EP1674375B1

EP1674375B1 - Method for determining a measure for evaluating the behaviour of a driver of a vehicle

Info

Publication number: EP1674375B1
Application number: EP20040030208
Authority: EP
Inventors: Wolfgang Birk; Mattias Brännström; Daniel Levin
Original assignee: Ford Global Technologies LLC
Current assignee: Ford Global Technologies LLC
Priority date: 2004-12-20
Filing date: 2004-12-20
Publication date: 2009-03-11
Anticipated expiration: 2024-12-20
Also published as: EP1674375A1; DE602004019953D1

Description

Field of the invention

The present invention relates to a method for determining a measure useful for evaluating the behaviour of a driver of a vehicle.

Background of the invention

Driver impairment due to e.g. drowsiness, illness or intoxication, is a major accident cause that craves a high number of lives each year. Although the number of safety features in a vehicle increases, the availability of systems that are able to reliably determine levels of impairment is still very limited. Although research has been going on for more than 40 years, there is still no off-the-shelf solution available. This is due to the fact that most systems are not reliable enough.
Several systems are known which attempt to evaluate a driver's behaviour in order to determine whether it is adequate or not, which in turn may be used to determine whether the driver is impaired due to e.g. drowsiness, illness, distraction or intoxication.
One type of current solution is based on driver monitoring with a camera where the eye-lid and head/gaze behaviour of the driver is analysed. Although this approach uses a direct measurement of the driver, it suffers from reliability issues due to light conditions, glasses and facial features of the driver.
Another type of current solution is based on analysis of the driver's performance in the lane keeping task.
US 5 745 031 describes a safety driving system for detecting a driver's alertness comprising an operation detecting section which updates monotonousness each time one of the driver-operated devices is operated, a steering section for deriving a steering amount and a running position detecting section for detecting a zigzag amount denoting a deviation of white lines detected based on an image of a road surface. Information from the different sections is used to determine the driver's alertness.
US 5 821 860 describes a driving condition-monitoring apparatus for monitoring a driving condition of a driver. A sensed yaw rate and speed are used to calculate a reference line of a lane along which the vehicle should travel. Also a parameter indicative of the deviation of the vehicle from this reference line is calculated, based on the sensed yaw rate and vehicle speed. The driving condition of the driver is calculated based on the deviation from the reference line.
US 6 335 689 describes a method in which the lateral displacement of a vehicle is detected consecutively and subjected to a frequency conversion to obtain a frequency component power. A driver's arousal level is judged based on an evaluation value obtained from performing integrations of the frequency component power.
US 5 925 082 describes a method in which the alertness of the vehicle operator is determined by collecting data on manual steering torque applied to the manual steering input means. The occurrences of abrupt corrective steering operations are counted and used to evaluate the alertness of the driver. In addition, when the alertness is low, a steering control system takes control over the steering of the vehicle.
From the paper P. H. Batavia, D. A. Pomerleau, C. E. Thorpe: "Predicting Lane Position for Roadway Departure Prevention" IEEE Intelligent vehicles 1998 Symposium, 28-30 Oct. 1998 Stuttgart, Germany, two methods are known for estimating time to lane crossing (TLC) wherein the first method is a kinematic projection of the present lateral position and velocity and the second is a memory based learning method.
The paper "Identification of driver state for lane keeping tasks" by T Pilutti, A. Galip, IEEE transactions on systems, man and cybernetics part A: systems and humans Vol. 29, No.5. Sept 1999, discloses the link between a mean lateral distance to lane boundaries and driver state.
A problem with prior art systems based on analysis of the driver's performance is that they do not provide a sufficiently high performance level, and may lead to non-conclusive results. Sometimes, these approaches suffer from misinterpretation of the driver's wobblings within the lane markings and do not show how well the driver actually behaves on the road.

Summary of the invention

In view of the above, it is an object of the present invention to provide a method for determining a measure useful for evaluating the impairment of a driver of a vehicle that is more reliable than prior art methods.
This object is achieved according to the invention by a method for determining a measure for evaluation of the behaviour of a driver of a vehicle, according to claim 1.
Thus, a new impairment measure namely the "planned deviation" is provided according to the invention. This impairment measure is intended to be used to evaluate slow impairment processes like drowsiness or fatigue.
The "state of the vehicle" may be described using variables indicating the way in which the driver controls the vehicle. The state of the vehicle may include the current speed and yaw rate, but also other variables such as acceleration or deceleration.
Since the planned path is the path that the vehicle would follow considering the state of the vehicle at the first instant, there is a possibility of adapting the system by selecting the variables used to describe the state. If for example only the speed and yaw rate of the vehicle are used to describe its state, the planned path will be the path followed if speed and yaw rate are held constant from the first instant. In a more complicated model, the acceleration of the vehicle could also be used, and the planned path would be calculated not using a constant speed, but assuming a constant acceleration.
The lateral position of the vehicle is the lateral position of the vehicle in relation to a road or lane. The lateral position may be determined using various lane tracking or GPS systems; particularly advantageous alternatives for determining the lateral position will however be described in the following.
The planned lateral position is the lateral position to which the vehicle would arrive a certain time/distance ahead in case the vehicle state is held constant from a current lateral position. The path followed by the vehicle if the vehicle state is held constant is referred to as the planned path. The planned path and the planned lateral position may be determined using a number of different methods. In the following, a method giving particularly good results will be described.
By comparing a current lateral position at a certain time with the planned lateral position a certain interval of time ahead if the vehicle were to continue on the planned path, the planned deviation being useful as a measure of the driver's ability to plan ahead is obtained. All drivers make steering corrections based on the upcoming road shape and traffic situation. Impaired drivers lose the ability to plan ahead to some extent. The inventors have discovered that the variability and/or magnitude of the planned deviation measure increases with increasing drowsiness of the driver. Accordingly, the variability and/or magnitude of the planned deviation may be studied to evaluate the driver's impairment.
Although in this application reference is made to a first and a second instant, and to a time interval, this is to be understood to be equivalent to referring to a first and a second distance, and to a distance interval. The person skilled in the art will readily translate the description being made using time as a variable so as to involve distance measures instead.
Calculations of the planned deviation measure may preferably be performed at subsequent instances having a selected frequency. This frequency would preferably selected such that the time gap between the instances is considerably less than the time period between the first and the second instant used in the planned deviation calculation. The frequency may for instance be selected in the order of magnitude of 10 Hz.
According to the invention, a number of planned deviation measures is calculated at subsequent instants and the deviation measures are used for the evaluation. the planned deviation measures are then compiled by forming a mean over the planned deviation measures, a sum thereof, obtaining a distribution thereof, a filtration or other compilation resulting in a measure representative of a number of planned deviation measures. In the following, this measure will be referred to as a Mean Planned Deviation measure (MPD), although it shall be understood from the above that this measure need not strictly be a mean of a number of planned deviation measures.
For impaired drivers, the MPD values tend to increase over time (as the driver gets more and more impaired), whereas for alert drivers the MPD values stay relatively constant, This property of the MPD signal can be used to detect impaired drivers in a very accurate way.
According to the invention, the MPD is compared with a threshold value in order to determine whether the driver is impaired or not.
The threshold value may be individually adapted for the driver. Advantageously, a threshold determining algorithm may be used wherein the driver's normal mean planned deviation is determined during a selected interval of a ride, and a threshold value for that driver is set in relation thereto. If desired, the threshold value may be updated during the ride.
As will be shown in the following examples, the Mean Planned Deviation with a self-adapting threshold algorithm has been found to be highly useful for drawing correct conclusions regarding the driver's state in numerous test drives performed.
The data representative of the state of the vehicle may advantageously comprise data representative of the vehicle' s speed and yaw rate. The planned path may be calculated using the vehicle's position, speed and yaw rate at the first instant t1 as indata. In this case, the planned path may advantageously be calculated using a linear bicycle model.
The lateral positions may advantageously be determined using a lane-tracking system including a camera.
In a particularly advantageous embodiment of the invention, the method comprises retrieving and storing data representative of the vehicle's state and retrieving and storing data representative of the surrounding environment at a first instant t1, retrieving and storing data representative of the vehicle's state and retrieving and storing data representative of the surrounding environment of the vehicle during an estimation period after said first instant, and using the data retrived and stored during the estimation period for determining the planned lateral position of the vehicle at the second instant.
This embodiment has the advantage that the actual road situation is accounted for in that data of the surrounding environment and vehicle's state during the estimation period is used in order to evaluate the behaviour of the driver at the start of the interval.
Thus, many disadvantages due to incorrect estimation of an upcoming road situation, as might occur in many prior art systems, are avoided. With the suggested method, the evaluation of the driver's behaviour cannot be made in real time, since an estimation period must elapse before the behaviour at the beginning of the time interval may be evaluated. For purposes of determining driver impairments such as drowsiness, however, this is not a problem. Driver drowsiness is a slow process, on the minutes-scale, and hence a delay of a few seconds in the evaluation is not crucial.
The estimation period Δt_TS may be at least 0.5 s, preferably at least 2 s, most preferred at least 5 s.
Further, the data representative of the surrounding environment may include information regarding surrounding movable objects, such as vehicles, and stationary objects. Accordingly, the surrounding traffic situation of the vehicle during the interval may also be taken into account when evaluating the driver's behaviour.
In a second aspect, the invention relates to a system for warning a driver using a method for determining a measure of a driver's behaviour as described above, and issuing or not issuing a warning based on said measure.
In a third aspect, the invention relates to a device having means for performing a method as described above.
Additional features and advantages of the invention will appear more clearly from the following detailed description of a preferred embodiment of the invention, which is given by way of non-limiting example only and with reference to the accompanying drawings.

Brief description of the drawings

Figs. 1a and 1b illustrate the current lateral position and the planned vehicle path of a vehicle.
Fig. 2 is a schematic picture of the planned vehicle path, from (x ₀, y ₀) to (x_p , y_p ) along a circle arc of length S with radius R.
Fig. 3 is an overview of the variables used for the planned lateral position calculation.
Fig. 4 is a diagram illustrating absolute Planned Deviation values for a driver when alert (top) and drowsy (bottom).
Fig. 5 is a flow chart illustrating an embodiment of a system using a method for estimating a traffic scenario according to the invention.

Detailed description of preferred embodiments of the invention

The method of the invention will now be described in relation to an embodiment where it is used in combination with a method for estimating a traffic situation. It is to be understood that the method for determining a measure for evaluation of the behaviour of a driver is not limited to this particular embodiment, although it has been found to be particularly advantageous.
An embodiment of a method for determining a measure for evaluation of the behaviour of a driver will first be described in relation to Figs. 1a and 1b.
The inventors have found that, generally, drivers being impaired by e. g. drowsiness, illness, distraction or intoxication seem to plan the path of the vehicle worse than alert drivers do. In other words, an impaired driver cannot estimate the upcoming road situation and/or cannot adapt his driving to the upcoming road situation as accurately as an alert driver can.
Thus, by calculating the planned path at the instant t1 for an interval ΔT_PD ahead of the vehicle, and comparing it to the estimated traffic scenario at the instant t1, the state of the vehicle and/or the behaviour of the driver at the first instant t1 may be evaluated.
The measure for evaluation of a driver makes use of the finding that the planned path 3 of impaired drivers often varies largely in relation to the lane markings 2. For alert drivers, this is not nearly as common. This suggests that impaired drivers seem to concentrate on keeping the vehicle on the road, and are not able to plan ahead in view of a traffic situation in the same way as alert drivers do.
The method for determining a measure for evaluation of a driver compares the lateral position LP1 of a vehicle 1 in relation to a lane at a first instant t1 with a planed lateral position PLP2 to which the vehicle would arrive at a second instant t2 if the planned path calculated from t1 were followed. Thereafter, a measure called "planned deviation" is determined, being the difference between the planned lateral position PLP2 at the second instant t2 and the lateral position LP1 at the first instant.
For calculating the planned lateral position PLP2 at the second instant t2, following a planned path from the first instant t1, the road shape following the first instant t1 must be estimated. For this estimation any suitable system may be used. However, the method for estimating a traffic scenario as described above has been found to be particularly advantageous and to provide accurate and useful results. In this case, the estimated traffic scenario may include estimation of the upcoming road shape only, or in combination with estimation of other scenario information such as the surrounding traffic situation, whether an overtaking took place etc., which could be used as additional variables for determining the ability of the driver to plan ahead.
In the following, the concept of calculating the planned path will be described in connection with the method for estimating a traffic situation. It is to be understood, that the method for determining a measure for evaluation of the behaviour of a driver is not restricted to be used with this particular method for estimating a traffic scenario. Instead, other methods for estimating traffic scenarios could be suitable for the application. However, the method for estimating a traffic scenario as described herein is particularly suitable, since it provides a relatively high accuracy as compared to previously known methods.
Thus, the embodiment for describing a method for estimating a traffic situation will now be described in relation to Fig. 1a. A vehicle 1 is illustrated at a first instant t1 in time. In this embodiment, the vehicle 1 is positioned in a lane, having lateral lane markings 2, at the first instant t1. Now, the method aims to estimate the traffic situation of the vehicle 1, more particularly, the traffic situation including objects that the vehicle will reach an estimation period Δt_TS forward in time.
In this context, the term "traffic situation" is to be understood as the situation surrounding the vehicle. It could include the curvature of the road upon which the vehicle is travelling, the location and speed of surrounding traffic, the location of stationary objects etc. When estimating the traffic situation, focus may be made on various aspects of the situation. Thus it is possible to select a number of different factors that should be evaluated in the situation, and to disregard others.
The estimation period Δt_TS is > 0, meaning that the method is not a method using real-time calculations. In other words, the estimated traffic situation at the first instant t1 is determined some time after the first instant t1, and is not, as in real-time systems, determined essentially at the same first instant t1. Depending on the situation, the estimation period Δt_TS may be > 0.5 s or even Δt_TS > 2 s.
Data representative of the vehicle's state and of the environment surrounding the vehicle is retrieved and stored at the first instant t1.
Data representative of the state of the vehicle are data giving information about the vehicle. Preferably, the vehicle's speed and yaw rate are included in the data regarding the state of the vehicle. However, other variables may also be included, such as current deceleration or acceleration of the vehicle..
The data representative of the surrounding environment may advantageously include information regarding surrounding dynamic objects, such as vehicles, and regarding surrounding stationary objects such as lane markings and/or road curves. Advantageously, the vehicle's position may be included in, or calculated from, the data representative of the environment of the vehicle, e.g. the vehicle's position in relation to a road or lane. Information regarding dynamic objects may in particular be used to take the surrounding traffic into account when estimating the traffic situation.
After the instant t1, the vehicle 1 continues its path 4 during the estimation period Δt_TS. Data representative of the surrounding environment and of the vehicle's state is retrieved and stored during the drive along the path 4. The data representative of the surrounding environment during the estimation period Δt_TS may include information regarding e.g. the shape of the road or lane, the surrounding traffic and/or weather conditions, depending on which characteristics of the traffic scenario are selected for the estimation. Thus, the data retrieved during the estimation period Δt_TS is used to estimate the traffic situation of the vehicle at the first instant t1. The traffic situation may be evaluated including information gathered during the entire period Δt_TS, or for parts thereof.
The estimated traffic situation at the first instant t1 may be used for a number of different purposes. Advantageously, the state of the vehicle at the first instant may be evaluated in view of the estimated traffic situation. Thus, it is possible to evaluate whether parameters such as the vehicle speed and yaw rate at the first instant t1 were suitable for the traffic situation or not.
In this application, the estimated traffic situation is used to evaluate the behaviour of a driver of a vehicle at the first instant t1. The behaviour of the driver is reflected in the state of the vehicle such as its speed and yaw rate as described above, but could alternatively or additionally include other aspects such as e.g. whether the driver was making a telephone call.
Further, the data retrieved during the estimation period may be used to estimate the traffic scenario during the estimation period. Herein, the term "traffic scenario" is used for describing the context of the traffic situation, e.g. if the driver is performing an overtaking, is in a line of vehicles, is giving way for another vehicle, is changing lanes or is selecting a side road etc. The traffic scenario estimation may be selected so as to focus primarily on selected aspects of different traffic scenarios.
The data representative of the surrounding environment as well as of the vehicle's state may include data from different systems and sensors. Advantageously, it may include data from a lane-tracking system, preferably including a camera.
When the above-described method for estimating a traffic situation is used together with the method for evaluating the state of the vehicle/behaviour of the driver, it will include calculation of a planned path of the vehicle from the first instant t1. The planned path is a path that the vehicle would be expected to follow from said instant t1, calculated using the data considering the state of the vehicle obtained at the first instant t1. The calculation may be made using different assumptions regarding the driver's behaviour. As an example, the planned path may be calculated using the presumption that the vehicle's speed and yaw rate are held constant from the instant t1. In another model, the acceleration of the vehicle at the first instant t1 may be taken into account, either assuming that the acceleration is held constant during the future drive, or that it forms part of some assumed driving pattern.
However, the planned path is determined using information obtained at the first instant t1, and not thereafter. In Fig. 1a, the planned path 3 is shown in a dotted line. As seen in Fig. 1a, the actual path 4 followed by the vehicle after the instant t1 need not be identical to the planned path 3, since the actual state of the vehicle after the first instant t1 may differ from the assumptions made due to measures taken by the driver along the way.
When calculating the planned deviation measure using the method for estimation of a traffic situation as described above, the data regarding the state of the vehicle at the first instant t1 includes the lateral position of the vehicle 1 in relation to a lane at the first instant t1, and e.g. the speed and yaw rate measured at that first instant. This data is stored. Thereafter, the vehicle travels along the road during an estimation period Δt_TS, and data representative of the vehicle's state and of the surrounding environment is gathered and stored. The data representative of the surrounding environment includes in particular information regarding the position of the lane markings 2 of the actual road.
Next, the planned path of the vehicle is calculated for an interval Δtp_D from the first instant t1 to a second instant t2. The planned path interval Δt_PD is in this case a portion of the estimation period used for the traffic situation estimation Δt_TS. The planned path 3 is calculated using the lateral position LP1 retrieved at the first instant t1 and the speed and yaw rate measured at the first instant t1.
Since a driver of a vehicle must adapt his driving behaviour to a traffic situation including e.g. stationary and dynamic objects which the vehicle will reach within a certain time period, the planned path interval Δt_PD may advantageously be selected so as to reflect a relevant time/distance of planning ahead. Thus, if the vehicle travels at a relatively high speed, the planned path interval Δt_PD may be selected differently as compared to when the vehicle is travelling with a lower speed.
The length of the planned path interval Δt_PD may be set in relation to one or more of vehicle or environment variables, such as the current speed of the vehicle, the surrounding traffic situation, whether the road is straight or curved, etc. Advantageously, the planned path interval Δt_PD may be updated in accordance with changes in the relevant variables.
The planned lateral position PLP2 at the second instant t2 is determined by comparing the planned path 3 with the estimated traffic situation obtained previously, including data about the road shape that have been collected during the estimation period Δt_TS. Accordingly, the planned lateral position PLP2 at the second instant t2 may be determined with very high accuracy.
Using the suggested method, the calculations of the planned lateral position can not be made in real time. For purposes of determining driver drowsiness, however, this is not a problem. Driver drowsiness is a slow process, on the minute-scale, and hence it is not a problem to have a delay of a few seconds in the calculations of the planned lateral position.
However, the planned lateral position may be determined in real time if real-time methods are used for the estimation of an upcoming traffic situation.
Information regarding the environment surrounding the vehicle may advantageously be gathered by a lane tracking system, preferably using a camera such as a forward looking monocamera, for determining the lateral position of the vehicle. However, also other sensors may be used to obtain information regarding the traffic situation, whether a take-over takes place etc.
Once the planned lateral position PLP2 has been determined, the planned deviation, PD, being the difference between the planned lateral position PLP2 and the lateral position LP1 at the first instant t1 may be determined. Thus, the PD values can be both positive and negative, depending on the current lateral position of the vehicle and how the driver "aims". It is in many cases, however, only the magnitude of the PD values that is interesting, which is why the absolute values of the PD values may advantageously be used for evaluation.
What the PD values could look like for a driver when alert and impaired, respectively, is shown in Fig 4.
For evaluating the behaviour of a driver, the planned deviation is advantageously determined at a selected frequency during a drive. The selected frequency is advantageously selected such that a new planned deviation calculation is initiated several times during the estimation period Δt_PD.
The planned deviation measures may be analysed for evaluation of the behaviour of a driver. In general, impaired drivers have higher planned deviation values than alert drivers. This can be explained by the impaired drivers tending to plan the future vehicle path worse than alert drivers.
According to the invention, a number of planned deviation measures are then compiled, e.g. by forming a mean over the planned deviation measures, a sum thereof, obtaining a distribution thereof, a filtration or other compilation resulting in a measure representative of a number of planned deviation measures. In the following, this measure will be referred to as a Mean Planned Deviation measure (MPD), although it shall be understood from the above that this measure need not strictly be a mean of a number of planned deviations measures.
For impaired drivers, the MPD values tend to increase over time (as the driver gets more and more impaired), whereas for alert drivers the MPD values stay relatively constant. This property of the MPD signal can be used to detect impaired drivers in a very accurate way.
Since different drivers have different driving patterns, a general limit indicating that a driver is impaired when the planned deviation measures or MPD exceeds the limit might be difficult to set. Instead, a threshold may be determined by gathering planned deviation measures or MPD for the driver during a selected portion of drive, such as the start of the drive, when the driver should generally not be impaired, or during a portion of the drive where the driver is found to drive particularly well. Thus, a threshold for the remaining part of the drive may be set in relation to the planned deviation measures/MPD during the selected portion. In addition, a warning system may be arranged for warning a driver of impairment if the planned deviation measures/MPD measures exceeds the threshold, or repeatedly exceeds the threshold during a certain time interval.
For determining the planned path of the vehicle, a linear bicycle model may be used, which will be described in the following with reference to Figs 2 and 3.
In order to determine the planned path from an instant, the yaw rate and the speed of the vehicle at that instant may be used to establish a good approximation. The radius R of the vehicle path can be calculated as $R = \frac{v}{\overset{}{Ψ}}$

where v is the vehicle speed [m/s] and Ψ̇ [rad/s] is the yaw rate rate of the vehicle.
In Fig. 2, the angle θ ₀ is the current angle of the vehicle in a global xy-coordinate system, v₀ is the current speed of the vehicle, R is the turn radius of the vehicle, S is the distance (along the circumference of a circle with radius R) between the current position of the vehicle and the position where the planned lateral position should be calculated, γ is the vehicle turning angle, (x ₀, y ₀) is the current position of the vehicle, (x_p, y_p ) is the planned position of the vehicle and, finally, (x_c , y_c ) is the centre of the circle that the vehicle is turning along. The angle θ ₀, the velocity v ₀, the radius R and the vehicle position (x ₀, y ₀) are known from the vehicle path calculation algorithm and the arc length S is predetermined, but the other variables must be calculated. First, the angle γ and the coordinates (x_c , y_c ) are calculated as $γ = \frac{S}{R} and (x_{c} y_{c}) = (x_{0} y_{0}) + R \cdot (- \sin θ_{0}, \cos θ_{0}) .$
Then, the coordinates (x_p, y_p ) can be calculated as $(x_{p} y_{p}) = (x_{0} y_{0}) + 2 R \cdot \sin γ \frac{}{2} \cdot [\cos (θ_{0} + \frac{γ}{2}), \sin (θ_{0} + \frac{γ}{2})] .$
Now that the planned position of the vehicle is known, the planned lateral position can be calculated. Since the lane markings next to the planned vehicle path are known from the lane tracking system, these can be used to obtain the planned lateral position. The tangent to the detected lane markings closest to the planned vehicle position (on the form y = k_lane · x + m_lane ) is calculated as $k_{lane} = \frac{y_{lane, k} - y_{lane, k - 1}}{x_{lane, k} - x_{lane, k - 1}}, m_{lane} = y_{lane, k} - k_{lane} \cdot x_{lane, k},$

where x_lane,k, x_lane,k-1 , y_lane,k and y_lane,k-1 are the coordinates of the two sampled points on the lane markings that are closest to (x_p, y_p ), this is displayed in Fig. 3.
The perpendicular direction from the vehicle to the lane markings can be calculated by using the points (x_p, y_p ) and (x_c, y_c ), to obtain an equation in the form y=k_normal · x + m_normal ^. This is done in the same way as when calculating the parameters k_lane and m_lane , $k_{normal} = \frac{y_{c} - y_{p}}{x_{c} - x_{p}}, m_{normal} = y_{p} - k_{normal} \cdot x_{p} .$
Using the parameters k_lane and m_lane together with k_normal and m_normal , the intersection of the two lines, (x_l, y_l ), which is the point that should be used together with (x_p, y_p ) to calculate the planned lateral position, can be calculated. $x_{l} = \frac{m_{lane} - m_{normal}}{k_{normal} - k_{lane}}, y_{l} = k_{lane} \cdot x_{l} + m_{lane}$
Now, the planned lateral position can finally be calculated. The distance d to the lane markings is calculated as the distance between the points (x_l , y_l ) and (x_p , y_p ). $d = \sqrt{{(x_{l} - x_{p})}^{2} + {(y_{l} - y_{p})}^{2}}$
Finally, it only remains to determine the sign of the lateral position, i. e. what side of the vehicle (in the global xy-coordinate system) the point (x_l, y_l ) is. This is needed, because the planned vehicle path might be so different from the road shape that the point (x_p, y_p ) is placed outside the lane markings.
To determine the sign of the lateral position, the two vectors P (which points in the direction of the planned vehicle path in the point (x_p , y_p )) and L (which points from point (x_p , y_p ) to point (x_l , y_l )) are calculated. By calculating the vector product, L×P, between these two vectors and studying the sign of this product, it is possible to determine what side of the vehicle that the point (x_l , y_l ) is. In order to perform the calculation of the vector product, a third dimension must be added to the vectors, i.e. a zero is added as a third element in the vectors L and P. If the sign is positive, that means that the lane markings are located to the right of the vehicle and a negative sign means that the lane markings are located to the left.
Fig. 5 is a flow schedule, illustrating an embodiment of a method for estimating a traffic situation when used in an embodiment of a method for determining a measure for the behaviour of a driver as described above. The traffic situation estimation is performed in the first block of the schedule, named "preprocessing":
In Fig. 5, the process illustrated within the square is referred to as the DIMON Controller, a Driver Impairment MONitor Controller. The DIMON controller may be influenced by the action of other processes, such as the exemplified DIMON Mode manager and DIMON HMI manager. The DIMON Mode manager may control the mode of the DIMON e.g. whether the DIMON Controller is in active state or not or whether a re-initialisation of parameters should be made, in view of information regarding other processes and states of the vehicle. The DIMON HMI manager may control whether it is suitable to display a possible warning from the DIMON Controller to or not to the driver, depending e.g. of other warnings or information that the driver need to pay attention to in a particular situation.
The DIMON Controller receives indata representative of the vehicle's state and of the surrounding environment. In the embodiment described, the DIMON Controller also receives interaction data regarding the interaction between the driver and the vehicle. Other data regarding the driver such as e.g. whether he receives a telephone call may also be received by the DIMON Controller. In the data preprocessing stage, the data is gathered and processed for performing the estimation of a traffic situation. In this case, indata is retrieved during estimation periods Δt_TS after subsequent instants t1 in order to estimate the traffic situations at the first instants t1. The instants t1 may be selected with a constant time delay being substantially shorter than the estimation period Δt_TS. A suitable frequency could be about 10 Hz. The indata is preprocessed so as to be suitable for the following calculations to be performed. In the DIMON Controller illustrated in Fig. 5, the planned path is calculated intervals Δt_PD being at least portions of the estimation periods Δt_TS.
The data from the data preprocessing stage may advantageously be used for adapting the time horizon, i.e. adapting Δt_PD for the planned path calculations. As explained previously, the relevant Δt_PD depends on the present traffic situation and could suitably be selected so as to reflect the time horizon in which an alert driver is normally capable to plan ahead.
In the illustrated embodiment, the planned path is used in an embodiment of a method for determining a planned deviation measure for evaluation of the driver. In this case, the lateral position of the vehicle in relation to a lane and the state of the vehicle at the first instant is suitably reflected in the processed data coming from the data preprocessing stage. Thus, in the planned deviation computation stage, the planned deviation, being the difference between the planned lateral position at the end of the planned path calculation interval Δt_PD and the lateral position at the first instant t1, is calculated, giving the planned deviation PD.
The resulting PD measures are evaluated by the forming of a Mean Planned Deviation value, using a number of PD measures. For estimating impairments such as drowsiness, the PD measures obtained during about a minute's time could be used to form the Mean Planned Deviation measure for evaluation. The Mean Planned Deviation value may advantageously be continuously updated.
In addition to the MPD computation, the PD value may be used in other alternative evaluation algorithms, e.g. filters distributions etc, as indicated in a dashed line in Fig. 5.
The MPD value may also be used for updating a threshold for whether a warning to the driver should be issued or not. As an example, the threshold adaptation may be made so as to always set the threshold in relation to the smallest MPD values obtained during a certain time period of a drive, i. e. a period during which the driver was planning his path well.
In a final step, the MPD value and/or other evaluation algorithm values are compared to the threshold in order to make a decision whether to issue a warning to the driver or not. The warning could preferably be a warning suitable for attracting the attention of an impaired driver, such as an audible warning. Even though the warning could advantageously function so as to temporary wake up a drowsy driver, the purpose of the warning is to indicate to the driver that he should stop driving as soon as possible and take a rest.
Although the described embodiment is of the preferred type where data representative of the state and the surrounding environment of the vehicle is retrieved and stored during an estimation period Δt_TS, and data from an interval Δt_PD is used to obtain a planned deviation measure, it is to be understood that the method of the invention may be used also in combination with systems where the situation within the time interval Δt_PD is estimated in advance, as an upcoming road situation from the first instant t1. Also, the variables used to determine the state of the vehicle and the planned path may be varied, and could include e.g. acceleration or deceleration. The lateral positions are believed to be most accurately determined if a lane-tracking system comprising a camera is used for determining the position of lane markings, but also other systems for lane tracking are possible. A system using the method according to the invention may optionally be integrated or used in combination with other systems such as an Adaptive Cruise Control system, a Lane Departure Warning system or a Lane Keeping Assist system. In addition, the person skilled in the art may readily imagine other alternative embodiments without departing from the invention as described in the specification and defined in the appended claims.

Claims

Method for evaluation of the behaviour of a driver of a vehicle comprising : calculating a number of planned deviation measures at subsequent instants, a planned deviation measure being calculated as follows:
a)- determining the lateral position of the vehicle in relation to a lane and the state of the vehicle at a first instant (t1),

b)- determining the planned path which the vehicle would be expected to follow from said first instant (t1) considering the state of the vehicle at said first instant (t1),

c)- determining the planned lateral position of the vehicle in relation to a lane at a second instant (t2) being a time interval (Δt_PD) after the first instant (t1), if said planned path were followed, and

d)- determining a planned deviation being the difference between the planned lateral position at the second instant (t2) and the lateral position at the first instant (t1),
compiling said planned deviation measures to form a Mean Planned Deviation measure (MPD), and
comparing said Mean Planned Deviation measure (MPD) with a threshold value to thereby classify the driver's ability.
Method according to claim 1, wherein said planned deviation measures are calculated at a selected frequency.
Method according to claim 1 or 2, comprising a threshold determining algorithm wherein the driver's planned deviation values are determined during a selected interval of a ride, and said threshold value for that driver is set in relation to the planned deviation values obtained during the selected interval.
Method according to any one of the previous claims, wherein the data representative of the state of the vehicle comprises data representative of the vehicle's speed and yaw rate.
Method according to claim 4, wherein said planned path is calculated using the vehicle's position, speed and yaw rate at the first instant (t1) as indata.
Method according to claim 5, wherein the planned path is calculated using a linear bicycle model.
Method according to any one of the previous claims, wherein a lane-tracking system including a camera is used to determine the lateral positions.
Method according to any one of the previous claims, comprising retrieving and storing data representative of the vehicle's state and retrieving and storing data representative of the surrounding environment of the vehicle at a first instant (t1), retrieving and storing data representative of the vehicle's state and retrieving and storing data representative of the surrounding environment of the vehicle during at least the interval (Δt_PD) after said first instant (t1),
using the data retrieved and stored during the interval (Δt_PD), for determining the planned lateral position of the vehicle at the second instant (t2).
Method according to any one of the previous claims, wherein the interval Δt_PD is >0 s, preferably > 0.5 s, most preferred > 2 s.
Method according to any one of the previous claims, wherein the interval Δt_PD is selected in relation to traffic situation factors such as e.g. the vehicle's speed, the shape of the road travelled or the surrounding traffic.
Method according to any one of the previous claims, wherein said data representative of the surrounding environment includes information regarding surrounding dynamic objects, such as surrounding traffic, and/or stationary objects.
System for warning a driver using a method for determining a measure of a driver's behaviour according to any one of the preceding claims, and issuing or not issuing a warning based on said measure.
Device for determining a measure for evaluation of the behaviour of a driver comprising means for performing the method according to any one of the claims 1 to 11.