A driver skill classification method and driving assistance system

Abstract

The present invention relates to a driver skill classification method and a driving assistance system. The method comprises: obtaining the current vehicle state and the current driver state in real time; calculating a vehicle stability margin based on the current vehicle state; and using the vehicle stability margin and the current driver state as inputs to a trained classification model to obtain a corresponding driver skill classification result. The vehicle stability margin is used to map the driver's situational awareness and vehicle control capabilities and is calculated based on an offline determined vehicle stability boundary, while the current driver state is used to map driving burden. Compared with existing technologies, the present invention has the advantages of high classification accuracy and reduced accident probability.

Description

Technical Field

[0001] The technology of the present invention mainly relates to the field of assisted driving, and in particular to a driver skill classification method and a driving assistance system. Background Art

[0002] Smart cars have attracted significant attention due to their revolutionary potential for traffic safety and mobility. A survey titled "A Survey of Personalization for Advanced Driver Assistance Systems" indicates that standardized driver assistance systems can cause discomfort to drivers, leading to a reluctance to activate them. This phenomenon is particularly pronounced among experienced drivers. This significantly reduces the use of driver assistance systems while driving, rendering them ineffective in reducing traffic accidents. Therefore, design methods for personalized driver assistance systems are urgently needed.

[0003] The paper "Changes in collision rates among novice drivers during the first months of driving" notes that novice female drivers and experienced young male drivers are the primary types of traffic participants involved in traffic accidents. The paper "Research on Classification and Recognition of Driving Styles Based on Feature Engineering" classifies drivers based on vehicle speed, acceleration, and throttle opening. This type of method can classify drivers based on their driving style, but it cannot characterize driver skills, which is one of the key indicators for personalized assisted driving design. The paper "APattern-Recognition Approach for Driving Skill Characterization" uses steering wheel information from a double lane-change test to classify drivers. The paper notes that experienced drivers have a more nuanced perception of vehicle status and make more frequent corrections under extreme conditions than novice drivers. Therefore, the accuracy of this method is severely affected in everyday driving conditions. Further research is needed to classify driving skills based on everyday driving conditions.

[0004] Driving skills are manifested in three aspects: understanding of driving scenarios, control of vehicles, and driving burden. Among them, the understanding of driving scenarios is reflected in the decision-making of driving behavior through situational awareness, such as the speed and deceleration point when entering a curve, which is the main cause of driving accidents. However, in an open environment, it is impossible to enumerate all traffic scenarios (including vehicle speed, curve curvature, road conditions, etc.). Therefore, a comprehensive indicator that can map various driving scenarios is needed. In addition, the driving burden that the driver can bear and the control accuracy of the vehicle are also the main factors for the driver to avoid risks in emergencies. Therefore, this field urgently needs an evaluation system that can comprehensively evaluate driver skills. Summary of the Invention

[0005] One of the purposes of the present invention is to overcome the defects of the above-mentioned prior art and provide a driver skill classification method with high classification accuracy.

[0006] The second purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a personalized driving assistance system with high usage rate and reduced probability of accidents.

[0007] The purpose of the present invention can be achieved by the following technical solutions:

[0008] A driver skill classification method, the method comprising:

[0009] Get the current vehicle status and current driver status in real time;

[0010] Calculating a vehicle stability margin based on the current vehicle state;

[0011] Using the vehicle stability margin and the current driver state as inputs to a trained classification model to obtain a corresponding driver skill classification result;

[0012] The vehicle stability margin is used to map the driver's situational awareness and vehicle control capabilities, and is calculated based on the vehicle stability boundary determined offline. The current driver state is used to map the driving burden.

[0013] Furthermore, the vehicle stability margin is obtained by calculating a minimum Euclidean distance between a current vehicle state and a vehicle stability boundary in the current state.

[0014] Furthermore, when the current vehicle state includes the vehicle body center of mass sideslip angle β and yaw angular velocity When , the calculation formula of vehicle stability margin is:

[0015]

[0016] Where D sis the vehicle stability margin in the current state, β is the current vehicle body center of mass side slip angle; is the current yaw angular velocity; D i is the Euclidean distance from the current state to the i-th point on the vehicle stability boundary, is a point on the vehicle's stable boundary.

[0017] Furthermore, the process of determining the vehicle stability boundary specifically includes:

[0018] Construct vehicle dynamics models;

[0019] Based on the vehicle dynamics model, obtaining a root locus diagram of vehicle state changes;

[0020] The state points in the root locus diagram are classified and trained by a support vector machine to obtain a classification hyperplane, which is used as a vehicle stability boundary.

[0021] Furthermore, the vehicle dynamics model adopts a two-degree-of-freedom vehicle model and a Dugoff tire model.

[0022] Furthermore, the root locus diagram is obtained by performing a third-order Runge-Kutta formula iteration on each set of states.

[0023] Furthermore, the vehicle stability boundaries in different states are stored in the form of a query data table.

[0024] Furthermore, the classification model is a Gaussian mixture-hidden Markov classification model, a neural network model or a deep neural network model.

[0025] Furthermore, the current driver state is a secondary task accuracy or a physiological acquisition signal.

[0026] The present invention also provides a personalized driving assistance system, comprising:

[0027] A driver skill classification module obtains corresponding driver skill classification results based on the above driver skill classification method and identifies the driver skill level online;

[0028] The assistance intensity determination module determines the corresponding assistance intensity based on the driver's skill level to achieve personalized assisted driving.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] 1. The present invention uses causal analysis to identify vehicle stability margin and current driver status as driving skill characteristics to achieve driving skill classification. Vehicle stability margin is used to map the driver's situational awareness and vehicle control capabilities. Because drivers all aim to quickly complete driving tasks, similar drivers will have similar vehicle stability margin distributions in driving scenarios of varying complexity. Current driver status (such as secondary task accuracy and physiological information) also shares similar principles. Therefore, the present classification method can accurately classify driving skills using a relatively small number of everyday driving scenarios, making the classification model easy to train and highly reliable.

[0031] 2. The present invention adopts a machine learning method to obtain the vehicle stability margin, which has the advantages of higher accuracy and less manual workload.

[0032] 3. The present invention realizes the assisted driving function based on the driving skill classification results, and can determine the corresponding assistance intensity according to the driver's status in a personalized manner, thereby improving the utilization rate of the assistance system and reducing the probability of accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a schematic diagram of the principle of the present invention;

[0034] Figure 2 This is a schematic diagram of the skill classification principle based on driving behavior and vehicle posture;

[0035] Figure 3 The schematic diagram of driver skill classification based on causal analysis;

[0036] Figure 4 Obtaining a schematic diagram for the vehicle stability boundary based on support vector machines;

[0037] Figure 5 When the front wheel angle is 0° and the longitudinal speed is 20 kmph root locus plot;

[0038] Figure 6 An aerial view of the test road. DETAILED DESCRIPTION

[0039] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0040] Example 1

[0041] Because the driving scenarios faced by vehicles operating in open environments are complex and ever-changing, exhaustive analysis is impossible. Furthermore, since routine driving conditions predominate during daily driving, evaluating driver skills based on performance under extreme conditions is also problematic. Therefore, driver skill classification methods that rely on driving scenarios are difficult to implement.

[0042] This embodiment provides a driver skill classification method, comprising: obtaining a current vehicle state and a current driver state in real time; calculating a vehicle stability margin based on the current vehicle state; and using the vehicle stability margin and the current driver state as inputs to a trained classification model to obtain a corresponding driver skill classification result.

[0043] In this method, a driving skill classification method based on the two characteristics of vehicle stability margin distribution and secondary task accuracy during driving is designed through causal analysis. This method includes two parts: offline calculation and online recognition and classification. The offline calculation is composed of two parts: vehicle stability margin acquisition and classification model training. Figure 1 This method uses the vehicle stability margin distribution to map the driver's situational awareness and vehicle control capabilities, while the secondary task accuracy maps the driving burden, enabling a comprehensive assessment of the driver's driving skills. Furthermore, because drivers all aim to quickly complete driving tasks, similar drivers will have similar vehicle stability margin distributions in driving scenarios of varying complexity. Therefore, this method offers the advantage of accurately classifying driving skills using a relatively small number of everyday driving scenarios.

[0044] The above method mainly includes the selection of driving skill features based on causal analysis, the acquisition of vehicle stability margin based on support vector machine, and the construction of classification model, which are described in detail as follows.

[0045] 1. Driving skill feature selection based on causal analysis

[0046] Driving skills are manifested in three aspects: understanding of driving scenarios, control of vehicles, and the ability to handle sudden tasks. Among them, the understanding of driving scenarios is reflected in the decision-making of driving behavior through situational awareness, such as the speed and deceleration point when entering a curve. This is the main cause of driving accidents. However, in an open environment, it is impossible to enumerate all traffic scenarios (including vehicle speed, curve radius, road conditions, etc.). Similarly, the evaluation of the ability to control the vehicle and handle sudden tasks also depends on the scenario, so it also faces the problem that driving scenarios cannot be enumerated. Through causal reasoning analysis, the skill classification principle based on driver behavior and vehicle posture is as follows: Figure 2As shown in the figure, driver behavior, vehicle posture, and driving environment are strongly correlated. Generally speaking, a set of driving environments corresponds to a set of driver behavior and vehicle posture. This type of method is highly accurate in specific driving environments, but its generalization is weak. The formula for calculating the probability of driving skill categories derived from these correlation factors is as follows:

[0047] P(y)=P(H w ,H b ,H s |x r ,x t ) (1)

[0048] Where: x r is the road environment, such as curve curvature, road adhesion coefficient, etc.; x t is the traffic flow environment, such as traffic flow density, types of traffic participants, etc.; H w It is the steering wheel signal under the driving environment, such as steering wheel angle range, speed, etc.; H b It is the pedal signal under the driving environment, such as pedal depth, pedal angle change speed, etc.; H s is the vehicle status in the driving environment, such as speed, acceleration, etc.; y is the driver category.

[0049] As shown in formula (1), the skill classification based on driver behavior and vehicle posture depends on the driving environment. Generally, the driver skill classification model should have different model coefficients in different driving environments. However, the driving environment faced by vehicles driving in a natural open environment is complex and changeable. It is unrealistic to achieve accurate classification of driving skills through exhaustive enumeration. To this end, the present invention proposes a driving skill classification method based on two evaluation features, vehicle stability margin distribution and secondary task accuracy during driving, based on causal reasoning. The driver's situational awareness ability and vehicle control ability are mapped through the vehicle stability margin distribution; the driving burden is mapped through the secondary task accuracy to characterize the driver's ability to deal with emergencies. The principle of driver skill classification based on causal analysis is as follows: Figure 3 shown.

[0050] First, the first part involves selecting the vehicle stability margin distribution. In simple driving conditions (low road curvature and high road adhesion), experienced drivers tend to drive faster. Conversely, in more complex driving environments, experienced drivers tend to reduce their speed. This approach aims to ensure that experienced drivers complete the driving task as quickly as possible while ensuring vehicle safety. This also leads to experienced drivers maintaining a specific stability margin distribution for their vehicles. Therefore, ignoring the influence of driver burden, the design of the vehicle stability margin distribution creates a chain-like independent condition in causal reasoning. In other words, the vehicle stability margin distribution can be used to infer the driver's skill category, regardless of the driving environment. The second part, focusing on secondary task accuracy, follows a similar principle. In dense traffic conditions, experienced drivers tend to reduce their speed, reducing their control burden and increasing their observation burden. The opposite is true in sparse traffic conditions. This approach aims to ensure that the driving task is completed as quickly as possible while still leaving enough time to detect and handle unexpected situations. This also leads experienced drivers to maintain a specific driving burden distribution. Similarly, if the influence of the vehicle stability margin distribution is ignored, the node of secondary task accuracy also forms a chain structure independent condition in causal reasoning. In other words, the driver's skill category can be deduced through the secondary task accuracy, regardless of the driving environment. Finally, the probability calculation method of the driving skill category is obtained through causal reasoning, as shown in formula (2). It realizes the comprehensive evaluation and classification of driving skills. At the same time, the driver skill classification system based on causal analysis classifies driving skills through the two evaluation indicators of vehicle stability margin distribution and secondary task accuracy, avoiding the problem of exhaustive driving environment, significantly reducing the cost and time of driver skill classification model development, and improving the generalization ability of the driver skill classification model.

[0051] P(y)=P(x vs ,x st ) (2)

[0052] Where: x vs is the vehicle stability margin distribution; x st is the secondary task accuracy; y is the driver category.

[0053] In addition to using the accuracy of the secondary task to evaluate the driver's driving burden, but not limited to this, in other implementations, physiological information (such as heart rate, skin electrical response, and blood flow, etc.) can also be used to represent it, which is easier to promote and industrialize. In this embodiment, the secondary task accuracy test is simply to assign a second secondary task to the driver in the driving task. If the secondary task accuracy is higher, it means that the burden of the main task (that is, the driving task) is lower, otherwise it is vice versa. The secondary task designed in this embodiment is to broadcast the task of changing the arrow direction on the central screen for 2 seconds. It is designed based on "Automotive Human Factors Engineering" and will not be explained further here.

[0054] 2. Vehicle stability margin acquisition based on support vector machine

[0055] First, the vehicle stability margin acquisition method based on support vector machine is to obtain the classification boundary of the root locus diagram - vehicle stability boundary through machine learning. This part is completed offline. Compared with the method of manually selecting approximate boundaries, it can greatly reduce labor costs while improving the acquisition of more accurate vehicle stability boundaries. Then, the distance from the current vehicle state to the stability boundary is used to obtain the vehicle stability margin of the current state. This is done online. It is mainly divided into four parts: vehicle dynamics construction, root locus diagram acquisition, vehicle stability boundary learning and vehicle stability margin acquisition. The structure is as follows: Figure 4 shown.

[0056] 1) Vehicle dynamics construction

[0057] This embodiment uses a two-degree-of-freedom vehicle model and the Dugoff tire model to construct the vehicle dynamics model. Simpler or more complex models can be used depending on the computing platform's capabilities, but are not limited to these. This serves as the basis for obtaining the root locus plot in the second section. The two-degree-of-freedom vehicle model and tire model are shown in Equations (3) and (4) below. Due to the maturity of the technology, detailed derivation is not provided here.

[0058]

[0059] Where: β is the side slip angle of the vehicle body center of mass; is the rate of change of the vehicle body's center of mass sideslip angle; is the yaw angular velocity; is the yaw angular acceleration; m is the vehicle mass; v is the longitudinal speed; F y is the front wheel cornering force; F yr is the rear wheel cornering force; δ f is the front wheel turning angle; L f is the distance from the vehicle's center of mass to the front axle; L r is the distance from the vehicle's center of mass to the rear axle; I z is the vehicle's moment of inertia.

[0060] F y =f·C y α

[0061]

[0062] Where: F y is the front wheel cornering force; C x ,C y are the longitudinal and lateral stiffness of the tire respectively; α is the tire slip angle; f is the tire cornering force deformation coefficient; F R is the deformation judgment coefficient; F z is the vertical load of the tire; s is the tire slip rate; μ e is the peak road adhesion coefficient. The above formula is also applicable to the calculation of rear wheel cornering force.

[0063] 2) Root locus diagram acquisition

[0064] This part is to obtain the root locus diagram. This embodiment adopts the method of iterating the 3rd order Runge-Kutta formula 100 times for each set of states. The state here consists of two parts. The first part is the input. From formula (3), we can see that the vehicle input is δ f The front wheel steering angle and v longitudinal speed. The front wheel steering angle range used here is ±20° with an accuracy of 1°; the longitudinal speed range is 0 to 120 kmph with an accuracy of 5 kmph. The second part is the initial state. From formula (3), we can see that the vehicle state is β body center of mass side slip angle and Yaw angular velocity. The center of mass slip angle range used here is ±1 radian, with an accuracy of 0.1 radian; the yaw angular velocity range is ±2 radians per second, with an accuracy of 0.1 radians per second. Under each set of inputs, all initial states should be traversed, and each set of initial states should be iterated 100 times through the 3rd-order Runge-Kutta formula. All the calculation results under a set of inputs are plotted on a graph to obtain the root locus. For example, the front wheel angle is 0° and the longitudinal speed v is 20kmph. Figure 5 shown.

[0065] 3) Vehicle stability boundary learning

[0066] The present invention adopts support vector machine to learn the vehicle stability boundary. The specific method is to judge whether the state belongs to a stable point based on whether the root trajectory in the state converges to the closed-loop pole. The classification task is then performed by the support vector machine. Since the classification principle of the support vector machine is to segment the data through a hyperplane, it is the same as the definition of the required vehicle stability boundary (that is, the state points outside the stable boundary are unstable points, and the points within the stable boundary are stable points). Therefore, the hyperplane obtained by the support vector machine to classify the state points in the root trajectory diagram is the vehicle stability boundary. The above 1) to 3) parts are all completed offline, and the final vehicle stability boundary will be saved in the controller as static data and will not be calculated again.

[0067] 4) Vehicle stability margin acquisition

[0068] The final vehicle stability margin is obtained by calculating the minimum Euclidean distance from the current state to the vehicle stability boundary in the current state. This part is calculated online in real time, and the calculation formula is as follows:

[0069]

[0070] Where: D s is the vehicle stability margin in the current state; β is the current vehicle body center of mass side slip angle; is the current yaw angular velocity; D i is the Euclidean distance from the current state to the i-th point on the vehicle stability boundary; is a point on the vehicle's stable boundary.

[0071] 3. Based on Gaussian mixture-hidden Markov classification model

[0072] The vehicle stability margin and secondary task accuracy obtained above serve as inputs to a Gaussian mixture-hidden Markov classification model, with the labels of the participating drivers used as recognition targets. Once the classification model's recognition error stabilizes (i.e., the proportion of data in the dataset where the classified labels are inconsistent with the actual labels no longer decreases), model training is stopped and the important coefficients of the classification model are saved. This part is also completed offline. In actual use, the assistance system program will directly call the trained model to classify driver skills, enabling online identification of the driver's skill level and providing personalized assistance intensity.

[0073] In other implementations, the classification model is not limited to the Gaussian mixture-hidden Markov classification model, and a neural network model, a deep neural network model, etc. may also be used.

[0074] If the above method is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program code.

[0075] Example 2

[0076] An overly conservative driving assistance system will make experienced drivers unwilling to turn on the assisted driving function, which results in the driving assistance system not being able to work when help is needed because it is not turned on, such as the active braking system. An overly aggressive assistance system will also be of little help to novice drivers, because novice drivers need more thinking time and tolerance for operating accuracy to avoid accidents due to their lack of driving skills. In order to achieve the goal of reducing traffic accidents through assisted driving functions, it is necessary to design a personalized assisted driving system. Among them, driving skills are one of the most important driver characteristics. This embodiment provides a personalized driving assistance system based on driver skills, including a driver skill classification module and an assistance strength determination module, wherein the driver skill classification module obtains the corresponding driver skill classification result based on the driver skill classification method as described in Example 1, and identifies the driver skill level online; the assistance strength determination module determines the corresponding assistance strength based on the driver skill level to achieve personalized assisted driving.

[0077] In one specific embodiment, the process of using the personalized driving assistance system can be described as including the following steps:

[0078] 1. Modify the parameters in the vehicle dynamics model according to the vehicle that needs to be equipped with auxiliary systems. The main parameters are: m is the vehicle mass; L f is the distance from the vehicle's center of mass to the front axle; L r is the distance from the vehicle's center of mass to the rear axle; I z is the moment of inertia of the vehicle; C x ,C y are the longitudinal and lateral stiffness of the tire, respectively.

[0079] 2. According to the root locus diagram acquisition method, the δf The front wheel angle and longitudinal vehicle speed v are input Root locus diagram and save it offline.

[0080] 3. Use support vector machine to The roots in the root locus diagram are classified and the classification hyperplane is obtained, which is the vehicle stability boundary, and is made into a f The data table for querying the front wheel angle and v longitudinal speed is saved offline.

[0081] 4. Set up the secondary task implementation environment. Assign the driver the arrow broadcast task by generating a random number every 2 seconds to change the direction of the arrow on the car's central screen.

[0082] 5. Design the vehicle driving environment. To ensure the safety of the test, a driving simulation is used. Gradual curvature roads are more likely to cause traffic accidents, and China's road law stipulates that the minimum turning distance on main roads is less than 20-30m. Therefore, a test road consisting of 6 straight lines and 6 curves with a radius of 20m is designed. Figure 6 shown.

[0083] 6. Recruit a number of drivers and fill out the driver rating form. Driving age is the main labeling indicator. Here, drivers with more than one year of driving experience are considered experienced drivers, and vice versa.

[0084] 7. Conduct driving tests and obtain vehicle stability margin in real time by calculating the minimum Euclidean distance from the current state to the vehicle stability boundary in the current state. Also, obtain secondary task accuracy by comparing the direction of the broadcast arrow with the direction shown on the display. These values ​​are recorded using data acquisition equipment.

[0085] 8. Set a 5s data observation window and use the vehicle stability margin data and secondary task accuracy data within the time window as Gaussian mixture-hidden Markov model input to train the driver skill classification model. When the recognition error of the classification model stabilizes, stop model training and save the important coefficients of the classification model.

[0086] The above workflow for the driver skill classification system, based on causal analysis, was completed before the personalized assistance system was released. After training the Gaussian mixture-hidden Markov classification model, only the classification model parameters and vehicle stability boundary parameters are used online in real time. During online real-time use, the current vehicle state is collected, and the vehicle stability margin is derived based on the vehicle stability boundary parameters. Secondary task accuracy data is also collected. The vehicle stability margin and the secondary task accuracy data are used together as inputs to the classification model to obtain the corresponding driver skill classification results. The driver's skill level is identified online, and the corresponding assistance intensity is determined based on the driver's skill level, achieving personalized assisted driving.

[0087] The above describes in detail the preferred embodiments of the present invention. It should be understood that those skilled in the art can make numerous modifications and variations based on the concepts of the present invention without inventive effort. Therefore, any technical solutions that can be derived by those skilled in the art through logical analysis, reasoning, or limited experimentation based on the concepts of the present invention and the prior art should be within the scope of protection defined by the claims.

Claims

1. A driver skill classification method, characterized in that: The method includes: Get the current vehicle status and current driver status in real time; Calculating a vehicle stability margin based on the current vehicle state; Using the vehicle stability margin and the current driver state as inputs to a trained classification model to obtain a corresponding driver skill classification result; The vehicle stability margin is used to map the driver's situational awareness and vehicle control capabilities, and is calculated based on the offline determined vehicle stability boundary. The current driver state is used to map the driving burden. The vehicle stability margin is obtained by calculating the minimum Euclidean distance between the current vehicle state and the vehicle stability boundary in the current state; When the current vehicle state includes the vehicle body center of mass sideslip angle β and yaw angular velocity When , the calculation formula of vehicle stability margin is: Where D s is the vehicle stability margin in the current state, β is the current vehicle body center of mass side slip angle; is the current yaw angular velocity; D i is the Euclidean distance from the current state to the i-th point on the vehicle stability boundary, is a point on the vehicle's stable boundary; The process of determining the vehicle stability boundary specifically includes: Construct vehicle dynamics models; Based on the vehicle dynamics model, obtaining a root locus diagram of vehicle state changes; Performing classification training on the state points in the root locus diagram by using a support vector machine to obtain a classification hyperplane, and using the classification hyperplane as a vehicle stability boundary; The vehicle dynamics model adopts a two-degree-of-freedom vehicle model and a Dugoff tire model; The root locus diagram is obtained by performing a third-order Runge-Kutta formula iteration on each set of states.

2. The driver skill classification method according to claim 1, characterized in that: The vehicle stability boundaries in different states are stored in the form of a query data table.

3. The driver skill classification method according to claim 1, characterized in that: The classification model is a Gaussian mixture-hidden Markov classification model, a neural network model or a deep neural network model.

4. The driver skill classification method according to claim 1, characterized in that: The current driver status is a secondary task accuracy or a physiological acquisition signal.

5. A personalized driving assistance system, characterized in that: include: a driver skill classification module, which obtains corresponding driver skill classification results based on the driver skill classification method according to claim 1 and identifies the driver skill level online; The assistance intensity determination module determines the corresponding assistance intensity based on the driver's skill level to achieve personalized assisted driving.

Images