WO2021164975A1

WO2021164975A1 - Method for configuring a computer, driving-assistance device comprising such a computer and motor vehicle equipped with such a device

Info

Publication number: WO2021164975A1
Application number: PCT/EP2021/051364
Authority: WO
Inventors: Stephanie AMBROISE
Original assignee: Renault S.A.S
Priority date: 2020-02-17
Filing date: 2021-01-21
Publication date: 2021-08-26
Also published as: FR3107380A1; EP4107051A1; FR3107380B1

Abstract

The invention relates to a method for configuring a computer (52) of a dynamic gauge representative of a predictable trajectory of at least one wheel of a motor vehicle (1), the method comprising - a step of compiling a database (4) associating, with each pair of steering wheel angle and vehicle speed values, at least two point coordinates (i_ini, i_fin) of a gauge (20) and a plurality of parameters of an approximation function of the shape of the gauge (20), - a step of training an algorithm to determine the dynamic gauge (20) depending on the steering wheel angle and the vehicle speed, the algorithm receiving the database (4) as training data, - saving the algorithm in the computer (52). The invention also relates to a driving-assistance device comprising a computer configured according to the invention, as well as a motor vehicle equipped with such a device.

Description

DESCRIPTION

TITLE OF THE INVENTION: PROCESS FOR SETTING A COMPUTER, DEVICE

DRIVING ASSISTANCE INCLUDING SUCH A COMPUTER AND MOTOR VEHICLE

EQUIPPED WITH SUCH A DEVICE TECHNICAL FIELD OF THE INVENTION

[0001] The present invention relates generally to aids to the driver of a motor vehicle.

[0002] It relates more particularly to a method for setting up a dynamic gauge calculator representative of a predictable trajectory of a motor vehicle. It also relates to a device for assisting in reversing maneuver comprising a computer configured in accordance with the aforementioned method. Finally, it relates to a motor vehicle equipped with such a device.

STATE OF THE ART

[0003] Currently, some motor vehicles are equipped with reversing cameras, that is to say cameras placed at the rear of vehicles and facing the rear.

[0004] Such a reversing camera is generally connected to a screen located on the dashboard of the vehicle so as to display, when the driver of the vehicle engages reverse gear, an image of the environment behind the vehicle.

[0005] In this image, it is customary to overlay a layer called a dynamic template.

[0006] The dynamic template generally comprises two arcs of circles which extend from the rear of the vehicle, over a given length (for example 3 meters), and which represent the foreseeable trajectories of the two rear wheels of the vehicle taking into account the steering angle of the steering wheel.

[0007] The dynamic template is thus intended to help the driver to park in reverse and in particular to appreciate the distance he has for parking. [0008] A conventional means of displaying this template comprises the use of a calculation unit capable of calculating point by point the dynamic template according to certain input parameters such as the angle at the steering wheel and the speed of the vehicle. This solution, described for example in the French patent application published under the number FR3031707, makes it possible to dispense with the recording of pre-established templates in the read-only memory of the computer.

[0009] However, a drawback of this method is the large number of operations per unit of time required for the complete calculation of the template, which involves the use of a powerful computer. Unfortunately, some motor vehicles do not have sufficient calculation capacity to perform these calculations, which limits the implementation of this solution to certain types of vehicles, in particular high-end vehicles.

PRESENTATION OF THE INVENTION [0010] In order to remedy the aforementioned drawback of the state of the art, a method is proposed for setting up a dynamic gauge calculator representative of a predictable trajectory of a motor vehicle, said method comprising:

a step of developing a database associating with each pair of values relating to the angle at the steering wheel and to the speed of the vehicle, at least two coordinates of points of a dynamic template and a plurality of parameters characterizing a approximation function of the shape of the dynamic template,

- a step of learning an algorithm to determine the dynamic template from the values relating to the steering wheel angle and the speed of the vehicle, the algorithm receiving said database as learning data, and

- a saving of the algorithm in said computer.

Thanks to the invention, we do away with the use of a computer configured to calculate the template point by point. The determination of the template is therefore faster and less expensive in terms of computation resources. It is therefore advantageously possible to use any type of computer to determine the dynamic template. Other advantageous and non-limiting characteristics of the process according to the invention, taken individually or in any technically possible combination, are as follows:

- the dynamic template may include a first curve representative of the foreseeable trajectory of a first rear wheel and a second curve representative of the foreseeable trajectory of a second rear wheel.

- the training step may include:

- training of a first regression model to predict a first of said coordinates of points of the dynamic template as a function of the vehicle speed and steering wheel angle values,

- training of a second regression model to predict a second of said coordinates of points of the dynamic template as a function of the values relating to the speed of the vehicle and to the steering wheel angle and of the first point coordinate;

- the training step can also include:

- training of a third regression model to predict an initial parameter of the approximation function as a function of the values relating to the speed of the vehicle and to the angle at the wheel, of the first point coordinate and of the second coordinate point,

- training, for each other parameter of the approximation function, of a respective regression model to determine said parameter of the approximation function as a function at least of the values relating to the speed of the vehicle and to the angle at steering wheel, steering wheel angle, first point coordinate and second point coordinate;

- the algorithm can be a neural network;

- the approximation function of the shape of the template can be a polynomial whose parameters are the coefficients of the monomials;

- the polynomial can be a second degree polynomial;

- Said at least two point coordinates can be half coordinates of two distinct points of the template. The term “half-coordinate” is understood to mean only one of the two constituent values of the coordinate of a point which are the abscissa and the ordinate. The invention also provides a device for assisting the maneuver in reverse of a motor vehicle, comprising a means for acquiring a value relating to the angle at the steering wheel of the vehicle, a means for acquiring a value relating to the speed of the vehicle, a reversing camera, a display screen adapted to display images acquired by the reversing camera, a computer configured according to the parameterization method described above and programmed to determine a dynamic template representative of a foreseeable trajectory of the motor vehicle, and a display module on the display screen of said dynamic template superimposed on the images acquired by the reversing camera.

[0014] Other advantageous and non-limiting characteristics of the device according to the invention, taken individually or in any technically possible combination, are as follows:

- the calculator can be configured to

- implement the first regression model to determine the first of the coordinates of points of the dynamic template from values relating to the speed of the vehicle and the angle at the wheel,

- implement the second regression model to determine the second of the point coordinates of the dynamic template from the values relating to the vehicle speed and the steering wheel angle and from the first point coordinate,

- implement the third regression model to determine an initial parameter of the approximation function as a function of the values relating to the speed of the vehicle and to the angle at the wheel, of the first point coordinate and of the second coordinate of point,

- implementing, for each other parameter of the approximation function, said respective regression model to determine said parameter as a function at least of the values relating to the speed of the vehicle and to the angle at the wheel, of the first coordinate of point, of the second coordinate of,

- calculate an approximation of the shape of the template using said function approximation.

[0015] The invention further provides a motor vehicle equipped with a maneuvering aid device according to the invention.

Of course, the different characteristics, variants and embodiments of the invention can be associated with each other in various combinations insofar as they are not incompatible or mutually exclusive.

DETAILED DESCRIPTION OF THE INVENTION

The description which will follow with regard to the appended figures, given by way of non-limiting examples, will make it clear what the invention consists of and how it can be carried out.

In the accompanying figures:

[0019] [Fig. 1] is a schematic top view of a motor vehicle, behind which is an image of a dynamic template.

[0020] [Fig. 2] illustrates a method for setting up a computer in accordance with the invention.

[0021] [Fig. 3] is a portion of an array forming a training database for an algorithm of a computer according to the invention.

In Figure 1 there is shown, seen from above, an embodiment of a motor vehicle 1 according to the invention.

Conventionally, this motor vehicle 1 comprises a frame which defines a passenger compartment for the driver of the vehicle. It also has four wheels, a computer, a system for controlling the steering of the rear wheels, and a steering wheel for controlling the orientation of the steered wheels to the right or to the left.

In the embodiment of Figure 1, the motor vehicle 1 comprises two front wheels 2 steered and two rear wheels 3 non-steerable (that is to say which do not participate in the steering of the motor vehicle 1 towards right or left). This example is not limiting and the invention can for example be applied to a vehicle with four steered wheels. At this stage, we can define a reference (O, C, U, Z) attached to the motor vehicle 1. Here, the reference considered is an orthonormal reference whose origin O is located at the center of the front axle of the motor vehicle 1, whose longitudinal axis X is oriented towards the rear of the motor vehicle, whose transverse axis Y is oriented towards the right of the motor vehicle and whose vertical axis Z is oriented upwards. Here, the terms "top", "bottom", "left" and "right" are to be understood from the point of view of a driver of the vehicle. Thus, the transverse axis Y which is oriented towards the right of the vehicle when one is placed from the point of view of the driver appears oriented towards the left of FIG. 1.

In the example illustrated, the motor vehicle 1 is equipped with a device for assisting in reversing maneuver comprising a reversing camera 50 and a display screen 51 adapted to display images acquired by the camera reversing camera 50. The reversing camera 50 is arranged at the rear of the chassis, for example above the vehicle registration plate, so that its lens opens out to the outside of the vehicle and is generally rearward and downward facing.

More specifically, the reversing camera 50 has an optical axis which is inclined:

- with respect to the plane (X, Y) downwards, with a non-zero pitch angle (greater than 5 degrees),

- relative to the plane (X, Z) to the left or to the right, by a yaw angle which may or may not be zero (which is here less than 10 degrees), and - around the longitudinal axis X, a roll angle which may or may not be zero

(which here is less than 10 degrees).

The display screen 51 is itself placed in the passenger compartment of the motor vehicle 1, for example on the dashboard of the latter. It therefore allows the driver to see the images acquired by the reversing camera 50. To control the various audiovisual devices on board the motor vehicle 1, in particular the reversing camera 50 and the display screen 51, provision is made for the device for assisting in reversing maneuvering to include a computer 52 comprising in particular a processor, random access memory, read only memory, analog-to-digital converters, and various input and output interfaces. By virtue of its input interfaces, the computer 52 is adapted to receive input signals from various sensors. Thanks to its analog-to-digital converters, the signals received by the computer 52 are sampled and digitized.

In its random access memory, the computer 52 thus continuously stores: - the (instantaneous) image acquired by the reversing camera 50, and transmitted to the computer 52,

- the gear ratio engaged by the driver of the vehicle,

- the steering angle (instantaneous) of the steering wheel, and

- the speed V (instantaneous) of the vehicle. By means of software stored in its read-only memory, the computer 52 is programmed to generate, for each operating condition of the vehicle, output signals.

Finally, thanks to its output interfaces, the computer 52 is suitable for transmitting these output signals to the audiovisual devices on board the motor vehicle 1, in particular to the display screen 51.

The computer 52 is then suitable for implementing a method of assisting the maneuver in reverse of the motor vehicle 1.

This method is implemented when the gear ratio engaged by the driver is reverse gear. This method consists in displaying a dynamic template superimposed on the images acquired by the reversing camera 50.

In Figure 1, there is shown an image of this dynamic template 20, seen in a plane parallel to the plane of the road on which the motor vehicle 1 travels (here in the plane (O, C, U)). The vehicle's reversing maneuver assistance device comprises a display module of this template in superimposition of the images acquired by the reversing camera so that it is visible to the driver.

The dynamic template 20 is a geometric figure which allows the driver to understand the predictable path that the motor vehicle 1 will take when it backs up. The dynamic template here comprises a first curve 21 representing the foreseeable trajectory of the first rear wheel, and a second curve 22 representing the foreseeable trajectory of the second rear wheel. This geometric figure has a shape that varies depending at least on the steering angle of the steering wheel. Here, the dynamic template 20 has a shape which is determined at each moment as a function of the input variables.

In particular here, provision is made to determine the shape of this dynamic template 20 as a function of the orientation angle of the steering wheel and of the speed V of the motor vehicle 1. [0042] According to a particularly advantageous characteristic of the 'invention, the software stored in the read-only memory of the computer 52 comprises an algorithm previously trained to predict the shape of the template as a function of the input signals, and in particular here as a function of the speed of the vehicle and of the angle at the wheel . FIG. 2 illustrates the steps of a process for setting the computer 52. The example described below illustrates the setting of the computer 52 to calculate the first curve 21. It should be noted that the setting of the computer 52 for determining the second curve 21 is done in a similar fashion and will not be described here for the purposes of simplifying the description. This setting consists in practice in determining an algorithm which will make it possible, once installed in the vehicle 1, to easily calculate the shape of the dynamic template 20. This setting operation is therefore carried out during the design of the vehicle, on a unit computer calculation separate from the motor vehicle. During a first step E1, this computer calculation unit calculates in a known manner, for example according to the manner described in the application FR3031707 cited above, for each pair of values of input parameters, here for each pair of steering wheel angle and vehicle speed values, the coordinates of the points of the first curve 20 in the plane (O, C, U). These coordinates are obtained and saved in the form of pairs of columns, a first column of which comprises the coordinates i of the points along X and a second column comprises the coordinates j of the points along Y.

This first step E1, considered in isolation, is known to those skilled in the art since it is already known to calculate the shape of the dynamic template analytically. During a second step E2, the computer calculation unit determines, for each pair of columns, the parameters of an approximation function of the shape of the curve 20. For example, it calculates the coefficients d 'a polynomial, here a second degree polynomial, making it possible to predict the coordinates j of the points of the curve along the Y axis, from the coordinates i of the points of the curve along the X axis.

During a third step E3, the computer calculation unit generates a training database 4 from the information obtained in the previous steps. Figure 3 is a representation of part of the database 4, here the first thirteen lines. Here, the database 4 is in the form of a table comprising, for each pair of steering wheel angle and speed values:

- a first coordinate ijni of a point of the first curve, here the abscissa (along the X axis) of the initial point, that is to say of the point closest to the rear face of the vehicle therefore having the lowest abscissa value, - a second i_fin coordinate of a second point of the template, here the abscissa of the end point, that is to say of the point of the curve furthest from the vehicle, for example previously chosen here to be located 3 meters from the initial point ijni along the curve, and

- the parameters of the approximation function, here the coefficients a2, a1, aO of the second degree polynomial. The polynomial making it possible to obtain the coordinates j is thus written in the form a2.i ² + a1 i + a0.

The following steps together form a learning phase during which the algorithm is taught to determine the first curve of the dynamic template 20 from the steering wheel angle and the speed of the vehicle. Here, the algorithm is configured to learn how to determine this curve from said database 4.

[0052] In particular here, the algorithm is formed from one or more neural networks. It comprises here five neural networks, which are preferably of the same type.

[0053] In this case, each neural network has an input layer which considers at least the values of driving angle and speed of the motor vehicle.

Each neural network has at least one hidden layer. In practice, it has two.

Finally, each neural network comprises an output layer which returns a single value.

The learning phase then consists in parameterizing the neurons of the different layers of the five networks, so that, when providing the algorithm (that is to say to the five neural networks) the values of steering wheel angle and speed of the motor vehicle, it provides an approximate value of the first coordinate ijni, of the second coordinate of point i_fin and of the coefficients a2, a1, aO of the second degree polynomial.

Each neural network thus generates a prediction function of a value (the first coordinate ijni, or the second coordinate of point i_fin or the coefficients a2, a1, aO of the second degree polynomial) obtained using d 'a regression model. Each neural network is built here using the free software “TensorFlow”.

We can now detail the way in which each neural network is obtained. During a fourth step E4, a first regression model M1 is built to predict the first point coordinate ijni as a function of the steering wheel angle and the speed of the vehicle. The first regression model M1 is trained in a learning environment from the database 4.

In this first step as well as in the following ones, the regression model constructed is a multivariate linear regression model, comprising several inputs and a single output.

In a fifth step E5, a second regression model M2 is constructed to predict the second point i_fin coordinate as a function of the steering wheel angle, the vehicle speed and the first point ijni coordinate. The second M2 regression model is trained in a learning environment from database 4.

During a sixth step E6, a third regression model M3 is built to predict a first parameter of the approximation function, here the maximum order coefficient a2 of the polynomial, from the angle at the steering wheel , the speed of the vehicle, the first coordinate of point ijni and the second coordinate of point i_fin. The third M3 regression model is trained in a learning environment from database 4.

During a seventh step E7, a fourth regression model M4 is built to predict a second parameter of the approximation function, here the coefficient of order one a1 of the polynomial, from the angle at the steering wheel , of the speed of the vehicle, of the first coordinate of point Mni, of the second coordinate of point Jin, and of the coefficient a2. The fourth M4 regression model is trained in a learning environment from database 4.

During an eighth step E8, a fifth regression model M5 is built to predict a third parameter of the approximation function, here the zero order coefficient aO of the polynomial, from the angle at the steering wheel , the speed of the vehicle, the first coordinate of point ijni, the second coordinate of point Jin, the coefficient a2 and the coefficient a1. The fifth regression model M5 is trained in a learning environment from the database 4.

Of course, as a variant, the regression models could be constructed differently. We could thus have started by predicting the values of the parameters of the polynomial before those of the x-coordinates ijni, i_fin.

Finally, during a ninth step E9, the algorithm comprising the five regression models M1 to M5 is saved on the computer 52 on board the motor vehicle.

The computer 52 is thus configured to determine the first curve of the template using the regression models M1 to M5 as follows:

[0068] ijni = M1 (angle_volant, speed)

[0069] i_fin = M2 (angle_volant, speed, ijni)

A2 = M3 (angle_volant, speed, Mni, Jfin)

A1 = M4 (angle_volant, speed, Mni, Min, a2)

AO = M5 (angle_volant, speed, Mni, Min, a2, a1)

Each coordinate point (i, j) of the curve of the dynamic template is then constructed using the following function, in which the calculation of the coordinate j depends on the value of the coordinate i, in particular such that :: [0074] j = a2 * i ² + a1 * i + aO

This curve is also well positioned and limited in length on the display screen thanks to the abscissas of the first and last points of the curve.

The invention therefore advantageously allows a determination of the dynamic template of the vehicle 1 in a simple manner, that is to say by implementing a limited number of operations per unit of time.

Various other modifications can be made to the invention within the scope of the appended claims. The present invention is in no way limited to the embodiments described and shown, but those skilled in the art will know how to provide any variant in accordance with the invention.

[0079] In particular, although a vehicle having two steered wheels has been described here, the invention is perfectly compatible with a vehicle comprising four steered wheels.

In addition, although it has been described here a second degree polynomial as an approximation function, the invention is compatible with any other type of polynomial as an approximation function, for example polynomials of degree 4 or 8.

And, the invention is in particular compatible with any other function suitable as an approximation function.

Claims

[Claim 1] Method for setting up a dynamic gauge calculator (52) representative of a predictable trajectory of at least one wheel of a motor vehicle (1), said method comprising:

- a step of developing a database (4) associating with each pair of values relating to the angle at the steering wheel and to the speed of the vehicle (1), at least two coordinates of points (ijni, i_fin) d 'a dynamic template (20) and a plurality of parameters characterizing an approximation function of the shape of the dynamic template (20),

- a step of learning an algorithm to determine the dynamic template (20) from the values relating to the angle at the steering wheel and to the speed of the vehicle (1), the algorithm receiving said database (4) as learning data, and

- a saving of the algorithm in the calculator (52).

[Claim 2] A method according to claim 1, in which the dynamic template (20) comprises a first curve (21) representative of the foreseeable trajectory of a first rear wheel and a second curve (22) representative of the foreseeable trajectory of a second rear wheel.

[Claim 3] A method according to claim 1 or 2, wherein the training step comprises:

- training of a first regression model (M1) to predict a first of said point coordinates (ijni) of the dynamic template (20) as a function of the vehicle speed (1) and steering wheel angle values,

- training of a second regression model (M2) to predict a second of said point coordinates (i_fin) of the dynamic template (20) as a function of the values relating to the speed of the vehicle (1) and to the angle at the steering wheel and the first point coordinate (ijni),

[Claim 4] A method according to any one of claims 1 to 3, wherein the training step also comprises:

- training of a third regression model (M3) to predict a initial parameter (a2) of the approximation function as a function of the values relating to the speed of the vehicle (1) and to the angle at the steering wheel, of the first point coordinate (ijni) and of the second point coordinate (i_fin ),

- training, for each other parameter (a1, aO) of the approximation function, of a respective regression model (M4, M5) to determine said parameter (a1, aO) as a function of at least the values relating to the vehicle speed (1) and at the steering wheel angle, of the first point coordinate (ijni), and of the second point coordinate (i_fin).

[Claim 5] A method according to any of claims 1 to 4, wherein the algorithm comprises at least one neural network.

[Claim 6] A method according to any one of claims 1 to 5, wherein the approximation function of the shape of the template is a polynomial whose parameters are the coefficients of the monomials.

[Claim 7] The method of claim 6, wherein the approximation function is a polynomial, such as a quadratic polynomial.

[Claim 8] A method according to any of claims 1 to 7, wherein said at least two point coordinates (ijni, i_fin) are half coordinates of two distinct points of the template (20).

[Claim 9] Device for assisting the maneuvering in reverse of a motor vehicle (1), comprising a means for acquiring a value relating to the angle at the steering wheel of the vehicle, a means for acquiring a value relating to the speed of the vehicle, a reversing camera (50), a display screen (51) adapted to display images acquired by the reversing camera, a computer (52) configured according to the method of one of the preceding claims and programmed to determine a dynamic template (20) representative of a predictable path of at least one wheel of the motor vehicle (1), and a display module on the display screen (51) of said template dynamic superimposition of the images acquired by the reversing camera.

[Claim 10] Device according to claim 9, in which the dynamic template (20) comprises a first curve (21) representative of the foreseeable trajectory of a first rear wheel and a second curve (22) representative of the foreseeable trajectory of a second rear wheel.

[Claim 11] Device according to claim 9 or 10, wherein the computer (52) is configured for:

- implement the first regression model (M1) to determine the first of the point coordinates (ijni) of the dynamic template from the values relating to the speed of the vehicle (1) and the angle at the steering wheel,

- implement the second regression model (M2) to determine the second of the coordinates of points (i_fin) of the dynamic template from the values relating to the speed of the vehicle (1) and the angle at the steering wheel and the first point coordinate (ijni), - implement the third regression model to determine an initial parameter (a2) of the approximation function as a function of the values relating to the speed of the vehicle (1) and to the angle at the steering wheel , the first point coordinate (ijni) and the second point coordinate (i_fin),

- implementing, for each other parameter (a1, aO) of the approximation function, said respective regression model (M4, M5) to determine said parameter as a function of at least the values relating to the speed of the vehicle (1) and at the steering wheel angle, the first point coordinate (ijni), and the second point coordinate (Jin),

- calculate an approximation of the shape of the template using said approximation function.

[Claim 12] A motor vehicle equipped with a maneuvering aid device according to any one of claims 9 to 11.