A sensor- and AI-modeled method for optimizing seat comfort

Abstract

This invention provides a seat comfort optimization calculation method based on sensors and AI modeling. The method includes: real-time acquisition of pressure distribution data of the occupant's sitting posture and spinal contact area using posture sensors and pressure sensors installed in the seat; dividing the seat support area into cervical, thoracic, lumbar, and sacral regions, and establishing a pressure distribution prediction model for each region; constructing an input vector based on average pressure, maximum pressure, pressure center of gravity, body shape parameters, and pathological weight factors, and using an artificial intelligence model to output the target support pressure value, pressure avoidance curve, and adjacent region migration ratio for each region; achieving dynamic adjustment through zoned airbags and drive units, and combining feedback monitoring for secondary fine-tuning and protection. The system includes a posture sensing module, a pressure sensing module, a data processing module, a model building module, a control execution module, and a feedback monitoring module, enabling adaptive control of different regions of the occupant's spine.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence and data processing technology, and in particular to a method for optimizing seat comfort based on sensors and AI modeling. Background Technology

[0002] With the widespread use of automobiles, long-distance driving has become a common occurrence. As the primary interface between the human body and the vehicle, the comfort of the seat directly impacts the spinal health of the occupant. Medical research indicates that improper posture and continuous uneven pressure distribution are significant factors contributing to spinal diseases such as cervical spondylosis, thoracic scoliosis, and lumbar disc herniation. For occupants with pre-existing spinal pathology, the vertebral bodies and surrounding soft tissues are more susceptible to abnormal stress in sensitive areas or affected segments. Prolonged exposure to concentrated pressure can easily exacerbate their condition or even cause secondary damage.

[0003] While some existing seats incorporate posture sensors and pneumatic adjustment modules, their adjustment targets are mostly limited to overall pressure balance, lacking integration with the occupant's spinal pathology information, making it difficult to achieve personalized pressure relief and protection for pathological occupants.

[0004] Therefore, we propose a method for optimizing seat comfort based on sensor data and AI modeling. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by providing a seat comfort optimization calculation method based on sensors and AI modeling, thereby solving the technical problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for optimizing seat comfort based on sensors and AI modeling includes the following steps:

[0008] S1. An array of posture sensors and pressure sensors is arranged in the seat back and seat cushion to collect the posture parameters and pressure distribution data of the occupant in real time. The collected data is subjected to spatiotemporal marking, seat detection, noise reduction and filtering smoothing, and characteristic parameters such as average pressure value, maximum pressure value, pressure center of gravity, left and right imbalance index and posture angle are extracted.

[0009] S2. Map the seat support area to the human spinal anatomical sub-region, identify the contact area and extract the force characteristics in each sub-region, establish a pressure distribution prediction model based on features such as posture angle, total support force and pressure center of gravity, and correct the model parameters through individualized calibration to ensure prediction accuracy.

[0010] S3. Obtain the occupant's spinal pathology information. Match the lesion segment with the anatomical sub-region through manual input, medical record import or detection signal acquisition, generate the pressure sensitivity level, pressure avoidance parameters and pathological weight factors, and input them into the subsequent model in a numerical manner.

[0011] S4. Construct an input vector that integrates the above feature parameters and pathological weight factors. Based on a multi-task neural network and a time-series prediction model, predict the target support pressure value, generate the pressure avoidance curve and the proportion of support migration in adjacent areas, and minimize the comfort score function under global constraints to output the optimized target pressure distribution.

[0012] S5. Based on the optimized target pressure value, generate control commands to drive the pneumatic support modules of each anatomical sub-region, perform independent adjustment and adjacent region gradient compensation, and achieve real-time pressure adjustment through dynamic transition curves and feedback closed-loop correction.

[0013] S6. Real-time monitoring of pressure threshold, pressure change rate and pressure gradient; identification and classification of abnormal trends; when abnormal stress is detected in the lesion area, the active protection mode is triggered to perform rapid pressure relief and adjacent area compensation, while generating and storing abnormal event records.

[0014] Preferably, S1 specifically involves: arranging an array of posture sensors and an array of pressure sensors inside the seat back and seat cushion, respectively, to collect posture parameters and pressure distribution data of the occupant in different sitting postures; marking the raw data collected by the sensors with time and space to distinguish between the occupant's sitting state and data from different sampling times; performing denoising and filtering smoothing processing on the raw data to eliminate data fluctuations caused by sensor noise and posture perturbations; extracting feature parameters such as average pressure value, maximum pressure value, pressure center of gravity position, left-right pressure imbalance index, and posture angle based on the filtered data; and using these feature parameters as the input data for subsequent prediction models and optimization algorithms.

[0015] Preferably, S2 specifically involves: dividing the support area of ​​the seat back and seat cushion into multiple sub-regions such as the cervical spine region, thoracic spine region, lumbar spine region, and sacral and coccygeal region according to the anatomical structure of the human spine; identifying the actual contact area between the occupant and the seat within each anatomical sub-region, and extracting the force characteristics such as the support area, local pressure peak, and pressure distribution uniformity of that area; constructing a pressure distribution prediction model for each anatomical sub-region based on parameters such as the posture angle, total support force, and pressure center of gravity; comparing the collected individualized data with the standard reference model, and correcting the model parameters to reduce prediction deviations caused by individual differences; and ensuring the accuracy and stability of the pressure distribution prediction results through the individualized calibrated prediction model.

[0016] Preferably, S3 specifically involves: acquiring the occupant's spinal pathology information, which is obtained through manual input, importing medical records, or collecting physiological signals; matching the lesion segments corresponding to the pathology information with the anatomical sub-regions of the seat support area to determine the location of the pathological influence; generating a pressure sensitivity level based on the lesion segments to characterize the pressure risk of different sub-regions; determining pressure avoidance parameters in combination with pathological characteristics and generating pathological weight factors for support adjustment; and quantifying the pressure sensitivity level, pressure avoidance parameters, and pathological weight factors as input variables for subsequent optimization models.

[0017] Preferably, S4 specifically involves: constructing a multi-dimensional input vector based on the feature parameters and pathological weight factors; inputting the input vector into a multi-task neural network and a time-series prediction model to predict the target support pressure value; generating a pressure relief curve based on the prediction results and calculating the support migration ratio of adjacent areas to coordinate the force on different anatomical sub-regions; establishing and minimizing a comfort scoring function under global constraints to optimize the overall seat support distribution; outputting the optimized target pressure distribution as the adjustment basis for the control execution module; and introducing a dynamic adjustment mechanism during the model prediction process to improve the real-time performance and robustness of the pressure distribution prediction.

[0018] Preferably, S5 specifically involves: generating control commands based on the optimized target pressure value and sending them to the pneumatic support modules of each anatomical sub-zone of the seat; driving the pneumatic support modules to perform independent adjustment operations so that the support force of the corresponding sub-zone reaches the target pressure value; calculating and applying adjacent zone gradient compensation during the sub-zone adjustment process to avoid discomfort caused by sudden pressure changes; using a dynamic transition curve to control the adjustment process, making pressure changes smooth and improving riding comfort; and combining real-time pressure feedback to construct a closed-loop correction mechanism to iteratively update the control commands to ensure the accuracy and stability of pressure adjustment.

[0019] Preferably, S6 specifically involves: real-time monitoring of pressure threshold, pressure change rate, and pressure gradient to identify potential abnormal trends; classifying the abnormal trends into mild, moderate, and severe risks; automatically triggering an active protection mode when an abnormal force is detected in the lesion area; performing a rapid depressurization operation in the active protection mode and applying compensating support to adjacent sub-areas to maintain overall force balance; generating an abnormal event record after the protection action is executed and storing the record in a database for subsequent analysis and tracking.

[0020] The beneficial effects of this invention are as follows:

[0021] This invention acquires occupant spinal pathology information (such as lesion segments, sensitivity levels, and pathological weighting factors) and combines it with seat pressure characteristic parameters to provide differentiated seat support optimization solutions for occupants with different pathological characteristics, thereby reducing stress on lesion sites and avoiding secondary injury. By integrating multi-dimensional features such as posture angle, pressure center of gravity, and left-right imbalance index, and constructing an AI prediction framework based on multi-task neural networks and time-series prediction models, it can achieve high-precision prediction of target pressure distribution, significantly improving the accuracy and stability of seat comfort optimization.

[0022] This invention generates control commands based on optimized target pressure values, driving the pneumatic support modules of each anatomical sub-zone to achieve independent adjustment and adjacent zone gradient compensation. Combined with dynamic transition curves and closed-loop feedback mechanisms, the seat can adjust in real time as the occupant's condition changes, ensuring continuous comfort and safety. During real-time monitoring of pressure thresholds and pressure change rates, when abnormal force is detected in the lesion area, an active protection mode is triggered, performing rapid depressurization and adjacent zone compensation, and generating an abnormal event record. This effectively prevents lesion aggravation and enhances the system's medical auxiliary value. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of a seat comfort optimization calculation method based on sensors and AI modeling according to the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Example 1: As Figure 1 As shown, this embodiment provides a seat comfort optimization calculation method based on sensors and AI modeling, including the following steps:

[0026] S1. Data Acquisition and Feature Extraction: An array of posture sensors and pressure sensors is arranged in the seat back and seat cushion to collect the occupant's posture parameters and pressure distribution data in real time. The collected data is then subjected to spatiotemporal labeling, seat detection, noise reduction and filtering smoothing, and feature parameters such as average pressure value, maximum pressure value, pressure center of gravity, left-right imbalance index and posture angle are extracted.

[0027] S2. Sub-region modeling and calibration: The seat support area is mapped to the human spinal anatomical sub-region. Contact areas are identified and force characteristics are extracted in each sub-region. A pressure distribution prediction model is established based on features such as posture angle, total support force and pressure center of gravity. The model parameters are corrected through individualized calibration to ensure prediction accuracy.

[0028] S3. Pathological data fusion: Obtain the pathological information of the occupant's spine, match the lesion segment with the anatomical sub-region through manual input, medical record import or detection signal acquisition, generate the pressure sensitivity level, pressure avoidance parameters and pathological weight factors, and input them into the subsequent model in a numerical manner.

[0029] S4. Artificial Intelligence Prediction and Optimization: Construct an input vector that integrates the above feature parameters and pathological weight factors. Based on a multi-task neural network and a time-series prediction model, predict the target support pressure value, generate the pressure avoidance curve and the proportion of support migration in adjacent areas, and minimize the comfort score function under global constraints to output the optimized target pressure distribution.

[0030] S5. Closed-loop execution and adjustment: Based on the optimized target pressure value, control commands are generated to drive the pneumatic support modules of each anatomical sub-region, perform independent adjustment and adjacent region gradient compensation, and achieve real-time pressure adjustment through dynamic transition curves and feedback closed-loop correction.

[0031] S6. Anomaly Monitoring and Protection: Real-time monitoring of pressure threshold, pressure change rate and pressure gradient; identification and classification of abnormal trends; when abnormal stress is detected in the lesion area, the active protection mode is triggered to perform rapid depressurization and adjacent area compensation, while generating and storing abnormal event records.

[0032] Step S1 specifically includes the following:

[0033] S110. Sensor deployment and signal acquisition: Attitude sensors and thin-film pressure sensor arrays are respectively deployed inside the backrest and seat cushion of the car seat.

[0034] The attitude sensor collects the pitch angle, lateral tilt angle and anterior tilt angle of the occupant's upper body, with a collection frequency of, for example, 100Hz. The attitude sensor preferably uses an inertial measurement unit composed of a three-axis accelerometer and a three-axis gyroscope, which are installed at the T12 position of the occupant's thoracic vertebra and the anterior superior iliac spine of the pelvis, respectively, to collect the occupant's upper body attitude information in real time.

[0035] Specifically, the acceleration and angular velocity signals output by the attitude sensor are fused by Kalman filtering to calculate the pitch angle (defined as the angle of inclination of the trunk's central axis relative to the vehicle's longitudinal plane), the roll angle (defined as the angle of inclination of the trunk's central axis relative to the vehicle's transverse plane), and the anterior tilt angle of the pelvis (defined as the angle between the line connecting the anterior superior iliac spine and the pubic symphysis and the horizontal plane).

[0036] First, regarding the calculation of the pitch angle, the three-axis acceleration signals output by the attitude sensor are denoted as follows: ,in Corresponding to the vehicle's forward and backward acceleration, Corresponding to lateral acceleration, The vertical acceleration corresponds to the acceleration signal. The acceleration and angular velocity signals are fused using Kalman filtering or complementary filtering to eliminate instantaneous interference and noise. Based on this, the pitch angle can be calculated using the following formula:

[0037]

[0038] This calculation can characterize the degree of forward and backward tilt of the spine's central axis relative to the vehicle's longitudinal plane.

[0039] Secondly, the calculation of the left and right tilt angles also utilizes the triaxial acceleration signals and the fused angular velocity signals. Gravity component decomposition eliminates the interference of vehicle acceleration or braking on the results. The left and right tilt angles can be calculated using the following formula:

[0040]

[0041] This angle is used to characterize the degree of left-right tilt of the spine's central axis relative to the vehicle's lateral plane.

[0042] Next, for the calculation of the anterior pelvic tilt angle, attitude sensors were placed at the anterior superior iliac spine and pubic symphysis on both sides of the pelvis to acquire three-dimensional coordinate data (x, y, z). The direction vector of the connecting line can be obtained from the coordinate difference.

[0043]

[0044] : Represents the position coordinates of the anterior superior iliac spine in a three-dimensional coordinate system; : Represents the three-dimensional coordinates of the pubic symphysis; subtracting the two gives a direction vector from the pubic symphysis to the anterior superior iliac spine. This vector can be understood as a spatial representation of the direction of anterior pelvic tilt;

[0045] in, Therefore, the angle between the line connecting the pelvis and the horizontal plane can be calculated:

[0046]

[0047] A positive angle indicates anterior pelvic tilt, while a negative angle indicates posterior pelvic tilt.

[0048] Through the above calculation process, the system can obtain the occupant's pitch angle, lateral tilt angle, and pelvic tilt angle in real time and stably, providing reliable posture characteristic parameter inputs for subsequent comfort modeling and seat optimization and adjustment.

[0049] The sampling frequency of the above posture parameters is set to 100Hz to ensure the real-time data acquisition and dynamic response characteristics, thereby providing accurate posture data for subsequent occupant posture recognition and seat support adjustment.

[0050] The pressure sensor array can be spaced at a density of 10mm-15mm, with each unit having a sampling frequency of 20Hz and a measurement range of 0-80kPa.

[0051] S120. Data Labeling and Time Series Alignment: To ensure consistency across multiple data sources, each data point is automatically labeled after collection.

[0052] Spatial labeling: Mark the location of the sensor on the seat (e.g., the upper edge of the cervical spine area, the midpoint of the lumbar spine area, the left rear edge of the seat cushion, etc.) to ensure that data from different sub-areas can be distinguished.

[0053] Timestamps: Use timestamps accurate to the millisecond (e.g., July 1, 2025, 13:20:30.123) to ensure all sensor data is aligned along a uniform timeline. All data is stored in the database in a "timestamp + spatial location" format.

[0054] S130. Seating Detection and Contact Area Confirmation:

[0055] When the system detects the total pressure value of the pressure array If the duration exceeds 0.5 seconds, it is determined that the occupant has taken their seat. Among these, Indicates the first The instantaneous pressure value (unit: Pa) detected by each pressure sensor unit. Indicates the first The force-bearing area corresponding to each pressure sensor unit (unit: m²) 2 ), the product of the two This indicates the force (in N) acting on the sensor unit. The pressure array consists of multiple independent pressure sensor units distributed across the seat back and cushion areas. Threshold The settings are based on the minimum static seating weight of an adult occupant (approximately 15kg of body weight), ensuring effective differentiation between a real occupant sitting down and small objects (such as mobile phones or miscellaneous items) placed on the seat. The duration threshold of 0.5s is used to filter out instantaneous impact or brief touch signals to avoid misjudgment.

[0056] After meeting the above criteria, the system will only record data from the effective contact area. The effective contact area is defined as follows: in the pressure array, if a sensor unit experiences a force greater than or equal to 5N, and at least three adjacent units simultaneously meet this condition, it is considered an effective contact area. During recording, the system only retains pressure distribution information within the effective contact area, ignoring invalid or isolated pressure units. This avoids invalid sampling under empty seating conditions or noise signals, ensuring the reliability of subsequent posture recognition and support adjustment data.

[0057] To ensure feasibility, the pressure sensor is preferably a resistive thin-film pressure sensor or a piezoelectric pressure sensor, with a measurement range of 0-100 N, a resolution preferably not less than 0.5 N, and a response time of less than 50 ms, ensuring real-time and accurate acquisition of pressure changes during occupant seating. The sensor is preferably arranged on a flexible circuit board beneath the seat surface, with an area preferably 2–4 cm². 2 This ensures spatial resolution and detection sensitivity.

[0058] S140, Data Denoising and Stability Calibration:

[0059] The raw data collected is typically affected by temperature drift, sensor noise, and environmental interference, thus requiring noise reduction and stability processing. Specifically, for the triaxial acceleration and triaxial angular velocity data collected by the inertial measurement unit (IMU), a first-order low-pass filter is used for processing. The cutoff frequency of this low-pass filter is set to 10Hz, and its discrete-time difference equation is as follows:

[0060]

[0061] in, Indicates the first Input raw data at any time, This represents the filtered output data. , The filtering time constant is The sampling period.

[0062] For pressure sensor array data, a 3×3 Gaussian kernel convolution smoothing method is used to remove isolated outliers. The process involves calculating a weighted average of the data from each pressure cell, centered on the data from its eight neighboring cells. The weighting coefficient matrix uses a normalized Gaussian kernel.

[0063]

[0064] The sum of the weighting coefficients is guaranteed to be 1. Furthermore, to identify faulty or drifting sensor units, a dead-zone determination rule is set for each sensor: if a sensor outputs data that remains unchanged for one consecutive second (e.g., 50 consecutive sampling points have the same value at a sampling rate of 50Hz), the sensor is marked as an outlier, and its data is removed from subsequent calculations. Through the above filtering, smoothing, and dead-zone determination processes, temperature drift and noise interference can be effectively suppressed, isolated outliers can be removed, and the stability and reliability of sensor data can be guaranteed.

[0065] S150, Data Feature Extraction and Storage:

[0066] After preprocessing, the system extracts key feature indicators, including: total support force (the total pressure value of the pressure array mentioned above). Average value of pressure distribution Maximum value Left-right balance index ;

[0067]

[0068] It is the first The pressure value measured by each pressure sensing unit; It is the first The force-bearing area of ​​each pressure sensing unit; It is the first The coordinates of each sensing unit in the horizontal direction (unit: m, usually with the geometric center of the seat as the zero point, negative to the left and positive to the right); n is the total number of sensing units; The LBI value is used to measure the degree of balance of the occupant's body pressure distribution in the left and right directions of the seat; if LBI is approximately equal to 0, it means that the forces on the left and right sides are balanced; if LBI is greater than 0, it means that the center of pressure is biased to the right; if LBI is less than 0, it means that the center of pressure is biased to the left; the larger the absolute value, the greater the degree of imbalance.

[0069] Attitude angle characteristics: pitch angle, left and right tilt angle and pelvic tilt angle. Attitude angle characteristics are obtained by integrating the three-axis angular velocity signals of the inertial measurement unit (IMU) and are used to describe the overall spatial attitude of the occupant's sitting posture.

[0070] Center of pressure ;

[0071]

[0072] in, These are the cell coordinates. COP is used to reflect the geometric center of the overall force distribution.

[0073] The data unit at each time point is saved as:

[0074] {Timestamp, Sensor Spatial Location, Attitude Data, Pressure Matrix Data, Feature Parameters}

[0075] For example, at 13:20:30 on July 1, 2025, the average pressure in the lumbar region was 8.2 kPa, the maximum pressure was 16.5 kPa, and the left-right balance index was −0.05. If the occupant weighs 70 kg, the average pressure in the seat area is approximately 9 kPa (corresponding to a support force of approximately 490 N), the average pressure in the backrest area is approximately 7 kPa (corresponding to a support force of approximately 190 N), and the total support force is approximately 680 N. This is basically consistent with the theoretical value corresponding to body weight, 70 × 9.81 m / s² ≈ 686 N, verifying the reliability of the extracted data results.

[0076] Step S2 specifically includes the following:

[0077] S210. Seat Area Division and Anatomical Sub-region Mapping: When mapping the seat support area to the anatomical sub-regions of the human spine, the junction of the seat back and seat cushion is used as the zero point of height, and the vehicle standard coordinate system is adopted, where: the X-axis points to the front of the vehicle, the Y-axis points to the left side of the vehicle, and the Z-axis points to the top of the vehicle. Measuring upwards along the Z-axis from the zero point, the seat back is divided into four anatomically corresponding sub-regions: the cervical spine region (C region), the thoracic spine region (T region), the lumbar spine region (L region), and the sacrococcygeal region (S region), with the following height ranges:

[0078] The sacrococcygeal region (S region): corresponds to the pelvis and coccyx area, located at the junction of the backrest and seat cushion, with a height range of approximately 0–120 mm; the lumbar region (L region): corresponds to the L1–L5 segments of the human spine, located in the lower middle part of the backrest, with a height range of approximately 120–250 mm; the thoracic region (T region): corresponds to the transition area from the scapula to the thoracic spine, with a height range of approximately 250–350 mm; the cervical region (C region): corresponds to the upper edge of the cervical spine, with a height range of approximately 350–450 mm.

[0079] The sections are connected continuously according to their height range, with no gaps or overlaps, thus ensuring a one-to-one correspondence between the support area and the anatomical structure of the human spine.

[0080] Those skilled in the art can make adaptive scaling adjustments based on the occupant's height and the total height of the backrest. The partition height can be calculated using the following formula: ,in, For standard height, The baseline height is (e.g., 170 cm). This is the scaled height. This refers to the occupant's height. The above height range is a numerical range based on a typical occupant's body type. Those skilled in the art can make adaptation adjustments based on the actual occupant's height or seat dimensions.

[0081] S220, Data Projection and Sub-region Contact Recognition:

[0082] The seat back is equipped with a pressure sensor array to form a two-dimensional pressure matrix. The system uses the line connecting the pelvic contact point and the scapular contact point as a reference central axis, and projects the pressure matrix onto the corresponding height range of each sub-zone along the central axis, thereby realizing contact recognition of each sub-zone.

[0083] Taking a backrest with a total height H=600mm as an example, the Z-axis range corresponding to the lumbar region is: Its sensing unit set is defined as follows: The set of effective contact points within the lumbar region can be determined by the following formula:

[0084]

[0085] in: This represents the two-dimensional coordinates of the pressure sensor on the backrest plane. This indicates the pressure value at the corresponding location; This is the contact detection threshold, used to filter out non-contact and noise points. For example, it can be set to... = 0.5.

[0086] Furthermore, with a total backrest height H=600mm, the height range of each anatomical sub-region can be specifically divided as follows: cervical spine region: 510-600mm; thoracic spine region: 270-510mm; lumbar spine region: 120-270mm; sacrococcygeal region: 0-120mm. Through the above mapping and determination method, the accurate correspondence between the two-dimensional pressure matrix and the anatomical sub-regions can be achieved, ensuring the accuracy of contact recognition.

[0087] S230. Sub-region Feature Extraction: Within each sub-region, extract feature parameters that reflect the stress state, including: mean pressure value. Maximum pressure value Center of Pressure (COP) Pressure variation range and standard deviation ;in, Sub-region Number of internal sampling points This represents the average pressure value. It can characterize the uniformity of the force in the sub-region; the larger the value, the more uneven the pressure distribution.

[0088] The symmetry index (pressure difference between left and right sub-regions) refers to the difference in average pressure values ​​between symmetrical areas of the human body (e.g., left and right lumbar spine areas, left and right thoracic spine areas, etc.) to characterize whether there is uneven stress distribution in body posture. When the average pressure values ​​of the left and right sub-regions are similar, it indicates that the stress distribution in that area is relatively symmetrical, and the occupant's posture is maintained in a natural state of balance. However, when the average pressure values ​​of the left and right sub-regions differ significantly, it indicates that one side is bearing excessive pressure, which often means that the occupant's posture is tilted, shifted, or that there is abnormal stress on one side.

[0089] For example, taking the lumbar region (L region) as an example, if the contact area is measured... Support The average pressure value for this area is:

[0090]

[0091] If the maximum pressure unit in this region is 16.5 kPa and the standard deviation is 3.2 kPa, it indicates that there is a significant localized concentrated stress phenomenon in the lumbar spine region. By extracting the above characteristic parameters, the pressure state of different anatomical sub-regions can be accurately quantified, reflecting both the overall stress situation and revealing local abnormalities or unevenness, thus providing reliable data support for subsequent posture analysis, fatigue assessment, and intelligent control.

[0092] S240. Sub-zone Pressure Distribution Prediction Model Establishment: To achieve subsequent intelligent control, a pressure distribution prediction model needs to be established for each sub-zone of the seat back. This model is used to predict the two-dimensional pressure distribution under different operating conditions based on the occupant's sitting posture characteristics and historical pressure data.

[0093] Input features: attitude angles (pitch, tilt), total support force, center of gravity (COP), average pressure, maximum pressure, and left-right imbalance index. Output target: a two-dimensional pressure distribution matrix, or an equivalent low-dimensional parametric representation (e.g., 5×5 DCT coefficients). Model form: For in-vehicle systems with limited computing power, linear regression can be used; for systems with high computing power, a lightweight convolutional neural network (CNN) can be used to predict pressure images. For example, in the lumbar region (L-zone), the input feature is: seat pitch angle. The occupant's weight was 70 kg, the COP offset was -12 mm, and the imbalance index was -0.08. The model predicted an average pressure of 8.5 kPa and a maximum pressure of 17.2 kPa, and the predicted image showed that the pressure in the left cell was 12% higher than that in the right cell. This result is consistent with the actual observation, indicating that the model has high accuracy.

[0094] S250. Calibration and Validation: To ensure the adaptability of the prediction model to individual differences, individualized calibration is required for each occupant.

[0095] Calibration process: When the occupant first sits down, the system automatically collects stable sitting posture data for 30–60 seconds to obtain the measured pressure matrix. Compare the measured data with the model prediction results. In comparison, if the error exceeds 10%, a calibration process is triggered. Calibration algorithm: The least squares method is used to adjust the model parameters to minimize the mean square error between the predicted and measured values. After calibration, the prediction error should converge to within ±5%.

[0096] For example, in the cervical spine region, the model predicted a pressure of 6.8 kPa, while the measured value was 7.5 kPa, resulting in an error of 0.7 kPa (≈10.3%). After least squares calibration, the model output was updated to 7.4 kPa, and the error decreased to 0.1 kPa (≈1.3%). This demonstrates that the calibration process can effectively improve the individualized accuracy of the model. Multiple linear regression, support vector machines, or convolutional neural networks are existing conventional modeling methods, while least squares is a common parameter estimation algorithm, which will not be discussed in detail here.

[0097] Step S3 specifically includes the following:

[0098] S310. Pathological Data Acquisition Channels: To enable the system to make targeted adjustments based on the occupant's spinal health status, it is necessary to first collect and integrate the occupant's pathological information. This involves collecting the occupant's spinal health status through multiple channels; for example: manual input: the occupant or medical personnel input the diagnostic results via the in-vehicle interface, tablet terminal, or mobile client. For example, inputting "L4-L5 lumbar disc herniation" and selecting the severity level (e.g., mild, moderate, severe).

[0099] Electronic medical record import: The system automatically reads passengers' previous diagnostic images or medical reports, such as MRI, X-ray, and CT reports, through an interface with the medical system database. The read results are then parsed by the natural language processing module and converted into structured text.

[0100] Example of fusion: If an occupant enters "L4-L5 intervertebral disc herniation" in the system interface and selects the "high-pressure sensitive" option, and at the same time the EMG test indicates that the erector spinae muscle RMS exceeds the threshold, the system will label the occupant as "L4-L5 lesion + high-stress sensitive".

[0101] After the above collection and fusion, the pathological data is transformed into standardized parameter inputs, such as: {segment: L4–L5, type: protrusion, severity: moderate, EMG_RMS: 0.35mV, sensitivity: high}.

[0102] S320. Lesion segment localization and data coding: After obtaining pathological information, the system needs to map the information to specific anatomical segments of the spine and perform standardized coding for subsequent modeling and control.

[0103] Spinal segment division: Cervical segments: C1–C7; Thoracic segments: T1–T12; Lumbar segments: L1–L5; Sacrococcygeal segments: S1–S5; Coding rules: Each diseased segment is identified by a combination of "segment + disease type" and stored in key-value pairs.

[0104] Example 1: "C5 intervertebral disc degeneration" is coded as {segment: C5, type: degeneration}; Example 2: "L4–L5 intervertebral disc herniation" is coded as {segment: L4–L5, type: herniation}; Lesion type set: The lesion types supported by the system include, but are not limited to: degeneration, herniation, spinal stenosis, fracture, slippage, and other clinically defined pathological changes;

[0105] Automatic matching and labeling: When a user inputs "L4–L5 lumbar disc herniation", the system automatically matches it to a "lumbar sub-region" and labels it as a "lesion area" in the two-dimensional pressure distribution model, so that differentiated intervention can be applied to this area in subsequent treatment. For example, if the user inputs "L4–L5 herniation, moderate", the system will ultimately store it as: {segment: L4–L5, type: herniation, severity: moderate, label: lesion area}.

[0106] S330, Pressure Sensitivity Grading: The pressure sensitivity grading module converts the pressure-bearing capacity of the lesion area into a quantifiable grading index, facilitating precise control in subsequent system adjustments. Based on pathological severity, physician recommendations, and clinical experience, this module categorizes the pressure sensitivity of the lesion area into five levels:

[0107] Level 0 (No abnormality): Pressure threshold ≥ 20 kPa; Level 1 (Mildly sensitive): Pressure threshold 15–20 kPa; Level 2 (Moderately sensitive): Pressure threshold 10–15 kPa; Level 3 (Severely sensitive): Pressure threshold 5–10 kPa; Level 4 (Extremely sensitive): Pressure threshold < 5 kPa. The thresholds for each level can be given by preset medical experience values, or dynamically adjusted based on clinical trial statistics or individual differences. For example, the grading standards can be personalized by combining information such as occupant body type, electromyography signals, and medical history.

[0108] During operation, the system automatically maps the affected area to the corresponding pressure sensitivity level based on the doctor's input or feedback from the detection equipment. When adjusting the seat pressure, the system compares the real-time detected local pressure with the grading threshold: if the detected value exceeds the upper limit threshold of the level, automatic pressure relief control is triggered to prevent the affected area from being subjected to excessive mechanical load.

[0109] For example, if a doctor's diagnosis indicates that "the L4-L5 intervertebral disc herniation area should avoid local pressure exceeding 10 kPa", the system will automatically mark the segment as "Level 3 Sensitive" and set the upper limit of the allowable pressure in the area to 10 kPa during subsequent seat support adjustments. Once the pressure is detected to exceed this threshold in real time, active pressure relief adjustment will be triggered to reduce the occupant's discomfort and the risk of worsening of the condition.

[0110] S340, Recommended Pressure Avoidance Parameter Generation: The recommended pressure avoidance parameter generation module automatically generates corresponding pressure avoidance control parameters based on the pressure sensitivity level of the lesion area, including:

[0111] Target pressure avoidance value: Defined as the maximum permissible average pressure in this area. For example, when the lesion area is marked as "Level 3 Sensitive", the system sets its pressure avoidance target to ≤8 kPa.

[0112] Buffer setting: Automatically increase the support force by 10%–20% in the upper and lower sub-regions adjacent to the lesion area to share the pressure load of the lesion area.

[0113] Adjustment rate parameter: For highly sensitive areas, the inflation and deflation rate of the seat airbag is required to be ≥20mmHg / s to ensure that the affected area can be quickly depressurized and protected when the occupant's posture changes suddenly.

[0114] During implementation, the system will dynamically compare the real-time detected pressure with the aforementioned pressure avoidance target value, and combine buffer transfer and adjustment rate for linkage control.

[0115] For example, if a doctor's diagnosis indicates "Level 3 sensitivity in the L4-L5 lumbar region," the system will automatically generate pressure avoidance parameters: target pressure ≤ 8 kPa, increased support force of adjacent upper lumbar vertebrae and sacrococcygeal segment by 12% each, and a balloon deflation rate ≥ 20 mmHg / s. This parameterized strategy can significantly reduce the peak pressure in the L4-L5 lesion area, thereby avoiding secondary local injury.

[0116] S350, Pathological Weight Factor Calculation: The system converts pathological information into numerical weights, which are used as input factors for the artificial intelligence model. Specifically, this includes:

[0117] Weighting factor definition: The system maps pathological parameters through functional relationships:

[0118]

[0119] in, Indicates the first Pathological weighting factors for each region, function It can be implemented using a linear weighted model, a piecewise function, or a machine learning regression model.

[0120] Parameter range settings: Lesion node area: assigned higher priority, weight coefficient range set to 1.5~2.0; Adjacent area: to maintain support continuity, weight coefficient range set to 0.8~1.2; Other areas can be set with default weight values ​​according to physiological differences, for example: 1.8 for lumbar spine, 1.1 for sacrococcygeal spine, and 0.9 for thoracic spine.

[0121] For example: Suppose the input pathological information is: "L4-L5 lumbar disc herniation, moderately sensitive (grade 2)". The system analysis results are as follows: Lesion area = lumbar region; Compression threshold = ≤12kPa; Compression avoidance parameter = Target pressure in the L4-L5 lumbar region ≤10kPa, with an increase of 15% in the support force of the adjacent sacrococcygeal region; Weighting factor = 1.5 for the lumbar region, 1.1 for the sacrococcygeal region, and 1.0 for the remaining regions.

[0122] Finally, the system inputs the above pathological information into the artificial intelligence model in the form of numerical vectors for subsequent seat support prediction and regulation in the S4 stage.

[0123] Step S4 specifically includes the following:

[0124] S410. Model Input Construction: In this embodiment, feature parameters for training and real-time inference are collected and input, including: average pressure value, maximum pressure value, pressure center coordinates, left-right imbalance index (LBI), weight, height, and sitting height. Simultaneously, to accommodate occupants with spinal abnormalities, a pathological weight factor is introduced, where the weight of the lesion area is set to greater than or equal to 1.5, the weight of adjacent areas is set between 0.8 and 1.2, and the weight of non-lesion areas is set to 1.0. The above feature data is constructed into an input vector for the artificial intelligence model to learn and predict.

[0125] S420 AI Prediction Mechanism: The artificial intelligence model can employ multi-task neural networks and temporal prediction models (such as LSTM and GRU). Input features are temporal feature vectors collected by attitude and pressure sensors, specifically including pitch angle, lateral tilt angle, pelvic tilt angle, average pressure value and pressure change rate of each seat zone, etc. The output target is the zone support pressure value and its change curve at the next moment, and further calculates the support migration ratio of adjacent zones. During training, mean squared error is used as the regression loss function to minimize the deviation between the predicted pressure curve and the measured pressure curve. During the inference phase, the model's computation response time for a single input vector is less than 100ms to ensure real-time use for occupant posture adjustment.

[0126] S430, Pressure Avoidance Curve Generation: In terms of pressure avoidance strategy, this embodiment introduces a pressure avoidance curve function, the expression of which is: ;

[0127] in, For a moment The support pressure value, As the initial pressure, This is the pressure relief rate coefficient (typically ranging from 0.5 to 1.0 s). -1 ), To ensure a lower limit of safety, the system further sets graded thresholds based on the sensitivity level of the spinal region. For example, for a level 3 sensitive area, the target pressure is limited to ≤10kPa, and the pressure relief rate is not less than 1.5kPa / s; for a level 4 sensitive area, the target pressure is limited to ≤5kPa, and the pressure relief rate is not less than 2.5kPa / s.

[0128] S440, Adjacent Zone Support Migration Strategy: Regarding adjacent zone support adjustment, when pressure in the affected area decreases, the system compensates according to the adjacent zone support migration strategy. It prioritizes increasing support in adjacent areas, with an increase range of +10% to +20%; while distant areas only undergo ±5% fine-tuning to ensure overall force balance. For example, when the lumbar spine experiences a 40N pressure reduction, the system will automatically distribute this 40N to the left and right adjacent zones, increasing support by approximately 20N in each, thereby achieving stable support.

[0129] S450, Global Adjustment Optimization: During the global optimization process, the system introduces smoothing constraints and dynamic response constraints. The smoothing constraint requires that the pressure gradient between adjacent sub-regions not exceed 2 kPa / 10 mm, and the dynamic response constraint is limited to ≤30 mmHg / s. Simultaneously, a comfort scoring function is defined.

[0130] Comfort rating function:

[0131] in: Indicates the first Real-time support pressure of each anatomical subregion; This indicates the target pressure value for this sub-region; This represents a weighting coefficient used to reflect the priority of different sub-regions in comfort and pathological correction, with a typical value range of 0.05 to 0.3. It represents the square of the deviation between the actual pressure and the target pressure, and is used to measure the degree of pressure deviation; An index representing the smoothness of pressure distribution can be defined, for example, as the sum of squares of the pressure differences between adjacent sub-regions:

[0132]

[0133] This represents a smoothing factor used to adjust the trade-off between comfort and stability, typically ranging from 0.1 to 1.0. It is achieved by minimizing the objective function. The system can ensure pressure avoidance in the affected area while maintaining a smooth pressure distribution and avoiding local abrupt changes, thereby improving the overall comfort and support stability of the occupants.

[0134] S460. During the output phase, the system generates target pressure values, pressure relief curve slopes, and adjacent area support migration ratios for each region, and drives the seat to perform adjustment actions accordingly. For example, in the initial state, the lumbar spine pressure is 12 kPa, the thoracic spine pressure is 10 kPa, the sacral and coccygeal spine pressure is 8 kPa, and the cervical spine pressure is 7 kPa. After 5 seconds of pressure relief and adjacent area compensation, the lumbar spine pressure decreases to 8 kPa, the thoracic and cervical spine pressures stabilize at 10 kPa and 7 kPa respectively, while the sacral and coccygeal spine pressure is dynamically adjusted to 9 kPa, thereby achieving effective pressure relief for the affected area and improving overall comfort.

[0135] Step S5 specifically includes the following:

[0136] S510, Pneumatic Support Module Structure Configuration: Independent airbag units are arranged in the cervical, thoracic, lumbar, and sacral regions. Each airbag is equipped with: an electrically controlled air valve with a response time ≤50ms; a pressure sensor with a range of 0–50kPa and an accuracy of ±0.5kPa; and a flow control unit with an adjustable inflation / deflation rate range of 0–40mmHg / s. For example, two independent airbags (left LL, right LR) are set in the lumbar region, each with a volume of approximately 0.8 L, which can be independently adjusted to balance the force difference between the left and right sides.

[0137] S520, Independent Control Strategy: Target Pressure Value for Each Anatomical Subregion From the AI ​​output of step S4, the system calculates the inflation / deflation command based on the target value:

[0138]

[0139] in, It is the first Inflation / deflation flow rate of the zone airbag (controlled variable); This is the effective force-bearing area of ​​the airbag. This refers to the airbag stiffness coefficient. It is the target pressure (set value); Current pressure (measured in real time by the sensor);

[0140] when The formula gives the positive flow rate. , indicates inflation; when The formula gives the negative flow rate. , indicates releasing air; when The formula is 0, the system remains balanced; single adjustment amplitude ≤ ±3kPa; adjustment cycle 0.5-1.0s; airbag inflation / deflation rate ≤ 30mmHg / s. For example: current lumbar spine pressure. Target value If so, it is calculated that approximately 0.09L of gas needs to be released, and the adjustment time is expected to be 4 seconds.

[0141] S530, Regional Linkage Gradient Response Mechanism: To avoid abrupt changes, the system introduces a gradient transition mechanism between the lesion area and the adjacent area: the pressure adjustment amplitude of the adjacent area = 10–30% of the adjustment amplitude of the lesion area; the adjustment of the adjacent area is delayed by 0.5–1.0s to form a buffer.

[0142] Global gradient constraint: Pressure difference between adjacent sub-regions ≤ 2 kPa. For example, if the lumbar region is decompressed by -3 kPa, the system will simultaneously decompress the thoracic region by -0.6 kPa and the sacrococcygeal region by -0.9 kPa to ensure the continuity of the transition curve.

[0143] S540, Dynamic Transition Curve Generation: The system generates a smooth curve based on the adjustment amplitude and response rate.

[0144]

[0145] :time The Zone airbag pressure; Initial pressure; Target pressure; This is an adjustment rate parameter that controls the speed of pressure convergence (typically 0.4–0.8), ensuring that 90% of the adjustment is completed within 5–8 seconds, especially for highly sensitive lesion areas. Ensure rapid pressure relief; for non-lesion areas, This ensures gradual adjustment. For example, adjusting the lumbar spine area from 12 kPa to 9 kPa using... It dropped to 9.3 kPa in 4 seconds and stabilized at 9.0 kPa in 6 seconds.

[0146] S550, Feedback Closed Loop and Real-time Correction: The airbag pressure sensor monitors the adjustment effect in real time: if the target deviation is > ±0.5kPa, a secondary fine adjustment is automatically triggered; if the pressure gradient of the adjacent area exceeds the limit (>2kPa / 10mm), the buffer delay is increased; if the lesion area still exceeds the threshold (>10% of the pressure avoidance target), a user prompt is issued or the active protection mode is entered (S6).

[0147] For example, if the target pressure in the lumbar region is 9 kPa, but the actual pressure stabilizes at 9.7 kPa, with a deviation of 0.7 kPa, the system will automatically release 0.02 L of gas for the second time, and after compensation, the deviation will be reduced to 0.2 kPa.

[0148] For example: Assume the initial conditions are: lumbar spine pressure 12 kPa (target 9 kPa); thoracic spine pressure 10 kPa (target 11 kPa); sacrococcygeal spine pressure 8 kPa (target 9 kPa). The system then adjusts as follows: the lumbar spine airbag gradually deflates over 6 seconds, reducing the pressure to 9 kPa; the thoracic spine airbag inflates by +1.0 kPa after a 0.5-second delay, reaching 11 kPa; the sacrococcygeal spine airbag inflates by +1 kPa, reaching 9 kPa; the transition between regions is smooth, with a gradient <2 kPa.

[0149] Step S6 specifically includes the following:

[0150] S610, Real-time Pressure Threshold Monitoring: The system uses pressure sensors deployed in key anatomical sub-regions of the spine to collect average pressure, maximum pressure, pressure change rate, and pressure gradient at monitoring points in real time. If the measured pressure value of a certain monitoring sub-region exceeds 10% of the target pressure value for that region and lasts for more than 3 seconds, or the pressure change rate is greater than 3 kPa / s, or the pressure gradient exceeds 2 kPa / 10 mm, then that region is determined to have entered an abnormal stress state.

[0151] S620. Abnormal Trend Identification and Classification: The system classifies abnormal states according to their magnitude and rate: Mild abnormality: pressure exceeds the target value by 10% to 20%, or the rate is 3 to 5 kPa / s; Moderate abnormality: pressure exceeds the target value by 20% to 40%, or the rate is 5 to 8 kPa / s; Severe abnormality: pressure exceeds the target value by more than 40%, or the rate exceeds 8 kPa / s; The abnormality level serves as the trigger condition for subsequent control.

[0152] S630 Active Protection Mode Trigger Mechanism: After detecting an anomaly, the system activates the active protection mechanism, which includes: automatically reducing the support stiffness of the abnormal area by 20% to 40% and continuously adjusting it at a rate of 2% per second until the pressure drops back to the target range; synchronously driving adjacent sub-areas to perform 20% to 40% compensation support to avoid global balance instability; when the anomaly is moderate or severe, the system prompts the occupants to enter protection mode through voice or visual interface.

[0153] S640, Dynamic Pressure Relief Compensation: When the system enters the active protection mode, for the detected diseased vertebral area, the control module first performs a rapid decompression operation, so that the airbag pressure in the diseased area gradually decreases along a preset dynamic function law; at the same time, the system performs smooth compensation inflation for adjacent vertebral areas, thereby forming a synergistic effect of dynamic pressure relief and comfort maintenance.

[0154] To achieve the above functions, this embodiment uses a first-order dynamic approach model to describe the pressure control process, and its mathematical expression is as follows:

[0155]

[0156] in, For time Pressure on the lesion area at all times As the initial pressure, For target pressure, To adjust the rate constant, a value between 1.0 and 1.5 is preferred to ensure that the adjustment process is completed within 3-5 seconds and maintains a smooth transition. For example, when the initial pressure in the lumbar region is 12 kPa and the target pressure is 8 kPa, the system gradually reduces it to the target value within 4 seconds; simultaneously, it compensates with +1.5 kPa in the sacral and thoracic regions, thereby ensuring the continuity of overall spinal support and comfort. This approach not only achieves rapid pressure relief in the affected area but also avoids abrupt changes in the overall stress on the spine through synergistic compensation of pressure in adjacent areas, effectively improving the protective effect and riding experience.

[0157] S650, User Feedback and Recording: After adjustment, the system automatically generates an abnormal event report. The report includes: the anatomical sub-area where the abnormality occurred, the type and level of the abnormality, the duration, the adjustment measures, and the final pressure recovery status. This report is stored in the local database in real time and can be synchronously uploaded to the cloud-based health record system for long-term tracking and medical reference. For example: The system prompts: "2025-08-17 10:30 Abnormal pressure in the lumbar spine area (level 3) was detected, triggering the protection mode, executing rapid pressure relief of -4kPa, and compensation of +3kPa to adjacent areas, with a total duration of 6 seconds." For example: Assume a passenger's lumbar spine lesion area (target 8kPa): initial pressure 9.5kPa; rises to 12.5kPa within 2 seconds (exceeding the threshold by +56%). System response: judged as level 3 abnormality; immediately triggers active protection mode; reduces pressure to 8.0kPa within 4 seconds; compensates +1.2kPa and +1.8kPa to the adjacent sacral and thoracic spine areas respectively; user interface prompts: "The lumbar spine area has entered rapid pressure relief protection."

[0158] Example 2: This example provides a seat comfort optimization calculation system based on sensors and AI modeling, including: a posture sensing module, which is installed in the seat back and seat cushion to collect the posture data of the occupant in real time;

[0159] The pressure sensing module includes pressure sensing arrays distributed in the cervical spine region, thoracic spine region, lumbar spine region and sacral coccygeal region, used to collect pressure distribution data of each region;

[0160] The data processing module is used to perform time-series calibration, filtering, and outlier removal on the collected data, and to extract average pressure, maximum pressure, pressure centroid, and distribution uniformity characteristic parameters.

[0161] The model building module is used to establish a pressure distribution prediction model based on the characteristic parameters of each zone, and to make corrections by combining the occupant's spinal pathology parameters and pressure sensitivity data to generate the target pressure value for each zone.

[0162] The control and execution module includes independently controlled electronic airbags, flow control units, and drive units, which are used to perform inflation and deflation regulation according to the target pressure value and control the support force of each zone;

[0163] The feedback monitoring module is used to detect the pressure status after regulation in real time, compare it with the target pressure value, and perform secondary fine-tuning or enter active protection mode when the deviation exceeds the set threshold.

[0164] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0165] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0166] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0167] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0168] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0169] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0170] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0171] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0172] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0173] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for optimizing seat comfort based on sensors and AI modeling, characterized in that, Includes the following steps: S1: An array of posture sensors and pressure sensors is arranged in the seat back and seat cushion to collect occupant posture parameters and pressure distribution data in real time. The collected data is subjected to spatiotemporal marking, seat detection, noise reduction and filtering smoothing to extract average pressure value, maximum pressure value, pressure center of gravity, left and right imbalance index and posture angle characteristic parameters. S2: Map the seat support area to the human spinal anatomical sub-region, identify the contact area and extract the force characteristics in each sub-region, establish a pressure distribution prediction model based on posture angle, total support force and pressure center of gravity characteristics, and correct the model parameters through individualized calibration; S3: Obtain the occupant's spinal pathology information, match the lesion segment with the anatomical sub-region, generate the pressure sensitivity level, pressure avoidance parameter and pathological weight factor, and input them into the subsequent model in a numerical manner; S4: Construct an input vector that integrates the above feature parameters and pathological weight factors, predict the target support pressure value based on a multi-task neural network, generate the pressure avoidance curve and the proportion of support migration in adjacent areas, minimize the comfort score function under global constraints, and output the optimized target pressure distribution; S5: Based on the optimized target pressure distribution, control commands are generated to drive the pneumatic support modules of each anatomical sub-region, perform independent adjustment and adjacent region gradient compensation, and achieve real-time pressure regulation through dynamic transition curves and feedback closed-loop correction.

2. The seat comfort optimization calculation method based on sensors and AI modeling according to claim 1, characterized in that, It also includes S6, real-time monitoring of pressure threshold, pressure change rate and pressure gradient, identification and classification of abnormal trends, and when abnormal force is detected in the lesion area, it triggers active protection mode, performs rapid pressure relief and adjacent area compensation, and generates and stores abnormal event records.

3. The seat comfort optimization calculation method based on sensors and AI modeling according to claim 1, characterized in that, S1 specifically refers to: An array of posture sensors and an array of pressure sensors are arranged inside the seat back and seat cushion, respectively, to collect the posture parameters and pressure distribution data of the occupant in different sitting positions. The raw data collected by the sensors are time-stamped and space-stamped to distinguish between the occupant's seated state and data from different sampling times; The raw data is denoised and filtered to smooth it out in order to eliminate data fluctuations caused by sensor noise and attitude perturbations. Based on the filtered data, the average pressure value, maximum pressure value, pressure center position, left-right pressure imbalance index, and attitude angle feature parameters are extracted. The feature parameters are used as the basic input data for subsequent prediction models and optimization algorithms.

4. The seat comfort optimization calculation method based on sensors and AI modeling according to claim 1, characterized in that, S2 specifically refers to: The support areas of the seat back and seat cushion are divided into the cervical spine region, thoracic spine region, lumbar spine region and sacrococcygeal region according to the anatomical structure of the human spine. Within each anatomical sub-region, the actual contact area between the occupant and the seat is identified, and the support area, local pressure peak, pressure distribution uniformity, and force characteristics of that area are extracted. Based on the posture angle, total support force, and pressure center of gravity parameters, a pressure distribution prediction model for each anatomical sub-region is constructed. The collected individualized data is compared with the standard reference model, and the model parameters are adjusted to reduce the prediction bias caused by individual differences. By using an individualized calibrated prediction model, the accuracy and stability of the pressure distribution prediction results are ensured.

5. The seat comfort optimization calculation method based on sensors and AI modeling according to claim 1, characterized in that, S3 specifically refers to: The occupant's spinal pathology information is obtained through manual input, import of medical records, or acquisition of physiological signals. The pathological segments corresponding to the pathological information are matched with the anatomical sub-regions of the seat support area to determine the location of the pathological influence; A pressure sensitivity level is generated based on the lesion segment to characterize the pressure risk of different sub-regions; The pressure avoidance parameters are determined by combining pathological characteristics, and pathological weighting factors for support regulation are generated. The pressure sensitivity level, pressure avoidance parameters, and pathological weighting factors are numerically processed and used as input variables for subsequent optimization models.

6. The method for optimizing seat comfort based on sensors and AI modeling according to claim 1, characterized in that, S4 specifically refers to: A multidimensional input vector is constructed based on the aforementioned feature parameters and pathological weighting factors; The input vector is fed into a multi-task neural network and a time-series prediction model to predict the target support pressure value; Based on the prediction results, a pressure relief curve is generated, and the proportion of support migration in adjacent areas is calculated to coordinate the stress on different anatomical sub-regions; Under global constraints, a comfort scoring function is established and minimized to optimize the overall seat support distribution; The optimized target pressure distribution is output as the basis for adjustment of the control execution module; A dynamic adjustment mechanism is introduced during the model prediction process to improve the real-time performance and robustness of pressure distribution prediction.

7. The method for optimizing seat comfort based on sensors and AI modeling according to claim 1, characterized in that, S5 specifically refers to: Based on the optimized target pressure value, control commands are generated and sent to the pneumatic support modules of each anatomical sub-area of ​​the seat; Drive the pneumatic support module to perform independent adjustment operations so that the support force in the corresponding sub-zone reaches the target pressure value; During subregion regulation, neighboring region gradient compensation is calculated and applied to avoid discomfort caused by sudden pressure changes; The dynamic transition curve control process makes pressure changes smooth and improves ride comfort. By combining real-time pressure feedback, a closed-loop correction mechanism is constructed to iteratively update control commands, thereby ensuring the accuracy and stability of pressure regulation.

8. The seat comfort optimization calculation method based on sensors and AI modeling according to claim 2, characterized in that, S6 specifically refers to: Real-time monitoring of pressure thresholds, pressure change rates, and pressure gradients is used to identify potential abnormal trends; The abnormal trends are classified into mild, moderate, and severe risks. When abnormal stress is detected in the lesion area, the active protection mode is automatically triggered; In the active protection mode, a rapid depressurization operation is performed, and compensating supports are applied to adjacent sub-regions to maintain overall force balance; After the protection action is executed, an abnormal event record is generated and stored in the database.

Images