A vehicle control method, apparatus, device, and medium

Abstract

The application discloses a vehicle control method, device, equipment and medium, comprising: decoupling the visual micro-motion signal from the vehicle body vibration, checking the periodic stability and phase coupling of the processed signal, and analyzing the signal passing the check to obtain a pure vital sign vector; calculating the waveform entropy and change gradient of the pure vital sign vector, inputting an evolution prediction model to obtain an evolution track, determining a physiological abnormality type based on the waveform entropy and change gradient, and determining the advance amount of entering an irreversible disability stage based on the evolution track and disability threshold state; inputting the abnormal probability, change gradient, waveform entropy change rate and typing weight into a risk energy accumulation function, introducing an active intervention intensity to construct an exponential annealing inhibition factor to dynamically update the risk energy, and triggering a takeover instruction when the threshold is exceeded, so as to accurately extract physiological signals, realize forward-looking prediction and provide a response time for takeover, realize smooth adjustment of the takeover willingness with the operation intensity of the driver, and guarantee driving safety in case of sudden disability.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of automotive technology, and in particular to a vehicle control method, device, equipment and medium. Background Technology

[0002] With the rapid development of intelligent vehicle technology and active safety systems, driver status monitoring and autonomous vehicle decision-making and control have become core issues in ensuring driving safety. Especially in manual driving mode, how to effectively deal with sudden physiological incapacitation of the driver (such as cardiac arrest, respiratory failure, deep coma, and other extreme scenarios) and achieve safe takeover and risk control of the vehicle is a key challenge that urgently needs to be overcome in the field of automotive safety technology.

[0003] Currently, existing technologies primarily capture driver facial features (such as the frequency of eye closure and the number of yawns) using in-vehicle cameras and combine this with physiological sensors (such as millimeter-wave radar and infrared sensors) to detect physiological indicators such as heart rate and respiration. Based on preset instantaneous thresholds, statistical determinations of fatigue or abnormal states are then made. Regarding minimum-risk maneuvers, existing solutions are typically applied in autonomous driving mode, focusing on searching for parking spots based on the vehicle's own state and a static environment map, and prioritizing them according to static risk.

[0004] However, the aforementioned existing technologies still have the following shortcomings. First, the statistical judgment based on instantaneous thresholds in existing solutions is a post-event detection mechanism, which cannot proactively predict the evolution trend of the driver's physiological indicators and is difficult to provide early warning before the driver's physiological functions are lost. Second, the vibration environment inside the vehicle severely interferes with the purity of visual micro-motion signals, and existing technologies lack effective vibration decoupling and signal authenticity verification mechanisms, making it difficult to guarantee the reliability of the extracted physiological signals. Third, the monitoring of the driver's physiological state and vehicle motion control are isolated from each other, lacking a mechanism to dynamically map physiological risk perception to vehicle control switching, and also failing to consider the impact of the driver's active intervention on the system takeover decision, thus failing to achieve a complete safety closed loop from physiological state perception to risk prediction and assessment, and then to autonomous takeover control. Summary of the Invention

[0005] This invention provides a vehicle control method, apparatus, device, and medium that can improve the reliability of physiological signal extraction to enable forward prediction, provide response time for safe takeover, and achieve smooth adjustment of the system's takeover intention according to the intensity of the driver's operation, thereby ensuring driving safety when the driver suddenly becomes incapacitated in manual driving mode.

[0006] In a first aspect, embodiments of the present invention provide a vehicle control method, comprising: The driver's visual micro-motion signals and vehicle driving status data are acquired. Based on the driving status data, the visual micro-motion signals are decoupled from the vehicle vibration. The periodic stability and phase coupling of the processed signals are verified by a physiological consistency constraint model. The verified signals are then analyzed to obtain the driver's pure vital sign vector. The waveform entropy and gradient of the pure vital signs vector within a continuous time window are calculated, and the waveform entropy and gradient are input into a preset evolution prediction model to obtain the evolution trajectory output by the model. Based on the waveform entropy and gradient, a preset abnormal classification dictionary is matched to determine the current physiological abnormality type of the driver. Based on the evolution trajectory and the preset disability threshold state, the advance amount before the driver enters the irreversible disability stage is determined. The abnormal probability corresponding to the physiological abnormality type, the change gradient, the rate of change of the waveform entropy, and the subtyping weight corresponding to the physiological abnormality type are input into a preset driver risk energy accumulation function. The driver's active intervention intensity is introduced to construct an exponential annealing suppression factor to dynamically update the risk energy. When the updated risk energy exceeds a preset safety threshold, the driver is determined to be incapacitated and an automatic vehicle takeover command is triggered.

[0007] Optionally, the method further includes: invoking a surround-view perception sensor in real time to acquire motion vectors of surrounding traffic participants; calculating the residual environmental risk entropy of the vehicle when adopting different candidate behaviors based on the motion vectors and a preset dynamic game risk assessment function, selecting the candidate behavior corresponding to the optimal solution of the residual environmental risk entropy as the minimum risk maneuver behavior, and controlling the vehicle to execute the minimum risk maneuver behavior; during the execution of the minimum risk maneuver behavior, marking the time anchor point that triggers the automatic takeover command of the vehicle, and simultaneously solidifying the physiological evolution evidence chain of the driver before disability, the environmental perception decision data frame, and the control execution log, so as to achieve a safety closed loop and responsibility traceability.

[0008] Optionally, the method further includes: performing time-frequency transformation on the signal after vehicle vibration decoupling processing, extracting the peak frequency within each time window, calculating the fluctuation value of each peak frequency relative to the average frequency, and determining that the signal satisfies periodic stability if the fluctuation value is less than a first preset threshold; extracting the respiratory phase and heartbeat phase in the signal, calculating the phase lock value between the respiratory phase and the heartbeat phase, and determining that the signal satisfies phase coupling if the phase lock value is greater than a second preset threshold; when the signal simultaneously satisfies periodic stability and phase coupling, determining that the signal is a real physiological oscillation signal and retaining it, otherwise discarding the signal data within the current time period.

[0009] Optionally, the method further includes: dividing a continuous time axis into multiple sliding time windows of fixed length; normalizing the pure vital sign vector within each sliding time window, constructing a probability distribution based on the absolute value of the normalized amplitude, and calculating the waveform entropy based on the probability distribution; the waveform entropy is used to characterize the stability of the vital sign signal within the time window; calculating the change gradient by dividing the difference between the pure vital sign vector at the start time and the pure vital sign vector at the end time of each sliding time window by the length of the sliding time window; the change gradient is used to characterize the dynamic change trend of the vital sign signal.

[0010] Optionally, the method further includes: multiplying the abnormal probability corresponding to the physiological abnormality type, the change gradient, the rate of change of the waveform entropy, and the subtyping weight corresponding to the physiological abnormality type by their respective preset risk gain weight coefficients and summing them to obtain a risk energy gain term; obtaining the driver's active intervention intensity; the active intervention intensity is determined based on the driver's current operating force on the vehicle control device; constructing an exponential annealing suppression factor with a natural constant as the base and the product of a negative exponential annealing coefficient and the active intervention intensity as the exponential annealing suppression factor; the exponential annealing suppression factor decreases as the active intervention intensity increases; multiplying the risk energy gain term by the exponential annealing suppression factor to obtain a risk energy increment; and adding the risk energy at the current moment to the risk energy increment to obtain the updated risk energy at the next moment.

[0011] Optionally, the method further includes: for each candidate behavior, predicting multiple possible behaviors of each surrounding traffic participant in the future time domain based on the motion vector of the surrounding traffic participants, combining the multiple possible behaviors of each surrounding traffic participant to obtain multiple candidate behavior combinations, and determining the probability of occurrence of each candidate behavior combination; For each candidate behavior combination, calculate the collision risk value, relative speed risk value, and road boundary risk value when the vehicle performs the current candidate behavior and the surrounding traffic participants adopt the current candidate behavior combination, and calculate the comprehensive risk value by weighted summation of the collision risk value, the relative speed risk value, and the road boundary risk value; Based on the occurrence probability of each candidate behavior combination and the corresponding comprehensive risk value, the residual entropy of environmental risk under the current candidate behavior is calculated; the residual entropy of environmental risk is positively correlated with the weighted sum of the entropy value of the occurrence probability and the comprehensive risk value; Iterate through all candidate behaviors and calculate the corresponding environmental risk residual entropy for each.

[0012] Secondly, embodiments of the present invention provide a vehicle control device, comprising: The multi-source signal perception and physiological verification module is used to acquire the driver's visual micro-motion signals and the vehicle's driving status data, and to perform vehicle vibration decoupling processing on the visual micro-motion signals based on the driving status data. The module also verifies the periodic stability and phase coupling of the processed signals through a physiological consistency constraint model, and analyzes the verified signals to obtain the driver's pure vital sign vector. The physiological state evolution trend prediction module is used to calculate the waveform entropy and change gradient of the pure vital sign vector within a continuous time window, input the waveform entropy and change gradient into a preset evolution prediction model to obtain the evolution trajectory output by the model, and determine the current physiological abnormality type of the driver based on the waveform entropy and change gradient matched with a preset abnormality classification dictionary, and determine the advance amount of the driver entering the irreversible disability stage based on the evolution trajectory and a preset disability threshold state. The nonlinear risk energy assessment and takeover decision module is used to input the abnormal probability corresponding to the physiological abnormality type, the change gradient, the change rate of the waveform entropy, and the classification weight corresponding to the physiological abnormality type into a preset driver risk energy accumulation function, and to introduce the driver's active intervention intensity to construct an exponential annealing suppression factor to dynamically update the risk energy. When the updated risk energy exceeds a preset safety threshold, the module determines that the driver is incapacitated and triggers an automatic vehicle takeover command.

[0013] Thirdly, embodiments of the present invention provide an electronic device, the electronic device comprising: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the vehicle control method provided in any embodiment of the present invention.

[0014] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the vehicle control method as provided in any embodiment of the present invention.

[0015] Fifthly, embodiments of the present invention provide a computer program product, including a computer program that, when executed by a processor, implements the vehicle control method provided in any embodiment of the present invention.

[0016] The technical solution of this invention acquires the driver's visual micro-motion signals and vehicle driving status data, and performs vehicle vibration decoupling processing on the visual micro-motion signals based on the driving status data. The periodic stability and phase coupling of the processed signals are verified using a physiological consistency constraint model, and the verified signals are analyzed to obtain the driver's pure vital sign vector. Vehicle vibration decoupling processing eliminates interference from the vehicle vibration environment on the signals, and dual verification of periodic stability and phase coupling ensures the reliability and purity of the extracted physiological signals, providing high-quality input data for subsequent processing. The waveform entropy and gradient of the pure vital sign vector within a continuous time window are calculated, and the waveform entropy and gradient are input into a preset evolution prediction model to obtain the evolution trajectory output by the model. Based on the waveform entropy and gradient, a preset anomaly classification dictionary is matched to determine the current physiological abnormality type of the driver. Based on the evolution trajectory and a preset disability threshold state, the advance amount before the driver enters the irreversible disability stage is determined. By quantifying the signal stability through waveform entropy and the deterioration rate through gradient, a forward-looking prediction is achieved in combination with the evolution prediction model. Accurate classification is achieved through the anomaly classification dictionary. The precise advance amount is calculated through the evolution trajectory and the disability threshold state, providing an early warning window for safe takeover. The abnormal probability corresponding to the physiological abnormality type, the change gradient, the rate of change of the waveform entropy, and the subtyping weight corresponding to the physiological abnormality type are input into a preset driver risk energy accumulation function. An exponential annealing suppression factor is constructed by introducing the driver's active intervention intensity to dynamically update the risk energy. When the updated risk energy exceeds a preset safety threshold, the driver is determined to be incapacitated and an automatic vehicle takeover command is triggered. By quantitatively mapping multi-dimensional physiological risks to a unified risk energy, combined with the exponential annealing suppression factor, the system's takeover intention is smoothly adjusted according to the driver's operation intensity. When the risk energy exceeds the safety threshold, takeover is automatically triggered, thereby ensuring driving safety when the driver suddenly becomes incapacitated in manual driving mode.

[0017] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1This is a flowchart of a vehicle control method provided in Embodiment 1 of the present invention; Figure 2 This is a flowchart of a vehicle control method provided in Embodiment 2 of the present invention; Figure 3 This is a schematic diagram of the structure of a vehicle control device provided in Embodiment 3 of the present invention; Figure 4 This is a schematic diagram of the structure of an electronic device that implements the vehicle control method of this invention. Detailed Implementation

[0020] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.

[0021] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0022] Example 1 Figure 1 This is a flowchart of a vehicle control method provided in Embodiment 1 of the present invention. This embodiment is applicable to situations where the driver experiences sudden physiological incapacitation in manual driving mode and the vehicle needs to take over autonomously. This method can be executed by a vehicle control device, which can be implemented in hardware and / or software and can be configured in an electronic device. For example... Figure 1 As shown, the method includes: S110. Acquire the driver's visual micro-motion signal and the vehicle's driving status data, and perform vehicle vibration decoupling processing on the visual micro-motion signal based on the driving status data. Verify the periodic stability and phase coupling of the processed signal through a physiological consistency constraint model, and analyze the verified signal to obtain the driver's pure vital sign vector.

[0023] In this embodiment of the disclosure, the driver can refer to a natural person who is in the driver's seat of the vehicle and is operating or preparing to operate the vehicle. Visual micro-motion signals can refer to weak motion signals captured by an image acquisition device, generated by changes in skin color difference and chest displacement caused by breathing and heartbeat in the driver's facial area. For example, a near-infrared camera can capture weak motion signals generated by changes in skin color difference and chest displacement caused by breathing and heartbeat in areas of interest such as the bridge of the nose, forehead, and upper edge of the collarbone. The vehicle can refer to a land-based transportation vehicle equipped with manual driving mode and / or automatic driving mode, such as a passenger car or commercial vehicle. Driving state data can refer to information reflecting the current motion state of the vehicle. Driving state data may include, but is not limited to, vehicle vibration base disturbances collected by the vehicle body inertial measurement unit and vehicle acceleration signals collected by the vehicle speed and acceleration information acquisition unit.

[0024] In this embodiment, vehicle vibration decoupling processing can refer to a method of removing low-frequency resonance interference components caused by vehicle road surface excitation from the original visual micro-motion signal by constructing a vibration decoupling model based on vehicle driving state data. For example, vehicle driving state data can be vehicle acceleration signals. The physiological consistency constraint model can refer to a mathematical model that uses prior knowledge of the inherent periodic stability and phase coupling of physiological signals such as human respiration and heartbeat to verify the authenticity of the extracted signal. The processed signal can refer to the signal obtained after vehicle vibration decoupling processing, which removes the low-frequency resonance interference caused by vehicle motion.

[0025] In this embodiment, periodic stability refers to the relatively stable periodic variation of physiological signals over time. This can be understood as the peak frequency of the signal fluctuating less than a preset threshold relative to the average frequency within a continuous time window. For example, the physiological signal can be a respiratory signal or a heartbeat signal. Phase coupling refers to the inherent phase modulation coupling relationship between the respiratory signal and the heartbeat signal. This can be understood as the lock-in value of the phase difference between the two being greater than a preset threshold. A verified signal can be a signal that simultaneously satisfies both the periodic stability condition and the phase coupling condition. A verified signal can be confirmed as originating from genuine physiological oscillations rather than random noise. A pure vital sign vector can be a combined vector of the driver's respiratory rate and heart rate obtained through bandpass filtering after the aforementioned vibration decoupling processing and physiological consistency verification.

[0026] Specifically, different types of real-time data are acquired using in-vehicle sensing modules. These modules include: a driver's face near-infrared camera unit, a steering wheel or seat micro-vibration sensing unit, a vehicle body inertial measurement unit, and a vehicle speed and acceleration information acquisition unit. The driver's face near-infrared camera unit captures subtle changes in skin color and micro-displacements of the chest. The steering wheel or seat micro-vibration sensing unit captures low-amplitude periodic movements induced by breathing and heartbeats. The vehicle body inertial measurement unit characterizes the vehicle body vibration base disturbance. These multiple signals are aligned using a unified timestamp.

[0027] A stable region of interest (ROI) is established on the driver's face, and the micro-displacement of each pixel is calculated using an optical flow algorithm or a phase difference amplification algorithm. The average value of the micro-displacements of all pixels within the ROI is taken to obtain the original visual micro-motion signal. Stable ROIs can be the bridge of the nose, forehead, or upper edge of the clavicle. Vehicle vibration decoupling processing is then performed. Since vehicle motion introduces low-frequency resonance interference, a vibration decoupling model is constructed to remove the noise influence caused by vehicle road excitation. For example, vehicle acceleration signals collected by the vehicle body inertial measurement unit and the vehicle speed and acceleration information acquisition unit are acquired, and the vibration transfer function is obtained by fitting using the least squares method. The original visual micro-motion signal is then subtracted from the convolution of the vibration transfer function and the vehicle acceleration signal to obtain the decoupled signal.

[0028] The decoupled signal is double-verified from two dimensions—periodic stability and phase coupling—using a physiological consistency constraint model. The verified real physiological oscillation signal is then separated by bandpass filtering, and the respiratory rate component and heart rate component are extracted and combined to form a pure vital sign vector.

[0029] This step effectively eliminates low-frequency interference from complex on-board vibration environments (such as road surface excitation and vehicle body resonance) on visual micro-motion signals through vehicle vibration decoupling. By employing a physiological consistency constraint model with dual verification of periodic stability and phase coupling, only signals that simultaneously meet both verification dimensions are recognized as genuine physiological oscillation signals and retained; otherwise, they are discarded. This mechanism significantly improves the reliability and purity of the extracted physiological signals, reduces the risk of misjudgment caused by vibration noise, and provides high-quality input data for subsequent prediction of physiological state evolution trends.

[0030] As an optional implementation of this disclosure, verifying the periodic stability and phase coupling of the processed signal through a physiological consistency constraint model may specifically include: performing time-frequency transformation on the signal after vehicle vibration decoupling processing, extracting the peak frequency within each time window, calculating the fluctuation value of each peak frequency relative to the average frequency, and determining that the signal satisfies periodic stability if the fluctuation value is less than a first preset threshold; extracting the respiratory phase and heartbeat phase from the signal, calculating the phase lock value between the respiratory phase and heartbeat phase, and determining that the signal satisfies phase coupling if the phase lock value is greater than a second preset threshold; when the signal simultaneously satisfies periodic stability and phase coupling, determining that the signal is a real physiological oscillation signal and retaining it, otherwise discarding the signal data within the current time period.

[0031] In this embodiment of the disclosure, time-frequency transformation can refer to an analysis method that converts a time-domain signal to a joint time-frequency domain. For example, short-time Fourier transform is used to observe the changes in the frequency components of a signal over time. A time window can refer to a finite-length signal segment selected when performing time-frequency analysis on a continuous signal. A peak frequency can refer to the frequency value corresponding to the point of maximum spectral amplitude after time-frequency transformation within a time window. An average frequency can refer to the average value of each peak frequency within multiple consecutive time windows. A fluctuation value can refer to a quantitative indicator of the degree of deviation of the peak frequency of each time window from the average frequency. The fluctuation value can be the average of the absolute values ​​of the differences between each peak frequency and the average frequency.

[0032] In this embodiment, the first preset threshold can refer to a pre-set critical value used to determine whether the stability of the signal cycle meets the requirements. When the fluctuation value is less than this threshold, it is determined to be a stable physiological cycle. The respiratory phase can refer to the cyclical position of the respiratory signal at a certain moment. The respiratory phase can be represented by an angle value (0 to 2π). The cyclical position can be the start of inspiration, the end of expiration, etc. The heartbeat phase can refer to the cyclical position of the heartbeat signal at a certain moment. The heartbeat phase can be represented by an angle value (0 to 2π). The cyclical position can be the peak position of the R wave.

[0033] In this embodiment, the phase lock value can be a quantitative indicator measuring whether a stable phase difference exists between the respiratory phase and the heartbeat phase. The phase lock value ranges from 0 to 1. The closer the phase lock value is to 1, the stronger the phase coupling between the two. The second preset threshold can be a pre-set critical value used to determine whether the signal phase coupling meets the requirements. When the phase lock value is greater than this threshold, it is determined that the signal originates from real physiological oscillations. Real physiological oscillation signals can refer to signals originating from the actual respiratory and heartbeat physiological activities of the human body, rather than environmental vibrations or random noise. The signal data within the current time period can refer to the continuous signal data of the segment currently being verified; if the verification fails, the entire segment is discarded.

[0034] Specifically, in terms of periodic stability verification, a short-time Fourier transform is performed on the signal after decoupling from vehicle vibration: Then, the peak frequency within each time window is extracted, and the fluctuation value of each peak frequency relative to the average frequency is calculated: If the fluctuation value is less than the first preset threshold ( If the signal satisfies periodic stability, then the signal is determined to be stable. Regarding phase coupling verification, respiration and heartbeat have a phase modulation coupling relationship; a phase-locked value is constructed as follows: .in, This is the respiratory phase. This refers to the heartbeat phase. If the phase lock value is greater than the second preset threshold ( If the signal satisfies both periodic stability and phase coupling, it is determined to be a genuine physiological oscillation and retained. Otherwise, the signal data for the current time period is discarded.

[0035] This optional implementation extracts peak frequencies and calculates fluctuation values ​​using short-time Fourier transform, achieving quantitative verification of the periodic stability of physiological signals. It also achieves quantitative verification of the phase coupling of physiological signals by calculating the phase lock values ​​between respiration and heartbeat. This dual verification mechanism, based on the inherent properties of physiological signals, effectively distinguishes between genuine physiological oscillations and random noise, significantly improving the accuracy of signal authenticity determination. When verification fails, the entire data segment is discarded, preventing unreliable signals from entering subsequent processing stages.

[0036] It should be noted that in the short-time Fourier transform formula, The time spectrum is obtained after short-time Fourier transform, representing the signal frequency. and time Energy distribution at a location. The denoised signal obtained after vehicle vibration decoupling processing is in time The amplitude at that point. A window function, used to capture the signal in time. Spectral analysis was performed on a local segment in the vicinity. For frequency variables, it represents a certain frequency component in the signal. The time integral variable represents the signal sampling time within the time range covered by the window function. The current analysis time represents the time position of the window function center. In the formula for peak frequency fluctuation, This represents the total number of time windows used to calculate the fluctuation value, i.e., the number of peak frequency samples included in the statistics. (Sum index) is the index number of the time window, with a value ranging from 1 to... , used to iterate through the peak frequencies of each time window. In the phase-locked value formula, N is used to calculate the total number of sampling points for the phase-locked value, i.e., the number of signal samples participating in the statistics. k is the index number of the sampling point, ranging from 1 to N, used to iterate through the phase values ​​of each sampling point. j is the imaginary unit, satisfying... , is used to represent real signals in complex form to calculate phase difference.

[0037] S120. Calculate the waveform entropy and gradient of the pure vital signs vector within a continuous time window, and input the waveform entropy and gradient into a preset evolution prediction model to obtain the evolution trajectory output by the model. Based on the waveform entropy and gradient, match the preset abnormal classification dictionary to determine the current physiological abnormality type of the driver, and determine the advance amount before the driver enters the irreversible disability stage based on the evolution trajectory and the preset disability threshold state.

[0038] In this embodiment, a continuous time window refers to multiple time intervals divided into continuously acquired pure vital sign vector time series according to a fixed time length. Waveform entropy refers to the value obtained by normalizing the vital sign signals within the time window, constructing a probability distribution based on the absolute value of the amplitude, and calculating the information entropy. Waveform entropy can be used to characterize the stability of vital sign signals within the time window. The smaller the waveform entropy, the more stable the signal; the larger the waveform entropy, the more disordered the signal. The change gradient refers to the rate of change of vital sign vectors between adjacent time windows. The change gradient can be calculated by dividing the difference in vital sign vectors between the start and end times of the time window by the length of the time window. The change gradient can be used to characterize the dynamic change trend of vital sign signals. The larger the absolute value of the change gradient, the more drastic the change in physiological state.

[0039] In this embodiment, the evolution prediction model can refer to a prediction model constructed using time series regression models, hidden Markov models, or recurrent neural network models, taking waveform entropy sequences and change gradients as inputs. The evolution prediction model can be used to iteratively predict physiological state estimates within multiple future time windows. The evolution trajectory can refer to the physiological state estimation sequence within multiple future time windows obtained through iterative prediction by the evolution prediction model. The anomaly classification dictionary can refer to a pre-established database containing feature templates for various abnormal physiological state types. Each template includes template waveform entropy, template change gradient, and a corresponding duration threshold.

[0040] In this embodiment, the current physiological abnormality type can refer to the abnormality type currently belonging to the driver, determined by similarity matching between the current feature vector and each abnormality type template in the abnormality classification dictionary. Physiological abnormality types can include, but are not limited to, respiratory attenuation, sudden drop in heart rate, heart rate instability, and respiratory disturbance. The disability threshold state can refer to a preset physical critical value for each physiological parameter in the pure vital signs vector. For example, the disability threshold state can be a heart rate below 30 beats / min, a respiratory rate below 8 breaths / min, etc. The irreversible disability stage can refer to the stage where the driver's physiological function has reached or exceeded the disability threshold state, and normal driving ability cannot be maintained through self-recovery. The lead time can refer to the remaining time from the current moment until the driver's physiological state is expected to reach the irreversible disability threshold state.

[0041] Specifically, multiple sliding time windows of fixed length are divided along a continuous time axis. Waveform entropy and gradient are calculated. During waveform entropy calculation, the input vital signs signal is normalized. Construct a probability distribution based on the normalized absolute values ​​of the amplitudes: Calculate waveform entropy: .in, For the first time window One sampling point, The mean, Standard deviation; Let N be the waveform entropy of the k-th time window, and N be the number of sampling points. When calculating the gradient, the difference between the vital sign vector at the start and end of each sliding time window is divided by the length of the sliding time window. .in, Represents the respiratory component and the heart rate component; It corresponds to the length of the time window. This represents the starting value of the k-th time window.

[0042] Waveform entropy and gradient change are used to construct the model input feature vector. The data is input into the evolution prediction model. Through the prediction function... Iterative calculations yielded the physiological evolution trajectory within multiple future time windows. Evolution prediction models can employ time series regression models, hidden Markov models, or recurrent neural network models.

[0043] The driver's current physiological abnormality type is determined by a pre-defined anomaly classification dictionary based on waveform entropy and gradient change matching. The anomaly classification dictionary contains multiple predefined anomaly type templates. Where H represents waveform entropy features, G represents physiological change gradient features, For duration. The current feature vector. Calculate the feature distance with each anomaly type template And convert it into probability form. Select the type with the highest probability and combine it with the duration threshold. After a second verification, the current physiological abnormality type is identified.

[0044] The lead time for a driver to enter the irreversible disability stage is determined based on the evolution trajectory and a preset disability threshold state. This is based on the predicted physiological evolution trajectory. and preset disability threshold status Find the time when the predicted trajectory first reaches this threshold. Calculate lead time .

[0045] This step quantifies the stability of vital signs by calculating waveform entropy within a sliding time window; an increase in waveform entropy indicates a loss of physiological regulation. The rate of physiological deterioration is quantified through gradient calculation. The waveform entropy and gradient are input into an evolution prediction model to iteratively predict future trajectories, achieving forward-looking prediction. Template matching from an anomaly classification dictionary enables precise classification of driver physiological abnormalities. Based on the evolution trajectory and disability threshold status, a precise lead time for the driver to enter the irreversible disability stage is calculated, providing an early warning window for system response and safe takeover.

[0046] It should be noted that in the waveform entropy calculation formula group, The amplitude of the i-th sample point after normalization is a dimensionless value obtained by subtracting the mean from the original sample value and then dividing by the standard deviation. `i` represents the proportion of the absolute amplitude of the i-th sample point after normalization to the sum of the absolute amplitudes of all sample points, used to construct the probability distribution. `i` is the index number of the sample point within the current time window, ranging from 1 to N. `j` is the summation index variable, used to sum the absolute amplitudes of all sample points in the denominator. In the gradient formula, This represents the starting time of the k-th sliding time window. The time length of the sliding time window, i.e., the duration covered by each time window (in seconds).

[0047] It should be noted that in the evolution prediction model formula, The feature vector is input to the model for the k-th time window, derived from the waveform entropy. and gradient of change composition. This represents the physiological state estimate for the (k+1)th time window output by the evolution prediction model. F is the prediction function of the evolution prediction model, used to infer the physiological state at the next moment based on the current physiological state and input features. n is the number of future time windows to be predicted, i.e., the number of steps to predict backward from the current moment. In the anomaly classification template and matching formula, This is the template for the i-th anomaly type in the anomaly classification dictionary, containing the standard waveform entropy value H, the standard change gradient value G, and the duration threshold τ corresponding to that anomaly type. In the disability advance formula, The moment when a driver's physiological state first reaches the disability threshold, i.e., the critical moment of disability. This is the current moment, i.e., the real-time time when the system calculates the lead time.

[0048] As an optional implementation of this disclosure, calculating the waveform entropy and gradient of a pure vital sign vector within a continuous time window may specifically include: dividing the continuous time axis into multiple sliding time windows of fixed length; normalizing the pure vital sign vector within each sliding time window, constructing a probability distribution based on the absolute value of the normalized amplitude, and calculating the waveform entropy based on the probability distribution; the waveform entropy is used to characterize the stability of the vital sign signal within the time window; calculating the gradient of change by dividing the difference between the pure vital sign vector at the start time and the pure vital sign vector at the end time of each sliding time window by the length of the sliding time window; the gradient of change is used to characterize the dynamic change trend of the vital sign signal.

[0049] In this embodiment, the continuous time axis can refer to the coordinate axis of a time series of vital sign vectors arranged chronologically. The fixed length can refer to the pre-defined time span of each sliding time window (e.g., 30 seconds or 60 seconds). The sliding time window can refer to a time interval on the continuous time axis divided by a fixed time length and continuously sliding forward over time. Adjacent time windows overlap to ensure real-time performance. The absolute amplitude value can refer to the absolute value of the normalized vital sign signal sampling points. The probability distribution can refer to the probability mass distribution obtained by dividing the absolute amplitude value of each normalized sampling point by the sum of the absolute amplitude values ​​of all sampling points. The dynamic change trend can refer to the direction and rate of change of the vital sign vector over time.

[0050] Specifically, multiple sliding time windows of fixed length are divided along a continuous time axis. For the pure vital sign vector within each sliding time window, normalization is first performed: Construct a probability distribution based on the normalized absolute values ​​of the amplitudes: Calculate waveform entropy: Waveform entropy is used to characterize the stability of vital signs signals within a time window. When vital signs are stable, waveform entropy is low. When abnormal physiological changes occur (such as heart rate instability or respiratory disturbances), waveform entropy increases significantly.

[0051] When calculating the gradient, the difference between the vital sign vector at the start and end of each sliding time window is divided by the length of the sliding time window. The gradient is used to characterize the dynamic changes in vital signs. The larger the absolute value of the gradient, the more drastic the change in physiological state.

[0052] This optional implementation uses a sliding window mechanism to capture the evolution trend of physiological signals while ensuring real-time performance. Normalization eliminates dimensional differences between different physiological indicators, making waveform entropy calculations comparable. Waveform entropy quantifies signal stability, providing a sensitive quantitative indicator for anomaly detection. The change gradient quantifies the rate of change of physiological states, providing directional information for trend prediction.

[0053] S130. Input the abnormal probability, change gradient, waveform entropy change rate and physiological abnormality type corresponding to the abnormality type into the preset driver risk energy accumulation function, and introduce the driver's active intervention intensity to construct an exponential annealing inhibition factor to dynamically update the risk energy. When the updated risk energy exceeds the preset safety threshold, determine that the driver is disabled and trigger the vehicle automatic takeover command.

[0054] In this embodiment of the disclosure, the anomaly probability can refer to the probability value obtained by feature distance transformation when the current feature vector matches each anomaly type template in the anomaly classification dictionary. Anomaly probability can be used to represent the likelihood that a driver is currently in a certain abnormal physiological state. The rate of change of waveform entropy can refer to the rate of change of the waveform entropy of the current time window relative to the waveform entropy of the previous adjacent time window. The rate of change of waveform entropy can be calculated by dividing the difference between the two by the time difference between the start times of the adjacent time windows. Classification weights can refer to pre-set weighting coefficients based on the degree of danger of the physiological abnormality type. Different types of abnormalities correspond to different classification weights.

[0055] In this embodiment, the driver risk energy accumulation function can be a function that dynamically updates the risk energy by using the probability of physiological abnormalities, the gradient of change, the rate of change of waveform entropy, and the classification weight as risk energy gain terms, and constructing an exponential annealing suppression factor based on the driver's active intervention intensity. The active intervention intensity can be a quantitative indicator determined based on the driver's current operational force on the vehicle control devices. The exponential annealing suppression factor can be an exponential function with a base of the natural constant and an exponent of the product of a negative exponential annealing coefficient and the active intervention intensity. The exponential annealing suppression factor can decrease as the active intervention intensity increases.

[0056] In this embodiment, risk energy can refer to a cumulative value used to quantify the current level of driver incapacity risk. Risk energy can be dynamically updated over time. A preset safety threshold can refer to a pre-set risk energy threshold used to determine whether the driver has entered a state of incapacity. Driver incapacity can refer to a state where the driver's physiological functions have severely declined, making it impossible to continue effectively operating the vehicle. An automatic vehicle takeover command can refer to a control command triggered by the system after determining driver incapacity, used to transfer vehicle control from the driver to the autonomous driving system.

[0057] Specifically, the driver's risk energy is dynamically updated according to the following formula: .in, This represents the driver's risk energy at time t. For the intensity of driver active intervention, This represents the probability of the current abnormal physiological state. This represents the gradient vector of changes in vital signs. The rate of change of entropy of vital sign waveforms. For physiological state typing weights, This is the risk gain weighting coefficient. This is the exponential annealing coefficient. The time step for risk energy updates represents the time interval between two consecutive risk energy calculations. The magnitude of the gradient vector of vital sign changes, i.e., the square root of the sum of the squares of the components of the gradient, is used to quantify the overall severity of physiological state changes.

[0058] The risk energy gain term is obtained by multiplying the anomaly probability, the magnitude of the gradient change, the waveform entropy change rate, and the fractal weight by their respective preset risk gain weight coefficients and then summing them. The driver's active intervention intensity is then obtained. The active intervention intensity is determined based on the driver's current operational force on the vehicle's control devices. An exponential annealing suppression factor is constructed with the natural constant as the base and the product of the negative exponential annealing coefficient and the active intervention intensity as the exponent. This factor decreases as the intensity of active intervention increases. Multiplying the risk energy gain term by the exponential annealing suppression factor yields the risk energy increment. Adding the current risk energy to the risk energy increment gives the updated risk energy for the next time step. When the risk energy satisfies... When the driver exceeds a preset safety threshold, the system determines that the driver has entered a disabled state and triggers automatic vehicle takeover control.

[0059] This step integrates anomaly probability, gradient change, waveform entropy change rate, and classification weights into a risk energy gain term, achieving a quantitative mapping of multi-dimensional physiological risks to a unified risk energy. By introducing the intensity of driver intervention to construct an exponential annealing inhibition factor, the system's takeover intention in human-machine co-driving mode is smoothly adjusted according to the intensity of driver operation. The stronger the driver's operation, the slower the risk energy increases. When the risk energy exceeds a preset safety threshold, a takeover command is automatically triggered, realizing a shift from passive alert to active takeover.

[0060] As an optional implementation of this disclosure, the abnormal probability, gradient of change, rate of change of waveform entropy, and classification weight corresponding to the physiological abnormality type are input into a preset driver risk energy accumulation function, and an exponential annealing suppression factor is constructed by introducing the driver's active intervention intensity to dynamically update the risk energy. Specifically, this includes: multiplying the abnormal probability, gradient of change, rate of change of waveform entropy, and classification weight corresponding to the physiological abnormality type by their respective preset risk gain weight coefficients and summing them to obtain a risk energy gain term; obtaining the driver's active intervention intensity; the active intervention intensity is determined based on the driver's current operating force on the vehicle control device; constructing an exponential annealing suppression factor with a natural constant as the base and the product of a negative exponential annealing coefficient and the active intervention intensity as the exponent; the exponential annealing suppression factor decreases as the active intervention intensity increases; multiplying the risk energy gain term by the exponential annealing suppression factor to obtain the risk energy increment; and adding the current risk energy to the risk energy increment to obtain the updated risk energy for the next moment.

[0061] In this embodiment of the disclosure, the preset risk gain weighting coefficient may refer to a pre-set coefficient used to adjust the contribution of each risk energy gain term to the risk energy, including... The risk energy gain term can be a value obtained by weighting and summing the anomaly probability, the magnitude of the gradient change, the waveform entropy change rate, and the fractal weight. The risk energy gain term can be used to represent the positive contribution of each risk factor to the risk energy at the current moment. Vehicle control devices can refer to the physical devices used by the driver to control the vehicle's driving state. Vehicle control devices can include, but are not limited to, steering wheels, brake pedals, and accelerator pedals.

[0062] In this embodiment of the disclosure, the current operating force can refer to the quantified value of the operating force or operating amplitude applied by the driver to the vehicle control device at the current moment. The exponential annealing coefficient can refer to a preset coefficient η used to control the influence of active intervention intensity on the exponential annealing suppression factor. The risk energy increment can refer to the product of the risk energy gain term and the exponential annealing suppression factor. The risk energy increment can be used to represent the increase in risk energy at the current moment. The risk energy at the current moment can refer to the cumulative risk energy value E(t) at time t. The risk energy at the next moment can refer to the risk energy value E(t+Δt) at time t+Δt after one update.

[0063] Specifically, the probability of anomalies , modulus of changing gradient Waveform entropy change rate and fractal weights Multiply by their respective preset risk gain weighting coefficients Summing the results, we obtain the risk energy gain term: Obtain the intensity of the driver's active intervention. This intensity is determined based on the driver's current operating force on at least one of the vehicle's control devices, such as steering wheel torque, brake pedal force, and accelerator pedal opening rate of change.

[0064] Using the natural constant e as the base, and a negative exponential annealing coefficient η as the active intervention intensity The product of these factors forms the exponential annealing inhibitor. This factor decreases as the intensity of active intervention increases. Multiplying the risk energy gain term by the exponential annealing inhibition factor yields the risk energy increment: Add the current risk energy E(t) to the risk energy increment ΔE to obtain the updated risk energy E(t+Δt) for the next moment: E(t+Δt) = E(t) + ΔE.

[0065] This optional implementation achieves the quantitative fusion of multi-dimensional physiological risk factors into a unified risk energy through weighted summation of risk energy gain terms. An exponential annealing suppression factor smoothly regulates the increase in risk energy based on the intensity of driver intervention. Risk energy is suppressed when the driver maintains effective operation. Risk energy increases rapidly when the driver loses operational capacity. This mechanism ensures that the system takes over promptly when the driver is truly incapacitated, but does not intervene rashly when the driver still has the ability to control the vehicle.

[0066] The technical solution of this invention acquires the driver's visual micro-motion signals and vehicle driving status data, and performs vehicle vibration decoupling processing on the visual micro-motion signals based on the driving status data. The periodic stability and phase coupling of the processed signals are verified through a physiological consistency constraint model, and the driver's pure vital sign vector is obtained by parsing the verified signals. The vehicle vibration decoupling processing eliminates the interference of the vehicle vibration environment on the signals, and the dual verification of periodic stability and phase coupling ensures the reliability and purity of the extracted physiological signals, providing high-quality input data for subsequent processing. The waveform entropy and gradient of the pure vital sign vector within a continuous time window are calculated, and the waveform entropy and gradient are input into a preset evolution prediction model to obtain the evolution trajectory output by the model. Based on the waveform entropy and gradient, a preset anomaly classification dictionary is matched to determine the current physiological abnormality type of the driver. Based on the evolution trajectory and the preset disability threshold state, the advance amount before the driver enters the irreversible disability stage is determined. The waveform entropy quantifies the signal stability and the gradient quantifies the deterioration rate. Combined with the evolution prediction model, forward-looking prediction is achieved. The anomaly classification dictionary achieves accurate classification. The evolution trajectory and the disability threshold state are used to calculate the accurate advance amount, providing an early warning window for safe takeover. The abnormal probability, gradient of change, rate of change of waveform entropy, and subtyping weight corresponding to the physiological abnormality type are input into a preset driver risk energy accumulation function. An exponential annealing suppression factor is constructed by introducing the driver's active intervention intensity to dynamically update the risk energy. When the updated risk energy exceeds a preset safety threshold, the driver is determined to be incapacitated and an automatic vehicle takeover command is triggered. By quantitatively mapping multi-dimensional physiological risks to a unified risk energy, combined with the exponential annealing suppression factor, the system's takeover intention is smoothly adjusted according to the driver's operation intensity. When the risk energy exceeds the safety threshold, takeover is automatically triggered, thereby ensuring driving safety in the event of sudden driver incapacity in manual driving mode.

[0067] Example 2 Figure 2 This is a flowchart of a vehicle control method provided in Embodiment 2 of the present invention. Based on the above embodiments, this embodiment describes in detail the process of game-theoretic decision-making and evidence chain consolidation in dynamic traffic flow after the vehicle is taken over. Explanations of terms that are the same as or corresponding to those in the above embodiments are not repeated here. Figure 2 As shown, the method includes: S210. Acquire the driver's visual micro-motion signal and the vehicle's driving status data, and perform vehicle vibration decoupling processing on the visual micro-motion signal based on the driving status data. Verify the periodic stability and phase coupling of the processed signal through a physiological consistency constraint model, and analyze the verified signal to obtain the driver's pure vital sign vector.

[0068] S220. Calculate the waveform entropy and gradient of the pure vital signs vector within a continuous time window, and input the waveform entropy and gradient into a preset evolution prediction model to obtain the evolution trajectory output by the model. Based on the waveform entropy and gradient, match the preset abnormal classification dictionary to determine the current physiological abnormality type of the driver, and determine the advance amount before the driver enters the irreversible disability stage based on the evolution trajectory and the preset disability threshold state.

[0069] S230. Input the abnormal probability, change gradient, waveform entropy change rate and physiological abnormality type corresponding to the abnormality type into the preset driver risk energy accumulation function, and introduce the driver's active intervention intensity to construct an exponential annealing inhibition factor to dynamically update the risk energy. When the updated risk energy exceeds the preset safety threshold, determine that the driver is incapacitated and trigger the vehicle automatic takeover command.

[0070] S240: In real time, call the surrounding perception sensor to obtain the motion vectors of surrounding traffic participants; calculate the environmental risk residual entropy of the vehicle when adopting different candidate behaviors based on the motion vectors and the preset dynamic game risk evaluation function, select the candidate behavior corresponding to the optimal solution of the environmental risk residual entropy as the minimum risk maneuver behavior, and control the vehicle to execute the minimum risk maneuver behavior.

[0071] In this embodiment of the disclosure, a panoramic sensing sensor can refer to a combination of sensors capable of sensing the environment within a 360-degree range around the vehicle. The panoramic sensing sensor may include, but is not limited to, cameras, millimeter-wave radar, and lidar. The panoramic sensing sensor can be used to acquire information such as the position, speed, and direction of movement of surrounding traffic participants. Surrounding traffic participants can refer to other road users moving within a certain range around the vehicle. Surrounding traffic participants may include, but are not limited to, other motor vehicles, non-motor vehicles, and pedestrians.

[0072] In this embodiment of the disclosure, a motion vector can refer to information describing the motion state of surrounding traffic participants. The motion vector can include speed magnitude and direction of motion. The dynamic game-theoretic risk assessment function can refer to an assessment function used to evaluate the collision risk, relative speed risk, and road boundary risk when the vehicle adopts a certain candidate behavior, considering various combinations of behaviors that surrounding traffic participants might take and their probability distributions. The candidate behavior can refer to the driving operations that the vehicle can choose when facing a dangerous scenario. Candidate behaviors can include, but are not limited to, lane keeping and deceleration, emergency braking, moving to the shoulder, changing lanes to avoid a collision, and lane keeping and coasting.

[0073] In this embodiment of the disclosure, residual environmental risk entropy can refer to the overall risk metric calculated by considering various possible combinations of behaviors of surrounding traffic participants, taking into account uncertainty and expected loss, for a candidate behavior. The smaller the residual entropy, the lower the overall environmental risk caused by the behavior in dynamic traffic flow. The optimal solution can refer to the solution with the smallest residual environmental risk entropy among all candidate behaviors. The candidate behavior corresponding to the optimal solution is the minimum risk maneuver behavior. The minimum risk maneuver behavior can refer to the driving operation with the smallest residual environmental risk entropy under the current dynamic traffic environment, taking into account the vehicle's state, driver incapacity, and possible reactions of surrounding traffic participants.

[0074] Specifically, after the system takes over the vehicle, it uses the surround-view perception sensors in real time to acquire the motion vectors of surrounding traffic participants. For each candidate action (such as lane keeping and deceleration), the system then... Emergency braking Move to the shoulder Changing lanes to avoid Maintain lane coasting Based on the motion vectors of surrounding traffic participants, predict multiple possible behaviors of each surrounding traffic participant in the future time domain. Combine the multiple possible behaviors of each surrounding traffic participant to obtain multiple candidate behavior combinations B, and determine the probability P(B) of each candidate behavior combination.

[0075] For each candidate behavior combination, calculate the collision risk value of this vehicle when it performs the current candidate behavior and surrounding traffic participants also adopt the current candidate behavior combination. Relative speed risk value and road boundary risk values The comprehensive risk value is calculated by weighting and summing the collision risk value, relative speed risk value, and road boundary risk value. .

[0076] Based on the probability of occurrence P(B) of each candidate behavior combination and the corresponding comprehensive risk value Calculate the residual entropy of environmental risk under the current candidate behavior: .in, The entropy value (uncertainty) represents the probability distribution of each candidate behavior combination. This represents the expected overall risk value (expected loss). Preset weighting coefficients are used. The residual entropy of environmental risk is positively correlated with the weighted sum of the entropy value of the probability of occurrence and the comprehensive risk value. All candidate behaviors are traversed, and their corresponding residual entropy is calculated. The candidate behavior with the minimum residual entropy is selected as the minimum risk maneuver, and the vehicle is controlled to execute it.

[0077] This step achieves real-time perception of the dynamic traffic environment by using a surround-view sensor to acquire the motion vectors of surrounding traffic participants. By predicting various candidate behavior combinations and their probability distributions of surrounding traffic participants, a multi-agent game is introduced into the decision-making process. Global risk assessment of candidate behaviors is achieved by calculating collision risk, relative speed risk, and road boundary risk and synthesizing them into environmental risk residual entropy. The candidate behavior with the minimum residual entropy is selected as the minimum-risk maneuver, ensuring global safety in dynamic traffic flow.

[0078] As an optional implementation of this disclosure, calculating the residual environmental risk entropy of a vehicle when it adopts different candidate behaviors based on motion vectors and a preset dynamic game risk assessment function may specifically include: for each candidate behavior, predicting multiple possible behaviors of surrounding traffic participants in the future time domain based on the motion vectors of surrounding traffic participants, combining the multiple possible behaviors of surrounding traffic participants to obtain multiple candidate behavior combinations, and determining the occurrence probability of each candidate behavior combination; for each candidate behavior combination, calculating the collision risk value, relative speed risk value, and road boundary risk value when the vehicle executes the current candidate behavior and surrounding traffic participants adopt the current candidate behavior combination, and calculating the comprehensive risk value by weighted summation of the collision risk value, relative speed risk value, and road boundary risk value; calculating the residual environmental risk entropy under the current candidate behavior based on the occurrence probability of each candidate behavior combination and the corresponding comprehensive risk value; the residual environmental risk entropy is positively correlated with the weighted sum of the entropy value of the occurrence probability and the comprehensive risk value; traversing all candidate behaviors and calculating the corresponding residual environmental risk entropy for each.

[0079] In this embodiment of the disclosure, the future time domain can refer to a time range extending into the future from the current moment. The future time domain can be used to predict the possible behaviors of surrounding traffic participants within a future time period. Possible behaviors can refer to various driving operations that surrounding traffic participants may take within the future time domain. For example, possible behaviors could be accelerating, decelerating, maintaining speed, changing lanes, yielding, etc. Candidate behavior combinations can refer to all possible scenarios obtained by permuting and combining the various possible behaviors of all surrounding traffic participants. Each combination represents a possible future traffic scenario. The probability of occurrence can refer to the actual likelihood P(B) of each candidate behavior combination occurring.

[0080] In this embodiment of the disclosure, the current candidate behavior may refer to a candidate driving operation of the vehicle currently being evaluated. The current candidate behavior combination can refer to a specific combination B of surrounding traffic participants currently being evaluated. The collision risk value can be a measure of the risk of a collision between the vehicle and surrounding traffic participants when the vehicle performs the current candidate behavior and the surrounding traffic participants adopt the current candidate behavior combination. Relative speed risk value can be defined as a measure of the risk posed by the relative speed between the vehicle and surrounding traffic participants when the vehicle performs the current candidate action and surrounding traffic participants adopt the same combination of candidate actions. .

[0081] In this embodiment of the disclosure, the road boundary risk value can refer to a risk measure of the vehicle deviating from the road or colliding with road boundary facilities when the vehicle performs the current candidate behavior. The comprehensive risk value can refer to the combined risk measure obtained by weighting and summing the collision risk value, relative speed risk value, and road boundary risk value. The entropy of the probability of occurrence can refer to the entropy of the probability distribution of each combination of candidate behaviors. The entropy value of the probability of occurrence can be used to quantify the degree of uncertainty in future traffic scenarios.

[0082] Specifically, for each candidate behavior Based on the motion vectors of surrounding traffic participants, multiple possible behaviors of each participant in the future time domain are predicted. These multiple possible behaviors are combined to obtain multiple candidate behavior combinations B, and the probability P(B) of each candidate behavior combination is determined. For each candidate behavior combination B, the probability of the vehicle executing the current candidate behavior is calculated. And the collision risk value when surrounding traffic participants adopt the current candidate behavior combination. Relative speed risk value and road boundary risk values The comprehensive risk value is calculated by weighting and summing the collision risk value, relative speed risk value, and road boundary risk value. . The preset weighting coefficients corresponding to the collision risk values ​​are used to adjust the contribution of collision risk to the overall risk. The preset weighting coefficients corresponding to the relative speed risk values ​​are used to adjust the contribution of relative speed risk to the overall risk. The preset weighting coefficients corresponding to the road boundary risk values ​​are used to adjust the contribution of road boundary risk to the overall risk.

[0083] Based on the probability of occurrence P(B) of each candidate behavior combination and the corresponding comprehensive risk value Calculate the residual entropy of environmental risk under the current candidate behavior: .in, Entropy is the probability of occurrence (representing uncertainty). This represents the expected value of the overall risk (indicating anticipated loss). Preset weighting coefficients are used to adjust the contribution of the expected comprehensive risk value to the residual entropy of environmental risk, balancing the relative importance of uncertainty and expected loss. The residual entropy of environmental risk is positively correlated with the weighted sum of the entropy value of the probability of occurrence and the comprehensive risk value. All candidate behaviors are traversed, and their corresponding residual entropy is calculated separately.

[0084] This optional implementation quantifies the multi-agent game relationship in dynamic traffic flow into a mathematically solvable form by predicting various candidate behavior combinations and their probability distributions of surrounding traffic participants. Through a triple-weighted comprehensive assessment of collision risk, relative speed risk, and road boundary risk, the risks of candidate behaviors across different dimensions are fully evaluated. Uncertainty and expected loss are uniformly incorporated into the evaluation system using environmental risk residual entropy. The entropy value characterizes the uncertainty of surrounding vehicle behavior, while the expected loss characterizes the severity of a physical collision. Selecting the candidate behavior with the minimum residual entropy means choosing the optimal strategy with the lowest overall risk even if the reactions of surrounding vehicles are uncertain.

[0085] S250: During the execution of minimum risk maneuvering behavior, mark the time anchor point that triggers the vehicle automatic takeover command, and simultaneously solidify the evidence chain of the driver's physiological evolution before incapacitation, environmental perception decision data frames, and control execution logs to achieve a safety closed loop and accountability.

[0086] In this embodiment, the time anchor point can refer to the moment when the system triggers the automatic vehicle takeover command. The time anchor point can be used to record the starting time point of the system's switch from normal driving mode to risk control mode. The physiological evolution evidence chain can refer to the data on the trajectory of vital sign vector changes and the process of determining the type of physiological abnormality within a preset time period tracing back from the takeover trigger moment. The physiological evolution evidence chain can be used to completely record the physiological state changes of the driver before incapacitation. The environmental perception decision data frame can refer to a data set formed by encapsulating the raw environmental data, target recognition results, and path planning trajectory collected by the surround-view perception sensors within a preset range before and after the takeover trigger moment according to a unified time series.

[0087] In this embodiment, the control execution log can refer to the serialized record of steering control commands, braking control commands, power output control commands, and vehicle status feedback information during the vehicle's execution of minimum-risk maneuvering behavior. The safety closed loop refers to the complete and traceable data link throughout the entire process from risk identification and decision triggering to control execution, achieving closed-loop safety management during vehicle operation. Liability tracing refers to providing objective evidence for accident analysis, system function verification, and liability determination through a fixed event data recording package.

[0088] Specifically, when the system determines that minimum risk maneuver control needs to be triggered based on the driver's physiological state assessment, the system first automatically generates and marks the time anchor point of the trigger takeover event to record the starting moment of the system's switch from normal driving mode to risk control mode. Simultaneously with marking the time anchor point, the system synchronously collects and stores multi-source data before and after the trigger event. For example, the moment the vehicle's automatic takeover command is triggered is used as the starting moment, and the system traces back the changes in vital sign vectors and the determination process data of physiological abnormality types within a preset time period to form a physiological evolution evidence chain. The raw environmental data, target recognition results, and path planning trajectories collected by the surround-view perception sensors within a preset range before and after the starting moment are encapsulated according to a unified time series to form an environmental perception decision data frame. The steering control commands, braking control commands, power output control commands, and vehicle status feedback information during the vehicle's execution of minimum risk maneuver behavior are serialized and recorded to form a control execution log. During data storage, the system synchronously marks and structurally encapsulates the above multi-source data according to a unified time series, forming a complete event data record package, which is stored in the vehicle's local secure storage unit.

[0089] This step precisely records the time nodes of system decisions by marking the time anchor points that trigger takeover. By solidifying the physiological evolution evidence chain, environmental perception decision data frames, and control execution logs, a complete data link is constructed from risk identification and decision triggering to control execution. Structured encapsulation and encrypted storage of unified time series data ensures data integrity and immutability. This data link provides objective evidence for accident analysis, system function verification, and liability determination, improving the interpretability of system operation and the traceability of responsibility.

[0090] The technical solution of this invention involves real-time acquisition of motion vectors of surrounding traffic participants using a panoramic perception sensor; calculation of environmental risk residual entropy when the vehicle adopts different candidate behaviors based on the motion vectors and a preset dynamic game risk evaluation function; selection of the candidate behavior corresponding to the optimal solution of environmental risk residual entropy as the minimum risk maneuver behavior; and control of the vehicle to execute the minimum risk maneuver behavior. Real-time perception of the dynamic traffic environment is achieved through panoramic perception. Multi-agent game theory is introduced into decision-making by predicting various combinations of surrounding vehicle behaviors and their probability distributions. Candidate behaviors are comprehensively evaluated through a triple weighted assessment of collision risk, relative speed risk, and road boundary risk. Uncertainty and expected losses are uniformly incorporated into the evaluation system through environmental risk residual entropy. The candidate behavior with the minimum residual entropy is selected as the optimal strategy, ensuring global safety during the minimum risk maneuver process. During the execution of minimum-risk maneuvers, the time anchor point that triggers the vehicle's automatic takeover command is marked, and the evidence chain of the driver's physiological evolution before incapacitation, environmental perception decision data frames, and control execution logs are simultaneously solidified to achieve a safety closed loop and accountability. By marking time anchor points to accurately record the system's decision-making moments, by solidifying the evidence chain of physiological evolution to fully record the changes in the driver's physiological state before incapacitation, by solidifying the environmental perception decision data frames to fully record the environmental information at the time of decision-making, and by solidifying the control execution logs to fully record the vehicle's control responses, the structured encapsulation and encrypted storage of the unified time series ensure the integrity and immutability of the data, thus achieving a full-link safety closed loop and accountability traceability from risk identification and game-theoretic decision-making to control execution.

[0091] The vehicle control device provided in the embodiments of the present invention can execute the vehicle control method provided in any embodiment of the present invention, and has the corresponding beneficial effects of executing the vehicle control method.

[0092] The following are embodiments of the vehicle control device provided in this invention. This device and the vehicle control methods in the above embodiments belong to the same inventive concept. For details not described in detail in the embodiments of the vehicle control device, please refer to the embodiments of the above vehicle control methods.

[0093] Example 3 Figure 3 This is a schematic diagram of a vehicle control device provided in Embodiment 3 of the present invention. Figure 3 As shown, the device includes: a multi-source signal sensing and physiological verification module 310, a physiological state evolution trend prediction module 320, and a nonlinear risk energy assessment and takeover decision module 330.

[0094] The multi-source signal perception and physiological verification module 310 is used to acquire the driver's visual micro-motion signals and vehicle driving state data, and to perform vehicle vibration decoupling processing on the visual micro-motion signals based on the driving state data. It then verifies the periodic stability and phase coupling of the processed signals through a physiological consistency constraint model, and analyzes the verified signals to obtain the driver's pure vital sign vector. The physiological state evolution trend prediction module 320 is used to calculate the waveform entropy and gradient of the pure vital sign vector within a continuous time window, and inputs the waveform entropy and gradient into a preset evolution prediction model to obtain the evolution trajectory output by the model. The driver's current physiological abnormality type is determined by matching the preset abnormality classification dictionary, and the advance amount before the driver enters the irreversible disability stage is determined based on the evolution trajectory and the preset disability threshold state. The nonlinear risk energy assessment and takeover decision module 330 is used to input the abnormality probability, change gradient, waveform entropy change rate and classification weight corresponding to the physiological abnormality type into the preset driver risk energy accumulation function, and introduce the driver's active intervention intensity to construct an exponential annealing suppression factor to dynamically update the risk energy. When the updated risk energy exceeds the preset safety threshold, the driver is determined to be disabled and the vehicle automatic takeover command is triggered.

[0095] The technical solution of this invention acquires the driver's visual micro-motion signals and vehicle driving status data, and performs vehicle vibration decoupling processing on the visual micro-motion signals based on the driving status data. The periodic stability and phase coupling of the processed signals are verified through a physiological consistency constraint model, and the driver's pure vital sign vector is obtained by parsing the verified signals. The vehicle vibration decoupling processing eliminates the interference of the vehicle vibration environment on the signals, and the dual verification of periodic stability and phase coupling ensures the reliability and purity of the extracted physiological signals, providing high-quality input data for subsequent processing. The waveform entropy and gradient of the pure vital sign vector within a continuous time window are calculated, and the waveform entropy and gradient are input into a preset evolution prediction model to obtain the evolution trajectory output by the model. Based on the waveform entropy and gradient, a preset anomaly classification dictionary is matched to determine the current physiological abnormality type of the driver. Based on the evolution trajectory and the preset disability threshold state, the advance amount before the driver enters the irreversible disability stage is determined. The waveform entropy quantifies the signal stability and the gradient quantifies the deterioration rate. Combined with the evolution prediction model, forward-looking prediction is achieved. The anomaly classification dictionary achieves accurate classification. The evolution trajectory and the disability threshold state are used to calculate the accurate advance amount, providing an early warning window for safe takeover. The abnormal probability, gradient of change, rate of change of waveform entropy, and subtyping weight corresponding to the physiological abnormality type are input into a preset driver risk energy accumulation function. An exponential annealing suppression factor is constructed by introducing the driver's active intervention intensity to dynamically update the risk energy. When the updated risk energy exceeds a preset safety threshold, the driver is determined to be incapacitated and an automatic vehicle takeover command is triggered. By quantitatively mapping multi-dimensional physiological risks to a unified risk energy, combined with the exponential annealing suppression factor, the system's takeover intention is smoothly adjusted according to the driver's operation intensity. When the risk energy exceeds the safety threshold, takeover is automatically triggered, thereby ensuring driving safety in the event of sudden driver incapacity in manual driving mode.

[0096] Based on the above technical solution, the device also includes: The dynamic environment game and vehicle behavior decision-making module is used to call the surrounding perception sensor in real time to obtain the motion vectors of surrounding traffic participants; based on the motion vectors and the preset dynamic game risk evaluation function, it calculates the environmental risk residual entropy of the vehicle when adopting different candidate behaviors, selects the candidate behavior corresponding to the optimal solution of the environmental risk residual entropy as the minimum risk maneuver behavior, and controls the vehicle to execute the minimum risk maneuver behavior. The control execution and evidence chain solidification module is used to mark the time anchor point that triggers the vehicle automatic takeover command during the execution of minimum risk maneuvering behavior, and simultaneously solidify the evidence chain of the driver's physiological evolution before incapacitation, environmental perception decision data frames, and control execution logs to achieve a safety closed loop and accountability.

[0097] Based on the above technical solution, the multi-source signal sensing and physiological verification module 310 is specifically used for: performing time-frequency transformation on the signal after vehicle vibration decoupling processing, extracting the peak frequency within each time window, calculating the fluctuation value of each peak frequency relative to the average frequency, and determining that the signal satisfies periodic stability if the fluctuation value is less than a first preset threshold; extracting the respiratory phase and heartbeat phase in the signal, calculating the phase lock value between the respiratory phase and heartbeat phase, and determining that the signal satisfies phase coupling if the phase lock value is greater than a second preset threshold; when the signal simultaneously satisfies periodic stability and phase coupling, determining that the signal is a real physiological oscillation signal and retaining it, otherwise discarding the signal data in the current time period.

[0098] Based on the above technical solution, the physiological state evolution trend prediction module 320 is specifically used for: dividing multiple sliding time windows of fixed length on a continuous time axis; normalizing the pure vital sign vector within each sliding time window, constructing a probability distribution based on the absolute value of the normalized amplitude, and calculating the waveform entropy based on the probability distribution; the waveform entropy is used to characterize the stability of the vital sign signal within the time window; calculating the change gradient by dividing the difference between the pure vital sign vector at the start time and the pure vital sign vector at the end time of each sliding time window by the length of the sliding time window; the change gradient is used to characterize the dynamic change trend of the vital sign signal.

[0099] Based on the above technical solution, the nonlinear risk energy assessment and takeover decision module 330 is specifically used for: multiplying the abnormal probability, change gradient, waveform entropy change rate, and physiological abnormality type corresponding to the physiological abnormality type by their respective preset risk gain weight coefficients and summing them to obtain the risk energy gain term; obtaining the driver's active intervention intensity; the active intervention intensity is determined based on the driver's current operating force on the vehicle control device; constructing an exponential annealing suppression factor with the natural constant as the base and the product of the negative exponential annealing coefficient and the active intervention intensity as the exponent; the exponential annealing suppression factor decreases as the active intervention intensity increases; multiplying the risk energy gain term by the exponential annealing suppression factor to obtain the risk energy increment; and adding the current risk energy to the risk energy increment to obtain the updated risk energy for the next moment.

[0100] Based on the above technical solution, the dynamic environment game and vehicle behavior decision-making module is specifically used for: for each candidate behavior, predicting multiple possible behaviors of surrounding traffic participants in the future time domain based on the motion vectors of surrounding traffic participants, combining the multiple possible behaviors of surrounding traffic participants to obtain multiple candidate behavior combinations, and determining the occurrence probability of each candidate behavior combination; for each candidate behavior combination, calculating the collision risk value, relative speed risk value, and road boundary risk value when the vehicle executes the current candidate behavior and surrounding traffic participants adopt the current candidate behavior combination, and calculating the comprehensive risk value by weighted summation of the collision risk value, relative speed risk value, and road boundary risk value; calculating the environmental risk residual entropy under the current candidate behavior based on the occurrence probability of each candidate behavior combination and the corresponding comprehensive risk value; the environmental risk residual entropy is positively correlated with the weighted sum of the entropy value of the occurrence probability and the comprehensive risk value; traversing all candidate behaviors and calculating the corresponding environmental risk residual entropy for each.

[0101] The vehicle control device provided in the embodiments of the present invention can execute the vehicle control method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the vehicle control method.

[0102] It is worth noting that in the above-described vehicle control embodiments, the various units and modules are divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the scope of protection of the present invention.

[0103] Example 4 Figure 4 A schematic diagram of an electronic device 10, which can be used to implement embodiments of the present invention, is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0104] like Figure 4As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0105] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0106] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as vehicle control methods.

[0107] In some embodiments, the vehicle control method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the vehicle control method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the vehicle control method by any other suitable means (e.g., by means of firmware).

[0108] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0109] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0110] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0111] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0112] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0113] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0114] This application also discloses a computer program product, which includes a computer program that, when executed by a processor, implements the vehicle control method provided in any embodiment of this application. This program product shares the same inventive concept as the vehicle control methods disclosed in the embodiments of this application, and therefore will not be described in detail here.

[0115] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0116] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A vehicle control method, characterized in that, include: The driver's visual micro-motion signals and vehicle driving status data are acquired. Based on the driving status data, the visual micro-motion signals are decoupled from the vehicle vibration. The periodic stability and phase coupling of the processed signals are verified by a physiological consistency constraint model. The verified signals are then analyzed to obtain the driver's pure vital sign vector. The waveform entropy and gradient of the pure vital signs vector within a continuous time window are calculated, and the waveform entropy and gradient are input into a preset evolution prediction model to obtain the evolution trajectory output by the model. Based on the waveform entropy and gradient, a preset abnormal classification dictionary is matched to determine the current physiological abnormality type of the driver. Based on the evolution trajectory and the preset disability threshold state, the advance amount before the driver enters the irreversible disability stage is determined. The abnormal probability corresponding to the physiological abnormality type, the change gradient, the rate of change of the waveform entropy, and the subtyping weight corresponding to the physiological abnormality type are input into a preset driver risk energy accumulation function. The driver's active intervention intensity is introduced to construct an exponential annealing suppression factor to dynamically update the risk energy. When the updated risk energy exceeds a preset safety threshold, the driver is determined to be incapacitated and an automatic vehicle takeover command is triggered.

2. The method according to claim 1, characterized in that, After determining driver incapacity and triggering the vehicle automatic takeover command, the method further includes: The motion vectors of surrounding traffic participants are obtained in real time by calling the surrounding perception sensor; the environmental risk residual entropy of the vehicle when adopting different candidate behaviors is calculated based on the motion vectors and the preset dynamic game risk evaluation function, and the candidate behavior corresponding to the optimal solution of the environmental risk residual entropy is selected as the minimum risk maneuver behavior, and the vehicle is controlled to execute the minimum risk maneuver behavior. During the execution of the minimum risk maneuver, the time anchor point that triggers the automatic takeover command of the vehicle is marked, and the evidence chain of the driver's physiological evolution before incapacitation, environmental perception decision data frames, and control execution logs are simultaneously solidified to achieve a safety closed loop and accountability.

3. The method according to claim 1, characterized in that, The verification of the periodic stability and phase coupling of the signal after processing through the physiological consistency constraint model includes: The signal after vehicle vibration decoupling is transformed by time and frequency, the peak frequency in each time window is extracted, the fluctuation value of each peak frequency relative to the average frequency is calculated, and if the fluctuation value is less than a first preset threshold, the signal is determined to meet the periodic stability. Extract the breathing phase and heartbeat phase from the signal, calculate the phase lock value between the breathing phase and the heartbeat phase, and if the phase lock value is greater than a second preset threshold, then determine that the signal satisfies phase coupling. When the signal simultaneously satisfies both periodic stability and phase coupling, the signal is determined to be a true physiological oscillation signal and is retained; otherwise, the signal data within the current time period is discarded.

4. The method according to claim 1, characterized in that, The calculation of the waveform entropy and gradient of the pure vital sign vector within a continuous time window includes: Divide a continuous time axis into multiple sliding time windows of fixed length; For each pure vital sign vector within the sliding time window, normalization is performed, a probability distribution is constructed based on the absolute value of the normalized amplitude, and the waveform entropy is calculated based on the probability distribution; the waveform entropy is used to characterize the stability of the vital sign signal within the time window. The gradient of change is calculated by dividing the difference between the pure vital sign vector at the start time and the pure vital sign vector at the end time of each sliding time window by the length of the sliding time window; the gradient of change is used to characterize the dynamic change trend of vital sign signals.

5. The method according to claim 1, characterized in that, The step of inputting the abnormal probability corresponding to the physiological abnormality type, the change gradient, the rate of change of the waveform entropy, and the subtyping weight corresponding to the physiological abnormality type into a preset driver risk energy accumulation function, and introducing the driver's active intervention intensity to construct an exponential annealing suppression factor to dynamically update the risk energy includes: The risk energy gain term is obtained by multiplying the abnormal probability corresponding to the physiological abnormality type, the change gradient, the rate of change of the waveform entropy, and the classification weight corresponding to the physiological abnormality type by their respective preset risk gain weight coefficients. The intensity of the driver's active intervention is obtained; the intensity of the active intervention is determined based on the driver's current operating force on the vehicle control device; An exponential annealing inhibition factor is constructed with the natural constant as the base and the product of the negative exponential annealing coefficient and the intensity of the active intervention as the exponential factor; the exponential annealing inhibition factor decreases as the intensity of the active intervention increases; Multiply the risk energy gain term by the exponential annealing suppression factor to obtain the risk energy increment; The risk energy at the current moment is added to the risk energy increment to obtain the updated risk energy for the next moment.

6. The method according to claim 2, characterized in that, The calculation of the residual environmental risk entropy of the vehicle when adopting different candidate behaviors, based on the motion vector and a preset dynamic game risk assessment function, includes: For each candidate behavior, multiple possible behaviors of the surrounding traffic participants in the future time domain are predicted based on the motion vectors of the surrounding traffic participants, and the multiple possible behaviors of the surrounding traffic participants are combined to obtain multiple candidate behavior combinations, and the occurrence probability of each candidate behavior combination is determined. For each candidate behavior combination, calculate the collision risk value, relative speed risk value, and road boundary risk value when the vehicle performs the current candidate behavior and the surrounding traffic participants adopt the current candidate behavior combination, and calculate the comprehensive risk value by weighted summation of the collision risk value, the relative speed risk value, and the road boundary risk value; Based on the occurrence probability of each candidate behavior combination and the corresponding comprehensive risk value, the residual entropy of environmental risk under the current candidate behavior is calculated; the residual entropy of environmental risk is positively correlated with the weighted sum of the entropy value of the occurrence probability and the comprehensive risk value; Iterate through all candidate behaviors and calculate the corresponding environmental risk residual entropy for each.

7. A vehicle control device, characterized in that, The device includes: The multi-source signal perception and physiological verification module is used to acquire the driver's visual micro-motion signals and the vehicle's driving status data, and to perform vehicle vibration decoupling processing on the visual micro-motion signals based on the driving status data. The module also verifies the periodic stability and phase coupling of the processed signals through a physiological consistency constraint model, and analyzes the verified signals to obtain the driver's pure vital sign vector. The physiological state evolution trend prediction module is used to calculate the waveform entropy and change gradient of the pure vital sign vector within a continuous time window, input the waveform entropy and change gradient into a preset evolution prediction model to obtain the evolution trajectory output by the model, and determine the current physiological abnormality type of the driver based on the waveform entropy and change gradient matched with a preset abnormality classification dictionary, and determine the advance amount of the driver entering the irreversible disability stage based on the evolution trajectory and a preset disability threshold state. The nonlinear risk energy assessment and takeover decision module is used to input the abnormal probability corresponding to the physiological abnormality type, the change gradient, the change rate of the waveform entropy, and the classification weight corresponding to the physiological abnormality type into a preset driver risk energy accumulation function, and to introduce the driver's active intervention intensity to construct an exponential annealing suppression factor to dynamically update the risk energy. When the updated risk energy exceeds a preset safety threshold, the module determines that the driver is incapacitated and triggers an automatic vehicle takeover command.

8. An electronic device, characterized in that, The electronic device includes: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the vehicle control method as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the vehicle control method as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the vehicle control method as described in any one of claims 1-6.

Images