Method for predicting driving assistance risk based on driver state monitoring and related device

Abstract

The application discloses an auxiliary driving risk prediction method based on driver state monitoring and related devices, and relates to the technical field of auxiliary driving risk prediction. In the application scheme, a clustering algorithm is used to carry out unsupervised clustering on historical driver physiological data and historical vehicle driving data, and the collision risk level of automobile operation is preliminarily divided. Then, according to a plurality of groups of historical driver physiological data, historical vehicle driving data and corresponding collision risk levels, a training data set and a test data set are constructed, and a trained collision risk prediction model is obtained. Finally, based on the measured vehicle driving data and the measured driver physiological data, the collision risk prediction model is used for risk prediction, so that the auxiliary driving risk prediction result is obtained. In addition, the auxiliary driving risk prediction result is also used to update the division of the collision risk level in combination with the actual situation. The application solves the problem that the traditional risk prediction method is only based on vehicle operation data, resulting in insufficient prediction accuracy, and improves the prediction accuracy of driving risk.

Description

Technical Field

[0001] This application relates to the field of assisted driving risk prediction technology, and in particular to an assisted driving risk prediction method and related device based on driver state monitoring. Background Technology

[0002] With the continuous development of intelligent driving technology, driver assistance systems (ADAS) are playing an increasingly significant role in improving driving safety and reducing traffic accidents. However, existing ADAS still have significant shortcomings when dealing with complex road conditions and emergencies. For example, their real-time monitoring and risk prediction capabilities for driver status are relatively weak. Traditional risk prediction methods typically rely solely on vehicle operating data, neglecting the crucial factor of driver physiological state, resulting in low accuracy in predicting emergencies. Moreover, during long drives or in complex road conditions, driver fatigue, inattention, and emotional fluctuations can significantly increase the risk of accidents. Therefore, there is an urgent need for a risk prediction system that can comprehensively integrate driver physiological state and vehicle operating data to more accurately determine the risk level during driving and improve the intelligence and safety performance of ADAS. Summary of the Invention

[0003] The purpose of this application is to provide a method and related device for predicting the risk of assisted driving based on driver status monitoring, which can accurately determine the risk level during the assisted driving process.

[0004] To achieve the above objectives, this application provides the following solution:

[0005] Firstly, this application provides a method for predicting assisted driving risks based on driver state monitoring, comprising the following steps:

[0006] Several sets of historical driver physiological data and corresponding historical vehicle driving data were collected during simulated vehicle driving within 5 seconds before the collision.

[0007] Clustering algorithms were used to perform unsupervised clustering on several sets of historical driver physiological data and historical vehicle driving data to preliminarily classify the collision risk levels of vehicle operation.

[0008] Based on several sets of historical driver physiological data and historical vehicle driving data, as well as the corresponding collision risk levels, a training dataset and a test dataset are constructed. The training dataset is used to train the collision risk prediction model, and the test dataset is used to test the performance of the collision risk prediction model.

[0009] Based on measured vehicle driving data and measured driver physiological data, a collision risk prediction model is used to predict risks and obtain assisted driving risk prediction results. The assisted driving risk prediction results are also used to update the collision risk level classification in combination with the actual situation.

[0010] Optionally, several sets of historical driver physiological data and corresponding historical vehicle driving data within 5 seconds before a collision are collected during simulated vehicle driving, specifically including the following steps:

[0011] The driver's pulse wave signal data was collected in the 5 seconds before the collision during simulated vehicle driving to serve as historical driver physiological data.

[0012] For any given historical driver's physiological data, the corresponding distance to the vehicle in front, vehicle travel time, and vehicle speed are collected as historical vehicle driving data.

[0013] Optionally, based on several sets of historical driver physiological data and historical vehicle driving data, as well as the corresponding collision risk levels, a training dataset and a test dataset are constructed, specifically including the following steps:

[0014] For any given historical driver physiological data, the driver's heart rate is calculated based on the historical driver physiological data.

[0015] For any given historical vehicle driving data, determine the maximum deceleration and the distance to the vehicle in front before the collision based on the historical vehicle driving data.

[0016] By taking any set of driver heart rate, corresponding maximum deceleration, distance from the vehicle in front before the collision, and vehicle speed as model inputs, and the corresponding collision risk level as a label, a sample data is constructed.

[0017] Several sample data points are divided into training datasets and test datasets according to a preset ratio.

[0018] Alternatively, the driver's heart rate can be calculated using the following formula:

[0019]

[0020] Where HR is the driver's heart rate, and t2-t1 represents the time difference between two consecutive peaks in the pulse wave data.

[0021] The maximum deceleration is determined by the following formula:

[0022]

[0023] Among them, a max For the maximum deceleration, v f v is the final velocity 5 seconds before the collision. iLet T be the initial velocity, and let T be the time required from the initial moment until 5 seconds before the collision.

[0024] The distance to the vehicle in front before the collision is determined using the following formula:

[0025] D i =D a -D b .

[0026] Among them, D i D represents the distance to the vehicle in front 5 seconds before the collision. a D represents the distance to the vehicle in front. b Distance to the main vehicle's position.

[0027] Optionally, a clustering algorithm is used to perform unsupervised clustering on several sets of historical driver physiological data and historical vehicle driving data to initially classify the collision risk level of vehicle operation. This includes the following steps:

[0028] Several sets of historical driver physiological data and historical vehicle driving data are input into the k-means model for training. Unsupervised learning is performed, and the model is trained under different numbers of clusters k to obtain the cluster centers and family assignments for each number of clusters k.

[0029] Based on the family allocation under each cluster number k, the target cluster number with the best clustering effect is selected, and the class centers obtained according to the target cluster number correspond to different collision risk levels.

[0030] Optionally, the collision risk prediction model is a model constructed and trained using the XGBoost algorithm. The training set data is input into the XGBoost algorithm model for training. Multiple decision trees are built iteratively, and the loss function is continuously optimized to gradually improve the overall prediction accuracy. After obtaining the trained collision risk prediction model, the test dataset is input into the collision risk prediction model to calculate the mean squared error loss. When the mean squared error loss is lower than a preset threshold, the collision risk prediction model is applied to the risk prediction of the actual vehicle driving data and the actual driver physiological data.

[0031] Secondly, this application provides an assisted driving risk prediction system based on driver state monitoring, including the following functional modules:

[0032] The historical driving simulation data acquisition module is used to collect several sets of historical driver physiological data and corresponding historical vehicle driving data within 5 seconds before a collision during simulated vehicle driving.

[0033] The preliminary collision risk level classification module is used to perform unsupervised clustering on several sets of historical driver physiological data and historical vehicle driving data using clustering algorithms to initially classify the collision risk level of vehicle operation.

[0034] The training and testing dataset construction module is used to build training and testing datasets based on several sets of historical driver physiological data and historical vehicle driving data, as well as the corresponding collision risk levels. The training dataset is used to train the collision risk prediction model, and the testing dataset is used to test the performance of the collision risk prediction model.

[0035] The assisted driving collision risk prediction module is used to predict risks based on actual vehicle driving data and actual driver physiological data using a collision risk prediction model, and obtain assisted driving risk prediction results; the assisted driving risk prediction results are also used to update the classification of collision risk levels in combination with actual conditions.

[0036] 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 state monitoring-based assisted driving risk prediction method described above.

[0037] 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-state monitoring-based assisted driving risk prediction method described above.

[0038] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the driver-state monitoring-based assisted driving risk prediction method described above.

[0039] According to the specific embodiments provided in this application, the following technical effects are disclosed:

[0040] This application provides a method and related device for predicting assisted driving risks based on driver state monitoring. The method first collects several sets of historical driver physiological data and corresponding historical vehicle driving data within 5 seconds before a collision occurs during simulated vehicle driving. Next, an unsupervised clustering algorithm is used to perform unsupervised clustering on these sets of historical driver physiological data and historical vehicle driving data to initially classify the collision risk levels of vehicle operation. Then, based on the sets of historical driver physiological data, historical vehicle driving data, and corresponding collision risk levels, training and testing datasets are constructed, resulting in a trained collision risk prediction model. Finally, based on actual vehicle driving data and actual driver physiological data, the collision risk prediction model is used to perform risk prediction, thereby obtaining the assisted driving risk prediction result. Furthermore, the assisted driving risk prediction result is also used to update the collision risk level classification based on actual conditions. This application simultaneously collects driver physiological data and vehicle driving data, achieving effective fusion of multi-dimensional data. It solves the problem of insufficient prediction accuracy in traditional risk prediction methods that rely solely on vehicle operation data, thus improving the accuracy of driving risk prediction. After obtaining the assisted driving risk prediction results, it also incorporates actual conditions into the collision risk level classification, enabling real-time updates of the collision risk level, optimizing the model's adaptability, and improving the intelligence level and safety performance of the assisted driving system. In summary, this application, by accurately judging collision risks during assisted driving, guides drivers to execute driving strategies under different traffic conditions, effectively improving following safety, traffic efficiency, and urban road traffic safety. Attached Figure Description

[0041] 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.

[0042] Figure 1 A flowchart illustrating an assisted driving risk prediction method based on driver state monitoring, provided as an embodiment of this application.

[0043] Figure 2 This is a flowchart of step S1 in an assisted driving risk prediction method based on driver state monitoring, provided as an embodiment of this application.

[0044] Figure 3 This is a flowchart of step S2 in an assisted driving risk prediction method based on driver state monitoring, provided as an embodiment of this application.

[0045] Figure 4This is a flowchart of step S3 in an assisted driving risk prediction method based on driver state monitoring, provided as an embodiment of this application.

[0046] Figure 5 This is a schematic diagram of the functional modules of an assisted driving risk prediction system based on driver status monitoring, provided in an embodiment of this application.

[0047] Figure 6 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0048] 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.

[0049] 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.

[0050] This application provides a method for predicting assisted driving risks based on driver state monitoring. In one exemplary embodiment, such as... Figure 1 As shown, it includes the following steps:

[0051] S1. Collect several sets of historical driver physiological data and corresponding historical vehicle driving data within 5 seconds before a collision during simulated vehicle driving. Driver physiological data is collected using a driver physiological signal intelligent measurement device, and vehicle driving data is collected using a vehicle trajectory recorder. In this embodiment, as... Figure 2 As shown, step S1 specifically includes the following steps:

[0052] S11. Collect pulse wave signal data of the driver within 5 seconds before the collision during simulated vehicle driving, as historical driver physiological data.

[0053] S12. For any historical driver's physiological data, collect the corresponding distance to the vehicle in front, vehicle travel time, and vehicle speed as historical vehicle driving data.

[0054] S2. Unsupervised clustering is performed on several sets of historical driver physiological data and historical vehicle driving data using a clustering algorithm to initially classify the collision risk level of vehicle operation. In this embodiment, as... Figure 3 As shown, step S2 specifically includes the following steps:

[0055] S21. Input several sets of historical driver physiological data and historical vehicle driving data into the k-means model for training, perform unsupervised learning, train under different cluster numbers k, and obtain the cluster centers and family assignments under each cluster number k.

[0056] S22. Based on the family allocation under each cluster number k, select the target cluster number with the best clustering effect, and divide each class center according to the target cluster number to correspond to different collision risk levels.

[0057] As an exemplary embodiment, a cluster number k of 3 is selected, which results in the optimal cluster allocation. The cluster centers are as follows: cluster0 represents a stable heart rate, large distance, and moderate speed, and is labeled as "low risk"; cluster1 represents a medium level and is labeled as "medium risk"; and cluster2 represents a high heart rate, fast speed, and short distance, and is labeled as "high risk".

[0058] S3. Based on several sets of historical driver physiological data and historical vehicle driving data, along with corresponding collision risk levels, a training dataset and a test dataset are constructed. The training dataset is used to train the collision risk prediction model; the test dataset is used to test the performance of the collision risk prediction model. In this embodiment, as... Figure 4 As shown, step S3 specifically includes the following steps:

[0059] S31. For any historical driver physiological data, calculate the driver's heart rate based on the historical driver physiological data. Specifically, calculate the driver's heart rate according to the following formula:

[0060]

[0061] Where HR is the driver's heart rate, and t2-t1 represents the time difference between two consecutive peaks in the pulse wave data.

[0062] S32. For any historical vehicle driving data, determine the maximum deceleration and the distance to the vehicle in front before the collision based on the historical vehicle driving data. In this embodiment, the maximum deceleration is determined according to the following formula:

[0063]

[0064] Among them, a max For the maximum deceleration, v f v is the final velocity 5 seconds before the collision. i Let T be the initial velocity, and let T be the time required from the initial moment until 5 seconds before the collision.

[0065] The distance to the vehicle in front before a collision can be determined using the following formula:

[0066] Di =D a -D b .

[0067] Among them, D i D represents the distance to the vehicle in front 5 seconds before the collision. a D represents the distance to the vehicle in front. b Distance to the main vehicle's position.

[0068] S33. Take any set of driver heart rate, corresponding maximum deceleration, distance from the vehicle in front before the collision, and vehicle speed as model inputs, and the corresponding collision risk level as labels to construct a sample data.

[0069] Specifically, the risk level assessment criteria are as follows:

[0070] High risk: Driver's heart rate ≥100 bpm, vehicle speed ≥80 km / h, distance from the vehicle in front ≤10m before collision, maximum deceleration ≥3.0 m / s² 2 .

[0071] Medium risk: Driver's heart rate is between 80-100 bpm, vehicle speed is between 40-80 km / h, the distance to the vehicle in front before the collision is 10-20m, and the maximum deceleration is 1.5-3.0 m / s². 2 .

[0072] Low risk: Driver's heart rate ≤ 80 bpm, vehicle speed ≤ 40 km / h, distance from the vehicle in front ≥ 20 m before collision, maximum deceleration ≤ 1.5 m / s² 2 .

[0073] S34. Divide several sample data sets into a training dataset and a test dataset according to a preset ratio. In this embodiment, the dataset is divided into a training dataset and a test dataset in a 4:1 ratio.

[0074] The training dataset is used to train the collision risk prediction model. Specifically, at the beginning of model training, initial prediction values ​​are set. (e.g., all sample labels y) i The mean of the values ​​is used as the baseline prediction result before training the first decision tree. At this point, the overall model's prediction result is... The prediction result of the t-th tree is the sum of the cumulative prediction results of the first t-1 trees and the prediction output of the t-th tree for sample i, as shown in the following formula:

[0075]

[0076] in, This represents the prediction results of the first t trees for sample i. f represents the cumulative prediction results of the first t-1 trees.t (X i ) represents the prediction output of the t-th tree for sample i.

[0077] Next, we construct the objective function. Based on the XGBoost algorithm structure defined in the scheme, the objective function includes a loss function and a regularization term, specifically:

[0078]

[0079] Where Obj(θ) is the objective function, θ are the model parameters, I represents the loss function, and y i Let Ω(f) be the true value of sample i. i ) is a regularization term used to control the complexity of the t-th decision tree (such as the depth of the tree and the number of leaf nodes), prevent the model from overfitting, and ensure reliability in practical applications. constant is a constant term in the XGBoost model.

[0080] In each iteration (round t), a new decision tree f is constructed. t The prediction error is reduced by optimizing the objective function. The specific process is as follows:

[0081] T1. Calculate the negative gradient (residual) to determine the learning objective of the new tree. Specifically, based on the current cumulative prediction value and the loss function, calculate the negative gradient for each sample, which serves as the target value that the t-th tree needs to fit. This negative gradient reflects the current model's prediction bias for sample i, and the training objective of the t-th tree is to minimize this bias.

[0082] T2. Generate candidate split points for the decision tree and construct the tree structure. Specifically, for the input feature vector X... i =[H i ,a i D i (i.e., heart rate, maximum deceleration, and distance between the two vehicles), the values ​​of each feature are sorted and candidate split points are generated (e.g., based on quantiles of feature values). All candidate split points are traversed, and the sum of squared residuals of the left and right subtrees after the split is calculated. The split point that maximizes the reduction of the sum of squared residuals after the split is selected, and the node structure of the decision tree is recursively constructed (e.g., splitting the "distance to the vehicle in front" feature first, then splitting the "heart rate" feature, etc.). In this embodiment, the regularization term Ω(f) is used. i Limit the depth of the tree and the number of leaf nodes (e.g., control the maximum depth and minimum number of split samples) to avoid overfitting due to an overly complex tree structure.

[0083] T3. Calculate the leaf node weights and optimize the output of the current tree. Specifically, once the decision tree structure is determined (i.e., the leaf nodes are partitioned), calculate the optimal weight w for each leaf node. j(j is the leaf node number), minimize the objective function. The t-th tree f t (x i The output of ) is the weight w of the leaf node to which sample i belongs. j .

[0084] T4. Update the cumulative prediction value to reduce the overall loss. Specifically, update the output f of the t-th tree. t (x i (Added to the previous cumulative forecast value) In the process, new predicted values ​​are obtained. At this point, the overall loss function value (mean squared error) decreases due to the reduction in residuals, thus improving prediction accuracy.

[0085] Repeat steps T1 to T4, continuously generating new decision trees (t = 1, 2, ..., T) until the iteration termination condition is met (e.g., reaching the preset maximum number of trees T, or the decrease in the loss function is less than a threshold). During training, model performance is optimized by adjusting hyperparameters (e.g., maximum tree depth, learning rate, regularization parameters, etc.). The scheme uses the training set to find the optimal hyperparameters to ensure the model's generalization ability on the test set and avoid overfitting.

[0086] After multiple iterations, an XGBoost model consisting of T decision trees is obtained. The final predicted value for sample i corresponds to the numerical label (0, 1, 2) of the risk level, combined with the risk level interval obtained from k-means clustering (e.g., ...). Compared with cluster centers), when For low risk, when At the time, it was medium risk, when When the risk level is high, the system can output predictions for low, medium, and high risks, enabling accurate prediction of driving risks.

[0087] The test dataset is used, specifically, to calculate the mean squared error function according to the following formula, to determine the accuracy of the model:

[0088]

[0089] when This indicates that the collision risk prediction model obtained through training is accurate and can be applied.

[0090] S4. Based on measured vehicle driving data and measured driver physiological data, a collision risk prediction model is used to predict the risk, resulting in an assisted driving risk prediction result. This result is also used to update the collision risk level classification based on actual conditions. The assisted driving risk prediction result is the predicted collision risk level. Based on the risk level predicted by the collision risk model, relevant indicators such as driver heart rate, distance between vehicles, vehicle speed, and driving speed are calculated. The reasonableness of the prediction result is then assessed in conjunction with actual conditions. If unreasonable, the relevant indicators can be adjusted; if reasonable, no changes are needed. In practical applications, the range of relevant indicators can also be determined based on local traffic conditions.

[0091] Compared with the prior art, the method provided in the above embodiments of this application has the following beneficial effects:

[0092] 1. In this embodiment, the data acquisition module simultaneously collects driver physiological data and vehicle driving data, realizing the effective fusion of multi-dimensional data. This solves the problem that traditional risk prediction methods, which rely solely on vehicle operation data, have insufficient prediction accuracy, thus improving the accuracy of driving risk prediction.

[0093] 2. The k-means unsupervised learning algorithm is used to classify collision risk levels, which can scientifically divide low, medium and high risk levels, achieve accurate classification of different risk levels, and balance the impact of different types of data on the overall model in the classification process.

[0094] 3. The XGBoost algorithm model was selected for risk prediction. This algorithm builds multiple decision trees iteratively and continuously optimizes the loss function, which helps to solve the overfitting problem that may occur during model training and ensures the reliability of the model in practical applications.

[0095] 4. By feeding back the prediction results from the risk prediction module to the risk level judgment module, the risk level is updated in real time, the adaptability of the model is optimized, and the intelligence level and safety performance of the assisted driving system are improved.

[0096] 5. The method provided in this application can more accurately determine the risk level during the driving process, guide drivers to implement driving strategies under different traffic conditions, effectively improve the safety of following other vehicles, and improve traffic efficiency and urban road traffic safety.

[0097] Based on the same inventive concept, this application also provides a system for implementing the aforementioned driver-state monitoring-based assisted driving risk prediction method. The solution provided by this system is similar to the implementation described in the above method. In an exemplary embodiment, such as... Figure 5As shown, a driver assistance risk prediction system based on driver state monitoring is provided, including:

[0098] The historical driving simulation data acquisition module is used to collect several sets of historical driver physiological data and corresponding historical vehicle driving data within 5 seconds before a collision during simulated vehicle driving.

[0099] The preliminary collision risk level classification module is used to perform unsupervised clustering on several sets of historical driver physiological data and historical vehicle driving data using clustering algorithms to initially classify the collision risk level of vehicle operation.

[0100] The training and testing dataset construction module is used to build training and testing datasets based on several sets of historical driver physiological data and historical vehicle driving data, as well as the corresponding collision risk levels. The training dataset is used to train the collision risk prediction model, and the testing dataset is used to test the performance of the collision risk prediction model.

[0101] The assisted driving collision risk prediction module is used to predict risks based on actual vehicle driving data and actual driver physiological data using a collision risk prediction model, and obtain assisted driving risk prediction results; the assisted driving risk prediction results are also used to update the classification of collision risk levels in combination with actual conditions.

[0102] certainly, Figure 5 The architecture shown is merely exemplary; it can be omitted as needed when implementing different functionalities. Figure 5 One or at least two components of the system shown.

[0103] 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 6 As shown, the computer device includes a processor, memory, input / output interfaces (I / O), 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 an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it can implement the driver state monitoring-based assisted driving risk prediction method provided in the previous embodiment.

[0104] Those skilled in the art will understand that Figure 6 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.

[0105] In one exemplary embodiment, a computer device is also 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.

[0106] 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.

[0107] 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.

[0108] 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.

[0109] Those skilled in the art will understand that all or part of the processes in the methods of 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 of the above methods. 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 (ReRAM), 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).

[0110] 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, distributed databases based on blockchain. The processors involved in the embodiments provided in this application may be, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc.

[0111] 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.

[0112] 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 predicting a driving risk based on driver state monitoring, characterized by, include: Collect several sets of historical driver physiological data and corresponding historical vehicle driving data within 5 seconds before the collision during simulated vehicle driving. The historical driver physiological data is the driver's pulse wave signal data, and the historical vehicle driving data is the corresponding distance to the vehicle in front, vehicle travel time, and vehicle speed. A clustering algorithm is used to perform unsupervised clustering on several groups of historical driver physiological data and historical vehicle driving data to preliminarily classify the collision risk level of vehicle operation; The clustering algorithm is a k-means model, specifically including: inputting several sets of historical driver physiological data and historical vehicle driving data into the k-means model for training, performing unsupervised learning, training at different numbers of clusters k, obtaining the cluster centers and family assignments for each number of clusters k, selecting the target number of clusters with the optimal clustering effect based on the family assignments for each number of clusters k, and assigning different collision risk levels to the cluster centers obtained based on the target number of clusters; the target number of clusters is 3, and the three corresponding cluster centers correspond to low risk, medium risk, and high risk levels, respectively. Low risk is defined as stable heart rate, large distance, and moderate speed; medium risk is defined as moderate level; and high risk is defined as high heart rate, high speed, and close distance. The criteria for high risk are: driver's heart rate ≥100 bpm, vehicle speed ≥80 km / h, distance to the vehicle in front ≤10m before collision, and maximum deceleration ≥3.0 m / s². 2 The criteria for determining medium risk are: driver's heart rate between 80-100 bpm, vehicle speed between 40-80 km / h, distance from the vehicle in front before the collision between 10-20 m, and maximum deceleration between 1.5-3.0 m / s². 2 The criteria for determining low risk are: driver's heart rate ≤ 80 bpm, vehicle speed ≤ 40 km / h, distance from the vehicle in front ≥ 20 m before collision, and maximum deceleration ≤ 1.5 m / s². 2 ; Based on several sets of historical driver physiological data and historical vehicle driving data, as well as the corresponding collision risk levels, a training dataset and a test dataset are constructed; the training dataset is used to train the collision risk prediction model; the test dataset is used to test the performance of the collision risk prediction model. Based on measured vehicle driving data and measured driver physiological data, the collision risk prediction model is used to predict risks and obtain assisted driving risk prediction results. The collision risk prediction model is a model built and trained using the XGBoost algorithm. The training dataset is input into the XGBoost algorithm model for training. By iteratively building multiple decision trees and continuously optimizing the loss function, the overall prediction accuracy is gradually improved. After obtaining the trained collision risk prediction model, the test dataset is input into the collision risk prediction model to calculate the mean squared error loss. When the mean squared error loss is lower than a preset threshold, the collision risk prediction model is applied to the risk prediction of measured vehicle driving data and measured driver physiological data. The assisted driving risk prediction results are also used to update the collision risk level classification based on actual conditions. Based on several sets of historical driver physiological data and historical vehicle driving data, along with corresponding collision risk levels, a training dataset and a test dataset are constructed, specifically including: For any of the aforementioned historical driver physiological data, the driver's heart rate is calculated based on the historical driver physiological data; the driver's heart rate is calculated according to the following formula: ； wherein, HR for the driver heart rate, t 2- t 1 represents the time difference between two consecutive peaks of the pulse wave data; For any of the aforementioned historical vehicle driving data, determine the maximum deceleration and the distance to the vehicle in front before the collision based on the historical vehicle driving data; determine the maximum deceleration according to the following formula: ； in, a max For maximum deceleration, v f The final velocity 5 seconds before the collision. v i The initial velocity, T This represents the time required from the initial moment to 5 seconds before the collision. The distance to the vehicle in front before the collision is determined using the following formula: ； in, D i The distance between the vehicle and the vehicle in front is 5 seconds before the collision. D a Distance to the vehicle in front D b Distance to the main vehicle's position; Using any set of driver heart rate, corresponding maximum deceleration, distance from the vehicle in front before the collision, and vehicle speed as model inputs, and the corresponding collision risk level as a label, a sample data is constructed. Several sample data points are divided into training datasets and test datasets according to a preset ratio. 2.The method of claim 1, wherein, Several sets of historical driver physiological data and corresponding historical vehicle driving data were collected during simulated vehicle driving within 5 seconds prior to a collision. Specifically, these included: The driver's pulse wave signal data was collected within 5 seconds before the collision during simulated vehicle driving to serve as historical driver physiological data. For any given historical driver's physiological data, the corresponding distance to the vehicle in front, vehicle travel time, and vehicle speed are collected as historical vehicle driving data.

3. A system for predicting the risk of assisted driving based on driver state monitoring, characterized by, include: The historical driving simulation data acquisition module is used to collect several sets of historical driver physiological data and corresponding historical vehicle driving data within 5 seconds before a collision during simulated vehicle driving. The historical driver physiological data is the driver's pulse wave signal data, and the historical vehicle driving data is the corresponding distance to the vehicle in front, vehicle travel time, and vehicle speed. The collision risk level preliminary classification module is used to perform unsupervised clustering on several groups of historical driver physiological data and historical vehicle driving data using a clustering algorithm to initially classify the collision risk level of vehicle operation. The clustering algorithm is a k-means model, specifically including: inputting several sets of historical driver physiological data and historical vehicle driving data into the k-means model for training, performing unsupervised learning, training under different cluster numbers k, obtaining the cluster centers and family assignments for each cluster number k, selecting the target cluster number with the best clustering effect based on the family assignments for each cluster number k, and assigning different collision risk levels to the cluster centers obtained based on the target cluster number; the target cluster number is 3, and the three corresponding cluster centers correspond to low risk, medium risk, and high risk levels, respectively, where low risk is characterized by stable heart rate, large distance, and moderate speed, medium risk is characterized by a moderate level, and high risk is characterized by high heart rate, high speed, and short distance; The training and testing dataset construction module is used to construct training datasets and testing datasets based on several sets of historical driver physiological data and historical vehicle driving data, as well as the corresponding collision risk levels; the training dataset is used to train the collision risk prediction model; the testing dataset is used to test the performance of the collision risk prediction model. The assisted driving collision risk prediction module is used to predict risks based on actual vehicle driving data and actual driver physiological data using the collision risk prediction model, and obtain assisted driving risk prediction results. The collision risk prediction model is a model built and trained using the XGBoost algorithm. The training dataset is input into the XGBoost algorithm model for training. By iteratively building multiple decision trees and continuously optimizing the loss function, the overall prediction accuracy is gradually improved. After obtaining the trained collision risk prediction model, the test dataset is input into the collision risk prediction model to calculate the mean squared error loss. When the mean squared error loss is lower than a preset threshold, the collision risk prediction model is applied to the risk prediction of actual vehicle driving data and actual driver physiological data. The assisted driving risk prediction results are also used to update the collision risk level classification based on actual conditions. The criteria for high risk are: driver's heart rate ≥100 bpm, vehicle speed ≥80 km / h, distance to the vehicle in front before the collision ≤10m, and maximum deceleration ≥3.0 m / s². 2 The criteria for determining medium risk are: driver's heart rate between 80-100 bpm, vehicle speed between 40-80 km / h, distance from the vehicle in front before the collision between 10-20 m, and maximum deceleration between 1.5-3.0 m / s². 2 The criteria for determining low risk are: driver's heart rate ≤ 80 bpm, vehicle speed ≤ 40 km / h, distance from the vehicle in front ≥ 20 m before collision, and maximum deceleration ≤ 1.5 m / s². 2 ; The training and testing dataset building module is specifically used for: For any of the aforementioned historical driver physiological data, the driver's heart rate is calculated based on the historical driver physiological data; the driver's heart rate is calculated according to the following formula: ； wherein, HR for the driver heart rate, t 2- t 1 represents the time difference between two consecutive peaks of the pulse wave data; For any of the aforementioned historical vehicle driving data, determine the maximum deceleration and the distance to the vehicle in front before the collision based on the historical vehicle driving data; determine the maximum deceleration according to the following formula: ； in, a max For maximum deceleration, v f The final velocity 5 seconds before the collision. v i The initial velocity, T This represents the time required from the initial moment to 5 seconds before the collision. The distance to the vehicle in front before the collision is determined using the following formula: ； in, D i The distance between the vehicle and the vehicle in front is 5 seconds before the collision. D a Distance to the vehicle in front D b Distance to the main vehicle's position; Using any set of driver heart rate, corresponding maximum deceleration, distance from the vehicle in front before the collision, and vehicle speed as model inputs, and the corresponding collision risk level as a label, a sample data is constructed. Several sample data points are divided into training datasets and test datasets according to a preset ratio.

4. 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 driver-assisted driving risk prediction method based on driver state monitoring as described in any one of claims 1-2.

5. A computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by the processor, it implements the driver assistance risk prediction method based on driver state monitoring as described in any one of claims 1-2.

6. A computer program product comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the driver assistance risk prediction method based on driver state monitoring as described in any one of claims 1-2.

Images