresponding to time steps and column vectors containing fused steady-state feature channels.

[0060] The transient disturbance component is obtained by calculating the residual of the original tensor and the reconstructed tensor, and the residual tensor is filtered by threshold filtering to extract significant transient components after removing noise: the millisecond-level impact pulse in the longitudinal acceleration residual corresponds to the wheel rolling over potholes, the high-frequency resonance cluster in the vertical vibration frequency spectrum residual reflects suspension resonance, the sharp peak in the heart rate RR interval residual marks the sudden sympathetic nervous activation, and the abnormal angular displacement in the eye movement trajectory residual represents the vestibular system disorder. These components are encoded as an abnormal disturbance identifier set, each identifier containing a four-tuple structure: disturbance type encoding, timestamp start position, spatial coordinate index, and intensity amplitude. The main feature matrix is transmitted to the feature screening unit, and the abnormal disturbance identifier set is transmitted to the abnormal feedback regulator.

[0061] In the matrix conversion stage, the timestamp alignment adopts the GPS synchronous clock signal, and the spatial coordinate synchronization establishes a right-hand coordinate system with the vehicle center of mass as the origin. In the standardization process, the sliding time window length is set to 200 milliseconds, and the window step length is 50 milliseconds, ensuring the balance between real-time and continuity. The one-hot encoded road surface type includes six predefined categories such as asphalt, gravel, ice, and snow, which are identified by the fusion of vehicle-mounted camera and radar data.

[0062] The rank constraint value R of tensor decomposition is determined through offline training: collect historical data under typical working conditions (urban roads, highways, and mountain bends), calculate the mutual information matrix between each feature channel, and take the rank corresponding to the sum of 95% energy of the singular value of the mutual information matrix as the reference value of R. In online operation, R value is adaptively expanded according to the feature dimension number of the current driving environment, and the expansion coefficient is adjusted by the visibility and humidity parameters in the environmental disturbance data. The factor matrix initialization adopts the random SVD method to accelerate convergence, and the iteration termination condition is set to less than 0.1% of the reconstruction error change rate or reaching the maximum iteration number of 50 times.

[0063] The residual filter adopts a double-threshold mechanism: first, calculate the root mean square value of each channel of the residual tensor, and set the residual below 30% of the root mean square value to zero; second, implement morphological filtering on the remaining residual to eliminate isolated pulses with a duration of less than 5 milliseconds. The intensity amplitude calculation of abnormal disturbance identification uses the normalized energy integration method: within the disturbance duration, the residual signal is squared and integrated, and then divided by the time window length. The identification of sudden road impact patterns is additionally indexed with spatial coordinates, and the impact source suspension partition is determined by the wheel position sensor.

[0064] Smoothness constraint is introduced during the reconstruction of steady-state feature basis: second-order difference regularization is applied to the time dimension factor matrix of the core tensor to suppress high-frequency jitter in the basis. Channel fusion is performed before the output of the principal feature matrix, which reduces the vehicle motion-related longitudinal / lateral / vertical features to three channels through principal component analysis, and the physiological features remain in independent channels. The abnormal disturbance identification set uses incremental compression storage, and continuous disturbance events at the same spatial position are merged into a single entry, and the event duration and maximum amplitude are recorded.

[0065] This implementation realizes the physical separation of steady-state control features and transient disturbance features under the premise of maintaining the spatio-temporal correlation of driving data through the mathematical properties of tensor decomposition. The construction of standardized feature tensors ensures the comparability of heterogeneous data in a unified mathematical space, the rank-constrained decomposition extracts the essential structure of the data through low-rank assumption, and the residual analysis focuses on the nonlinear response components of the system. The principal feature matrix provides stable input for shock absorption control, and the abnormal disturbance identification set establishes a precise triggering mechanism for dynamic tuning.

[0066] Example 2: refer to Figure 3The main feature matrix is input to the correlation graph construction module of the feature screening unit. This module extracts five types of feature vectors from the matrix: vehicle longitudinal acceleration features are obtained by low-pass filtering the time-domain signals collected by the chassis accelerometer, lateral roll angle features are calculated by integrating the roll angular velocity measured by the gyroscope, vertical vibration frequency spectrum features are extracted by fast Fourier transform of the three-axis data collected by the seat acceleration sensor, passenger heart rate variability features are calculated by the standard deviation of the RR interval sequence recorded by the electrocardiogram monitoring device, and eye movement trajectory features are obtained by converting the pupil position coordinates collected by the infrared eye tracker to angular velocity curves. After aligning these feature vectors by timestamp, they are input into a graph convolution network to construct a multi-modal feature correlation graph. Each node in the graph corresponds to a feature vector, and the node attributes include feature type and time window statistics; the edge weights between nodes are determined by calculating the Pearson correlation coefficient of the feature pairs in the sliding time window, with a window length of 300 milliseconds and a step size of 50 milliseconds. The longitudinal acceleration and lateral roll angle form a strong negative correlation edge (weight -0.82), the vertical vibration frequency spectrum and the heart rate variability establish a positive correlation edge (weight 0.76), and the eye movement trajectory and the lateral roll angle present a nonlinear correlation edge (weight 0.68).

[0067] The entropy calculation module performs dynamic correlation entropy analysis based on the multi-modal feature correlation graph. For each feature node, the information entropy in the last 20 time windows is calculated: first, the numerical distribution histogram of the feature in the time window is calculated, and the Shannon entropy value is calculated based on the probability distribution. At the same time, the joint conditional entropy of the feature relative to other features is calculated: the top three features connected by the edge weight are selected as the conditional variables, and the conditional probability distribution is calculated by kernel density estimation. The dynamic correlation entropy is finally represented by the weighted sum of the information entropy and the conditional entropy, and the weight coefficient is determined according to the feature sensitivity table in the vehicle dynamics parameter library. In the vehicle emergency braking condition, the information entropy of the longitudinal acceleration feature increases to 2.8 bits, the joint conditional entropy of the feature relative to the lateral roll angle reaches 1.5 bits, and the comprehensive value of the dynamic correlation entropy is 4.3 bits; the dynamic correlation entropy of the vertical vibration frequency spectrum maintains at 0.9 bits during uniform cruising.

[0068] The sequence generation module receives the dynamic correlation entropy sequence of all features, sorts them in descending order, and processes them according to the pre-set screening rules. The feature dimension table stored in the vehicle dynamics parameter library specifies that the top 3 features are selected under urban road conditions and the top 5 features are selected under highway conditions. The screening process introduces a redundancy suppression mechanism: if the entropy difference between adjacent features is less than 0.3 bits, they are combined into a feature group. The final core feature sequence uses a three-dimensional data structure: timestamp, feature type code, and feature quantization value. For example, when driving on a bumpy road, the sequence contains three features: vertical vibration frequency spectrum (amplitude 0.8g), heart rate variability (standard deviation 35ms), and eye movement trajectory (angular velocity 42deg / s).

[0069] The core feature sequence inputs a strategy mapping module of the shock absorption strategy generator, which retrieves a shock absorption benchmark curve library from a vehicle dynamics parameter library, the library containing control curves of 120 standard working conditions, each curve being defined by a suspension displacement-damping force relationship matrix. The matching process uses a dynamic time warping algorithm: the core feature sequence is nonlinearly aligned with the feature templates corresponding to the benchmark curves, and the minimum path cumulative distance is calculated. When the vehicle passes through an S-shaped curve, the lateral roll angle (peak 15 deg) in the core feature sequence and the "high-speed cornering" template (feature template ID #07) in the benchmark curve library achieve a 95% matching degree, mapping out a basic shock absorption control curve, which contains camber compensation parameters (front wheel -1.2 deg, rear wheel +0.8 deg) and roll stiffness parameters (increased by 25%).

[0070] The multi-level optimization module receives the basic shock absorption control curve and performs hierarchical decomposition according to the vehicle suspension topology structure. The main suspension level control strategy is independently designed for the four wheels: the stiffness parameters in the basic curve are input into the LQR optimizer, combined with the sprung mass, unsprung mass, and suspension geometry parameters, to solve the optimal damping force sequence. For example, the front left wheel suspension generates a damping force curve in the braking condition: the initial segment is 500N (compression stroke), the peak segment is 1200N (rebound stroke), and the decay segment is 800N (steady state stroke). The auxiliary suspension level control strategy is designed for the active stabilizer bar: the compensation parameters in the basic curve are input into the PID optimizer, combined with the vehicle body attitude sensor data, to generate the front stabilizer bar torque (0-120Nm linearly increasing) and the rear stabilizer bar actuating force (50-200N step change). The final multi-level cooperative shock absorption control strategy adopts a hierarchical instruction structure: the top layer is the main suspension level damping force matrix, and the bottom layer is the auxiliary suspension level actuating force vector, both of which are synchronously coupled through time stamp.

[0071] In the feature correlation graph updating mechanism, the edge weight is recalculated every 100 milliseconds, and the node attribute is refreshed every 200 milliseconds. The dynamic correlation entropy calculation uses sliding window recursive update to avoid repeated operations. When generating the core feature sequence, the selected features are normalized: each feature value is mapped to the [0, 1] interval, and the normalized benchmark value comes from the feature extreme value table in the vehicle dynamics parameter library. The strategy mapping module sets a matching confidence threshold (80%), and when the threshold is lower, it starts a multi-template weighted fusion mechanism, for example, fusing the "high-speed cornering" template (weight 0.6) with the "lane changing and overtaking" template (weight 0.4) to generate a new curve. The state weight matrix in the LQR optimization is dynamically adjusted according to the driving mode: the roll angular velocity weight is increased by 3 times in the sport mode, and the vertical acceleration weight is reduced by 50% in the comfort mode.

[0072] The embodiment reveals the coupling relationship between vehicle motion and physiological response through multi-modal feature correlation graph, quantifies the contribution of feature to system uncertainty with dynamic correlation entropy, and focuses on the most critical control variables through the selection of core feature sequence. Strategy mapping converts complex feature combination into physically executable control curve, and hierarchical optimization decomposition makes the control strategy adapt to the dynamic characteristics of different suspension actuators. The LQR control of the main suspension level ensures the global optimality of the system, and the PID optimization of the auxiliary suspension level realizes local rapid response, which cooperates to improve the accuracy and robustness of the damping control.

[0073] Example 3: refer to Figure 4 The disturbance pattern recognition module of the abnormal feedback regulator receives the transient disturbance component from the data separation engine. The component contains a residual signal in the form of a time series, with a sampling frequency of 1 kHz and a duration that varies dynamically according to the disturbance event. The module first analyzes the input signal in time and frequency domain: short-time Fourier transform is used to analyze the frequency domain characteristics, Hanning window is selected as the window function, window length is 256 points, overlap rate is 50%, and frequency spectrum is generated; at the same time, continuous wavelet transform is carried out, Morlet wavelet is selected as the mother wavelet, the scale parameter a is in the range of 1-100, and the time-frequency distribution matrix is generated. The identification of sudden road impact pattern is based on the sudden increase of high-frequency energy: in the frequency band of 800Hz-1kHz, if the energy amplitude increases more than the threshold value within 5 milliseconds (background noise standard deviation), and the wavelet coefficient presents local maximum in time domain, it is marked as impact event. The detection of continuous resonance mode focuses on the sustained energy peak value in a specific frequency band: in the frequency band of 100-200Hz, if the energy is sustained more than 2.5 times the baseline level and the duration is greater than 50 milliseconds, and the wavelet coefficient presents periodic fluctuation, it is determined as resonance event. The identification of occupant physiological instability mode is analyzed by heart rate variability coefficient: the heart rate signal residual is extracted from the transient component, the coefficient of variation is calculated, if the coefficient of variation of continuous 3 sampling periods exceeds the threshold value 0.35, and is accompanied by high-frequency jitter of eye movement signal, it is classified as physiological instability event.

[0074] All identified events are encoded into structured data format, including event type, start timestamp, duration, peak intensity, main frequency band and spatial location information. The tuning parameter generation module retrieves the preset tuning rule library according to the identification results. The rule library is stored in a hierarchical structure: the first layer is the disturbance type index, the second layer is the vehicle working condition parameter (vehicle speed, load, road type), and the third layer is the tuning parameter mapping table. For sudden road impact mode, query suspension stiffness correction coefficient table to obtain correction function based on impact intensity and duration:

[0075]

[0076] wherein:​​​ is the adjusted suspension stiffness coefficient, is the reference stiffness value, is the impact peak intensity (unit: g), is the impact duration (unit: ms), is the intensity sensitivity factor (value range 0.1-0.3), is the duration weight coefficient (value range 0.05-0.2). For continuous resonance mode, query the damping compensation gradient table to obtain the compensation amount based on the resonance frequency and energy level:

[0077] ,

[0078] where: is the damping compensation amount (unit: Ns / m), is the reference damping value, is the resonance energy amplitude, is the reference energy value, is the resonance frequency (unit: Hz), is the system natural frequency, and is the frequency-dependent adjustment coefficient. For the occupant physiological instability mode, a parameter attenuation strategy is adopted: query the physiological instability rule table to obtain the control parameter attenuation curve, and reduce the suspension actuation intensity according to the exponential law.

[0079] The strategy recalculation module receives the tuning parameters and performs real-time strategy update. This module maintains the currently effective multi-level cooperative shock absorption control strategy, and its data structure includes the main suspension level parameter matrix and the auxiliary suspension level parameter vector. For the stiffness correction coefficient , the stiffness parameters in the main suspension level control strategy are updated by linear interpolation method: read the stiffness value in the current strategy, calculate the new value , where is the interpolation rate factor (default value 0.8). For the damping compensation amount , the damping parameters are updated by gradient descent method: set the learning rate η = 0.05, and iterate 3 times to obtain the optimized damping value. For the attenuation strategy triggered by the physiological instability mode, a first-order inertial decay is implemented with a time constant τ = 200 ms: where is the actuation intensity at time t, is the initial value. All parameter updates are completed within strict time constraints, and the end-to-end delay from disturbance identification to strategy recalculation does not exceed 20 milliseconds.

[0080] The signal conversion module of the control instruction distributor converts the recalculated strategy into driving signals. This module uses pulse width modulation technology to discretize continuous control parameters into square wave signals with adjustable duty cycles. The conversion process is based on a pre-calibrated mapping relationship: suspension stiffness parameter K is mapped to pulse width where and are the minimum and maximum pulse widths (typical values 1 ms and 5 ms), and are the stiffness parameter range. The damping parameter C is mapped to pulse frequency where is the reference frequency (100 Hz) and κ is the scaling factor (0.8 Hz·m / Ns). The partitioned distribution module distributes signals according to the position encoding of the suspension actuators: the main suspension controller (located near the wheels) receives pulse width modulation signals, transmitted through the CAN bus, with data frames containing controller ID, pulse width value, and effective timestamp; the auxiliary suspension controller (located in the vehicle body stability system) receives pulse frequency modulation signals, transmitted through the FlexRay bus, with data frames containing frequency value, phase offset, and duration. Each controller address strictly corresponds to a physical location, with the front left wheel controller ID = 0x10, the front right wheel ID = 0x11, the rear left wheel ID = 0x12, the rear right wheel ID = 0x13, and the active stabilizer bar controller ID = 0x20.

[0081] By accurately capturing the essential characteristics of disturbance events through time-frequency analysis, the tuning rule library provides parameter mapping based on physical models, and strategy recalculation ensures the real-time adaptability of the control system. The signal conversion and partitioned distribution mechanism converts digital control strategies into precise actions of physical actuators, forming a closed-loop response from disturbance perception to execution control. The entire process ensures real-time performance while maintaining the stability and reliability of the control system.

[0082] Embodiment 4: The cloud-side co-processor receives cross-vehicle coordination identifiers in the transient disturbance component through the on-board V2X communication module. These identifiers contain the vehicle's latitude and longitude coordinates (WGS84 format), current velocity vector, acceleration amplitude, and disturbance type code. When the vehicle is traveling at 85 km / h on a highway, a continuous resonance pattern (code RES_102) is detected, and a coordination request frame is immediately generated, containing the vehicle's location (longitude 116.46°, latitude 39.92°), velocity vector (eastward 23.6 m / s, northward 0.3 m / s), and resonance frequency 123 Hz. The request is broadcast to the cloud server through 5G-V2X, and the server retrieves neighboring vehicles within a radius of 500 meters, filtering out three vehicles (license plates Beijing A12345, Beijing B67890, and Beijing C24680) that have shock absorption coordination capabilities. The cloud extracts the latest 5-minute main feature matrix compressed version of these vehicles from the historical database, including 12 dimensions such as longitudinal acceleration smoothing value, vertical vibration fundamental frequency amplitude, suspension actuation state flag, etc. The data is transmitted to the requesting vehicle in binary stream format, with a transmission delay controlled within 35 milliseconds.

[0083] After the coupling strategy optimizer receives the shock absorption feature matrix of the neighboring vehicles, it first performs time alignment and coordinate conversion. The data of each vehicle is uniformly converted to a coordinate system with the vehicle's center of mass as the origin, and the time axis is synchronized with the GPS clock. Then, the coupling constraint term is constructed: the phase difference between the vehicle and each neighboring vehicle in the vertical vibration phase is calculated to generate a phase synchronization constraint boundary; the vibration energy transfer relationship is analyzed to establish an amplitude coordination constraint range. The optimization goal is set to minimize the vibration transmission difference between vehicles, and a distributed optimization algorithm is used for iterative calculation. For the phase synchronization constraint, the vertical vibration phase difference between the vehicle and the neighboring vehicles is required to be within ±15 degrees; for the amplitude coordination constraint, the vehicle's vibration amplitude is required to be within ±20% of the average amplitude of the neighboring vehicles. After 3 rounds of iterative calculation, the final coupling constraint term set is generated, including phase adjustment, amplitude scaling factor, and time sequence offset parameters.

[0084] The feature tracing module continuously monitors the amplitude changes of the transient disturbance component, and when it detects that the vertical vibration residual amplitude exceeds the threshold value of 0.4g for 5 consecutive sampling periods, it triggers the historical data backtracking mechanism. The system automatically loads the multi-dimensional driving data set within the past 30 seconds, extracts the feature vector related to the current disturbance: including road roughness index, suspension hydraulic pressure value, vehicle body vertical acceleration spectral features, passenger heart rate coefficient of variation, etc. 18 dimensions. The backtracking time window is dynamically adjusted according to the disturbance duration, and the longest can trace back to the historical data 120 seconds ago.

[0085] The disturbance clustering module receives the correlation feature vectors obtained by backtracking, and first adds a vehicle working condition label to each data point. The label contains three types of information: driving mode, road type, and load state. An improved K-means clustering algorithm is used for processing: the data points with significant working condition differences are preferentially selected as the initial cluster centers; the weighted Euclidean distance is used for distance measurement, and higher weight is given to vehicle motion-related features; the number of clusters is dynamically adjusted during the iteration process, and a maximum of 5 clusters is allowed. The clustering results generate two types of device-level disturbance clusters and environment-level disturbance clusters. The device-level disturbance cluster mainly contains suspension system wear features and sensor drift features; the environment-level disturbance cluster contains road roughness features, climate disturbance features, and traffic flow features.

[0086] Table 1: Feature distribution table of disturbance clustering results

[0087]

[0088] Referring to Table 1, the device-level disturbance cluster is output to the suspension health diagnosis unit, which uses a deep learning-based predictive maintenance algorithm to analyze the performance degradation trend of the suspension system and generate a maintenance recommendation report. The environment-level disturbance cluster is output to the environment adaptation learning unit, which uses an incremental learning mechanism to continuously update the road feature library and climate disturbance model, and optimizes the environmental adaptability of the shock absorption control system.

[0089] During the entire implementation process, end-to-end encryption mechanism is used for data transmission, and the cloud collaborative processor uses the national SM4 algorithm to encrypt the interaction data. After the coupling constraint term is generated, it needs to pass through the runtime verification of the vehicle controller to check whether the constraint boundary is within the physical limit range of the vehicle, so as to avoid generating unfeasible control instructions. The history data backtracking uses a circular buffer storage mechanism, with the latest data covering the oldest data, to ensure storage efficiency. The clustering algorithm re-trains the cluster center every 24 hours to adapt to the slow changes of the vehicle system. The analysis results of the device-level disturbance cluster generate a health status report every week, and the learning model of the environment-level disturbance cluster is updated comprehensively once a month. Through inter-vehicle collaborative perception, the cognitive range of the system to environmental disturbances is expanded, and the coupling optimization improves the shock absorption effect when multiple vehicles are driving cooperatively. The historical data backtracking mechanism provides context information for disturbance analysis, and the intelligent clustering accurately distinguishes between internal device problems and external environmental factors.

[0090] Example 5: Edge computing partition adopts heterogeneous computing architecture to realize real-time processing area and non-volatile storage area with physical isolation. The real-time processing area is built based on programmable logic devices, with dual-core processing units and dedicated hardware accelerators. The operation of the data separation engine is realized through a pipeline mechanism: the input multi-dimensional driving data set enters the first level cache, the matrix conversion module calls the standardization processing coprocessor, the acceleration data is subjected to sliding window normalization, and the physiological data is subjected to baseline drift correction, and the processed data is written into the second level cache. The tensor decomposition module activates the tensor operation accelerator, loads the preset rank constraint parameters, and completes the feature basis extraction through the parallel matrix multiplication unit. The feature screening unit is deployed in the second core, and the graph construction module calls the graph computing engine to generate a multi-modal feature correlation graph in the dedicated memory area; the entropy calculation module uses the hardware entropy calculation unit to complete the real-time update of the dynamic correlation entropy. The data processing delay of the entire real-time processing area is controlled within 8 milliseconds, and the clock frequency is stabilized at 450 MHz.

[0091] The non-volatile storage area adopts a flash memory array with power failure protection function, which is divided into parameter library area and rule library area. The parameter library area stores vehicle dynamics parameter data set, including structured record of suspension geometry parameters: front suspension kingpin caster angle range, rear suspension trailing arm installation angle, stabilizer bar torsional stiffness coefficient, etc.; the mass distribution matrix is dynamically stored according to vehicle configuration, including three-dimensional moment of inertia of sprung mass and wheel center coordinates of unsprung mass; the shock response curve library stores force-displacement characteristic curves under different working conditions, each curve containing 256 sampling points. The rule library area stores the tuning rule mapping table, which adopts a hierarchical index structure: the first index is classified by disturbance type, the second index is divided by vehicle speed, and the third index stores specific tuning parameters. All data is written with ECC check, read with parity check, and data update uses write-ahead logging mechanism to prevent power failure damage.

[0092] The secure isolation repeater establishes a bidirectional encrypted channel between the edge computing partition and the cloud end processor. During channel initialization, bidirectional authentication is performed: the edge end generates an authentication request based on the vehicle VIN code, and the cloud end returns a digital certificate; both parties exchange session keys through the SM2 algorithm. Data transmission adopts a hybrid encryption mechanism: control instructions are encrypted in groups using the SM4 algorithm with a key length of 256 bits; data packets are attached with a message authentication code based on the SM3 algorithm. The encrypted channel maintains a heartbeat detection mechanism, exchanging timestamp verification packets every 200 milliseconds, and automatically disconnecting the connection if the timeout is 300 milliseconds.

[0093] The coupling constraint items issued by the cloud are verified at runtime, and the verification process includes three levels: format verification checks whether the data packet structure conforms to the predefined protocol specification; signature verification uses the preset public key of the cloud to verify the digital signature; and physical constraint verification detects whether the parameter value is within the vehicle safety range. For the phase adjustment amount, it is verified whether it is within the range of ± 30 degrees; for the amplitude scaling factor, it is checked whether it is within the interval of 0.5 to 2.0; the timing offset parameter needs to meet the requirement that the delay difference between the front and rear wheels is less than 50 milliseconds. When the verification fails, three levels of responses are triggered: a slight abnormality is logged and a default value is adopted; a moderate abnormality sends a retransmission request to the cloud; and a serious abnormality starts a safety mode and isolates the cloud control.

[0094] The fault recovery mechanism of the real-time processing area adopts a dual backup design. When the main processing unit is abnormal, the backup unit takes over within 5 milliseconds and loads the latest valid state snapshot from the non-volatile storage area. The snapshot is updated every 50 milliseconds and includes the intermediate tensor results of the data separation engine and the latest associated entropy values of the feature screening unit. The wear leveling algorithm of the non-volatile storage area dynamically adjusts the storage location, and the flash memory block is automatically marked as read-only when the number of erase times exceeds 100,000. The safety isolation relay maintains a session state machine, records the encryption parameters of the last 10 communications, and supports fast session recovery after connection interruption.

[0095] This embodiment ensures the real-time performance of critical control processes through hardware-level isolation, and ensures the persistent availability of the parameter library through non-volatile storage design. The encryption channel and the runtime verification mechanism build a defense-in-depth system, supporting cloud collaboration while maintaining the integrity and security of the local control system. The fault recovery strategy minimizes the risk of system interruption, forming a highly reliable vehicle-mounted edge computing environment.

[0096] It should be noted that, in the present document, relational terms such as first and second and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0097] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A car sickness prevention intelligent shock absorption control system, characterized in that, The method comprises the following steps: A multi-dimensional sensing array is used to collect vehicle driving state data, passenger physiological index data and environmental disturbance data in real time, and the vehicle driving state data, passenger physiological index data and environmental disturbance data are time-stamped and synchronized with spatial coordinates to generate a multi-dimensional driving data set with time and space alignment; A data separation engine is used to perform feature decoupling reconstruction processing on the multi-dimensional driving data set with time and space alignment, and separate a main feature matrix for shock absorption control and an abnormal disturbance identification set for feedback tuning; A feature screening unit is used to perform multi-dimensional feature correlation analysis based on the main feature matrix and a pre-set vehicle dynamics parameter library to screen a core feature sequence for shock absorption decision-making; A shock absorption strategy generator is used to generate a multi-level coordinated shock absorption control strategy according to the core feature sequence; An abnormal feedback regulator is used to analyze the abnormal disturbance identification set, identify sudden disturbance events in the vehicle driving state data or abnormal fluctuation patterns in the passenger physiological index data, and trigger dynamic tuning of the shock absorption control strategy; A control instruction distributor is used to convert the tuned shock absorption control strategy into a driving signal and distribute it to a suspension actuator cluster; The data separation engine comprises: A matrix conversion module is used to convert the multi-dimensional driving data set with time and space alignment into a standardized feature tensor; A tensor decomposition module is used to perform rank-constrained decomposition on the standardized feature tensor to obtain a steady-state feature basis and a transient disturbance component; The steady-state feature basis constitutes the main feature matrix, and the transient disturbance component constitutes the abnormal disturbance identification set; The feature screening unit comprises: An association graph construction module is used to extract vehicle longitudinal acceleration features, lateral roll angle features, vertical vibration frequency spectrum features, passenger heart rate variability features and eye movement trajectory features from the main feature matrix to construct a multi-modal feature correlation graph; An entropy calculation module is used to calculate the dynamic correlation entropy of each feature based on the multi-modal feature correlation graph; A sequence generation module is used to arrange and screen the top K features in descending order of the dynamic correlation entropy to generate the core feature sequence; The shock absorption strategy generator comprises: A strategy mapping module is used to match the core feature sequence with a pre-set vehicle dynamics parameter library to map out a basic shock absorption control curve; A multi-level optimization module is used to perform hierarchical optimization decomposition on the basic shock absorption control curve according to the vehicle suspension topology to generate the multi-level coordinated shock absorption control strategy including a main suspension level control strategy and an auxiliary suspension level control strategy; The abnormal feedback regulator comprises: A disturbance pattern recognition module is used to perform time-frequency feature analysis on the transient disturbance component to identify sudden road impact patterns, continuous resonance patterns or passenger physiological instability patterns; A tuning parameter generation module is used to retrieve a pre-set tuning rule library according to the identified disturbance pattern to generate a suspension stiffness correction coefficient or a damping compensation gradient; A strategy recalculation module is used to inject the suspension stiffness correction coefficient or the damping compensation gradient into the multi-level coordinated shock absorption control strategy for real-time recalculation; The control instruction distributor comprises: The signal conversion module is configured to convert the recalculated multi-level cooperative shock absorption control strategy into a pulse width modulation signal. The partition delivery module is configured to distribute the pulse width modulation signal to a corresponding main suspension controller or auxiliary suspension controller according to position coding of the suspension actuator.

2. The anti-motion sickness intelligent damping control system of claim 1, wherein, The system further comprises: The cloud-side cooperative processor is configured to receive cross-vehicle cooperative identification in the transient disturbance component, and dispatch a shock absorption feature matrix of a neighboring vehicle from the cloud side. The coupling strategy optimizer is configured to generate a coupling constraint term based on the shock absorption feature matrix of the neighboring vehicle, and implement cooperative optimization on the multi-level cooperative shock absorption control strategy.

3. The anti-motion sickness intelligent damping control system of claim 2, wherein, The system further comprises: The feature tracing module is configured to trace an associated feature vector in a historical driving data set when the transient disturbance component exceeds a preset threshold. The disturbance clustering module is configured to cluster and group the associated feature vector according to a vehicle working condition label, and generate a device-level disturbance cluster and an environment-level disturbance cluster. The device-level disturbance cluster is input to a suspension health diagnosis unit, and the environment-level disturbance cluster is input to an environment adaptation learning unit.

4. The anti-motion sickness intelligent damping control system of claim 1, wherein, The system further comprises: The edge computing partition includes a real-time processing area and a non-volatile storage area which are isolated from each other. The real-time processing area runs the data separation engine and the feature screening unit, and the non-volatile storage area solidifies the preset vehicle dynamics parameter library and the tuning rule library.

5. The anti-motion sickness intelligent damping control system of claim 4, wherein, The system further comprises: The secure isolation repeater is configured to establish an encrypted channel between the edge computing partition and the cloud-side cooperative processor, and implement runtime verification on the delivered coupling constraint term.

Images