A motion sickness comfort real-time prediction and adaptive adjustment system and method

Abstract

This invention relates to the field of motion sickness comfort technology, specifically to a real-time prediction and adaptive adjustment system and method for motion sickness comfort. The method includes: loading static features of occupants and calculating an initial dynamic baseline threshold based on these features; collecting vehicle motion data, occupant physiological behavior data, and cabin environment data; performing feature extraction and vector construction to form vehicle motion feature vectors, occupant real-time physiological behavior feature vectors, and occupant static feature vectors; dynamically weighting the feature vectors based on the occupant static features; and outputting a real-time motion sickness risk index based on the weighted feature vectors using a risk prediction model; comparing the risk index with the dynamic baseline threshold to determine the risk level and the type of risk trigger; and generating an adjustment instruction package from a preset multimodal adjustment strategy matrix based on the motion sickness risk level and the type of risk trigger. This solution can more accurately predict occupant motion sickness risk and provide targeted, personalized intervention measures.

Description

Technical Field

[0001] This invention relates to the field of motion sickness comfort technology, specifically to a real-time prediction and adaptive adjustment system and method for motion sickness comfort. Background Technology

[0002] With the rapid development of the intelligent electric vehicle industry, consumers' demands for driving comfort have shifted from simply improving ride quality to ensuring a personalized and intelligent all-scenario comfort experience. Electric vehicles, with their technological advantages such as fast torque response and flexible energy recovery mechanisms, demonstrate significant competitiveness in power performance. However, due to their dynamic characteristics, such as excessively large jerk rate of acceleration and dramatic pitch angular velocity, the probability of motion sickness is significantly higher than in traditional gasoline vehicles, becoming one of the core bottlenecks restricting the improvement of the driving experience of electric vehicles.

[0003] A series of studies have been conducted on the evaluation and optimization of motion sickness comfort. Regarding assessment and threshold calibration, existing technologies can set personalized motion sickness warning thresholds based on the static physiological characteristics of occupants, achieving certain results in offline calibration during the development phase. In terms of risk prediction, schemes based on vehicle motion parameters to construct motion sickness evaluation models have emerged, enabling a certain degree of quantitative assessment of motion sickness risk. Meanwhile, the rapid maturation of smart cockpit technology has enabled subsystems such as seat posture adjustment, zoned air conditioning control, ambient lighting, fragrance release, and entertainment displays to possess more refined and faster controllability, providing a hardware foundation for alleviating motion sickness symptoms through multi-sensory channel coordinated adjustment.

[0004] However, analyzing the development trajectory of the above technologies reveals that the transition from offline optimization during the development phase to real-time proactive protection during driving still faces technical bottlenecks. The root cause lies in the contradiction between the complex and dynamic nature of motion sickness mechanisms and the static and discrete nature of existing technical architectures.

[0005] Specifically, motion sickness is not solely determined by external vehicle movement, but rather by the continuous coupling and cumulative effect of external motion stimuli and the occupant's internal physiological and psychological state over time. The sensitivity threshold to motion stimuli for the same occupant shifts in real time depending on their fatigue level, attention level, emotional state, and even the sequence of road conditions they experience. Existing technologies, based on personalized thresholds set according to the occupant's static characteristics, essentially provide a "snapshot" calibration of motion sickness susceptibility at a specific moment or in a specific scenario. Their underlying logic assumes that motion sickness susceptibility remains relatively constant after calibration, which fundamentally conflicts with the strong time-cumulative effect and immediate state dependence of motion sickness. The direct consequence is that the system cannot promptly identify the escalating risk trend when the occupant's actual susceptibility increases, nor can it avoid unnecessary false alarms when the occupant is in good condition. The accuracy and timeliness of warnings are structurally constrained.

[0006] Furthermore, current technology only provides visual alerts via the dashboard or screen when predicting excessive risk. This passive information delivery is significantly less effective for passengers who are resting or focused on in-vehicle screens, and the alerts themselves do not offer any substantial symptom relief. The few attempts at proactive intervention typically limit themselves to switching the vehicle to a "comfort mode" to globally restrict power response and reduce motion stimulation, failing to precisely adjust for the different sensitivity channels of various occupants. Summary of the Invention

[0007] The purpose of this invention is to propose a real-time prediction and adaptive adjustment system and method for motion sickness comfort, which can more accurately predict the risk of motion sickness in passengers and provide targeted personalized intervention measures.

[0008] To achieve the above objectives, in a first aspect, the present invention proposes a method for real-time prediction and adaptive adjustment of motion sickness comfort, comprising: Load the static characteristics of the occupants, and calculate the initial dynamic baseline threshold based on the static characteristics; Collect vehicle motion data, occupant physiological behavior data, and cabin environment data; extract features and construct vectors to form vehicle motion feature vectors, occupant real-time physiological behavior feature vectors, and occupant static feature vectors. Based on the static characteristics of the occupants, the feature vector is dynamically weighted, and the risk prediction model outputs a real-time motion sickness risk index based on the weighted feature vector. The risk index is compared with a dynamic baseline threshold to determine the risk level and the type of risk trigger. Based on the motion sickness risk level and risk trigger type, an adjustment instruction package is generated from the preset multimodal adjustment strategy matrix.

[0009] Beneficial effects of the basic solution: This solution constructs an initial dynamic baseline threshold by combining occupant static characteristics, integrates multi-dimensional data on vehicle motion, real-time occupant physiological behavior, and cabin environment to build a feature vector, and completes dynamic weight allocation based on occupant static characteristics, taking into account individual differences in physical condition and tolerance among different occupants, avoiding prediction bias caused by using a uniform threshold. Simultaneously, the weighted feature vector is input into the risk prediction model to output a real-time motion sickness risk index, replacing the traditional static threshold and single vehicle parameter judgment method, capturing the real-time motion sickness state of occupants, and realizing the shift from generalized risk prediction to individualized real-time perception, improving the real-time performance and accuracy of motion sickness risk prediction.

[0010] This solution compares the real-time motion sickness risk index with a dynamic baseline threshold, classifying risk levels and identifying risk trigger types through interpretable features. This overcomes the limitations of existing technologies that can only determine the level of risk but cannot pinpoint the trigger. It can distinguish between different sources of motion sickness, such as vehicle motion stimuli, occupant physiological states, and cabin environmental disturbances, providing a clear basis for subsequent adjustment strategies. Based on the motion sickness risk level and trigger type, it generates a customized adjustment instruction package from a pre-set multimodal adjustment strategy matrix. This package can match corresponding multi-dimensional collaborative adjustment methods for different triggers, replacing traditional single-intervention methods and achieving personalized intervention.

[0011] The solution is based on the static characteristics of occupants and uses real-time dynamic data as its core. It dynamically adjusts weights and baseline thresholds and combines them with real-time risk indices to dynamically match adjustment strategies. It can adapt to changes in motion sickness status of different occupants, different driving conditions, and different cabin environments. It enables the adjustment strategy to dynamically and adaptively adjust according to the real-time status of occupants and external conditions, ensuring the long-term stability and universality of motion sickness comfort adjustment. This transforms motion sickness prevention and control from static passive early warning to dynamic proactive protection.

[0012] As a feasible preferred approach, static characteristics of the occupants are loaded, and an initial dynamic baseline threshold is calculated based on these static characteristics, specifically including the following: Methods for verifying passenger identity include biometric identification or device binding; The static characteristics also include age, gender, BMI, and personal adjustment preferences; The formula for calculating the initial dynamic baseline threshold is:

[0013] in, The threshold value is the general threshold, and α is the sensitivity amplification factor. S MSSQ Assess the MSSQ score of the occupants. The average MSSQ score of the population. This is the highest score for MSSQ.

[0014] As a feasible preferred embodiment, the vehicle motion data includes at least one of triaxial acceleration, jerk, yaw rate and pitch rate; the occupant physiological behavior data includes at least one of blink frequency, pupil diameter change, heart rate, heart rate variability, respiratory rate and seated center of gravity shift; the cabin environment data includes at least one of temperature and CO2 concentration.

[0015] As a feasible and preferred option, dynamic weight allocation includes weight adjustment for highly susceptible groups and adjustment for age decay factors; After weight allocation, the weighted eigenvector synthesis at time... Input feature vector Represented as:

[0016] in, For vehicle motion characteristics, These are real-time physiological and behavioral characteristics. S For the static characteristics of the occupants, Weights for vehicle motion sickness features. Adjusted by age decay factor, The weighting coefficients represent the weights of real-time physiological and behavioral characteristics. Adjusted from highly susceptible groups, The fixed embedding weights are for static features.

[0017] As a feasible preferred embodiment, the dual-layer LSTM fusion timing processing includes: The first LSTM layer captures short-term dependencies:

[0018] in This is the first hidden state, used to extract the coupling features between instantaneous motion and physiological response; The second-layer LSTM combines an attention mechanism to capture long-term dependence and highlight the impact of key dizziness-inducing moments; Calculate attention score:

[0019] in, It is the second layer of LSTM in the 1st... The hidden state at all times These are trainable parameters; Normalized weights:

[0020] in, For retrospective time windows; Context vector generation:

[0021] context vector Input to a fully connected layer, output time step through the Sigmoid activation function. Real-time motion sickness risk index :

[0022] in, For the Sigmoid function, and These are the output layer weights and biases.

[0023] As a feasible preferred embodiment, the method also includes executing the adjustment command package to adjust at least one of the smart seat, the cabin environment, or the infotainment display; The risk levels include low risk, medium risk, and high risk; When the risk is low, a level one adjustment strategy is adopted to adjust the cabin environment and screen display. When the risk is medium risk or the first-level adjustment is ineffective, the second-level adjustment strategy is adopted, and active intervention of the seat is initiated. When the risk is high or the aforementioned methods are ineffective, intervention is made in the vehicle's underlying control to adjust the chassis damping and torque output.

[0024] As a feasible preferred solution, when it is determined to be induced by longitudinal movement, the lumbar support airbag is controlled to slightly inflate / deflate before the vehicle accelerates / decelerates to provide pre-support; When the movement is determined to be induced by lateral motion, the control side wing support provides a sense of envelopment and drives the vibration motor to generate a slight pulse vibration in the opposite direction to the centrifugal force. When the visual conflict is identified as the cause, the visual stabilization mode is activated, dynamically adjusting video playback parameters or switching screen display modes.

[0025] As a feasible preferred solution, it also includes collecting passenger physiological behavior data again within a set time window after the adjustment action is performed, calculating the adjustment effect score; when the score is greater than the threshold, the strategy is deemed effective and given priority recommendation; when the score is less than the threshold, the strategy is upgraded or switched; and the dynamic baseline threshold is updated according to the cumulative travel time to simulate the cumulative fatigue effect.

[0026] As a feasible preferred option, the calculation of the adjustment effect score is based on the weighted sum of the rate of change of heart rate variability before and after adjustment, the rate of change of head posture stability index, and the rate of change of blink frequency. The formula for updating the dynamic baseline threshold is:

[0027] in, This is the fatigue accumulation coefficient. This is the time increment.

[0028] Secondly, this invention also proposes a real-time prediction and adaptive adjustment system for motion sickness comfort, which utilizes the aforementioned real-time prediction and adaptive adjustment method for motion sickness comfort, including: The multi-source sensing and data fusion layer is used to capture information from people, vehicles, and the environment. Intelligent computing and decision-making layer, used for risk prediction and adjustment strategy generation; A multimodal execution and regulation layer is used to execute the regulation instruction package; The effect evaluation and closed-loop optimization layer is used to collect passenger physiological behavior data again within a set time window after the adjustment action is executed, calculate the adjustment effect score, determine the effectiveness of the strategy based on the score and upgrade or switch the strategy, and update the dynamic baseline threshold based on the cumulative riding time. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of the architecture of a time-series embedding generation system for multi-task knowledge retrieval of a large lithium battery model.

[0030] Figure 2 This is a logical diagram of a temporal embedding generation method for multi-task knowledge retrieval of a large lithium battery model. Detailed Implementation

[0031] To make the technical solution and advantages of this application clearer, the technical solution of the present invention will be further described in detail below with reference to the accompanying drawings. It is understood that the specific embodiments described herein are only some embodiments of the present invention, and are only used to explain this application, not to limit it. It should be noted that the technical features or combinations of technical features described in the following embodiments should not be considered isolated; they can be combined with each other to achieve better technical effects. The same reference numerals appearing in the accompanying drawings of the following embodiments represent the same features or components, and can be applied to different embodiments.

[0032] Furthermore, unless otherwise defined, the technical or scientific terms used in this invention description shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains.

[0033] The present invention will now be described in further detail with reference to the accompanying drawings.

[0034] Reference Figure 1 This disclosure provides a real-time prediction and adaptive adjustment system for motion sickness comfort, the hardware structure of which includes an intelligent cockpit domain controller, a multi-source sensing module, an on-board actuator, and an on-board communication network.

[0035] In this embodiment, the intelligent cockpit domain controller is integrated inside the vehicle's dashboard or below the center console, and includes a high-performance processor (CPU / GPU), memory (RAM / ROM), and I / O interfaces. It connects to the sensing module and actuators via an in-vehicle Ethernet or CAN-FD bus, serving as the system's computing core, responsible for running the motion sickness prediction model and generating adjustment commands.

[0036] In this embodiment, the multi-source sensing module includes an in-cabin visual sensor (typically an infrared camera), installed on the inside of the A-pillar or at the ceiling reading light, facing the occupant's face. It is used to collect facial micro-expressions (such as paleness and blinking frequency) and gaze direction.

[0037] In this embodiment, the bio-radar sensor is integrated inside the seat headrest or the inner side of the B-pillar. It is used for non-contact monitoring of the occupant's heart rate variability (HRV) and respiratory rate.

[0038] In this embodiment, the seat pressure sensor array is laid under the seat cushion foam layer to detect changes in the occupant's center of gravity and body fidgeting.

[0039] In this embodiment, the vehicle motion sensor, i.e. the vehicle's original IMU (Inertial Measurement Unit), is installed near the vehicle's center of gravity (usually on the center console or chassis) to collect triaxial acceleration and angular velocity.

[0040] The vehicle-mounted actuators include intelligent seat actuators, air conditioning and fragrance modules, screen display modules, and chassis control modules.

[0041] The software structure of a motion sickness comfort real-time prediction and adaptive adjustment system includes a multi-source perception and data fusion layer, an intelligent computing and decision-making layer, a multi-modal execution and adjustment layer, and an effect evaluation and closed-loop optimization layer. Data interaction and command coordination are carried out through vehicle Ethernet or high-speed bus (such as CAN-FD), forming an organic whole with the cockpit domain controller (or central computing platform) as the core.

[0042] The multi-source perception and data fusion layer is used to capture information about people, vehicles, and the environment in real time, including the occupant status perception module and the vehicle motion and environment perception module.

[0043] The occupant status perception module includes non-invasive visual sensors (such as in-cabin cameras DMS / OMS), bio-radar / millimeter-wave radar, and piezoelectric / capacitive sensors integrated into the seat.

[0044] Specifically, the visual sensor uses image analysis algorithms to extract occupant facial features, eye movement parameters (such as saccade speed, blink frequency, and pupil diameter), and head posture (pitch and yaw angles) in real time. The bio-radar emits and receives weak electromagnetic waves to detect subtle movements in the occupant's chest cavity without contact, thereby calculating heart rate and its variability, and respiratory rate. The seat sensor monitors the occupant's postural stability (the trajectory of the body's center of pressure) and involuntary micro-movements. These signals constitute a dynamic feature vector characterizing the occupant's real-time physiological arousal, visual fatigue, and vestibular-visual conflict state.

[0045] The vehicle motion and environment perception module reuses the vehicle's existing sensor network, including the onboard inertial measurement unit, body domain controller, and temperature / humidity / air quality sensors.

[0046] Specifically, the IMU provides raw motion data such as vehicle three-dimensional acceleration, jerk, and yaw / pitch angular velocity. The body domain controller provides contextual information such as vehicle speed, steering angle, and current driving mode. Environmental sensors provide data such as cabin temperature and humidity, and CO2 concentration. These signals constitute the external environmental feature vector that induces motion sickness.

[0047] The intelligent computing and decision-making layer is used for risk prediction and adjustment strategy generation, including occupant characteristics and model library, motion risk prediction model, and adaptive strategy decision engine.

[0048] Occupant data is stored in onboard storage or in the cloud, including a static feature library (occupant age, gender, BMI, basic MSSQ score, personal regulation preferences) and a dynamic feature buffer (real-time physiological behavior time series data).

[0049] Real-time motion sickness risk prediction models are pruned and quantized edge computing optimization models (such as TinyLSTM and Mini Transformer). Their innovation lies in the fact that the model's input is no longer a single vehicle motion parameter, but a fusion of vehicle motion feature vectors, occupant real-time physiological behavior feature vectors, and occupant static feature vectors. Through this fusion, the model can learn an individual's physiological response patterns under specific motion stimuli, thereby outputting a dynamically changing real-time motion sickness risk index between 0 and 1. This index more accurately reflects the occupant's cumulative motion sickness risk "at this moment."

[0050] The adaptive strategy decision engine comprises a risk assessment unit and a multimodal adjustment strategy matrix. The risk assessment unit compares a real-time risk index with a dynamic baseline threshold (adaptively adjusted based on static characteristics and cumulative travel time to simulate the cumulative effect of fatigue) to determine the risk level (low, medium, or high). Simultaneously, it identifies the primary risk trigger types (e.g., longitudinal motion-induced, lateral motion-induced, visual conflict-induced, or a combination thereof) through model interpretability analysis (e.g., feature contribution). The decision engine then uses the risk level and trigger type as a composite key to query a predefined or online-learned multimodal adjustment strategy matrix, generating a set of specific, coordinated cabin subsystem adjustment command packages.

[0051] The multimodal execution and adjustment layer is used to precisely and collaboratively execute adjustment commands, including the intelligent seat adjustment module, the cabin environment adjustment module, and The intelligent seat adjustment module features active lumbar support, side wing support, and a seat vibration motor. Specifically, when the system determines that longitudinal movement poses a risk, it can control the lumbar support airbag to slightly inflate / deflate before the vehicle accelerates / decelerates, providing pre-support to counteract inertial forces; when it determines that lateral movement poses a risk, it can control the side wing support to provide a sense of envelopment and drive the vibration motor to generate slight pulse vibrations in the opposite direction of centrifugal force, providing tactile cues to alleviate vestibular sensory conflict.

[0052] The cabin environment control module includes multi-zone intelligent air conditioning, an integrated fragrance generator, an ambient lighting system, and a color-changing panoramic sunroof. Specifically, when the risk increases, a fresh air mode can be activated to increase ventilation in the passenger area and switch to external circulation to reduce CO2 concentration; release refreshing mint or citrus fragrances to improve subjective feelings; adjust the ambient lighting to a cooler color temperature, or direct the light towards the apex of a curve to provide visual spatial anchoring.

[0053] The infotainment and display adjustment module includes the central control screen, the passenger-side screen, and the rear entertainment screen. To address the risks induced by visual conflicts, a visual stabilization mode can be activated to dynamically adjust the field of view or image scrolling speed of video playback to better match the vehicle's movement; or the screen can be automatically switched to a blue light filtering, low dynamic content mode to reduce visual stimulation.

[0054] The effect evaluation and closed-loop optimization layer is used to enable the system's self-learning and continuous improvement, including feedback data collection and strategy effect evaluator.

[0055] Within a short time window (e.g., 10-30 seconds) after the adjustment action is performed, the system collects the occupant's physiological behavior data again (e.g., whether heart rate variability has leveled off, whether head posture is more stable). By comparing the changes in risk indices and physiological characteristics before and after adjustment, the effectiveness of the current adjustment strategy is quantitatively evaluated. Effective strategies will be reinforced and recorded for priority recommendation in similar scenarios; ineffective or poorly performing strategies will trigger the decision engine to upgrade the strategy (e.g., add adjustment dimensions) or switch strategies (e.g., from visual adjustment to tactile adjustment), thus forming an online adaptive optimization closed loop.

[0056] The effect evaluation and closed-loop optimization layer is used to enable the system's self-learning and continuous improvement, including feedback data collection and strategy effect evaluator.

[0057] Within a short time window (e.g., 10-30 seconds) after the adjustment action is performed, the system collects the occupant's physiological behavior data again (e.g., whether heart rate variability has leveled off, whether head posture is more stable). By comparing the changes in risk indices and physiological characteristics before and after adjustment, the effectiveness of the current adjustment strategy is quantitatively evaluated. Effective strategies will be reinforced and recorded for priority recommendation in similar scenarios; ineffective or poorly performing strategies will trigger the decision engine to upgrade the strategy (e.g., add adjustment dimensions) or switch strategies (e.g., from visual adjustment to tactile adjustment), thus forming an online adaptive optimization closed loop.

[0058] Reference Figure 2 This disclosure also provides a method for real-time prediction and adaptive adjustment of motion sickness comfort, including the following steps.

[0059] Step S100, Occupant identification and system initialization, includes: After the vehicle is started, the system enters real-time monitoring mode, as follows.

[0060] Step S101, Occupant Identification: When the vehicle starts, the domain controller verifies the occupant's identity using one of the following methods: Method 1: Using biometrics (such as facial recognition), in-cabin visual sensors capture images of occupants' faces and compare them with pre-stored facial features of authorized users; Method 2: Device binding. When a passenger's smartphone connects to the vehicle via Bluetooth, their account information is automatically linked.

[0061] Step S102, Personalized Feature Loading: After identity verification, the domain controller loads the occupant's personalized feature data from the vehicle's onboard storage or the cloud, including: Static characteristics include age, gender, BMI, basic MSSQ (Motion Sickness Susceptibility Questionnaire) score, and personal conditioning preferences (such as sensitivity to temperature, tolerance to vibration, etc.). The dynamic feature buffer is cleared and initialized to store real-time physiological behavior time-series data collected during this ride.

[0062] Step S103: Dynamic baseline threshold initialization. The system calculates the initial dynamic baseline threshold based on the occupant's static characteristics. Specifically, let the occupant's MSSQ score be... S MSSQ Then the initial dynamic baseline threshold T base The calculation formula is:

[0063] in, The threshold value is the general threshold, and α is the sensitivity amplification factor. The average MSSQ score of the population. This is the maximum MSSQ score. This threshold will be adaptively adjusted downwards in subsequent steps based on cumulative travel time to simulate the cumulative effect of fatigue.

[0064] Step S200, real-time synchronous acquisition and feature extraction of multimodal data, including: Step S201: Parallel data acquisition. The domain controller acquires vehicle motion data, occupant physiological behavior data, and cabin environment data in parallel at a uniform sampling frequency of 10Hz.

[0065] The sensors used to acquire vehicle motion data include an IMU (Induction Unit), and the data types include triaxial acceleration. a x , a y , a z (Unit: m / s²) jerk j x , j y , j z (Unit: m / s³) Yaw angular velocity ω z (Unit: ° / s) Pitch angular velocity ω y (Unit: ° / s).

[0066] Sensors used to collect occupant physiological and behavioral data include visual sensors, bio-radar, and pressure sensors. Data types include blink frequency. fblink (Unit: times / minute), Pupil diameter change Δ d pupil (Unit: mm) Heart rate HR (Unit: bpm), Heart rate variability HRV (Unit: ms) Respiratory rate f breath (Unit: times / minute), Seated center of gravity shift Δp CoP (Unit: mm)

[0067] The sensors used to acquire cabin environmental data include temperature / humidity / air quality sensors, and the data types include temperature. T cabin (Unit: °C) CO 2 concentration C CO2 (Unit: ppm).

[0068] Step S202, Time Synchronization and Alignment: Due to slight deviations in the sampling clocks of different sensors, the domain controller uses IEEE 1588 PTP (Precise Time Protocol) to timestamp all data, ensuring that multimodal data at the same moment correspond strictly on the time axis, and controlling the time synchronization error within 1ms.

[0069] Step S203, Feature Extraction and Vector Construction: The domain controller filters and extracts features from the original signal. Extract heart rate waveforms from radar signals and calculate time-domain metrics of HRV (such as RMSSD). Extract eye movement trajectories from images and calculate saccadic velocity and pupil diameter change rate; The standard deviation of the seated center of gravity movement trajectory is calculated from the pressure sensor data.

[0070] The system generates vehicle motion feature vectors, occupant real-time physiological behavior feature vectors, and occupant static feature vectors.

[0071] Step S300, personalized real-time dynamic prediction of motion sickness risk, including: To achieve truly personalized prediction, this invention employs a dynamic weight allocation mechanism based on occupant static profiles, as detailed below: Step S301, let the static feature vector of the occupant be:

[0072] in Scoring of motion sickness susceptibility questionnaire.

[0073] Define dynamic weight vector The calculation logic is as follows: Weighting of highly susceptible groups: If (High susceptibility) increases the weighting of real-time physiological behavioral characteristics (such as heart rate variability HRV and skin conductance EDA). The expression is:

[0074] in, The sensitivity coefficient (preferably set to 0.15) Based on the weights.

[0075] Age decay factor: if (For the elderly population), reduce vehicle motion characteristics (such as high-frequency acceleration). Weighting coefficients Because older adults typically have a higher threshold for perceiving high-frequency vibrations, the expression is:

[0076] in It is the attenuation constant (preferably set to 0.05).

[0077] Step S302, after weight allocation, the weighted feature vector synthesis is completed at time... Input feature vector Represented as:

[0078] in, For vehicle motion characteristics, These are real-time physiological and behavioral characteristics. The fixed embedding weights are for static features.

[0079] Step S303: Dual-layer LSTM fusion temporal processing. Given that motion sickness has a significant time accumulation effect, this invention uses a dual-layer LSTM network combined with an attention mechanism to fuse temporal data.

[0080] Short-time dependency capture (first layer LSTM) input First, the system enters the first layer of the LSTM network to capture dynamic changes within a short window (e.g., the past 10 seconds):

[0081] in This is the first hidden state, used to extract the coupling features between instantaneous motion and physiological response.

[0082] Long-term dependency and attention-weighted approach (second-layer LSTM + Attention): The output sequence of the first layer is fed into the second-layer LSTM. To highlight the impact of critical dizziness moments, an attention mechanism is introduced: Calculate attention score:

[0083] in It is the second layer of LSTM in the 1st... The hidden state at all times These are trainable parameters.

[0084] Normalized weights:

[0085] in, Set a backtracking time window (preferably set to 300 seconds, i.e. 5 minutes).

[0086] Context vector generation: .

[0087] Step S304, final risk output, will be the context vector. Input to a fully connected (dense) layer, output time step through a sigmoid activation function. Real-time motion sickness risk index :

[0088] in, For the Sigmoid function, and For output layer weights and biases. Risk index. A higher value indicates a higher risk of motion sickness.

[0089] Step S305, Risk Level Determination and Cause Classification: The domain controller will display the real-time risk index. With dynamic baseline threshold T base ( t Compare: like R t < T low ， Determined to be low risk; like T low ≤ R t < T high ， Determined to be of medium risk; like R t ≥ T high ， It has been determined to be high-risk.

[0090] Step S400: Based on risk and incentive-based intelligent policy decision-making, the adaptive policy decision engine in the domain controller generates an optimal adjustment instruction package from a predefined multimodal adjustment policy matrix according to a composite key combining risk level and incentive type, including: In step S401, when the risk level is determined to be low, a primary adjustment strategy is adopted. When the risk is low but shows an upward trend, the system initiates fine-tuning at the environmental and information levels. The controller sends commands via the CAN bus to control the air conditioning damper motor. The blower rotates under the motor's drive, increasing airflow, especially targeting the air vents near the face, to introduce fresh air and reduce CO2 concentration. The micro-atomizing motor is activated to release refreshing mint or citrus fragrances. The screen brightness is automatically reduced by 10%-15%, and the UI theme is switched to a cool color scheme (such as light blue) to reduce visual stimulation.

[0091] In step S402, if the risk level is determined to be medium or if the risk value does not decrease after the first-level adjustment, a second-level adjustment strategy is adopted. Active seat intervention is activated, and the controller sends a signal to the lumbar support air pump. The air pump motor operates to build up air pressure (preferably within the range of 0.1-0.3 MPa), inflating the lumbar support airbag to provide physical support for the occupant and reduce body swaying. Simultaneously, the massage motor (preferably with a vibration frequency of 40-60 Hz) is activated to divert attention through vibration. For low-frequency swaying that causes motion sickness, the seat's fore-and-aft adjustment motor is controlled to perform reciprocating micro-movements (in millimeters, preferably ±5 mm).

[0092] Taking the seat's fore-and-aft adjustment as an example, its movement is as follows: The controller outputs a PWM signal to the seat's fore-and-aft adjustment motor. The motor's output shaft begins to rotate clockwise, driving the worm gear to rotate; the worm gear meshes with the reduction gear, driving the lead screw to rotate; the nut on the lead screw is restricted by the slide rail and cannot rotate, thus converting the rotational motion into the linear motion of the seat slide rail. By precisely controlling the motor's start-stop frequency (preferably 1-2Hz) and stroke, the seat produces a slight fore-and-aft displacement, using vestibular stimulation to counteract motion sickness.

[0093] If determined to be induced by longitudinal motion, the lumbar support airbag is slightly inflated / deflated 0.5 seconds before the vehicle accelerates / decelerates to provide pre-support to counteract inertial forces; If the movement is determined to be induced by lateral motion, the control side wing support motor provides a sense of envelopment (the side wings retract inward by 5-10mm), and drives the vibration motor to generate a slight pulse vibration (preferably 20-30Hz) in the opposite direction to the centrifugal force, providing tactile cues to alleviate vestibular sensory conflict. If determined to be caused by visual conflict: activate visual stabilization mode, dynamically adjust the field of view of video playback or the scrolling speed of the image to better match the movement of the vehicle, or automatically switch the screen to a blue light filtering and low dynamic range content mode.

[0094] In step S403, if the risk is deemed high or the aforementioned measures are ineffective, the system intervenes in the vehicle's underlying control, sending a command to the CDC damper controller to increase the damping force (preferably by 30%-50%) to suppress vehicle pitch and roll. A request is sent to the VCU (Vehicle Controller Unit) to smooth the torque output curve and limit acceleration and deceleration during rapid acceleration and deceleration (Jerk limit, preferably within ±5m / s³). The ambient lighting is adjusted to a cooler color temperature (preferably above 6500K), or the lights are directed towards the apex of the curve during cornering to provide visual spatial anchoring.

[0095] In step S500, multi-system coordinated smooth adjustment is executed. The domain controller distributes the adjustment command packets generated in step S400 to the actuators of each subsystem via the CAN-FD bus or Ethernet. To ensure the precision and coordination of the adjustment actions, the system adopts a gradual execution strategy: all adjustment commands are not completed instantaneously, but rather linearly ramped up over 0.5-2 seconds to avoid additional discomfort caused by sudden changes; the adjustment actions of multiple subsystems are staggered in time (with an interval of 0.3-0.5 seconds) to avoid perceptual conflicts caused by simultaneous actions; the adjustment range strictly corresponds to the risk level: the adjustment range for low risk is 20%-30% of the maximum value, for medium risk it is 50%-60%, and for high risk it is 80%-100%.

[0096] Taking seat back angle adjustment as an example, its mechanical movement is as follows: Power source component: Backrest adjustment motor (DC permanent magnet motor).

[0097] Transmission components: worm gear mechanism, planetary gear reducer, angle adjuster.

[0098] Movement: When the controller determines that the backrest angle (reclining) needs to be increased, it applies a positive current to the motor. The motor output shaft begins to rotate counterclockwise, driving the worm gear at the front end to rotate synchronously. The worm gear meshes with the worm wheel, and since the worm wheel axis is perpendicular to the motor axis, the rotational motion of the worm wheel is converted into a direction perpendicular to the seat plane. The worm wheel drives the cam plate inside the adjuster to rotate, driving the locking mechanism to release. Finally, under the action of gravity or an electric push rod, the backrest frame rotates backward around a fixed axis, achieving angle adjustment.

[0099] Step S600, real-time evaluation of adjustment effects and strategy iteration, including: Step S601, effect monitoring, within a short time window after the adjustment action is executed (preferably T). eval =10 (30 seconds) The system collects the occupants' physiological behavior data again, including heart rate variability. (and pre-adjustment heart rate variability) (Comparison) Head posture stability index (Standard deviation of head posture angle), blink frequency wait.

[0100] Step S602, the domain controller calculates the adjustment effect score. E The formula is as follows:

[0101] in, , , These are the weighting coefficients for each indicator.

[0102] Step S603: If the adjustment effect score is greater than the adjustment effect score threshold, the current adjustment strategy is determined to be effective, and the strategy is recorded as an effective strategy. It will be given priority recommendation in similar scenarios (same risk level + trigger type). If the adjustment effect score is less than the adjustment effect score threshold, the current strategy is determined to be invalid, triggering the decision engine to upgrade the strategy (such as adding adjustment dimensions, upgrading from single seat adjustment to joint adjustment of seat, air conditioning and screen) or switch the strategy (such as changing from visual adjustment to tactile adjustment), and the original strategy is marked as to be optimized.

[0103] Step S604: Dynamic baseline threshold update. At regular intervals, the system updates the dynamic baseline threshold based on the cumulative travel time.

[0104] in, This is the fatigue accumulation coefficient. This update uses a time increment. It simulates the cumulative effect of occupant fatigue, enabling the system to gradually lower the warning threshold during long journeys and detect risks in advance.

[0105] The above content is merely an embodiment of the present invention. Commonly known structures and characteristics of the solutions are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can improve and implement this solution based on the guidance provided in this application and their own capabilities. Typical well-known structures or operating methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A method for real-time prediction and adaptive adjustment of motion sickness comfort, characterized in that, include: Load the static characteristics of the occupants, and calculate the initial dynamic baseline threshold based on the static characteristics; Collect vehicle motion data, occupant physiological behavior data, and cabin environment data; extract features and construct vectors to form vehicle motion feature vectors, occupant real-time physiological behavior feature vectors, and occupant static feature vectors. Based on the static characteristics of the occupants, the feature vector is dynamically weighted, and the risk prediction model outputs a real-time motion sickness risk index based on the weighted feature vector. The risk index is compared with a dynamic baseline threshold to determine the risk level and the type of risk trigger. Based on the motion sickness risk level and risk trigger type, an adjustment instruction package is generated from the preset multimodal adjustment strategy matrix.

2. The method for real-time prediction and adaptive adjustment of motion sickness comfort according to claim 1, characterized in that, Load the occupant's static characteristics, and calculate the initial dynamic baseline threshold based on the static characteristics, specifically including the following: Methods for verifying passenger identity include biometric identification or device binding; The static characteristics also include age, gender, BMI, and personal adjustment preferences; The formula for calculating the initial dynamic baseline threshold is: in, The threshold value is the general threshold, and α is the sensitivity amplification factor. S MSSQ Assess the MSSQ score of the occupants. The average MSSQ score of the population. This is the highest score for MSSQ.

3. The method for real-time prediction and adaptive adjustment of motion sickness comfort according to claim 2, characterized in that, The vehicle motion data includes at least one of three-axis acceleration, jerk, yaw rate and pitch rate; the occupant physiological behavior data includes at least one of blink frequency, pupil diameter change, heart rate, heart rate variability, respiratory rate and seated center of gravity shift; the cabin environment data includes at least one of temperature and CO2 concentration.

4. The method for real-time prediction and adaptive adjustment of motion sickness comfort according to claim 1, characterized in that, Dynamic weight allocation includes weight adjustments for highly susceptible groups and adjustments for age decay factors; After weight allocation, the weighted eigenvector synthesis at time... Input feature vector Represented as: in, For vehicle motion characteristics, These are real-time physiological and behavioral characteristics. S For the static characteristics of the occupants, Weights for vehicle motion sickness features. Adjusted by age decay factor, The weighting coefficients represent the weights of real-time physiological and behavioral characteristics. Adjusted from highly susceptible groups, The fixed embedding weights are for static features.

5. The method for real-time prediction and adaptive adjustment of motion sickness comfort according to claim 4, characterized in that, The dual-layer LSTM fusion timing processing includes: The first LSTM layer captures short-term dependencies: in This is the first hidden state, used to extract the coupling features between instantaneous motion and physiological response; The second-layer LSTM combines an attention mechanism to capture long-term dependence and highlight the impact of key dizziness-inducing moments; Calculate attention score: in, It is the second layer of LSTM in the 1st... The hidden state at all times These are trainable parameters; Normalized weights: in, For backtracking time windows; Context vector generation: context vector Input to a fully connected layer, output time step through the Sigmoid activation function. Real-time motion sickness risk index : in, For the Sigmoid function, and These are the output layer weights and biases.

6. The method for real-time prediction and adaptive adjustment of motion sickness comfort according to claim 1, characterized in that, It also includes executing the adjustment command package to adjust at least one of the smart seat, cabin environment, or infotainment display; The risk levels include low risk, medium risk, and high risk; When the risk is low, a level one adjustment strategy is adopted to adjust the cabin environment and screen display. When the risk is medium risk or the first-level adjustment is ineffective, the second-level adjustment strategy is adopted, and active intervention of the seat is initiated. When the risk is high or the aforementioned methods are ineffective, intervention is made in the vehicle's underlying control to adjust the chassis damping and torque output.

7. The method for real-time prediction and adaptive adjustment of motion sickness comfort according to claim 1, characterized in that, When determined to be induced by longitudinal movement, the lumbar support airbag is slightly inflated / deflated before the vehicle accelerates / decelerates to provide pre-support; When the movement is determined to be induced by lateral motion, the control side wing support provides a sense of envelopment and drives the vibration motor to generate a slight pulse vibration in the opposite direction to the centrifugal force. When the visual conflict is identified as the cause, the visual stabilization mode is activated, dynamically adjusting video playback parameters or switching screen display modes.

8. The method for real-time prediction and adaptive adjustment of motion sickness comfort according to claim 1, characterized in that, It also includes collecting passenger physiological behavior data again within a set time window after the adjustment action is executed, calculating the adjustment effect score; when the score is greater than the threshold, the strategy is deemed effective and recommended first; when the score is less than the threshold, the strategy is upgraded or switched; and the dynamic baseline threshold is updated according to the cumulative travel time to simulate the cumulative fatigue effect.

9. The method for real-time prediction and adaptive adjustment of motion sickness comfort according to claim 1, characterized in that, The adjustment effect score is calculated based on the weighted sum of the rate of change of heart rate variability before and after adjustment, the rate of change of head posture stability index, and the rate of change of blink frequency. The formula for updating the dynamic baseline threshold is: in, This is the fatigue accumulation coefficient. For time increments.

10. A real-time prediction and adaptive adjustment system for motion sickness comfort, characterized in that, The method employs a real-time prediction and adaptive adjustment method for motion sickness comfort as described in any one of claims 1-8, comprising: The multi-source sensing and data fusion layer is used to capture information from people, vehicles, and the environment. Intelligent computing and decision-making layer, used for risk prediction and adjustment strategy generation; A multimodal execution and regulation layer is used to execute the regulation instruction package; The effect evaluation and closed-loop optimization layer is used to collect passenger physiological behavior data again within a set time window after the adjustment action is executed, calculate the adjustment effect score, determine the effectiveness of the strategy based on the score and upgrade or switch the strategy, and update the dynamic baseline threshold based on the cumulative riding time.

Images