Driver takeover performance quantification method, system, device and medium in assisted driving environment

Abstract

The application discloses a driver takeover performance quantification method, system, device and medium in an assisted driving environment, relates to the field of assisted driving data processing, and comprises the following steps: acquiring multi-modal physiological signal data of a driver in real time; constructing a feature vector based on the multi-modal physiological signal data; constructing an index prediction model; inputting the feature vector into the index prediction model to obtain a comprehensive index prediction value; determining a safety level based on the comprehensive index prediction value; and generating a driver takeover suggestion based on the safety level. The application can comprehensively monitor the state of the driver, realize real-time and accurate evaluation and quantification of the state of the driver, and further provide a basis for a personalized takeover strategy, so that the driving safety is significantly improved.

Description

Technical Field

[0001] This application relates to the field of driver assistance data processing, and in particular to a method, system, device and medium for quantifying driver takeover performance in a driver assistance environment. Background Technology

[0002] Traditional driver takeover methods typically rely on data from a single sensor. Data from a single sensor is prone to errors or noise, and cannot comprehensively and accurately reflect the driving environment and the driver's state. For example, cameras may be affected by external conditions such as lighting and weather, while radar may be interfered with by obstacles. Furthermore, relying on data from a single sensor cannot achieve multi-angle, multi-dimensional information acquisition, and its data processing capabilities are also limited.

[0003] To address the aforementioned issues, while known technologies employ deep learning, big data analytics, and other techniques to process and analyze sensor data in real time to improve data accuracy and reliability, the complexity and variability of the driving environment mean that data from a single sensor cannot fully reflect environmental changes and thus cannot comprehensively monitor the driver's condition. Consequently, it is impossible to achieve real-time and accurate quantitative assessment of the driver's condition. Summary of the Invention

[0004] The purpose of this application is to provide a method, system, device, and medium for quantifying driver takeover performance in assisted driving environments. This method can comprehensively monitor the driver's state, achieve real-time and accurate quantification of the driver's state, and thus provide a basis for personalized takeover strategies to significantly improve driving safety.

[0005] To achieve the above objectives, this application provides the following solution:

[0006] Firstly, this application provides a method for quantifying driver takeover performance in an assisted driving environment, including:

[0007] Real-time acquisition of multimodal physiological signal data of drivers;

[0008] Feature vectors are constructed based on the aforementioned multimodal physiological signal data;

[0009] Construct an indicator prediction model;

[0010] The feature vector is input into the indicator prediction model to obtain the comprehensive indicator prediction value;

[0011] The security level is determined based on the predicted values ​​of the comprehensive indicators.

[0012] A driver takeover recommendation is generated based on the safety level.

[0013] Optionally, the multimodal physiological signal data includes physiological data and eye-tracking data; a feature vector is constructed based on the multimodal physiological signal data, including:

[0014] The physiological data is filtered to obtain filtered physiological data;

[0015] Time-domain and frequency-domain features were obtained from filtered physiological data;

[0016] The eye-tracking data is subjected to sliding window normalization to obtain standard eye-tracking data;

[0017] The feature vector is constructed based on time-domain and frequency-domain features and standard eye-tracking data.

[0018] Optionally, the time-domain features include the MeanHR value of heart rate variability and the SDNN value of heart rate variability; the frequency-domain features include the LF norm value of the frequency-domain features of heart rate variability and the HF norm value of the frequency-domain features of heart rate variability.

[0019] Standard eye-tracking data include: horizontal dispersion distance between fixations, average pupil diameter, blink frequency, and average number of saccades.

[0020] Optionally, the physiological data is filtered to obtain filtered physiological data, including:

[0021] The physiological data were filtered using a Butterworth filter to obtain filtered physiological data.

[0022] Optionally, an indicator prediction model is constructed, including:

[0023] Acquire raw data; the raw data includes: driver's multimodal historical physiological signal data, and driver reaction data and vehicle state history data corresponding to the multimodal historical physiological signal data; the driver's multimodal historical physiological signal data includes historical physiological data and historical eye movement data; the vehicle state history data includes: maximum historical lateral offset, minimum historical collision time and maximum historical longitudinal deceleration;

[0024] The raw data is preprocessed, and a historical feature vector is constructed based on the preprocessed multimodal historical physiological signal data.

[0025] Using historical feature vectors as input samples and preprocessed driver reaction data and vehicle state history data corresponding to multimodal historical physiological signal data as output samples, a sample dataset is obtained.

[0026] The sample dataset is divided into multiple mutually exclusive subsets;

[0027] A multi-objective regression prediction model is constructed using the XGBoost algorithm;

[0028] One subset is selected as the validation set, and the remaining subset is used as the training set. The multi-objective regression prediction model is trained and validated using the ten-fold cross-validation method to obtain a multi-objective regression prediction model that meets the prediction requirements.

[0029] The multi-objective regression prediction model that meets the prediction requirements is used as the indicator prediction model.

[0030] Optionally, during the acquisition of raw data, the vehicle control signal in the historical vehicle status data is used as the reference time axis, and a dynamic time warping algorithm is used to align the historical physiological data and historical eye-tracking data in time sequence.

[0031] Optionally, determining the security level based on the predicted value of the comprehensive index includes:

[0032] Takeover time threshold, lateral offset threshold, collision time threshold, and longitudinal deceleration threshold;

[0033] When the driver takeover reaction time, maximum lateral deviation, minimum collision time, and maximum longitudinal deceleration all meet the set conditions, the safety level is determined to be Level 1.

[0034] When any one or two of the driver takeover reaction time, maximum lateral deviation, minimum collision time, and maximum longitudinal deceleration fail to meet the set conditions, the safety level is determined to be Level 2.

[0035] The safety level is determined to be Level 3 if any three or all of the following conditions are not met: driver takeover reaction time, maximum lateral deviation, minimum collision time, and maximum longitudinal deceleration.

[0036] Secondly, this application provides a system for quantifying driver takeover performance in an assisted driving environment, comprising:

[0037] The data acquisition module is used to acquire multimodal physiological signal data of the driver in real time;

[0038] The data processing module is used to construct feature vectors based on the multimodal physiological signal data;

[0039] The model prediction module is used to construct an indicator prediction model and input the feature vector into the indicator prediction model to obtain a comprehensive indicator prediction value.

[0040] The feedback execution module is used to determine the safety level based on the predicted value of the comprehensive index, generate a driver takeover suggestion based on the safety level, and trigger a graded warning.

[0041] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the driver takeover performance quantification method in the assisted driving environment provided above.

[0042] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the driver takeover performance quantification method provided above in an assisted driving environment.

[0043] According to the specific embodiments provided in this application, this application has the following technical effects:

[0044] This application provides a method, system, device, and medium for quantifying driver takeover performance in assisted driving environments. By acquiring multimodal physiological signal data of the driver in real time, it can comprehensively monitor the driver's state, solving the problem of insufficient accuracy when relying on data from a single sensor. Furthermore, by constructing a feature vector based on the real-time acquired multimodal physiological signal data of the driver, and using it as input to an indicator prediction model, a comprehensive indicator prediction value is obtained to determine the safety level. This enables real-time and accurate quantification of driver state assessment. Based on the safety level, driver takeover suggestions are generated, providing a basis for personalized takeover strategies to significantly improve driving safety. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 A flowchart illustrating a method for quantifying driver takeover performance in an assisted driving environment, provided as an embodiment of this application;

[0047] Figure 2 A schematic diagram illustrating the implementation architecture of a method for quantifying driver takeover performance in an assisted driving environment, provided in an embodiment of this application;

[0048] Figure 3 A schematic diagram of a driver takeover performance quantification system in an assisted driving environment is provided as another embodiment of this application;

[0049] Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0050] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0051] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0052] In one exemplary embodiment, this application provides a method for quantifying driver takeover performance in an assisted driving environment. This method is executed by a computer device, specifically a terminal or server, or both. In this embodiment, the method is described using a server as an example. Figure 1 As shown, the method includes:

[0053] Step 100: Acquire the driver's multimodal physiological signal data in real time. Multimodal physiological signal data includes physiological data and eye movement data. Physiological data refers to physiological electrical signals, including heart rate.

[0054] In practical applications, photoplethysmography (PPG) sensors and electrocardiogram (ECG) sensors can be used to wirelessly collect physiological electrical signals during driver takeover activities. Eye movement data can be tracked and collected using an eye tracker.

[0055] Step 101: Construct feature vectors based on multimodal physiological signal data.

[0056] Step 102: Construct an indicator prediction model.

[0057] Step 103: Input the feature vector into the index prediction model to obtain the comprehensive index prediction value. The comprehensive index prediction value is expressed as: , For takeover response time, This represents the maximum lateral offset. TTCmin is the minimum collision time. This represents the maximum longitudinal deceleration.

[0058] Step 104: Determine the safety level based on the predicted value of comprehensive indicators.

[0059] Step 105: Generate driver takeover recommendations based on safety levels.

[0060] By implementing steps 100-105 above, this application can comprehensively monitor the driver's state, solve the problem of insufficient accuracy of data from a single sensor, and achieve real-time and accurate quantitative assessment of the driver's state, providing a basis for personalized takeover strategies to significantly improve driving safety.

[0061] In another exemplary embodiment of this application, the implementation process of step 101 described above may include:

[0062] Step 11: Filter the physiological data to obtain filtered physiological data. For example, use a Butterworth filter to filter the physiological data to remove noise from the physiological signals.

[0063] Step 12: Obtain time-domain and frequency-domain features based on filtered physiological data. Time-domain features include the MeanHR value of heart rate variability (HRV) and the SDNN value of heart rate variability. Frequency-domain features include the LF norm value and the HF norm value of heart rate variability frequency domain features.

[0064] Among them, (1) the MeanHR value of heart rate variability (HRV) is the average heart rate value, which is used to measure the overall frequency of cardiac activity, and its formula is:

[0065] .

[0066] In the formula, MeanHR value refers to heart rate variability (HRV). Heart rate (bpm) is a basic indicator for measuring cardiac activity. N This represents the total number of RR intervals, i.e., the total number of normal heartbeats within a specified time period. It is an integer from 1 to N, representing the sequence number of the RR interval. For the first indivual The interval refers to the distance between two R waves.

[0067] (2) The SDNN value of heart rate variability (HRV) is the standard deviation of the normal RR interval. It is used to assess the overall variation and dispersion of HRV and is a sensitive indicator for assessing the activity of the autonomic nervous system. The formula is:

[0068] .

[0069] In the formula, The SDNN value represents the heart rate variability (HRV). For all within the specified time period The average value of the interval.

[0070] (3) The LF norm value of the frequency domain characteristics of heart rate variability, representing the normalized low-frequency power, is given by the following formula:

[0071] .

[0072] In the formula, LF norm value represents the frequency domain characteristics of heart rate variability, and TP represents the total power in the range of 0-0.4Hz, reflecting the overall variability of the signal. ULF is the ultra-low frequency band (0-0.0033Hz). VLF is the extremely low frequency band (0.0033-0.04Hz). LF is the low frequency band (0.04-0.15Hz). HF is the high frequency band (0.15-0.4Hz).

[0073] (4) The HF norm value of the frequency domain characteristics of heart rate variability, representing the normalized high-frequency power, is given by the following formula:

[0074] .

[0075] In the formula, This represents the HF norm value of the frequency domain characteristics of heart rate variability.

[0076] Step 13: Standardize the eye-tracking data using a sliding window (e.g., window length of 3 seconds, step size of 0.01 seconds) to obtain standard eye-tracking data. Standard eye-tracking data includes: horizontal dispersion distance between fixations, average pupil diameter, blink frequency, and average number of saccades.

[0077] Step 14: Construct a feature vector based on time-domain and frequency-domain features and standard eye-tracking data. The feature vector is represented as follows: :

[0078] .

[0079] In the formula, MeanHR value is the heart rate variability (HRV). The SDNN value represents the heart rate variability (HRV). The LF norm value is the frequency domain characteristic of heart rate variability. The HF norm value is the frequency domain characteristic of heart rate variability. The horizontal dispersion distance between fixation points. The average pupil diameter, The frequency of blinking. This represents the average number of scans.

[0080] In another exemplary embodiment of this application, in order to obtain the optimal prediction result, the implementation process of step 102 above may include:

[0081] Step 21: Within a preset time window before the assisted driving takeover request is triggered, continuously collect raw data. The raw data includes: the driver's multimodal historical physiological signal data, and the corresponding driver reaction data and vehicle status history data. The driver's multimodal historical physiological signal data includes historical physiological data and historical eye-tracking data. The driver reaction data includes: historical takeover reaction time, which represents the time interval from the moment the takeover request is triggered to the driver's fastest possible manipulation (accelerator, brake, steering wheel). The vehicle status history data includes: maximum historical lateral offset, minimum historical collision time, and maximum historical longitudinal deceleration.

[0082] Among them, the maximum historical lateral offset represents the maximum lateral positional deviation of the vehicle controlled by the driver after a takeover request has been issued, from the moment of the takeover request, characterizing the stability of lane-changing operations after the driver takes over control. The minimum historical collision time refers to the shortest time before the vehicle collides with a vehicle or obstacle ahead; it is a metric for alternative safety and controllability. If a collision occurs, the minimum collision time is zero. The maximum historical longitudinal deceleration represents the maximum longitudinal deceleration of the vehicle after the driver takes over control; it is often used to evaluate longitudinal stability after takeover.

[0083] In practical applications, the acquisition method for drivers' multimodal historical physiological signal data can be the same as described above. Driver reaction data and vehicle status history data can be acquired through vehicle sensors.

[0084] Furthermore, during the acquisition of raw data, the vehicle control signal in the historical vehicle status data can be used as the reference time axis, and the dynamic time warping (DTW) algorithm can be used to align the historical physiological data and historical eye-tracking data in time sequence.

[0085] Step 22: Preprocess the raw data and construct historical feature vectors based on the preprocessed multimodal historical physiological signal data.

[0086] Step 23: Using historical feature vectors as input samples, and preprocessed driver reaction data and vehicle state history data corresponding to multimodal historical physiological signal data as output samples, obtain the sample dataset (corresponding to...). Figure 2 (Feature dataset in the dataset).

[0087] Step 24: Divide the sample dataset into multiple mutually exclusive subsets. For example, randomly divide the sample dataset into 10 mutually exclusive subsets (D1~D2). 10 This ensures that the data distribution of each subset remains consistent.

[0088] Step 25: Construct a multi-objective regression prediction model using the XGBoost algorithm.

[0089] Step 26: Select one subset as the validation set (corresponding to...) Figure 2 The test dataset), and the remaining subset as the training set (corresponding to...). Figure 2 The multi-objective regression prediction model is trained and validated using a 10-fold cross-validation method on the training dataset to obtain a model that meets the prediction requirements. For example, the training and validation process is iterated 10 times, selecting one subset as the validation set each time, and combining the remaining 9 subsets into the training set. The multi-objective regression prediction model is fitted on each training set, and the mean of the target index is calculated on the validation set. The results of the 10 validations are aggregated to obtain a robust estimate of the performance of the multi-objective regression prediction model. The multi-objective regression prediction model is then retrained using all the training data, and the parameters of the final multi-objective regression prediction model are determined by retraining with all the training data.

[0090] Step 27: Use the multi-objective regression prediction model that meets the prediction requirements as the indicator prediction model.

[0091] Based on the above description, the loss function of the multi-objective regression prediction model provided in this application is expressed in the form of weighted mean square error, and is as follows:

[0092] .

[0093] In the formula, For loss function, For the first The true value of a target For the first The predicted value of each target, All are weights. .

[0094] Furthermore, to explain the performance of the multi-objective regression prediction model, in this embodiment, the contribution of each feature to the prediction result can be calculated using the SHAP value, as follows:

[0095] .

[0096] In the formula, For feature set, A subset of the feature set, For model prediction function, Features The SHAP value.

[0097] The SHAP value analysis results are used to quantify the contribution of each eigenvector to the prediction results and calculate its importance value, so as to explain the effect of the multi-objective regression prediction model.

[0098] Furthermore, the training of the multi-objective regression prediction model can employ a transfer learning strategy. The pre-training data can be historical driving datasets from multiple vehicle brands, including driver takeover reaction time, maximum lateral offset, TTCmin, maximum longitudinal deceleration, etc.

[0099] In another exemplary embodiment of this application, the implementation process of step 104 includes:

[0100] Step 41: When the driver takeover reaction time, maximum lateral deviation, minimum collision time, and maximum longitudinal deceleration all meet the set conditions, the safety level is determined to be Level 1. For example, if... , , , If so, the safety level is determined to be "good". Among them, For takeover response time, This is the maximum lateral offset. For TTCmin, For the maximum longitudinal deceleration, The threshold for the takeover reaction time. The maximum lateral offset threshold. The TTCmin threshold, Maximum longitudinal deceleration threshold.

[0101] Step 42: If any one or two of the following conditions are not met: driver takeover reaction time, maximum lateral deviation, minimum collision time, and maximum longitudinal deceleration, the safety level is determined to be Level 2. For example, if any two conditions are not met, the safety level is determined to be "moderate".

[0102] Step 43: If any three or all of the following conditions are not met: driver takeover reaction time, maximum lateral deviation, minimum collision time, and maximum longitudinal deceleration, the safety level is determined to be Level 3. For example, if three or more conditions are not met, the safety level is determined to be "deficient".

[0103] In another exemplary embodiment of this application, the safety level can be fed back in real time through the in-vehicle human-machine interface, and the driver can be informed of the current safety level and the driver takeover suggestion.

[0104] Based on the above description, the specific implementation process of the driver takeover performance quantification method in the assisted driving environment provided in this application may include: real-time collection of the driver's physiological data (PPG, ECG), eye-tracking data (horizontal gaze distance, average pupil diameter, blink frequency, average number of saccades), and driving state data within 30 seconds before the autonomous driving takeover request is issued; and construction of a temporal feature set through preprocessing and feature extraction. A multi-objective regression prediction model (corresponding to...) is established using the XGBoost algorithm. Figure 2 The XGBoost model is used to predict comprehensive indicators after takeover (takeover reaction time, lateral offset, minimum collision time TTCmin, and maximum deceleration), and the SHAP value method is used to perform interpretability analysis on the feature contribution. The prediction results are mapped to three levels: "good," "average," and "deficient," generating three novel safety alert methods. Based on this, in practical applications, the implementation architecture of the method provided in this application is as follows: Figure 2 As shown.

[0105] In summary, this application solves the problem of insufficient accuracy in traditional reliance on single sensor data by quantifying the dynamic correlation between physiological and behavioral data. It can assess the driver's state in real time, provide a basis for personalized takeover strategies, and significantly improve driving safety.

[0106] Based on the same inventive concept, this application also provides a system for quantifying driver takeover performance in an assisted driving environment, used to implement the aforementioned method for quantifying driver takeover performance in an assisted driving environment. The solution provided by this system is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the system for quantifying driver takeover performance in an assisted driving environment provided below can be found in the limitations of the method for quantifying driver takeover performance in an assisted driving environment described above, and will not be repeated here.

[0107] In one exemplary embodiment, such as Figure 3 As shown, a driver takeover performance quantification system in an assisted driving environment is provided, comprising: a data acquisition module 300, a data processing module 301, a model prediction module 302, and a feedback execution module 303.

[0108] The data acquisition module 300 is used to acquire multimodal physiological signal data of the driver in real time.

[0109] The data processing module 301 is used to construct feature vectors based on multimodal physiological signal data.

[0110] The model prediction module 302 is used to construct an indicator prediction model and input feature vectors into the indicator prediction model to obtain a comprehensive indicator prediction value.

[0111] The feedback execution module 303 is used to determine the safety level based on the predicted value of comprehensive indicators, generate driver takeover suggestions based on the safety level, and trigger graded warnings. These graded warnings can be triggered through a combination of visual, auditory, and tactile means, such as by integrating a HUD display, voice prompts, and seat vibration to deliver takeover signals.

[0112] As an optional implementation, the driver takeover performance quantification system provided in this application for assisted driving environment can support communication with a cloud server for online updates and data synchronization of indicator prediction models.

[0113] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 4 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores quantitative data on driver takeover performance in assisted driving environments. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a method for quantifying driver takeover performance in assisted driving environments.

[0114] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0115] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0116] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0117] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0118] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0119] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (RRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0120] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0121] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0122] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for quantifying driver takeover performance in an assisted driving environment, characterized in that, include: Real-time acquisition of multimodal physiological signal data of drivers; Feature vectors are constructed based on the aforementioned multimodal physiological signal data; Construct an indicator prediction model; The feature vector is input into the indicator prediction model to obtain the comprehensive indicator prediction value; The safety level is determined based on the predicted values ​​of the comprehensive indicators, including: driver intervention time threshold, lateral offset threshold, collision time threshold, and longitudinal deceleration threshold; when the driver intervention reaction time, maximum lateral offset, minimum collision time, and maximum longitudinal deceleration all meet the set conditions, the safety level is determined to be Level 1; when any one or two of the driver intervention reaction time, maximum lateral offset, minimum collision time, and maximum longitudinal deceleration do not meet the set conditions, the safety level is determined to be Level 2; when any three or all of the driver intervention reaction time, maximum lateral offset, minimum collision time, and maximum longitudinal deceleration do not meet the set conditions, the safety level is determined to be Level 3. A driver takeover recommendation is generated based on the safety level.

2. The method for quantifying driver takeover performance in an assisted driving environment according to claim 1, characterized in that, Multimodal physiological signal data includes physiological data and eye-tracking data; Based on the multimodal physiological signal data, a feature vector is constructed, including: The physiological data is filtered to obtain filtered physiological data; Time-domain and frequency-domain features were obtained from filtered physiological data; The eye-tracking data is subjected to sliding window normalization to obtain standard eye-tracking data; The feature vector is constructed based on time-domain and frequency-domain features and standard eye-tracking data.

3. The method for quantifying driver takeover performance in an assisted driving environment according to claim 2, characterized in that, Temporal features include MeanHR values ​​and SDNN values ​​of heart rate variability; Frequency domain features include: the LF norm value of the heart rate variability frequency domain feature and the HF norm value of the heart rate variability frequency domain feature; Standard eye-tracking data include: horizontal dispersion distance between fixations, average pupil diameter, blink frequency, and average number of saccades.

4. The method for quantifying driver takeover performance in an assisted driving environment according to claim 2, characterized in that, The physiological data is filtered to obtain filtered physiological data, including: The physiological data were filtered using a Butterworth filter to obtain filtered physiological data.

5. The method for quantifying driver takeover performance in an assisted driving environment according to claim 1, characterized in that, Constructing an indicator prediction model includes: Acquire raw data; the raw data includes: driver's multimodal historical physiological signal data, and driver reaction data and vehicle state history data corresponding to the multimodal historical physiological signal data; the driver's multimodal historical physiological signal data includes historical physiological data and historical eye movement data; the vehicle state history data includes: maximum historical lateral offset, minimum historical collision time and maximum historical longitudinal deceleration; The raw data is preprocessed, and a historical feature vector is constructed based on the preprocessed multimodal historical physiological signal data. Using historical feature vectors as input samples, and preprocessed driver reaction data and vehicle state history data corresponding to multimodal historical physiological signal data as output samples, a sample dataset is obtained. The sample dataset is divided into multiple mutually exclusive subsets; A multi-objective regression prediction model is constructed using the XGBoost algorithm; One subset is selected as the validation set, and the remaining subset is used as the training set. The multi-objective regression prediction model is trained and validated using the ten-fold cross-validation method to obtain a multi-objective regression prediction model that meets the prediction requirements. The multi-objective regression prediction model that meets the prediction requirements is used as the indicator prediction model.

6. The method for quantifying driver takeover performance in an assisted driving environment according to claim 5, characterized in that, During the acquisition of raw data, the vehicle control signal in the historical vehicle status data is used as the reference time axis, and the dynamic time warping algorithm is used to align the historical physiological data and historical eye-tracking data in time sequence.

7. A system for quantifying driver takeover performance in an assisted driving environment, characterized in that, A system for implementing the method for quantifying driver takeover performance in an assisted driving environment as described in any one of claims 1-6; the system comprises: The data acquisition module is used to acquire multimodal physiological signal data of the driver in real time; The data processing module is used to construct feature vectors based on the multimodal physiological signal data; The model prediction module is used to construct an indicator prediction model and input the feature vector into the indicator prediction model to obtain a comprehensive indicator prediction value. The feedback execution module is used to determine the safety level based on the predicted value of the comprehensive index, generate a driver takeover suggestion based on the safety level, and trigger a graded warning.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the method for quantifying driver takeover performance in an assisted driving environment 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 executed by a processor, the computer program implements the method for quantifying driver takeover performance in an assisted driving environment as described in any one of claims 1-6.

Images