Method for dynamically adjusting rigidity of intelligent seat in fatigue driving state

Abstract

The invention relates to the technical field of man-machine engineering intelligent control for improving comfort and safety of automobile cabins, and discloses a dynamic adjusting method for rigidity of an intelligent seat in a fatigue driving state. A fixed adjustment period is constructed, original physiological data such as eyelid closing duration are collected, and a stable fatigue index is obtained through time window constraint, abnormal value truncation and dimension unified processing. And combining the initial rigidity and the fatigue index of the seat, dynamically calculating the target rigidity according to a monotone mapping relation, forming error measurement with the difference value of the actual rigidity, determining the adjusting direction and the self-adaptive stride, and ensuring the response speed and the adjustment stability. Through intermediate stiffness interval projection, stride factor adjustment and physical boundary constraint, the accuracy and safety of stiffness adjustment are improved. Furthermore, the update rigidity is mapped into an execution instruction, a continuous scheduling track is generated, and sudden adjustment and hardware damage are effectively avoided in combination with a rate inspection and maintenance mechanism.

Description

Technical Field

[0001] This invention relates to the field of human-machine engineering intelligent control technology for improving the comfort and safety of automotive cabins, specifically a method for dynamically adjusting the stiffness of an intelligent seat during fatigued driving. Background Technology

[0002] With the increasing intelligence of modern automobiles, the comfort and health of driver seats, as the core component of human-machine interaction, are receiving increasing attention. Current passenger and commercial vehicles generally employ mechanical or electric seat structures, allowing drivers to adjust parameters such as seat tilt, height, backrest support, or lumbar support via simple buttons or knobs to meet basic comfort needs. Some high-end models have begun to integrate seat heating, ventilation, and massage functions, but their core adjustment modes remain primarily manual or preset programs, failing to achieve truly on-demand adaptive adjustment. Current mainstream seat stiffness adjustment technologies mostly use fixed-gear mechanical structures or simple logic switching based on limited sensor input, unable to dynamically optimize stiffness parameters according to the driver's real-time physiological or fatigue state, resulting in often lagging and coarse adjustment effects.

[0003] Furthermore, existing intelligent seat systems are mostly passively sensing, responding only to static parameters such as vehicle driving status or driver weight, lacking the ability to assess and adjust for driver fatigue or health risks in real time. For example, some vehicles on the market estimate the driver's posture distribution using pressure or acceleration sensors, but struggle to accurately capture the driver's fatigue level and actual needs for seat support stiffness. Even models claiming "adaptive seats" often rely on preset fixed curves or static matching based on finite variables, failing to achieve continuous stiffness adjustment that updates over time and with driving conditions. More significantly, existing seat stiffness adjustment methods often employ empirical rules or simple threshold strategies, making it difficult to balance comfort, safety, and active protection needs. When drivers are in prolonged driving, inattentive, or experiencing eyelid fatigue, traditional seat structures cannot actively adjust support stiffness to effectively alleviate fatigue, suppress dangerous micro-posture changes, or sudden physiological instability, exhibiting significant drawbacks such as slow response, limited adjustment range, and poor adaptability. Furthermore, traditional stiffness adjustment devices struggle to reconcile the conflict between physical limits, safety protection, and user experience. Under extreme conditions, they are prone to over-adjustment of stiffness or lag in response, affecting driving safety and passenger experience.

[0004] Therefore, this case aims to propose a method for dynamic adjustment of the stiffness of an intelligent seat under fatigue driving conditions. The method first extracts a dimensionless fatigue index from physiological indicators such as eyelid closure time and sets a corresponding emergency threshold. Then, based on the mapping relationship between fatigue level and preset initial stiffness, the stiffness adjustment target is determined. Next, the actual stiffness deviation and construction error metric are calculated, and an intermediate stiffness is generated step by step according to the error reduction direction and adaptive stride factor. Then, the intermediate stiffness is projected onto the physically permissible range to obtain the feasible stiffness. Finally, the stiffness is converted into actuator commands using linear calibration mapping, generating a continuous time trajectory and completing rate verification and issuance. Summary of the Invention

[0005] This invention provides a method for dynamically adjusting the stiffness of an intelligent seat during fatigued driving, thereby helping to solve the problems mentioned in the background art.

[0006] This invention provides the following technical solution: a method for dynamically adjusting the stiffness of an intelligent seat during fatigued driving, comprising: A fixed adjustment cycle is established, and raw data such as eyelid closure duration are collected from the eye movement sensing system at each cycle. Time window constraints, outlier truncation and dimension unification are performed on the data to obtain a dimensionless fatigue index, and an emergency threshold related to the reference duration is established. Based on the preset initial stiffness of the seat and the fatigue index, the target stiffness for the current cycle is calculated according to a monotonic mapping relationship; the mapping provides a target stiffness that does not decrease and remains bounded as the fatigue index increases. The current actual stiffness is read from the seat stiffness control system, the difference between the target stiffness and the actual stiffness is constructed as the error, and a squared error metric is formed based on this. Using the direction of error reduction as the adjustment direction, a stride factor that varies with the fatigue index and is less than the unit amplitude is constructed. The physical allowable range and bandwidth of seat stiffness are preset, the range clamping rules of the initial stiffness are set, and the emergency threshold of error and the protective stiffness used for handling are set. The intermediate stiffness value without boundary projection is calculated according to the adjustment direction and step factor, and the current actual stiffness remains unchanged when the error is zero. The intermediate stiffness is projected within the physically permissible range to obtain the updated stiffness for the current period as the final feasible stiffness. The updated stiffness is mapped to the target command of the actuator, projected within the allowable range of the command, and a time-continuous linear scheduling trajectory is generated within an adjustment cycle. After the rate reachability test, the trajectory is executed or recalculated according to the test result. When the mapping relationship is unavailable, maintenance events are recorded. At the end of the cycle, the command is issued and the next cycle begins.

[0007] Optionally, the process involves establishing a fixed adjustment cycle, collecting raw data such as eyelid closure duration from the eye-tracking system at each cycle point, performing time window constraints, outlier truncation, and dimensionless processing on the data to obtain a dimensionless fatigue index, and establishing an emergency threshold related to the reference duration. Specifically, this includes: Set a fixed adjustment period time interval and construct a discrete adjustment time series; At each adjustment moment, the cumulative duration of eyelid closure in the previous cycle is obtained by the eye-tracking sensor system as the original fatigue quantity; Set a closed reference duration and obtain the dimensionless fatigue index by the ratio of the original quantity to the reference duration. When the cumulative duration of the previous cycle closure is zero, the fatigue index is set to zero; when the reference duration is zero, the adjustment is terminated and an alarm is triggered. When the acquisition error causes the cumulative duration of closure to exceed the upper limit of the time window, the time window is truncated according to the upper limit, and the fatigue index is recalculated according to the truncation result. The emergency fatigue threshold is set based on the ratio of the time window to the reference duration; Write the fatigue index and emergency fatigue threshold into the calculation cache of the current cycle.

[0008] Optionally, the step of calculating the target stiffness for the current cycle based on a preset initial seat stiffness and the fatigue index according to a monotonic mapping relationship specifically includes: Set the initial stiffness of the seat, ensuring that the initial stiffness is positive and within the physically feasible range; A monotonic mapping relationship from fatigue index to target stiffness is established, wherein the relationship provides a target value of not less than the initial stiffness as the fatigue index increases and converges to a finite upper bound. Substituting the fatigue index of the current cycle into the aforementioned relationship yields the target stiffness of the current cycle. Record the target stiffness in the current cycle cache.

[0009] Optionally, the step of reading the current actual stiffness from the seat stiffness control system, constructing the difference between the target stiffness and the actual stiffness as an error, and forming a squared error metric based on this error specifically includes: Read the actual stiffness value at the start of the current cycle from the seat stiffness control system; The difference between the target stiffness and the actual stiffness is calculated as the stiffness error, and the positive and negative directions of the error are marked. A square-shaped error metric is constructed based on the stiffness error; The actual stiffness, target stiffness, and error metric are recorded simultaneously to form the state triplet for the current cycle.

[0010] Optionally, the step factor, which varies with the fatigue index and is less than the unit amplitude, is constructed using the direction of error reduction as the adjustment direction. Specifically, it includes: presetting the physical allowable range and bandwidth of seat stiffness, setting the initial stiffness range clamping rules, and setting the emergency threshold and protective stiffness used for error handling. The adjustment direction is determined based on the decreasing direction of the error measurement, so that the adjustment direction is pointed in the direction of reducing the error; Construct a step size factor that monotonically increases with the error magnitude and is less than one, and set the factor to zero when the error is zero; Set the physical minimum and maximum stiffness, and calculate the bandwidth; set the initial stiffness within the range of the minimum and maximum stiffness, and perform interval clamping if it exceeds the limit; A dimensionless step factor is constructed based on the bandwidth, so that the step size has a suppressive trend as the absolute value of the error increases, thus limiting the intensity of single-cycle adjustment. Set the error limit as the emergency threshold, and set the protection stiffness as the midpoint between the minimum stiffness and the maximum stiffness; Output adjustment direction, stride gene, dimensionless stride factor, emergency threshold, and protection stiffness.

[0011] Optionally, the step of calculating the intermediate stiffness value without boundary projection according to the adjustment direction and step factor, and keeping the current actual stiffness unchanged when the error is zero, specifically includes: A linear update is performed on the current actual stiffness according to the adjustment direction and the dimensionless step factor to obtain the unprojected intermediate stiffness value. When the error is zero, the intermediate stiffness is kept consistent with the current actual stiffness; Write the intermediate stiffness into the cache and wait for interval projection.

[0012] Optionally, the step of performing interval projection on the intermediate stiffness within the physically permissible range to obtain the updated stiffness of the current period as the final feasible stiffness specifically includes: Confirm the minimum and maximum stiffness within the physically permissible range; Establish interval projection rules: take the minimum stiffness for values ​​less than the minimum stiffness, keep the values ​​within the range unchanged, and take the maximum stiffness for values ​​greater than the maximum stiffness. The intermediate stiffness is projected according to the interval projection rule to obtain the updated stiffness, which is used as the final stiffness that can be executed in this cycle. Write the updated stiffness to the execution cache and record a projection event for out-of-bounds cases.

[0013] Optionally, the step of mapping the updated stiffness to the target command of the actuator, projecting it within the allowed range of the command, generating a time-continuous linear scheduling trajectory within an adjustment cycle, executing the trajectory based on the rate reachability check result or recalculating the trajectory according to the maximum reachable rate after the rate reachability check, recording maintenance events when the mapping relationship is unavailable, and issuing commands at the end of the cycle and entering the next cycle, specifically includes: Calibration phase: Two different stiffness calibration points and their corresponding actuator control command values ​​are selected to form two sets of reachable calibration pairs; Based on the two sets of calibration points, linear mapping parameters are calculated to obtain a one-to-one linear relationship between stiffness and control commands. The shortest full-stroke response time is measured under the same deployment conditions, and the maximum achievable average command change rate of the equipment is calculated accordingly. Online mapping and limitations: When any emergency condition is met, i.e. the fatigue index is not lower than the emergency threshold or the absolute value of the error exceeds the error limit, the update stiffness will be reset to the protection stiffness. The updated stiffness is mapped to the target command according to the linear relationship, and the target command is subjected to interval constraints to obtain the target command within the interval; Trajectory generation and rate verification: Read the actual instruction value at the start of the current cycle and execute the interval limit; Generate a linear time scheduling trajectory from the starting actual instruction to the target instruction within the interval in the current cycle time domain, and obtain the execution instruction at the end of the cycle; Calculate the average instruction rate required for this cycle; when the required rate is not greater than the maximum achievable average rate of the device, keep the linear trajectory unchanged; When the required rate exceeds the maximum achievable average rate of the device, the achievable end-of-cycle instruction is calculated based on the difference between the target and the current instruction, using the maximum achievable average rate of the device. An interval limit is applied to this instruction, and the linear time scheduling trajectory is regenerated with the new end-of-cycle instruction. Stiffness prediction and anomaly labeling: Based on the known stiffness-command linear relationship, the time scheduling trajectory is mapped to the predicted stiffness trajectory; When the linear relationship degenerates and the stiffness becomes uncontrollable by the command, the predicted stiffness is set to the current actual stiffness and the maintenance event is recorded. Command issuance and cycle rotation: Continuous time-varying instructions are issued to the actuator controller according to the aforementioned time scheduling trajectory; At the end of the current cycle, the actuator reaches the end-of-cycle instruction and then enters the next adjustment cycle.

[0014] The present invention has the following beneficial effects: 1. This solution employs a combination of fixed-period and multi-stage data preprocessing in the fatigue index acquisition stage. The total eyelid closure time is constrained by a time window, truncated for anomalies, and standardized in terms of dimensions to obtain a dimensionless fatigue index. A dynamic emergency threshold is then established based on a reference duration. Compared to existing technologies that rely heavily on single-instantaneous detection or simple threshold judgment, this method filters out occasional false alarms, eliminates acquisition noise, and standardizes fatigue indices onto a unified scale, enhancing measurement stability and comparability. Furthermore, the introduction of a reference duration-related threshold allows the system to provide rapid warnings when fatigue accumulates to a dangerous level while maintaining stable adjustments within normal fluctuation ranges. This solves the problem of false alarms or missed alarms associated with traditional fixed thresholds, improving driving safety and user experience.

[0015] 2. In the stiffness target determination stage, this solution innovatively proposes a monotonic mapping relationship between the fatigue index and initial stiffness. This ensures that as fatigue levels increase, seat stiffness at least does not decrease and gradually approaches its upper limit, avoiding abrupt changes or drops. Unlike the linear weighted or preset adjustment commonly found in existing technologies, this mapping relationship considers both initial comfort and automatically strengthens support as fatigue intensifies. This achieves a refined and hierarchical stiffness adjustment strategy without increasing system complexity. It maximizes the soft experience for the driver when fatigue is low while actively strengthening support as fatigue accumulates, reducing uneven stress on muscles and the spine, thereby reducing fatigue, delaying physical discomfort, and truly balancing comfort and safety.

[0016] 3. This scheme proposes using the difference between the target stiffness and the actual stiffness as the direct basis for adjustment in the error construction stage, and further introduces a squared error metric to enhance sensitivity to significant deviations. Unlike traditional approaches that only focus on the sign or proportional relationship of the deviation, the squared metric allows large deviations to receive greater weight, making the adjustment step or direction more explicit and effectively avoiding problems of slow convergence or back-and-forth oscillations. In addition, the system records the three-dimensional state of actual stiffness, target stiffness, and error metric in real time, providing complete historical data for subsequent step setting and anomaly detection, achieving transparency and traceability of closed-loop feedback.

[0017] 4. In the adjustment direction and stride setting stage, this scheme combines the error reduction direction with the adaptive stride gene, and integrates the dimensionless stride function with the physically permissible bandwidth to form a multi-level stride control strategy. Unlike existing technologies that use fixed stride or coarse proportional adjustment, this method not only dynamically determines the adjustment direction based on the error direction, but also slightly suppresses the stride as the error increases through monotonic mapping of the error amplitude and bandwidth suppression, ensuring that each adjustment step can respond quickly without excessive fluctuations. Furthermore, by setting an error emergency threshold and protection stiffness, the system can automatically enter a protection mode under abnormal or extreme conditions to prevent uncontrolled adjustment or entry into unsuitable ranges, thereby improving the safety and robustness of the adjustment process.

[0018] 5. In the intermediate stiffness calculation stage, the innovative approach uses linear updates by multiplying the adjustment direction by an adaptive stride factor to directly obtain the unprojected intermediate stiffness value, maintaining constant stiffness when the error is zero. Compared to the incremental accumulation or threshold jumps commonly used in existing technologies, this approach ensures continuity and controllability through simple linear updates, helping to avoid stride imbalance or accumulated errors caused by multiple truncations or stride limits. Simultaneously, the mechanism of writing intermediate values ​​to a cache provides data support for subsequent projection and diagnostics, ensuring that projection is performed only after a thorough evaluation of errors and boundary conditions in each adjustment cycle.

[0019] 6. During the interval projection and stiffness update phases, this solution introduces a standardized projection function, strictly constraining the intermediate stiffness within the physically permissible minimum and maximum range, forming a concise rule of "taking the lower limit for low stiffness, the upper limit for high stiffness, and maintaining the middle range." Compared to traditional multiple judgments through conditional branches, this projection function not only simplifies the control logic but also ensures that the execution stiffness always operates within the safe boundaries, eliminating instruction outputs beyond hardware limits. Furthermore, projection events are logged, providing a basis for subsequent maintenance and safety audits, further improving system reliability. This innovation effectively prevents damage or malfunctions in the seat mechanism due to instruction overflow, enhancing the long-term stability and safety of the product.

[0020] 7. In the final execution phase, the solution establishes a one-to-one linear correspondence between stiffness and execution commands through the two-point mapping parameter calculation during the calibration phase. Combined with the device's maximum achievable average speed, it performs reachability checks and dynamic recalculation on the continuously time-varying command trajectory. Unlike the single-point calibration or constant-speed interpolation commonly used in existing technologies, this design ensures the accuracy of command output while maintaining execution continuity and reachability even under speed constraints by regenerating trajectories that meet hardware capabilities. The solution also automatically records maintenance events when mapping fails and resets commands to the protection level in abnormal branches, enhancing the system's self-diagnosis and fault warning capabilities. Overall, the innovations in this phase not only make execution more reliable but also provide dual protection for hardware lifespan and human-machine collaborative safety. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0022] 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.

[0023] Example, refer to Figure 1 A method for dynamically adjusting the stiffness of an intelligent seat under fatigue driving conditions, comprising: A fixed adjustment cycle is established, and raw data such as eyelid closure duration are collected from the eye movement sensing system at each cycle. Time window constraints, outlier truncation and dimension unification are performed on the data to obtain a dimensionless fatigue index, and an emergency threshold related to the reference duration is established. Based on the preset initial stiffness of the seat and the fatigue index, the target stiffness for the current cycle is calculated according to a monotonic mapping relationship; the mapping provides a target stiffness that does not decrease and remains bounded as the fatigue index increases. The current actual stiffness is read from the seat stiffness control system, the difference between the target stiffness and the actual stiffness is constructed as the error, and a squared error metric is formed based on this. Using the direction of error reduction as the adjustment direction, a stride factor that varies with the fatigue index and is less than the unit amplitude is constructed. The physical allowable range and bandwidth of seat stiffness are preset, the range clamping rules of the initial stiffness are set, and the emergency threshold of error and the protective stiffness used for handling are set. The intermediate stiffness value without boundary projection is calculated according to the adjustment direction and step factor, and the current actual stiffness remains unchanged when the error is zero. The intermediate stiffness is projected within the physically permissible range to obtain the updated stiffness for the current period as the final feasible stiffness. The updated stiffness is mapped to the target command of the actuator, projected within the allowable range of the command, and a time-continuous linear scheduling trajectory is generated within an adjustment cycle. After the rate reachability test, the trajectory is executed or recalculated according to the test result. When the mapping relationship is unavailable, maintenance events are recorded. At the end of the cycle, the command is issued and the next cycle begins.

[0024] This paper proposes a complete dynamic closed-loop adjustment process from fatigue index acquisition to command execution, achieving adaptive matching of intelligent seat stiffness through multi-stage orderly connection. First, a fixed adjustment cycle is set and raw physiological data such as eyelid closure duration are collected, solving the problem of single-instantaneous sampling being susceptible to noise and occasional motion interference in existing technologies. By using time window constraints and outlier truncation, the raw signals are unified to the same dimension, obtaining a stable fatigue index and providing a reliable foundation for subsequent calculations. Second, the fatigue index is mapped to a stiffness target, avoiding the abrupt switching of traditional preset adjustment levels and ensuring the continuity and comfort of adjustment. Third, by constructing the difference between the target stiffness and the actual stiffness and generating an error metric, precise closed-loop feedback is achieved, overcoming the low accuracy problem of relying solely on experience or on / off control. Subsequently, the adjustment direction is determined based on the direction of error reduction, and an adaptive step factor is designed in conjunction with the error amplitude, linking the adjustment step size to the actual difference. This ensures both response speed and avoids excessive oscillation, overcoming the contradiction between response and stability in traditional fixed step size control. In the intermediate stiffness calculation and interval projection stages, by strictly projecting the updated stiffness to the physically permissible range, the safe operation of the hardware is ensured, and mechanical damage caused by out-of-bounds commands is prevented. This method is simpler and more efficient than the existing method that requires manual setting of lower and upper limits. Finally, feasible stiffness is mapped to the target commands of the actuator and issued according to the linear time scheduling trajectory. Through reachability rate verification and recalculation mechanisms, the problem of loss of control or sudden stop when speed is limited is solved. At the same time, maintenance events are recorded when mapping fails, providing a basis for subsequent system diagnosis. The overall solution avoids abrupt adjustments, false alarms and misjudgments, and hardware damage. It introduces multiple adaptive and fault-tolerant measures to improve the accuracy, smoothness, and safety of seat stiffness adjustment, providing a comprehensive system-level solution for the driving experience under fatigue driving conditions.

[0025] The process involves establishing a fixed adjustment cycle, collecting raw data such as eyelid closure duration from the eye-tracking system at each cycle point, applying time window constraints, outlier truncation, and dimensionless processing to the data to obtain a dimensionless fatigue index, and establishing an emergency threshold related to the reference duration. Specifically, this includes: Set a fixed adjustment period time interval and construct a discrete adjustment time series; At each adjustment moment, the cumulative duration of eyelid closure in the previous cycle is obtained by the eye-tracking sensor system as the original fatigue quantity; Set a closed reference duration and obtain the dimensionless fatigue index by the ratio of the original quantity to the reference duration. When the cumulative duration of the previous cycle closure is zero, the fatigue index is set to zero; when the reference duration is zero, the adjustment is terminated and an alarm is triggered. When the acquisition error causes the cumulative duration of closure to exceed the upper limit of the time window, the time window is truncated according to the upper limit, and the fatigue index is recalculated according to the truncation result. The emergency fatigue threshold is set based on the ratio of the time window to the reference duration; Write the fatigue index and emergency fatigue threshold into the calculation cache of the current cycle.

[0026] Further specific implementation steps include: Set the adjustment cycle time interval as Seconds, constructing adjustment time series ;in, Number the discrete time steps; It is the set of natural numbers that includes 0; For the first The start time of each time step; At each adjustment time point The total duration of eyelid closure is collected by the eye-tracking sensor system and recorded as follows: ;in, In the first The cumulative duration of eyelid closure within each cycle, in seconds; Set the closure reference duration as ; It is a reference value for eye-closing time, which is the total duration of eyelid closure of the driver within a certain period of actual detection. Convert to "dimensionless fatigue index" In this way, the closure performance of different drivers, different time windows, and different individuals can all be unified on the same numerical benchmark, ensuring that the algorithm is comparable and consistent in different scenarios. It is also a watershed moment for the state of "extreme fatigue." For example, when... (If the eyes are closed throughout the entire cycle), then . The larger the value, the smaller the total duration of eye closure will be. The system's sensitivity to short-term eye closure decreases; conversely, if the value is too small, even brief eye closure may be judged as high fatigue, leading to excessively frequent or large stiffness adjustments. The target stiffness and stride functions in the algorithm directly depend on... Value, therefore A smaller value results in a faster increase in the stiffness gain function, leading to a more drastic response to abnormal eye closure. While this may reduce comfort, it improves safety. Conversely, a value that is too large may result in a sluggish response, decreasing the algorithm's timeliness in protecting against dangerous fatigue. As mentioned above, The smaller the value, the easier it is for the system to trigger protection under extreme fatigue conditions, making the system more sensitive; the larger the value, the lower the probability of extreme protection actions, which may cause some risk scenarios to be missed. Basis for Values: Statistical analysis of eye-closing behavior in healthy drivers: This is typically based on long-term actual driving measurements of a large number of healthy drivers (non-fatigued, without abnormalities), statistically analyzing the mean or high quantile (95th percentile) of the total eye-closing duration per unit time (e.g., within 10 or 60 seconds) under normal attention levels. Driving Conditions and Usage Scenarios: For long-distance driving or nighttime driving conditions, it is recommended to refer to the upper limit of the eye-closing distribution before "decreased alertness". For specific high-risk positions, a more conservative reference standard can be considered. Adjustment and Follow-up: Initially, minor adjustments can be made through online calibration and user experience feedback. To balance comfort and protection. Recommended values: Common engineering recommendations: Use a 10-second cycle; generally recommended. The recommended time is between 0.5 and 1.5 seconds. For example, a healthy adult driver normally closes their eyes for about 0.2 to 0.6 seconds every 10 seconds. This increases to more than 1 second at night or during mild fatigue, therefore, a shorter time is recommended. Take 1.0 second. Adaptive / Personalized Extension: For users with personalized needs or special groups (such as older drivers or those who frequently change shifts), dynamic adaptive settings can be configured (e.g., the system automatically recommends the optimal setting based on the user's historical average eye-closing time). Actual value selection strategy: When initially setting values, it is recommended to refer to the average value of normal driving statistics plus 1 standard deviation to avoid excessive false alarms; subsequent adjustments can be made periodically based on operational data.

[0027] Calculate the fatigue index at the current time step: ;in, For the first Fatigue index of time step; like Then let ;like If the condition is not met, the adjustment will be terminated and an alarm signal will be issued. If due to data acquisition error Then, the time window will be truncated based on its upper bound. ; Constructing an emergency fatigue threshold .

[0028] The process of collecting and normalizing fatigue indicators has been refined, achieving highly reliable fatigue assessment through clear steps. First, a fixed period is set and a time series of adjustment is constructed, resolving the problem of disordered data acquisition timing caused by the lack of a periodic concept in existing systems. This allows subsequent analysis to unfold across discrete and uniform time grids, improving the comparability of the indicators. Then, at each adjustment time point, the eye-tracking system is invoked to obtain the total closure duration of the previous period, avoiding misjudgments that are easily influenced by brief blinks when judging fatigue based solely on instantaneous frequency, ensuring the sensitivity of the indicators to cumulative fatigue. The ratio of the collected closure duration to the reference duration is processed to generate a dimensionless fatigue index, fundamentally solving the problem of inconsistent dimensions between different devices and drivers, enhancing the portability and versatility of the algorithm in various environments. For zero-value cases and cases with acquisition errors, the fatigue index is either set to zero or truncated, and an alarm is triggered when the reference duration is zero. This redundant judgment and protection mechanism effectively prevents system crashes caused by division by zero and data loss, making it more robust and reliable compared to the single-path processing in existing technologies. Finally, the emergency threshold is dynamically set based on the ratio of the time window to the reference duration, providing a real-time threshold for subsequent stiffness switching and safety control.

[0029] The calculation of the target stiffness for the current cycle based on the preset initial stiffness of the seat and the fatigue index, according to a monotonic mapping relationship, specifically includes: Set the initial stiffness of the seat, ensuring that the initial stiffness is positive and within the physically feasible range; A monotonic mapping relationship from fatigue index to target stiffness is established, wherein the relationship provides a target value of not less than the initial stiffness as the fatigue index increases and converges to a finite upper bound. Substituting the fatigue index of the current cycle into the aforementioned relationship yields the target stiffness of the current cycle. Record the target stiffness in the current cycle cache.

[0030] Further specific implementation steps include: Set the initial stiffness scalar of the seat as The unit is ; Based on fatigue indicators Calculate the target stiffness value at the current time step, specifically: ;in, For the first Target stiffness at the time step.

[0031] By establishing a "stiffness target determination" step, a preset initial stiffness is linked to the fatigue index through a monotonic mapping relationship, achieving refined adaptive human-machine interface characteristics. First, a positive initial stiffness is set to ensure the adjustment starting point aligns with the human comfort range, solving the problem of traditional "uniform settings" failing to accommodate individual differences. Then, the target stiffness is calculated based on the fatigue index, shifting the adjustment strategy from a "switch-on" to a "continuous" approach, avoiding discomfort caused by sudden stiffness changes when the fatigue threshold is triggered. The mapping relationship is characterized by "bounded ends and a smooth middle." When fatigue is mild, the stiffness adjustment is small, preserving ample comfort space and meeting the body's need for softness during mild fatigue. As fatigue worsens, the stiffness gradually increases to its upper limit, enhancing spinal and lumbar support and actively alleviating fatigue accumulation, overcoming the limitation of traditional systems failing to provide effective support during severe fatigue. This mapping strategy balances comfort and support, and allows for flexible adjustment of the initial stiffness and mapping curve according to different user groups or application scenarios. It is superior to existing methods of rotating multiple stiffness levels or frequent manual adjustments, improving user experience while reducing the frequency of manual intervention and maintenance costs.

[0032] The process of reading the current actual stiffness from the seat stiffness control system, constructing the difference between the target stiffness and the actual stiffness as the error, and forming a squared error metric based on this error specifically includes: Read the actual stiffness value at the start of the current cycle from the seat stiffness control system; The difference between the target stiffness and the actual stiffness is calculated as the stiffness error, and the positive and negative directions of the error are marked. A square-shaped error metric is constructed based on the stiffness error; The actual stiffness, target stiffness, and error metric are recorded simultaneously to form the state triplet for the current cycle.

[0033] Further specific implementation steps include: The actual output value of the current stiffness adjuster is read from the seat stiffness control system and recorded as follows: ;in, For the first The actual seat stiffness measured at the start of the time step; The error value between the target stiffness and the actual stiffness is constructed as follows: ;in, For the first Stiffness error at the time step; Construct the squared error loss function: ;in, For the first The loss function value at each time step.

[0034] Focusing on the "current stiffness and error construction" stage, this approach reads the actual stiffness and constructs the error and its metric, providing accurate real-time feedback for subsequent adaptive adjustment. First, the actual output value is obtained from the stiffness control system, directly reflecting the current state of the actuator and resolving the error accumulation problem caused by relying solely on theoretical targets or sensor estimations. Then, the difference between the target stiffness and the actual stiffness is calculated, explicitly defining the deviation as the error value, eliminating the information asymmetry of traditional methods that only focus on the target or the actual value. Furthermore, a squared error metric is introduced, giving higher weight to large deviations, allowing significant deviations to trigger larger-amplitude responses more quickly, avoiding slow convergence or prolonged oscillations caused by fixed strides in the same direction. In addition, the actual stiffness, target stiffness, and error metric are recorded simultaneously to form a state triplet, supporting accurate calculation of the subsequent stride factor and providing detailed system operation logs, which is helpful for later fault analysis and performance optimization. Compared to simple proportional error or switching error control in existing technologies, this scheme highlights the importance of abnormal deviations with a squared metric, effectively improving adjustment accuracy and convergence speed, while providing data assurance for system safety and stable operation.

[0035] The adjustment direction is determined by reducing the error. A stride factor that varies with the fatigue index and is less than the unit amplitude is constructed. The physical allowable range and bandwidth of the seat stiffness are preset. Initial stiffness range clamping rules are set, and emergency thresholds and protective stiffness for error handling are established. Specifically, this includes: The adjustment direction is determined based on the decreasing direction of the error measurement, so that the adjustment direction is pointed in the direction of reducing the error; Construct a step size factor that monotonically increases with the error magnitude and is less than one, and set the factor to zero when the error is zero; Set the physical minimum and maximum stiffness, and calculate the bandwidth; set the initial stiffness within the range of the minimum and maximum stiffness, and perform interval clamping if it exceeds the limit; A dimensionless step factor is constructed based on the bandwidth, so that the step size has a suppressive trend as the absolute value of the error increases, thus limiting the intensity of single-cycle adjustment. Set the error limit as the emergency threshold, and set the protection stiffness as the midpoint between the minimum stiffness and the maximum stiffness; Output adjustment direction, stride gene, dimensionless stride factor, emergency threshold, and protection stiffness.

[0036] Further specific implementation steps include: Calculate the partial derivative of the loss function with respect to the current stiffness. ; Let the negative gradient direction be the adjustment direction: ;in, For the first The adjustment direction vector of the time step; Constructing normalized step basis functions ;in, For the first The step size basis function for the time step; Follow Monotonically increasing, and because ,have ; The preset allowable range of seat stiffness is a fixed constant range: ;in, , These are the minimum and maximum allowable stiffness, respectively; Therefore, the stiffness range bandwidth is set. ; System deployment settings ; If the above interval is empty or does not meet the requirements, then direct clamping of the physical interval will be used: ; Constructing a dimensionless step function: , ;in, For the first The actual stride factor of the time step; when hour, ; Set the emergency error threshold to ; Set the protection stiffness to .

[0037] In the "Adjustment Direction and Step Size Setting" stage, a multi-level adaptive control strategy is proposed. Through multiple mechanisms including error guidance, bandwidth limitation, step size function, and protection threshold, the stability and robustness of the adjustment are comprehensively improved. First, the adjustment direction is determined based on the direction of error reduction, ensuring that each adjustment step is directed towards error convergence. This solves the problem of traditional methods failing to distinguish between increasing and decreasing stiffness, ensuring that the adjustment is always effective. Then, a step size gene that varies with the error amplitude is constructed, making the step size strongly correlated with the deviation. This allows for automatic adjustment of the step size according to different deviation magnitudes, fully balancing response speed and adjustment smoothness, overcoming the drawbacks of fixed step sizes, such as slow convergence with large deviations and excessive oscillations with small deviations. Bandwidth detection within physically permissible ranges and initial stiffness interval clamping rules are introduced, allowing for parameter validity verification during system startup or deployment. This avoids execution risks caused by parameter drift or exceeding limits during unattended operation. Furthermore, an error emergency threshold and protection stiffness are set. When the error exceeds the limit or enters emergency mode, the system can quickly switch to protection stiffness, preventing long-term error accumulation or safety hazards caused by unforeseen circumstances. These innovations effectively combine fault self-protection, boundary constraints, and adaptive adjustment, improving system reliability and operational safety.

[0038] The intermediate stiffness value without boundary projection is calculated according to the adjustment direction and step factor, and the current actual stiffness remains unchanged when the error is zero. Specifically, this includes: A linear update is performed on the current actual stiffness according to the adjustment direction and the dimensionless step factor to obtain the unprojected intermediate stiffness value. When the error is zero, the intermediate stiffness is kept consistent with the current actual stiffness; Write the intermediate stiffness into the cache and wait for interval projection.

[0039] Further specific implementation steps include: According to the current adjustment direction with dimensionless stride Calculate the stiffness adjustment value before the update (unprojected): ;in, This represents the intermediate stiffness value to be applied in the next time step, but before boundary projection; if ,but .

[0040] A concise and efficient linear update strategy is adopted for the "intermediate stiffness calculation" stage, achieving a balance between accuracy and continuity. The intermediate stiffness is directly calculated using the aforementioned adjustment direction and adaptive step factor, avoiding computational delays and accumulated errors caused by complex recursive or integral processes. This ensures that stiffness adjustments within each adjustment cycle are based on the latest state. Simultaneously, the stiffness is explicitly kept constant when the error is zero, eliminating unnecessary calculations and instruction issuance under indiscriminate conditions, reducing system power consumption and actuator wear. Furthermore, the intermediate stiffness is written to a cache, providing intermediate data support for interval projection and fault diagnosis, enhancing process transparency and traceability. Compared to the common "direct projection followed by calculation" or "global optimization" strategies in existing technologies, this scheme performs calculations in segments, simplifying the control logic while ensuring computational accuracy. This provides a reliable and complete numerical foundation for subsequent projection stages, improving overall adjustment efficiency and system stability.

[0041] The step of performing interval projection of the intermediate stiffness within the physically permissible range to obtain the updated stiffness of the current period as the final feasible stiffness specifically includes: Confirm the minimum and maximum stiffness within the physically permissible range; Establish interval projection rules: take the minimum stiffness for values ​​less than the minimum stiffness, keep the values ​​within the range unchanged, and take the maximum stiffness for values ​​greater than the maximum stiffness. The intermediate stiffness is projected according to the interval projection rule to obtain the updated stiffness, which is used as the final stiffness that can be executed in this cycle. Write the updated stiffness to the execution cache and record a projection event for out-of-bounds cases.

[0042] Further specific implementation steps include: Set stiffness adjustment allowable range ; Constructing stiffness value projection function Specifically: ;in, As the independent variable; Calculated stiffness value The updated stiffness is obtained by inputting the projection function: ;in, This is the final feasible stiffness before execution.

[0043] The "interval projection and stiffness update" process was standardized, introducing a simple and clear projection function to strictly constrain intermediate stiffness within the physically permissible range, ensuring that the final executed stiffness never exceeds the mechanism's limits or falls below the safety lower limit. This projection strategy handles all overflow cases through unified rules, avoiding control confusion and execution delays caused by multiple condition judgments and branching logic in existing technologies. Simultaneously, logging is performed at projection events, providing detailed evidence for equipment maintenance and fault analysis, improving system maintainability and safety inspection efficiency. Furthermore, the clear projection rules make the control algorithm easy to verify and authenticate, meeting the high reliability requirements of engineering applications. Compared to the traditional approach of only taking emergency stops or triggering manual alarms in case of faults, this solution embeds boundary protection within the normal process, reducing the risk of hardware damage and system downtime. All these improvements ensure that stiffness updates accurately match requirements while always remaining within safe and permissible ranges, effectively extending the lifespan of the actuator and improving ride comfort.

[0044] The process of mapping updated stiffness to target instructions for the actuator, projecting within the allowable instruction range, generating a time-continuous linear scheduling trajectory within an adjustment cycle, executing based on the rate reachability check result or recalculating the trajectory according to the maximum reachable rate, recording maintenance events when the mapping relationship is unavailable, and issuing instructions at the end of the cycle and entering the next cycle specifically includes: Calibration phase: Two different stiffness calibration points and their corresponding actuator control command values ​​are selected to form two sets of reachable calibration pairs; Based on the two sets of calibration points, linear mapping parameters are calculated to obtain a one-to-one linear relationship between stiffness and control commands. The shortest full-stroke response time is measured under the same deployment conditions, and the maximum achievable average command change rate of the equipment is calculated accordingly. Online mapping and limitations: When any emergency condition is met, i.e. the fatigue index is not lower than the emergency threshold or the absolute value of the error exceeds the error limit, the update stiffness will be reset to the protection stiffness. The updated stiffness is mapped to the target command according to the linear relationship, and the target command is subjected to interval constraints to obtain the target command within the interval; Trajectory generation and rate verification: Read the actual instruction value at the start of the current cycle and execute the interval limit; Generate a linear time scheduling trajectory from the starting actual instruction to the target instruction within the interval in the current cycle time domain, and obtain the execution instruction at the end of the cycle; Calculate the average instruction rate required for this cycle; when the required rate is not greater than the maximum achievable average rate of the device, keep the linear trajectory unchanged; When the required rate exceeds the maximum achievable average rate of the device, the achievable end-of-cycle instruction is calculated based on the difference between the target and the current instruction, using the maximum achievable average rate of the device. An interval limit is applied to this instruction, and the linear time scheduling trajectory is regenerated with the new end-of-cycle instruction. Stiffness prediction and anomaly labeling: Based on the known stiffness-command linear relationship, the time scheduling trajectory is mapped to the predicted stiffness trajectory; When the linear relationship degenerates and the stiffness becomes uncontrollable by the command, the predicted stiffness is set to the current actual stiffness and the maintenance event is recorded. Command issuance and cycle rotation: Continuous time-varying instructions are issued to the actuator controller according to the aforementioned time scheduling trajectory; At the end of the current cycle, the actuator reaches the end-of-cycle instruction and then enters the next adjustment cycle.

[0045] Further specific implementation steps include: S701, Calibration point determination mapping parameters: During the system deployment phase, select two sets of reachable calibration pairs. and ;in, These are two different stiffness calibration points; , The numerical value of the control command executed by the mechanism is denoted by unit. For example, current or displacement ; Calculate the linear mapping parameters: , ;in, The linear slope from stiffness to command; The intercept of the linear relationship between stiffness and command; The shortest full-stroke response time was measured under the same deployment conditions. Based on this, the maximum achievable average speed of the device is given as follows: ;in, To calibrate the shortest full-stroke response time obtained from the experiment; The maximum average rate of change of instructions achievable by the device; S702, Target Instruction Calculation and Instruction Range Limitation: Before instruction mapping, if any of the following emergency conditions are met: Then first Then Mapped to the target instruction of the actuator: ;in, The target instruction value for the next time step; Set the allowable range of actuator instructions to and satisfy Construct the instruction projection function: The target instruction within the interval is obtained: ;in, These represent the lower and upper physical limits of the command, respectively, in units of... ; For the interval projection function of the instruction quantity, As the independent variable; The target instruction is within the interval; S703, Generation of time-scheduled trajectory within a cycle: Read the actual instruction value of the current time step start point And perform interval projection: ;in, For the first The current actual instruction value at the time step; In the interval Internally generated linear time scheduling trajectory: ;in, The instruction trajectory is continuous in time; independent variable For continuous time; Therefore, the execution instructions at the end of the cycle can be obtained: ; S704, Rate Reachability Check and Branch Processing: Calculate the average rate required for this period: ;in, This represents the average instruction rate required in this cycle. Set symbol function ; like Then the trajectory obtained in step S703 is retained. Execute unchanged; like The endpoint can be reached under computational speed limitations: And perform interval projection to obtain new cycle end instructions. Subsequently, in the same form as S703, with the updated Recalculate And execute; S705, Predicting Stiffness Trajectory and Degradation Boundary: like The corresponding predicted stiffness trajectory can be obtained from the command trajectory: ;in, The predicted stiffness trajectory is calculated based on the known command trajectory; If it appears Then the mapping pair If uncontrollable, record and set a maintenance flag as follows: and report Maintenance events; S706, Command issuance and execution completed: Will As a time-varying instruction within the cycle, it is issued to the actuator controller; exist At that time, the implementing agency reached Then it enters the next cycle.

[0046] In the "command mapping, scheduling, and execution" phase, a complete closed-loop execution mechanism was constructed, encompassing calibration, mapping, trajectory generation, and reachability verification. This comprehensively addresses the discrepancies between stiffness and executed commands, as well as the limitations on speed. Firstly, during the deployment phase, a linear mapping parameter from stiffness to commands is generated through two-point calibration to ensure command output accuracy. Under the same operating conditions, the shortest response time is measured, and the maximum achievable average speed is calculated, laying the foundation for dynamic trajectory generation and speed verification. For emergency conditions triggered by fatigue or significant deviations, the system can automatically reset stiffness to the protection value and remap commands, promptly switching to a safe mode to avoid the dangers of uncontrolled execution. During linear time scheduling trajectory generation, a smooth trajectory is generated using continuous-time interpolation by reading the current actual command and the target command, eliminating the mechanical shocks caused by conventional discrete-time pre-step control. Subsequently, a comparison verification between the average speed and the maximum achievable speed is performed. When limits are exceeded, the trajectory is dynamically recalculated to ensure a smooth transition within the hardware's capabilities. Finally, maintenance events are recorded when mapping fails or degrades, enhancing the system's self-diagnostic capabilities and providing empirical data for subsequent maintenance. Compared with the traditional simple control method that only performs a single mapping or does not perform rate verification, this solution improves execution controllability and reliability, ensuring that the seat stiffness adjustment can meet comfort requirements and strictly comply with mechanical performance constraints and safety specifications throughout its entire life cycle.

[0047] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0048] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for dynamically adjusting the stiffness of an intelligent seat during fatigued driving, characterized in that, include: A fixed adjustment cycle is established, and raw data such as eyelid closure duration are collected from the eye movement sensing system at each cycle. Time window constraints, outlier truncation and dimension unification are performed on the data to obtain a dimensionless fatigue index, and an emergency threshold related to the reference duration is established. Based on the preset initial stiffness of the seat and the fatigue index, the target stiffness of the current cycle is calculated according to the monotonic mapping relationship. The mapping provides a target stiffness that does not decrease and remains bounded as the fatigue exponent increases. The current actual stiffness is read from the seat stiffness control system, the difference between the target stiffness and the actual stiffness is constructed as the error, and a squared error metric is formed based on this. Using the direction of error reduction as the adjustment direction, a stride factor that varies with the fatigue index and is less than the unit amplitude is constructed. The physical allowable range and bandwidth of seat stiffness are preset, the range clamping rules of the initial stiffness are set, and the emergency threshold of error and the protective stiffness used for handling are set. The intermediate stiffness value without boundary projection is calculated according to the adjustment direction and step factor, and the current actual stiffness remains unchanged when the error is zero. The intermediate stiffness is projected within the physically permissible range to obtain the updated stiffness for the current period as the final feasible stiffness. The updated stiffness is mapped to the target command of the actuator, projected within the allowable range of the command, and a time-continuous linear scheduling trajectory is generated within an adjustment cycle. After the rate reachability test, the trajectory is executed or recalculated according to the test result. When the mapping relationship is unavailable, maintenance events are recorded. At the end of the cycle, the command is issued and the next cycle begins.

2. The method for dynamically adjusting the stiffness of an intelligent seat in a fatigued driving state according to claim 1, characterized in that, The process involves establishing a fixed adjustment cycle, collecting raw data such as eyelid closure duration from the eye-tracking system at each cycle point, applying time window constraints, outlier truncation, and dimensionless processing to the data to obtain a dimensionless fatigue index, and establishing an emergency threshold related to the reference duration. Specifically, this includes: Set a fixed adjustment period time interval and construct a discrete adjustment time series; At each adjustment moment, the cumulative duration of eyelid closure in the previous cycle is obtained by the eye-tracking sensor system as the original fatigue quantity; Set a closed reference duration and obtain the dimensionless fatigue index by the ratio of the original quantity to the reference duration. When the cumulative duration of the previous cycle closure is zero, the fatigue index is set to zero; when the reference duration is zero, the adjustment is terminated and an alarm is triggered. When the acquisition error causes the cumulative duration of closure to exceed the upper limit of the time window, the time window is truncated according to the upper limit, and the fatigue index is recalculated according to the truncation result. The emergency fatigue threshold is set based on the ratio of the time window to the reference duration; Write the fatigue index and emergency fatigue threshold into the calculation cache of the current cycle.

3. The method for dynamically adjusting the stiffness of an intelligent seat in a fatigued driving state according to claim 2, characterized in that, The calculation of the target stiffness for the current cycle based on the preset initial stiffness of the seat and the fatigue index, according to a monotonic mapping relationship, specifically includes: Set the initial stiffness of the seat, ensuring that the initial stiffness is positive and within the physically feasible range; A monotonic mapping relationship from fatigue index to target stiffness is established, wherein the relationship provides a target value of not less than the initial stiffness as the fatigue index increases and converges to a finite upper bound. Substituting the fatigue index of the current cycle into the aforementioned relationship yields the target stiffness of the current cycle. Record the target stiffness in the current cycle cache.

4. The method for dynamically adjusting the stiffness of an intelligent seat in a fatigued driving state according to claim 3, characterized in that, The process of reading the current actual stiffness from the seat stiffness control system, constructing the difference between the target stiffness and the actual stiffness as the error, and forming a squared error metric based on this error specifically includes: Read the actual stiffness value at the start of the current cycle from the seat stiffness control system; The difference between the target stiffness and the actual stiffness is calculated as the stiffness error, and the positive and negative directions of the error are marked. A square-shaped error metric is constructed based on the stiffness error; The actual stiffness, target stiffness, and error metric are recorded simultaneously to form the state triplet for the current cycle.

5. The method for dynamically adjusting the stiffness of an intelligent seat in a fatigued driving state according to claim 4, characterized in that, The adjustment direction is determined by reducing the error. A stride factor that varies with the fatigue index and is less than the unit amplitude is constructed. The physical allowable range and bandwidth of the seat stiffness are preset. Initial stiffness range clamping rules are set, and emergency thresholds and protective stiffness for error handling are established. Specifically, this includes: The adjustment direction is determined based on the decreasing direction of the error measurement, so that the adjustment direction is pointed in the direction of reducing the error; Construct a step size factor that monotonically increases with the error magnitude and is less than one, and set the factor to zero when the error is zero; Set the physical minimum and maximum stiffness, and calculate the bandwidth; set the initial stiffness within the range of the minimum and maximum stiffness, and perform interval clamping if it exceeds the limit; A dimensionless step factor is constructed based on the bandwidth, so that the step size has a suppressive trend as the absolute value of the error increases, thus limiting the intensity of single-cycle adjustment. Set the error limit as the emergency threshold, and set the protection stiffness as the midpoint between the minimum stiffness and the maximum stiffness; Output adjustment direction, stride gene, dimensionless stride factor, emergency threshold, and protection stiffness.

6. The method for dynamically adjusting the stiffness of an intelligent seat in a fatigued driving state according to claim 5, characterized in that, The intermediate stiffness value without boundary projection is calculated according to the adjustment direction and step factor, and the current actual stiffness remains unchanged when the error is zero. Specifically, this includes: A linear update is performed on the current actual stiffness according to the adjustment direction and the dimensionless step factor to obtain the unprojected intermediate stiffness value. When the error is zero, the intermediate stiffness is kept consistent with the current actual stiffness; Write the intermediate stiffness into the cache and wait for interval projection.

7. The method for dynamically adjusting the stiffness of an intelligent seat in a fatigued driving state according to claim 6, characterized in that, The step of performing interval projection of the intermediate stiffness within the physically permissible range to obtain the updated stiffness of the current period as the final feasible stiffness specifically includes: Confirm the minimum and maximum stiffness within the physically permissible range; Establish interval projection rules: take the minimum stiffness for values ​​less than the minimum stiffness, keep the values ​​within the range unchanged, and take the maximum stiffness for values ​​greater than the maximum stiffness. The intermediate stiffness is projected according to the interval projection rule to obtain the updated stiffness, which is used as the final stiffness that can be executed in this cycle. Write the updated stiffness to the execution cache and record a projection event for out-of-bounds cases.

8. The method for dynamically adjusting the stiffness of an intelligent seat in a fatigued driving state according to claim 7, characterized in that, The process of mapping updated stiffness to target instructions for the actuator, projecting within the allowable instruction range, generating a time-continuous linear scheduling trajectory within an adjustment cycle, executing based on the rate reachability check result or recalculating the trajectory according to the maximum reachable rate, recording maintenance events when the mapping relationship is unavailable, and issuing instructions at the end of the cycle and entering the next cycle specifically includes: Calibration phase: Two different stiffness calibration points and their corresponding actuator control command values ​​are selected to form two sets of reachable calibration pairs; Based on the two sets of calibration points, linear mapping parameters are calculated to obtain a one-to-one linear relationship between stiffness and control commands. The shortest full-stroke response time is measured under the same deployment conditions, and the maximum achievable average command change rate of the equipment is calculated accordingly. Online mapping and limitations: When any emergency condition is met, i.e. the fatigue index is not lower than the emergency threshold or the absolute value of the error exceeds the error limit, the update stiffness will be reset to the protection stiffness. The updated stiffness is mapped to the target command according to the linear relationship, and the target command is subjected to interval constraints to obtain the target command within the interval; Trajectory generation and rate verification: Read the actual instruction value at the start of the current cycle and execute the interval limit; Generate a linear time scheduling trajectory from the starting actual instruction to the target instruction within the interval in the current cycle time domain, and obtain the execution instruction at the end of the cycle; Calculate the average instruction rate required for this cycle; when the required rate is not greater than the maximum achievable average rate of the device, keep the linear trajectory unchanged; When the required rate exceeds the maximum achievable average rate of the device, the achievable end-of-cycle instruction is calculated based on the difference between the target and the current instruction, using the maximum achievable average rate of the device. An interval limit is applied to this instruction, and the linear time scheduling trajectory is regenerated with the new end-of-cycle instruction. Stiffness prediction and anomaly labeling: Based on the known stiffness-command linear relationship, the time scheduling trajectory is mapped to the predicted stiffness trajectory; When the linear relationship degenerates and the stiffness becomes uncontrollable by the command, the predicted stiffness is set to the current actual stiffness and the maintenance event is recorded. Command issuance and cycle rotation: Continuous time-varying instructions are issued to the actuator controller according to the aforementioned time scheduling trajectory; At the end of the current cycle, the actuator reaches the end-of-cycle instruction and then enters the next adjustment cycle.

Images