Method for assisting driving behavior decision based on driving style and light physiological signal

Abstract

The application discloses an auxiliary driving behavior decision-making method based on driving style and light physiological signals and belongs to the field of intelligent traffic and driving assistance, and comprises the following steps: S1, collecting vehicle motion, driver eye movement and skin electricity data, preprocessing through time stamp alignment, abnormal value elimination, normalization and sliding window construction; S2, calculating a driving style index based on driving behavior characteristics; S3, calculating a concentration index by fusing eye movement and skin electricity data, and adapting to the situation of sensor data loss; S4, mapping longitudinal target time interval, lateral lane changing threshold and other parameters in combination with the two, and setting a grading coefficient according to vehicle type and speed; S5, performing safety check through minimum collision time, triggering a fallback mechanism, outputting control parameters to a controller and synchronously prompting an HMI; the application does not change the main sensing system, reduces the calibration difficulty and system complexity, reduces the dependence on high-cost sensing, and improves safety redundancy and user experience.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation and driver assistance, and in particular to a method for driver assistance behavior decision-making based on driving style and light physiological signals. Background Technology

[0002] Currently, mass-produced Advanced Driver Assistance Systems (ADAS) mostly use uniform parameters for longitudinal following (target distance, acceleration / jerk limits) and lateral lane changing (feasibility criteria, minimum clearance), making it difficult to take into account the significant differences among different drivers in terms of acceleration / deceleration preferences, distance selection, and lane-changing tendencies. In practice, drivers generally exhibit a conservative-conventional-aggressive style spectrum; if the same strategy is adopted, it is easy to cause human-machine conflict ("the system is too conservative or too aggressive"), a decline in experience, and insufficient acceptance.

[0003] Secondly, while introducing complex multi-channel physiological systems (such as EEG, ECG, and EMG) can improve the recognition of driving states, the high deployment cost and engineering complexity limit its implementation in vehicles. For engineering implementation needs, in-vehicle scanning eye-tracking and electrical conduction analysis (EDA) are low-cost, lightweight alternative physiological signal sources that can supplement "attention" information, providing a feasible path for transitioning from uniform parameters to personalized and interpretable parameter mapping.

[0004] Therefore, proposing a personalized behavioral decision-making method based on driving style and assisted by lightweight physiological signals, without changing the main sensing system, with minor modifications and easy calibration, has become a practical need to improve the quality of human-machine collaboration, safety redundancy and user acceptance. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide an assisted driving behavior decision-making method based on driving style and lightweight physiological signals. While keeping the system complexity controllable, it can achieve personalized adaptive adjustment of core strategy parameters such as longitudinal target distance and lateral lane change threshold, and output assisted driving suggestions or control commands under simple safety criteria constraints, thereby reducing human-machine conflict, improving safety redundancy and user experience.

[0006] To achieve the above objectives, this invention provides an assisted driving behavior decision-making method based on driving style and light physiological signals, comprising the following steps:

[0007] S1. Collect vehicle motion data, eye movement data, and skin conductance data and perform preprocessing;

[0008] S2. Calculate the driving style index;

[0009] S3. Calculate the driver's attention index;

[0010] S4. Map the parameters of the assisted driving strategy based on the driving style index calculated in S2 and the driver attention index calculated in S3.

[0011] S5. Control strategy security verification and result output.

[0012] Preferably, step S1 specifically includes:

[0013] S101. Install scanning eye trackers and skin conductance collectors on the vehicle and read the vehicle's driving data;

[0014] S102. Establish a unified timestamp and use a unified system clock to record the timestamps of each channel;

[0015] S103. Detect eye-tracking frame drops, blink occlusion, and skin conductance motion artifacts;

[0016] S104. Median filtering was used to remove outliers in acceleration / time interval, thresholding was used to remove abnormal jumps in eye movement, first-order high-pass filtering was used to detrend electrodermal activity, and finally each feature was scaled and normalized according to the in-subject z-score.

[0017] S105. Construct a sliding window: Create a sliding window with duration W and step size Δ to extract style and readiness features.

[0018] Preferably, step S2 specifically includes:

[0019] S201. Construct driving style characteristics and calculate the standard deviation of longitudinal acceleration. The average absolute value of jerk The distance from the front of the car to the 25th percentile is 1 / HD. P25 Number of lane changes per unit distance , forming vector x style ;

[0020] S202, x style Normalization style Eliminate the differences between vehicle models and driving scenarios;

[0021] S203. Driving style index is obtained by monotonic mapping using the Sigmoid function. i The expression is as follows:

[0022] ;

[0023] In the formula, For the Sigmoid function; w This is the calibration coefficient vector, used to represent the linear weights of each normalized driving style feature on the driving style index; The calibration bias parameters are used to adjust the overall baseline level of the model output; style This is a characteristic of normalized style;

[0024] First-order low-pass filter is used for driving style index i The smoothing process is performed using the following expression:

[0025] ;

[0026] In the formula, For smoothed i value; Smoothed from the previous time i value; β For smoothing coefficients, The original mapping value for the current window;

[0027] S204. Set hysteresis and boundary conditions, and introduce up / down switching hysteresis. d θ ;

[0028] S205, Smoothed Driving Style Index Stored in the loop cache for S4 parameter mapping and HMI hints.

[0029] Preferably, step S3 specifically includes:

[0030] S301. Extract driver physiological characteristics, extract eye-tracking data PERCLOS and average fixation duration. Extract skin conductance signal data, including skin conductivity level (SCL) and skin conductance response peak number (SCR_rate).

[0031] S302. Perform outlier removal and normalization on the physiological characteristic data in S301 to obtain the indicators. ;

[0032] S303. Combine the indicators that have completed outlier removal and normalization to calculate the driver attention index. R The expression is as follows:

[0033] ;

[0034] In the formula, α 1. α 2. α 3 represents the indicators respectively. The corresponding weighting coefficients; α 1, α 2, α 3≥0, and α 1+ α 2+ α3≤1; This is a truncation function;

[0035] S304. Detect the confidence level of eye-tracking data or skin conductance signals. If the confidence level is low, the weights are linearly reduced according to the effective coverage. Detect any sensor value. If data from one sensor is missing, the weights are calculated based on single-source focus. R eye or R eda Replacement; if data from both sensors is missing, then... R Set to a neutral value;

[0036] S305, Focus Indicators R Perform a moving average process when R continuously below the threshold R low Once the set time is reached, a "declining attention" prompt will be triggered, and the system will enter a conservative control strategy mode.

[0037] Preferably, step S4 specifically includes:

[0038] S401, Adjust the longitudinal target time interval, according to and R Calculate the longitudinal target time interval The expression is as follows:

[0039] ;

[0040] In the formula, T 0 is the baseline target time interval; a The coefficient characterizing the sensitivity of the driving style index to target time-distance adjustments. To characterize the degree of influence of road traffic flow conditions on target time-distance adjustment, and a , b ≥0;

[0041] S402, Adjust the lateral lane change threshold, according to and R Calculate the lateral lane change threshold The expression is as follows:

[0042] ;

[0043] In the formula, The baseline threshold; c This is a coefficient characterizing the influence of the driving style index on the lateral lane change threshold. d A coefficient characterizing the degree to which road traffic flow conditions affect the lateral lane change threshold, and c , d ≥0;

[0044] S403. Perform amplitude limiting and time smoothing processing on the adjusted longitudinal target time distance and lateral lane change threshold, as shown in the following expression:

[0045] ;

[0046] In the formula, This represents the minimum permissible time distance for longitudinal targets. This represents the maximum permissible time distance for longitudinal targets. This is the minimum permissible value for the lateral lane change threshold; This is the maximum permissible value for the lateral lane change threshold;

[0047] A first-order low-pass filter is applied to the longitudinal target time distance and lateral lane change threshold after amplitude limiting and time smoothing, as shown in the following expression:

[0048] ;

[0049] In the formula, γ is the smoothing coefficient; For the first t The final parameter values ​​after smoothing are calculated once; For the first t -1 calculation of the smoothed parameter values; This represents the original, unsmoothed parameter values ​​at the current moment;

[0050] S404. Adjust the calibration coefficient by setting the gearing coefficient according to different vehicle speed ranges. a , b , c , d；

[0051] S405. Use a combination of icons and text to provide HMI prompts and explanations.

[0052] Preferably, step S5 specifically involves the following steps:

[0053] S501, Calculate the minimum conflict time (TTC) for the safety alternative indicator;

[0054] S502, Setting linked safety thresholds based on driver focus ;

[0055] S503, The minimum conflict time (TTC) of the safety alternative indicator calculated according to S501 and the linkage safety threshold set in S502. Determine whether to trigger rollback and motion suppression mechanisms;

[0056] S504, Send final control parameters and record key data.

[0057] Preferably, the expression for calculating the minimum conflict time (TTC) of the safety alternative index in S501 is as follows:

[0058] ;

[0059] In the formula, d Forward spacing; Δ v It is the relative velocity; To prevent the elimination of zero items.

[0060] Preferably, a linkage safety threshold is set in S502. The expression is as follows:

[0061] ;

[0062] In the formula, Basic minimum collision time threshold; k This represents the influence coefficient on focus.

[0063] Preferably, the specific content of determining whether to trigger the rollback and motion suppression mechanism in S503 is as follows:

[0064] If the current collision time is detected to meet the condition When this happens, the system triggers the following security response procedure:

[0065] Based on preset safety rules, the following distance threshold is adaptively adjusted, and acceleration and jerk are limited, as shown in the following expression:

[0066] ;

[0067] In the formula, Minimum necessary headway under safety constraints.

[0068] If the current collision time is not met. In this case, the system will not trigger the "safety rollback mechanism" and will perform normal parameter sending and recording.

[0069] Preferably, the specific content of sending the final control parameters and recording key data in S504 is as follows:

[0070] The final parameters of the control strategy are sent to the longitudinal and lateral controllers, and the output is submitted to the HMI system and recorded. , R, Th, TTC and rollback events.

[0071] Therefore, the present invention employs the above-described method, which has the following beneficial effects:

[0072] By establishing a single, defined mapping framework and combining a minimum TTC (Traffic Time To Call) linkage backoff mechanism with a rapid response strategy of speed gradation and event windowing, online personalized parameter tuning is achieved without altering the main sensing system. Compared to existing technologies, this method overcomes the problems of human-machine conflict and low acceptance caused by uniform parameters, significantly improving the safety redundancy of longitudinal following and the decision stability of lateral lane changing, while ensuring real-time performance and interpretability, and adapting to continuous scene switching in complex traffic environments. This innovation reduces calibration difficulty and system complexity with fixed parameters and a unique process, reduces reliance on high-cost sensors, and provides a safer, more comfortable, and efficient solution for the mass production of ADAS (Advanced Driver Assistance Systems) with human-vehicle collaboration.

[0073] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0074] Figure 1 This is an overall flowchart of the assisted driving behavior decision-making method based on driving style and lightweight physiological signals of the present invention;

[0075] Figure 2 This is a flowchart of the rollback and action inhibition mechanism of the assisted driving behavior decision-making method based on driving style and light physiological signals of the present invention. Detailed Implementation

[0076] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the embodiments of the present invention and are not intended to limit the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of this application. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout.

[0077] It should be noted that the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, such as a process, method, system, product, or server that includes a series of steps or units, not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such process, method, product, or device.

[0078] The following is combined Figure 1-Figure 2 The embodiments of the present invention will be described in detail below.

[0079] Example

[0080] S1. Acquisition and processing of vehicle motion data, eye-tracking data, and skin conductance data;

[0081] S101. In a specific implementation scenario, the data acquisition hardware configuration is as follows:

[0082] Eye-tracking data acquisition: A non-contact infrared scanning eye tracker is installed on the A-pillar on the driver's side of the vehicle, with its camera facing the driver's face. This eye tracker captures the driver's eye movements, blinks, and gaze direction at a sampling rate of 60Hz, ensuring accurate capture of subtle eye movements.

[0083] Skin conductance data acquisition: Two conductive electrodes are integrated at the "nine o'clock" and "three o'clock" positions on the steering wheel. Skin conductance response (EDA) signals can be collected when the driver's hands are on the steering wheel. The signals are collected at a sampling rate of 15Hz, which is higher than the typical skin conductance signal variation frequency (0.05-0.1Hz), meeting the analysis requirements. Alternatively, data can be collected using a smart wristband worn by the driver.

[0084] Vehicle driving data reading: Vehicle driving data is read in real time at a frequency of 50Hz via the vehicle's CAN (Controller Area Network) bus interface. This data includes vehicle speed provided by the ABS / ESP unit, acceleration provided by the engine control unit (ECU) or motor controller, headway relative to the vehicle ahead provided by radar or camera sensors, and lane change events calculated by the steering angle sensor and wheel speed sensor. (Acceleration...) jerk The result is obtained by performing a first-order difference calculation on the acceleration signal.

[0085] S102. The central processing unit assigns a unified, high-precision system-level timestamp to each received frame of data (eye tracking, skin conductance, vehicle data). Considering the potential fixed delays of different sensors and buses, the system pre-measures these delays (the transmission delay of CAN bus signals) during the calibration phase and performs time offset compensation during data processing to ensure precise alignment of all data on the time axis.

[0086] S103. Within a sliding window of length W, the system calculates the coverage of valid data. For eye-tracking data, a valid frame refers to a frame in which the pupil is successfully detected; for electrodermal data, a valid frame refers to a frame in which the signal amplitude is within the normal physiological range. If the valid coverage of any data source is lower than a preset threshold, the window is marked as low confidence, and its calculation result will be downweighted in subsequent smoothing processing to prevent decision-making errors caused by sensor occlusion or signal interference.

[0087] S104. A median filter with a window size of 5 was applied to the acceleration and headway time-distance sequences to filter out sudden spike noise. A thresholding method was used on the eye-tracking data to remove jumps outside the normal visual field. A first-order Butterworth high-pass filter with a cutoff frequency of 0.05 Hz was applied to the skin conductance signal to eliminate slow baseline drift caused by changes in body temperature, etc., while retaining the rapidly changing components (SCR) related to emotion and cognitive load. To eliminate individual physiological differences, each driver's physiological characteristics (such as blink frequency and skin conductance level) and driving behavior characteristics were z-score normalized in their individual historical data distribution, and then scaled to [0,1] using the Sigmoid function. This within-subject normalization method ensured the model's universality across different drivers.

[0088] S105. To achieve online calculation, the system constructs a sliding window with a duration W = 60 seconds and a step size Δ = 5 seconds. In each step period, the window slides forward, and the driving style index and attention index are calculated once based on the data within the window, thus achieving near real-time online updates. The longer window W ensures the stability of the style assessment, while the shorter step size Δ ensures the system's response speed.

[0089] S2, Driving Style Index Calculation;

[0090] S201. Driving style characteristics construction, statistical standard deviation of longitudinal acceleration. accelerometer ( jerk () average absolute value The distance from the front of the car to the 25th percentile is 1 / HD. P25 Number of lane changes per unit distance Equal to each other, forming a vector x style ;

[0091] S202, x style Normalization is performed to obtain the normalized style vector. style Eliminate the impact of differences between vehicle models and driving scenarios;

[0092] S203. The driving style index is obtained by monotonically mapping the multidimensional feature vector using the Sigmoid function. i :

[0093] ;

[0094] In the formula, is the Sigmoid function; w is the calibration coefficient vector, used to represent the linear weights of each normalized driving style feature on the driving style index; The calibration bias parameters are used to adjust the overall baseline level of the model output; "style" refers to normalized style characteristics.

[0095] The resulting original style index will be smoothed using a first-order low-pass filter:

[0096] ;

[0097] In the formula, β The smoothing coefficient is set to 0.1 to suppress rapid fluctuations in the style index, making its changes smoother and more consistent with human driving habits. This is the original mapping value for the current window.

[0098] S204. To avoid frequent changes in style categories near the critical point, a hysteresis mechanism is introduced for switching between styles. d θ When there are insufficient valid samples, i Revert to the last valid value;

[0099] S205, Smoothed Driving Style Index Stored in the loop cache for S4 parameter mapping and HMI hints.

[0100] S3. Calculate the driver's attention index R. This step aims to assess the driver's current level of alertness and focus. The higher the value, the more focused the driver is (R is close to 1).

[0101] S301. Extract driver physiological characteristics, including PERCLOS (eye-closed ratio) and average fixation duration from eye-tracking data. Skin conductance signal data were extracted to determine the skin conductivity level (SCL) and the number of skin conductance response peaks (SCR_rate).

[0102] S302. Feature fusion: The driver physiological feature data obtained in step S301 is subjected to outlier removal and normalization to obtain indicators. , , ;

[0103] S303. Perform feature fusion on the indicators that have completed outlier removal and normalization to calculate the driver attention index R:

[0104] ;

[0105] In the formula, α 1, α 2, α 3≥0, and α 1+ α 2+ α 3≤1, This is the cutoff function. In a typical long-distance highway driving scenario, PERCLOS is the primary criterion for judging fatigue, and its weight a1 might be set to 0.7; while in complex urban traffic environments, the weights a2 and a3 of SCR_rate and SCL, which reflect cognitive load, might be appropriately increased. The clip() function ensures that the R value is always within the range.

[0106] S305, Low Attention Warning: The attention index R is processed by a sliding average. When the smoothed R value remains below the preset threshold R_low=0.3 for 10 consecutive seconds, the system will trigger an "attention decline" prompt (sound and visual alarms via HMI) and automatically switch to a more conservative control strategy.

[0107] S4: Based on the calculated driving style index i Personalized driver assistance strategy parameters are mapped to the driver attention index R.

[0108] S401, Personalized adjustment of longitudinal target time interval, according to Calculate the longitudinal target time interval using R:

[0109] ;

[0110] In the formula, is the baseline target time distance, and 'a' is a coefficient characterizing the sensitivity of the driving style index to adjustments in the target time distance. A coefficient characterizing the degree of influence of road traffic flow conditions on target time-distance adjustment. Parameters a The impact of adjustment style: aggressive driver ( i →1) The target time distance is close to the reference, while the conservative driver ( i →0) will increase the time distance. The effect of parameter b on concentration: Regardless of the style, as long as the driver's concentration decreases (R→0), the system will forcibly increase the safety redundancy.

[0111] S402, Personalized adjustment of lateral lane change threshold, according to Calculate the lateral lane change threshold with R :

[0112] ;

[0113] In the formula, is the baseline threshold, c is a coefficient characterizing the influence of the driving style index on the lateral lane change threshold, and d is a coefficient characterizing the degree of adjustment of the road traffic flow state on the lateral lane change threshold, and c, d ≥ 0.

[0114] In the formula, This is the minimum safe clearance time required for a baseline lane change. Parameter cAdjustment style influence: Aggressive driver ( i →1) Smaller lane change clearance is acceptable. Parameters d Adjusting the impact of focus: When focus decreases (R→0), the system will require a larger lane change gap and may even suppress lane change suggestions to ensure safety.

[0115] S403. Perform amplitude limiting and time smoothing processing on the personalized adjusted longitudinal target time distance and lateral lane change threshold:

[0116] ;

[0117] In the formula, This represents the minimum permissible time distance for longitudinal targets. This represents the maximum permissible time distance for longitudinal targets. This is the minimum permissible value for the lateral lane change threshold; This is the maximum permissible value for the lateral lane change threshold;

[0118] First-order low-pass filter used:

[0119] ;

[0120] In the formula, γ is the smoothing coefficient; For the first t The final parameter values ​​after smoothing are calculated once; For the first i -1 calculation of the smoothed parameter values; This represents the original, unsmoothed parameter values ​​at the current moment;

[0121] S404. Adjust the calibration coefficient by setting the gearing coefficient according to different vehicle speed ranges. a , b , c , d The parameter settings are shown in Table 1:

[0122] Table 1. Coefficient Calibration

[0123]

[0124] S405, Human-Machine Interface (HMI) output: The system displays the adjusted strategy parameters to the driver through a graphical interface and concise text on the instrument panel or central control screen, such as "The following distance has been adjusted to a more comfortable distance" or "Your attention has been detected to be reduced, and the safe distance has been increased", thus making the decision-making process explainable.

[0125] S5. Control strategy security verification and result output;

[0126] S501, The system calculates the minimum time to collision (TTC) with the vehicle in front in real time:

[0127] ;

[0128] Where d is the forward spacing and Δv is the relative speed; To eliminate zero terms.

[0129] S502. Set the linked safety threshold in combination with driving concentration:

[0130] ;

[0131] Where TTC0 is the absolute safety baseline, k is the adjustment coefficient. The less concentrated the driver is, the higher the TTC threshold for the system to trigger emergency avoidance, and the earlier the intervention.

[0132] S503. At any time, if it is detected that TTC < TTC_min(R), the system immediately triggers the safety fallback mechanism. This mechanism will ignore the personalized parameters calculated in step S4 and force the target time headway to be set to a preset safety value , and strictly limit the acceleration and jerk to increase the distance from the vehicle ahead in the fastest and smoothest way. During this period, all automatic lane change or lane change suggestion functions will be suppressed:

[0133] ;

[0134] Where The minimum necessary time headway under safety constraints.

[0135] S504. The final control parameters ( , ) after safety verification are sent to the longitudinal and lateral motion controllers of the vehicle (such as ACC and LKA / LCK controllers) to perform the final driving assistance actions. At the same time, all key parameters ( , R, Th, , TTC) and fallback events will be recorded by the system for subsequent auditing and algorithm iteration optimization.

[0136] Therefore, this invention adopts the above-mentioned personalized assisted driving behavior decision-making method based on driving style and lightweight physiological signals. Without changing the main sensing system, it realizes personalized adaptive adjustment of assisted driving strategy. By calculating the driving style index from vehicle motion data and combining it with the driver attention index extracted from eye movement and skin conductance signals, it dynamically maps core strategy parameters such as longitudinal following distance and lateral lane change threshold. This overcomes the problems of human-machine conflict and low user acceptance caused by the use of uniform parameters in traditional assisted driving systems, and significantly improves the human-machine collaboration quality, safety redundancy and user experience of advanced driver assistance systems (ADAS).

[0137] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for assisted driving behavior decision-making based on driving style and lightweight physiological signals, characterized in that, Includes the following steps: S1. Collect vehicle motion data, eye movement data, and skin conductance data and perform preprocessing; S2. Calculate the driving style index; S3. Calculate the driver's attention index; S4. Map the parameters of the assisted driving strategy based on the driving style index calculated in S2 and the driver attention index calculated in S3. S5. Control strategy security verification and result output; Step S2 specifically includes: S201. Construct driving style characteristics and calculate the standard deviation of longitudinal acceleration. The average absolute value of jerk The distance from the front of the car to the 25th percentile is 1 / HD. P25 Number of lane changes per unit distance , forming vector x style ; S202, x style Normalization style Eliminate the differences between vehicle models and driving scenarios; S203. Driving style index is obtained by monotonic mapping using the Sigmoid function. θ The expression is as follows: ； In the formula, For the Sigmoid function; w is the calibration coefficient vector, used to represent the linear weight of each normalized driving style feature to the driving style index; is the calibration bias parameter, used to adjust the overall baseline level of the model output; style This is a characteristic of normalized style; First-order low-pass filter is used for driving style index θ The smoothing process is performed using the following expression: ； In the formula, For the current smoothed θ value; Smoothed from the previous time θ value; β For smoothing coefficients, The original mapping value for the current window; S204. Set hysteresis and boundary conditions, and introduce up / down switching hysteresis. δ θ ; S205, Smoothed Driving Style Index Stored in the loop cache for S4 parameter mapping and HMI hints; Step S4 specifically includes: S401, Adjust the longitudinal target time interval, according to Calculate the longitudinal target time distance with R The expression is as follows: ； In the formula, T 0 represents the baseline target time interval; a The coefficient characterizing the sensitivity of the driving style index to target time-distance adjustments. b To characterize the degree of influence of road traffic flow conditions on target time-distance adjustment, and a , b ≥0; S402, Adjust the lateral lane change threshold, according to and R Calculate the lateral lane change threshold The expression is as follows: ； In the formula, R Indicators of driver attention level The baseline threshold; c This is a coefficient characterizing the influence of the driving style index on the lateral lane change threshold. d A coefficient characterizing the degree to which road traffic flow conditions affect the lateral lane change threshold, and c , d ≥0; S403. Perform amplitude limiting and time smoothing processing on the adjusted longitudinal target time distance and lateral lane change threshold, as shown in the following expression: ； In the formula, This represents the minimum permissible time distance for longitudinal targets. This represents the maximum permissible time distance for longitudinal targets. This is the minimum permissible value for the lateral lane change threshold; This is the maximum permissible value for the lateral lane change threshold; A first-order low-pass filter is applied to the longitudinal target time distance and lateral lane change threshold after amplitude limiting and time smoothing, as shown in the following expression: ； In the formula, γ is the smoothing coefficient; For the first t The final parameter values ​​after smoothing are calculated once; For the first t -1 calculation of the smoothed parameter values; This represents the original, unsmoothed parameter values ​​at the current moment; S404. Adjust the calibration coefficient by setting the gearing coefficient according to different vehicle speed ranges. a , b , c , d； S405. Use a combination of icons and text to provide HMI prompts and explanations.

2. The assisted driving behavior decision-making method based on driving style and lightweight physiological signals according to claim 1, characterized in that, Step S1 specifically includes: S101. Install scanning eye trackers and skin conductance collectors on the vehicle and read the vehicle's driving data; S102. Establish a unified timestamp and use a unified system clock to record the timestamps of each channel; S103. Detect eye-tracking frame drops, blink occlusion, and skin conductance motion artifacts; S104. Median filtering was used to remove outliers in acceleration / time interval, thresholding was used to remove abnormal jumps in eye movement, first-order high-pass filtering was used to detrend electrodermal activity, and finally each feature was scaled and normalized according to the in-subject z-score. S105. Construct a sliding window: Create a sliding window with duration W and step size Δ to extract style and readiness features.

3. The assisted driving behavior decision-making method based on driving style and lightweight physiological signals according to claim 2, characterized in that, Step S3 specifically includes: S301. Extract driver physiological characteristics, extract eye-tracking data PERCLOS and average fixation duration. Extract skin conductance signal data, including skin conductivity level (SCL) and skin conductance response peak number (SCR_rate). S302. Perform outlier removal and normalization on the physiological characteristic data in S301 to obtain the indicators. ; S303. Combine the indicators that have completed outlier removal and normalization to calculate the driver attention index. R The expression is as follows: ； In the formula, α 1. α 2. α 3 represents the indicators respectively The corresponding weighting coefficients; α 1, α 2, α 3≥0, and α 1+ α 2+ α 3≤1; This is a truncation function; S304. Detect the confidence level of eye-tracking data or skin conductance signals. If the confidence level is low, the weights are linearly reduced according to the effective coverage. Detect any sensor value. If data from one sensor is missing, the weights are calculated based on single-source focus. R eye or R eda Replacement; if data from both sensors is missing, then... R Set to a neutral value; S305, Focus Indicators R Perform a moving average process when R continuously below the threshold R low Once the set time is reached, a "decreased attention" prompt will be triggered, and the system will enter a conservative control strategy mode.

4. The assisted driving behavior decision-making method based on driving style and lightweight physiological signals according to claim 3, characterized in that, Step S5 is as follows: S501, Calculate the minimum conflict time (TTC) for the safety alternative indicator; S502, Setting linked safety thresholds based on driver focus ; S503, The minimum conflict time (TTC) of the safety alternative indicator calculated according to S501 and the linkage safety threshold set in S502. Determine whether to trigger rollback and motion suppression mechanisms; S504, Send final control parameters and record key data.

5. The assisted driving behavior decision-making method based on driving style and lightweight physiological signals according to claim 4, characterized in that, The expression for calculating the minimum conflict time (TTC) in S501 is as follows: ； In the formula, d Forward spacing; Δ v It is the relative velocity; To prevent the elimination of zero items.

6. The assisted driving behavior decision-making method based on driving style and lightweight physiological signals according to claim 5, characterized in that, The linkage safety threshold set in S502 The expression is as follows: ； In the formula, Basic minimum collision time threshold; k This represents the influence coefficient on focus.

7. The assisted driving behavior decision-making method based on driving style and lightweight physiological signals according to claim 6, characterized in that, The specific details of S503 regarding whether to trigger the rollback and motion suppression mechanism are as follows: If the current collision time is detected to meet the condition When this happens, the system triggers the following security response procedure: Based on preset safety rules, the following distance threshold is adaptively adjusted, and acceleration and jerk are limited, as shown in the following expression: ； In the formula, Minimum necessary headway under safety constraints.

8. The assisted driving behavior decision-making method based on driving style and light physiological signals according to claim 7, characterized in that, The specific details of sending final control parameters and recording key data in S504 are as follows: The final parameters of the control strategy are sent to the longitudinal and lateral controllers, and the output is submitted to the HMI system and recorded. R , TTC and rollback events.

Images