CN102795225B - Method for detecting disturbance state of driver by utilizing driver-side longitudinal control model - Google Patents

Abstract

The present invention provides a method for detecting driver interference state by using a driver model, the method comprising: obtaining the driving environment information of the vehicle by an external sensor, and judging the driving intention according to the relationship between the position of the vehicle and the boundary of the lane; when the driving intention is to keep the lane , the driver model calculates the steering wheel angle change amount according to the driving environment information and vehicle state information, calculates and outputs the accelerator opening according to the current vehicle speed and the expected longitudinal acceleration determined by the interval time between the vehicle and the vehicle in front; when the driving intention is to change lanes , the driver model calculates the steering wheel angle change amount according to the driving environment information and state information, calculates and outputs the throttle opening according to the current vehicle speed and the expected longitudinal acceleration calculated from the lateral acceleration; the steering wheel angle and throttle opening calculated by the driver model Compare with the real driver's control data obtained by internal sensors to determine whether the driver is disturbed; repeat the above steps until the car stops.

Description

A Method of Detecting Driver's Interference State Using Driver's Side Longitudinal Control Model

technical field

The invention relates to a method for detecting a driver's interference state by using a driver's side longitudinal comprehensive control model. In terms of detecting driver interference, the method uses the driver model to simulate the driver control data (steering wheel angle and throttle opening) in the current driving environment in real time, and compares the control data of the driver model with the control data of the real driver. And identify whether the driver is disturbed according to the comparison result. The method has low hardware cost and high recognition accuracy, and is especially suitable for developing an auxiliary driving system for vehicles.

Background technique

The interference of the driver (for example, distracted, eating, drinking water/beverage, making a phone call, chatting with passengers, and using the vehicle system, etc.) is an important cause of traffic accidents, so it is necessary to develop detection methods to detect the driver's state of mind in real time. Interference state, in order to remind the driver in time in combination with traffic conditions, reduce or avoid the occurrence of accidents.

At present, the detection of driver interference mainly focuses on the use of the driver's physiological characteristics, such as continuously tracking the movement of the eyes to determine whether the driver is affected by visual interference. Although the driver's physiological representation can reflect the driver's disturbed state, it is also very deceptive. For example, when the driver is cognitively disturbed (such as thinking about problems while driving), the driver's facial expression may Nothing has changed, and such interference is difficult to detect with existing methods. On the contrary, the driver's control behavior on the vehicle can truly reflect the driver's state. Research shows that the driver's control data under normal conditions and the control data under disturbance are very different.

Contents of the invention

The purpose of the present invention is to provide a new method for detecting driver interference, which uses a driver model to simulate the driver's control behavior in real-time in the current driving environment, and compares the control data of the real-time simulation of the driver model with the real driver's By controlling the data, it is possible to accurately determine whether the driver is being disturbed.

According to the purpose of the present invention, a driver model includes: a path planning module, which receives the driving environment information from the vehicle's external sensors, decides the driver's driving intention (change lane or keep the lane), plans the corresponding driving trajectory, and real-time according to The position of the vehicle body is constantly corrected; the preview module uses the driver's preview characteristics to obtain the expected trajectory data according to the data from the path planning module; the prediction module obtains the predicted trajectory data according to the state information from the vehicle dynamics model; the first The comparison module receives the expected trajectory data and the predicted trajectory data, and obtains the lateral deviation data through the comparison of the two trajectory data; the lateral control module calculates the steering wheel angle change and the lateral acceleration according to the lateral deviation data And output it to the second comparison module and the longitudinal control module respectively; The longitudinal control module calculates the expected longitudinal acceleration according to the lateral acceleration, and then calculates the throttle opening value according to the expected longitudinal acceleration and outputs it to the vehicle dynamics model; The second comparison module calculates the final steering wheel angle according to the change amount of the steering wheel angle and the current steering wheel angle of the vehicle dynamics model and outputs it to the vehicle dynamics model.

According to the purpose of the present invention, a method for detecting the driver's interference state using the above-mentioned driver model is proposed, the method includes: step a, obtaining the driving environment information of the vehicle through the vehicle external sensor, and comparing the current position of the vehicle with the lane in which the vehicle is located The relationship between the boundaries, judging the driving intention of the driver to change lanes or keep the lane; step b, if it is judged that the driving intention is to keep the lane, the driver model uses PD control to calculate the amount of change in steering wheel angle according to the driving environment information and state information of the vehicle , and calculate and output the accelerator opening value according to the current vehicle speed and the expected longitudinal acceleration determined by the expected interval time (safety time) between the vehicle and the vehicle in front; step c, if it is judged that the driving intention is to change lanes, the driver model is based on the vehicle's The driving environment information and state information use PD control to obtain the amount of change in the steering wheel angle, and calculate and output the accelerator opening value according to the current speed of the vehicle and the desired longitudinal acceleration determined by the lateral acceleration; step d, step b or c The steering wheel angle and accelerator opening data of the vehicle are compared with the corresponding control data of the real driver obtained from the vehicle's internal sensors to determine whether the driver is disturbed; step e, repeat steps a-d until the vehicle stops.

The method also includes: before step a, connecting the vehicle external sensor and the vehicle internal sensor with the computer, debugging the vehicle external sensor and the vehicle internal sensor, and initializing the driver model.

The criterion for judging whether the driver is disturbed uses a time window of 1s and an update amount of 0.25s for data processing. Among them, the criterion for judging whether the driver is disturbed uses the steering wheel angle, accelerator opening value and the data of the real driver within 1s The cumulative difference of is used as the classification feature, and the discriminant function for judging whether the driver is disturbed is a support vector machine (SVM) whose kernel function is a Gaussian function.

Input the classification features calculated above into the support vector machine model, if the result of the model is greater than 0, it can be judged that the driver is disturbed, otherwise, there is no disturbance.

The driver model of the present invention is based on the existing queuing network cognitive system. While tracking the expected trajectory, it can accurately simulate the driving characteristics and physiological limitations of real drivers, and can reflect the driving behavior of skilled drivers . The environmental perception sensors outside the vehicle obtain the current driving environment information in real time, and output this information to the driver model in real time. The driver model calculates driver control data in real time based on information from environmental perception sensors. The real-time control data of the real driver is obtained through the internal sensors of the vehicle, and by comparing the driving data calculated by the driver model with the control data of the real driver, it is possible to accurately determine whether the driver is disturbed.

The present invention has the advantages of: proposing a method for detecting whether the driver is disturbed by comparing the simulated control data of the driver model and the control data of the real driver in real time, the detection process itself will not interfere with the driver, and The hardware is easy to implement and the cost is low; the driving data can truly reflect the driving state of the driver, and the difference between the simulation data of the driver model and the driving data of the real driver can be used to identify whether the driver is disturbed, which can greatly reduce the error of interference detection. Judgment rate, improve the accuracy of detection, and promote the intelligence of the assisted driving system.

Description of drawings

Figure 1 is a schematic diagram of driver interference detection using a driver model.

Fig. 2 is a structure diagram of the driver model in Fig. 1 .

Fig. 3 is a flowchart of a method for detecting driver interference by using a driver model.

Detailed ways

A method for detecting a driver's disturbance state according to the present invention will be described in detail below with reference to the accompanying drawings. In the present invention, to simplify the description, it is assumed that the driver model is on the same expected trajectory as the real driver, although the present invention is not limited thereto.

Figure 1 is a schematic diagram of driver interference detection using a driver model. As shown in Figure 1, the principle of driver interference detection using the driver model is as follows: (1) Real-time acquisition of driving environment information (road type (straight or curved), vehicle current lane position information, vehicle speed) through the vehicle’s external sensors , the distance to the front car and the rear car in the same lane, the speed of the vehicle in the adjacent lane (the front car and the rear car) and the distance to the vehicle); (2) The driver model receives data from the external sensor on the vehicle, simulates The driving behavior of a skilled driver simulates the driving behavior in the current driving environment in real time, and outputs the driver control data to the vehicle dynamics model to realize the closed-loop correction of the simulated driving behavior. The driving control data output from the driver model is used as (3) While performing step (2), obtain the control data of the real driver on the vehicle through the vehicle sensor in real time; (4) The comparison module receives the driver model control data from step (2) and the control data from the step ( 3) The real driver's control data, by comparing the driver model and the real driver's control data, it can be accurately judged whether the driver is disturbed.

The driver model shown in Figure 1 is based on the existing queuing network cognitive system. While tracking the expected trajectory, it can accurately simulate the driving characteristics and physiological limitations of real drivers, and can reflect the skillful driving driver's driving behavior. The driver model will be described in detail below with reference to FIG. 2 .

As shown in Figure 2, the driver model includes a path planning module, a preview module, a prediction module, a comparison module 1, a comparison module 2, a lateral control module, and a longitudinal control module.

On the one hand, the driving environment information from the vehicle’s external sensors is input to the path planning module, and the path planning module determines the driver’s driving intention (change lane or keep the lane) and plans the corresponding driving trajectory, and constantly corrects it according to the position of the car body in real time. The aiming module obtains expected trajectory data by using the driver's preview characteristics according to the data from the path planning module.

On the other hand, the vehicle state information (S _n (x, y, a _x , a _y , v _x , v _y , yaw) from the vehicle dynamics model, where x is the vehicle body side coordinates, y is the vehicle body longitudinal coordinates, a _x is the longitudinal acceleration, a _y is the lateral acceleration, v _x is the longitudinal velocity, v _y is the lateral velocity, yaw is the vehicle body yaw angle) input to the prediction module, and the prediction module outputs the predicted trajectory according to the state information data. Both the expected trajectory data and the predicted trajectory data are input into the comparison module 1, so that the comparison module 1 can compare the two trajectory data to obtain lateral deviation data E (described below). The lateral control module calculates the steering wheel angle change Δδ (described below) according to the lateral deviation data E from the comparison module 1 and outputs it to the comparison module 2 and the longitudinal control module respectively. The longitudinal control module calculates the final throttle opening value α (to be described below) and outputs it to the vehicle dynamics model, and the comparison module 2 is based on the steering wheel angle change Δδ from the lateral control module and the current steering wheel angle of the vehicle dynamics model ( (to be described later) calculate and output the final steering wheel angle δ (to be described below) and output it to the vehicle dynamics model. This enables a closed-loop correction of the driving behavior of the driver model.

Since the realization of the comparison/comparison of the sensor detection-comparison module/comparison module/comparison module-control module (for example, software mode, hardware mode, etc.) involved in the description of Fig. 1 and Fig. 2 belongs to the prior art, here The description will not be repeated.

The method for detecting driver interference by using the driver model will be described in detail below with reference to FIG. 3 .

The process of driver interference detection using the driver model is as follows:

In steps 301 and 302, debug the vehicle external sensors and vehicle internal sensors to make them work normally, initialize each module of the driver model and ensure that the clocks of each module are consistent, and at the same time ensure that the real driver's driving time is consistent with each sensor and driving time. The activation time of the employee model is the same. The various sensors are connected to a computer (eg, an on-board computer) to communicate with the driver model.

In step 303, the driver drives the vehicle forward, the vehicle's external sensors obtain the current driving environment information in real time, and the vehicle's internal sensors obtain the vehicle's state information in real time. Of course, individual sensors may be interrupted by conditions (eg, sensor failure, vehicle parking, etc.).

In steps 304 and 305, the driver model obtains the driving environment information of the vehicle according to the vehicle’s external sensors, and judges the driver’s driving behavior by comparing the relationship between the current position of the vehicle (that is, the position in the lane) and the boundary of the lane. Intent (change lane or keep lane).

According to the driving intention (change lane or keep lane) obtained in step 305, the driver model performs path planning according to the driving safety criteria based on the current driving environment information obtained from the vehicle’s external sensors, and performs continuous corrections in real time according to the vehicle’s position information . The planned path obtained according to the driving intention is used as the expected trajectory of the driver model, and the driver model tracks the expected trajectory to obtain the corresponding driving behavior benchmark. In this way, the expected trajectory obtained through path planning can ensure that the driver model is on the same or approximately the same expected trajectory as the real driver.

According to the driving intention obtained in step 305, the driver model selects different control modes to track the expected trajectory obtained in step 305 according to the corresponding driving intention (lane changing or lane keeping).

If the driving intention obtained in step 305 is to keep the lane, then in steps 306, 307, 308 and 309, the distance d _n between the two vehicles (the vehicle and its front) and the speed of the front vehicle v _hn and the vehicle state information S _n (x _n , y _n , a _xn , a _yn , v _xn , v _yn , yaw _n ) acquired by the vehicle internal sensors are sent to the driver model. The preview module of the driver model obtains the expected trajectory point P _n (x _n , y _n ) within the current preview time (T _p = 1.5s) through the planned path obtained in step 305, and the prediction module obtains the vehicle internal sensor State information S _n (x _n ,y _n ,a _xn ,a _vn ,v _xn ,v _vn ,yaw _n ) and predict the location coordinates of the vehicle within the preview time T _p From this, the lateral position error E _n of the expected trajectory and the predicted trajectory can be obtained. In order to accurately track the expected trajectory, it is necessary to adjust the steering wheel angle to reduce the lateral position deviation. In the driver model, the PD control is used to obtain the amount of change in the steering wheel angle. In the vertical aspect, in order to ensure the safety of driving and prevent rear-end collision accidents, it is necessary to ensure that the distance between the two vehicles is within the safe range (or greater than the safe time). Therefore, the interval time t _cn between the two vehicles is calculated, and the expected longitudinal acceleration a _xn is proportional to the difference between the interval time t _cn and the safety time t _f (t _f =4s) between the two vehicles, thus according to the vehicle dynamics Relationship Get the desired longitudinal acceleration and the accelerator opening corresponding to the current speed (positive value represents acceleration, negative value represents braking), the formula is as follows:

${\mathrm{<span\; class="notranslate">\; E.</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}^{\mathrm{<span\; class="notranslate">\; q</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; yn</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; \&Center\; Dot;</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; E.</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; yn</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; yn</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; yn</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}}{{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>$

Δδ _n =k _p a _yn +k _d a' _yn (4)

δ _n = δ _n-1 + Δδ _n (5)

${\mathrm{<span\; class="notranslate">\; t</span>}}_{\mathrm{<span\; class="notranslate">\; cn</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; [</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; xn</span>}}<span\; class="notranslate">\; \&CenterDot;</span>{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 0.5</span>}<span\; class="notranslate">\; \&CenterDot;</span>{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; hn</span>}}{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ]</span><span\; class="notranslate">\; /</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; xn</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; xn</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 6</span>}<span\; class="notranslate">\; )</span>$

$\begin{array}{c}{{\mathrm{<span\; class="notranslate">\; a</span>}}^{\mathrm{<span\; class="notranslate">\; d</span>}}}_{\mathrm{<span\; class="notranslate">\; xn</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; K</span>}<span\; class="notranslate">\; \&Center\; Dot;</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; t</span>}}_{\mathrm{<span\; class="notranslate">\; cn</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; t</span>}}_{\mathrm{<span\; class="notranslate">\; f</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 7</span>}<span\; class="notranslate">\; )</span>\\ {\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; xn</span>}}<span\; class="notranslate">\; ,</span>{{\mathrm{<span\; class="notranslate">\; a</span>}}^{\mathrm{<span\; class="notranslate">\; d</span>}}}_{\mathrm{<span\; class="notranslate">\; xn</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 8</span>}<span\; class="notranslate">\; )</span>\end{array}$

According to the formula (1), the lateral position deviation E _n of the nth step is calculated by subtracting the lateral coordinates of the expected trajectory points from the lateral coordinates of the predicted trajectory points. According to the formula (2), the n-th step lateral velocity v _vn is obtained according to the internal sensor of the vehicle, and the lateral acceleration a _yn to reach the expected position is calculated. According to the formula (3), the derivative a′ _yn of the nth step lateral acceleration can be obtained by dividing the difference between the nth step lateral acceleration and the n-1th step lateral acceleration by the preview time. According to the formula (4), through the PD control, the amount of change in the steering wheel angle of the nth step Δδ _n is obtained. According to the formula (5), the steering wheel angle of step n-1 is added to the steering wheel angle change Δδ _n to obtain the final steering wheel angle δ _n . In terms of calculating the longitudinal control parameters (i.e. accelerator or brake opening), the interval time t _cn between the two vehicles at the nth step is calculated by formula (6), and the expected longitudinal acceleration is obtained by formula (7) Finally, the final accelerator opening value α _n is obtained from the look-up table provided by formula (8), where f(v _xn , a ^d _xn ) is a dynamic function expression about the vehicle engine, and different longitudinal accelerations and longitudinal velocities correspond to different Throttle opening.

Since how to implement PD control belongs to the prior art, it will not be described here.

From the above description, it can be seen that the driver model obtains the expected trajectory according to the external sensors on the vehicle, obtains control commands by simulating the driving characteristics of skilled drivers under normal driving conditions, and then outputs them to the vehicle dynamics model.

If the driving intention obtained in step 305 is to change lanes, then in steps 310, 311, 312 and 313, the driving environment information of the adjacent lane obtained through the vehicle external sensor and the vehicle state information S obtained through the vehicle internal sensor _n (x _n ,y _n ,a _xn ,a _vn ,v _xn ,v _vn ,yaw _n ) is sent to the driver model. The calculation methods of the preview module, the prediction module and the steering wheel angle in the driver model are the same as steps 306, 307, 308 and 309, the difference lies in the longitudinal acceleration. In order to simulate the physiological limitations of real drivers, when changing lanes, the expected longitudinal acceleration is determined according to the lateral acceleration, and the accelerator opening corresponding to the expected longitudinal acceleration is obtained according to the relationship of vehicle dynamics (positive values represent acceleration, negative values represent brake), the formula is as follows:

a ^d _{x n} =K·a _yn +a _d (9)

α _n =f(v _xn ,a ^d _xn ) (10)

The steering wheel angle needed to track the desired trajectory is calculated from the above equations (1)-(5). The desired longitudinal acceleration is calculated by formula (9), where a _yn is the lateral acceleration calculated by formula (2), and K, a _d are constants. The throttle opening value at the current speed and desired longitudinal acceleration is calculated from the look-up table provided by equation (10).

From steps 306-309 or steps 310-313, the driving behavior data M (δ _Mn , α _Mn ) of the driver model simulation under normal driving conditions (for example, keeping the lane or changing lanes) can be obtained, and the real driving The driver’s real-time control data D (δ _Dn , α _Dn ), and then judge whether the driver is disturbed by other tasks through the comparison module (see Figure 1). The judgment criterion is shown in the following formula:

${\mathrm{<span\; class="notranslate">\; RMSE</span>}}_{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}<span\; class="notranslate">\; =</span>\sqrt{\frac{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; D.</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; no</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 11</span>}<span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; RMSE</span>}}_{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}<span\; class="notranslate">\; =</span>\sqrt{\frac{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; D.</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; no</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 12</span>}<span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; SVM</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; RMSE</span>}}_{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; RMSE</span>}}_{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}<span\; class="notranslate">\; )</span>$

In this judgment criterion, a time window of 1s and an update amount of 0.25s are used. Calculate the cumulative difference between the steering wheel angle and accelerator opening value of the driver model and the real driver’s data within 1s from the formulas (11) and (12) respectively, and use the Gaussian kernel function data to convert the formula (11 ), (12) The calculated eigenvalues are input into the support vector machine function (SVM). According to the calculation results, if the output value is greater than 0, it can be judged whether the driver is affected by the disturbance (step 314) (that is, the above calculated The classification feature is input to the support vector machine model, if the result of the model is greater than 0, it can be judged that the driver is disturbed, otherwise, the driver is not disturbed).

Vehicle external sensors and vehicle internal sensors acquire driving environment information and vehicle internal state information in real time, and continuously transmit the information to the driver model, that is, repeat the above steps 304-315 until parking.

Claims (7)

1. A method utilizing a driver model to detect driver disturbance state, said method comprising: Step a, obtain the driving environment information of the vehicle through the vehicle external sensor, and judge the driving intention of the driver to change lanes or keep the lane by comparing the relationship between the current position of the vehicle and the boundary of the lane; Step b, if the driving intention is to keep the lane, the driver model uses PD control to calculate the amount of change in the steering wheel angle according to the driving environment information and state information of the vehicle, and according to the current vehicle speed and the desired time determined by the interval between the vehicle and the vehicle in front Calculate the longitudinal acceleration and output the throttle opening value; Step c, if the driving intention is to change lanes, the driver model uses PD control to calculate the steering wheel angle change amount according to the driving environment information and state information of the vehicle, and calculates the desired longitudinal direction according to the current speed of the vehicle and the lateral acceleration Acceleration calculation and output throttle opening value; Step d, comparing the steering wheel angle change amount and accelerator opening value in step b or c with the real driver's control data obtained from the vehicle's internal sensors, and judging whether the driver is disturbed; Step e, repeat steps a-d until parking, Among them, the driver model includes: The path planning module receives the driving environment information from the vehicle's external sensors, decides the driver's driving intention to change lanes or keep the lane, plans the corresponding driving trajectory, and continuously corrects it in real time according to the position of the vehicle body; The preview module utilizes the preview feature of the driver to obtain expected trajectory data according to the data from the path planning module; Prediction module, calculates forecast track data according to vehicle state information; The first comparison module receives expected trajectory data and predicted trajectory data, and obtains lateral deviation data by comparing the two trajectory data; The lateral control module calculates the steering wheel angle change and lateral acceleration according to the lateral deviation data and outputs them to the second comparison module and the longitudinal control module respectively; The longitudinal control module calculates the accelerator opening value according to the desired longitudinal acceleration and the current speed of the vehicle and outputs it to the vehicle dynamics model; The second comparison module calculates the final steering wheel angle according to the change amount of the steering wheel angle and the current steering wheel angle of the vehicle dynamics model and outputs it to the vehicle dynamics model, thereby realizing closed-loop correction of the driving behavior of the driver model. 2. The method according to claim 1, further comprising: before step a, connecting the vehicle external sensors and the vehicle internal sensors with the computer, debugging the vehicle external sensors and the vehicle internal sensors, and initializing the driver model. 3. The method of claim 2, wherein the computer is an onboard computer. 4. The method according to claim 1, wherein, according to step a, the driver model performs path planning according to driving safety criteria to obtain an expected trajectory. 5. The method according to claim 1, wherein the driving environment information includes the distance between two vehicles and the speed of the vehicle in front, and the state information is obtained by a vehicle internal sensor, Among them, the driver model compares the expected trajectory point within the current preview time with the position coordinates to be reached by the vehicle to obtain the lateral position error between the expected trajectory and the predicted trajectory, and calculate the steering wheel angle change. 6. The method according to claim 1, wherein the criterion for judging whether the driver is disturbed adopts a time window of 1s, an update amount of 0.25s, Among them, the criterion for judging whether the driver is disturbed is to use the cumulative difference between the steering wheel angle and accelerator opening value obtained by the model and the corresponding control data of the real driver within 1 second as the classification feature, Wherein, the function for judging whether the driver is disturbed is a support vector machine (SVM) whose kernel function is a Gaussian function. 7. The method according to claim 6, wherein the above-mentioned calculated classification features are input into the support vector machine model, if the result of the model is greater than 0, it can be judged that the driver is disturbed, otherwise, it is not disturbed.