Method for assisting the reversing of a motor vehicle towing a trailer

Abstract

The invention relates to a method for assisting the reversing of a motor vehicle (10) to which a trailer (11) is attached. According to the invention, the method comprises: - acquiring the relative position of the motor vehicle with respect to a target zone, and - determining a steering angle command enabling the motor vehicle to reverse into said target zone, by means of an artificial neural network to which the relative position of the target zone with respect to the motor vehicle is provided. Figure for the abstract: Fig. 3

Description

Title of the invention: Method for assisting the reversing of a motor vehicle towing a trailer Technical field of the invention

[0001] The present invention relates generally to driving aids for motor vehicles.

[0002] It relates more particularly to a method of assisting in reversing a motor vehicle to the rear of which a trailer is attached.

[0003] It also relates to a method of setting the parameters of a reversing assistance computer of a motor vehicle.

[0004] It also relates to a motor vehicle equipped with such a computer. State of the art

[0005] When equipped with a coupling system, motor vehicles are capable of traveling with a trailer.

[0006] While forward maneuvers are easy to perform, the same cannot be said for reverse maneuvers. This is especially true when the trailer is small, as it will tend to quickly become overhanging in relation to the vehicle (i.e., "jackknifed") as soon as the driver turns the steering wheel with too large an angle.

[0007] A solution is then known from document US20200001920 allowing the determination of the maximum steering angle that can be applied to the steering wheel without risking putting the trailer in overhang relative to the vehicle.

[0008] Despite this solution, reversing maneuvers remain complex, particularly because they are counterintuitive, requiring the steering wheel to be turned in one direction to move in the other, in order to position the trailer on the desired trajectory. Presentation of the invention

[0009] It is in this context that the present invention proposes a method for providing a steering angle command (to an actuator of the vehicle or to the driver of the vehicle), so that the vehicle can reach a target area without this posing any difficulty for the driver.

[0010] The invention more particularly proposes a method for assisting the reversing of a motor vehicle to which a trailer is attached, the method comprising: - a step of acquiring the relative posture of the motor vehicle in relation to a target area, and - a step of determining a steering angle command allowing the motor vehicle to reverse to said target zone, by means of a computer in which an artificial neural network is stored, the relative position of the target zone with respect to the motor vehicle being provided as input to said artificial neural network and said steering angle command being generated as output to said artificial neural network.

[0011] Thus, thanks to the invention, the neural network will be able, throughout the reverse maneuver, to determine the steering angle to be applied at each moment to reach the target area as simply and safely as possible.

[0012] This solution then allows the driver to easily park his vehicle in any parking space.

[0013] In particular, it allows the vehicle to follow a complex trajectory when necessary, i.e. a trajectory requiring a variation in the steering angle.

[0014] This solution is also inexpensive since its implementation requires only a limited number of components. In particular, it requires the use of a single reversing camera.

[0015] Preferably, the target area is a parking space previously selected by a driver of the motor vehicle.

[0016] Preferably, at the determination stage, the artificial neural network generates as output a speed command enabling the motor vehicle to reverse to said target zone.

[0017] Preferably, after the determination step, the steering angle command is displayed or given vocally to the driver of the motor vehicle, or it is transmitted to an actuator controlling the steering wheel orientation of the motor vehicle.

[0018] The invention also relates to a method for setting the parameters of a reversing assistance computer for a motor vehicle towing a trailer, said method comprising: - a step involving the development of a database associating each initial position of a reference vehicle relative to a target area with trajectory data enabling the reference vehicle to reverse to said target area, - a neural network training step, which makes it possible to determine the variations over time of a steering angle to be applied to the steering wheel of the motor vehicle to reach any target area as a function of an initial posture of the motor vehicle relative to the target area, said neural network receiving said database as training data, - a backup of the neural network in the computer.

[0019] Preferably, the database is developed using, as a reference vehicle, a virtual model of the motor vehicle equipped with a trailer in a virtual environment comprising a plurality of parking spaces, a physical person driving the virtual model so as to park it in reverse in said parking spaces.

[0020] Preferably, to drive the virtual model, the driver uses a view taken from the virtual motor vehicle model and oriented towards the rear, and a top view of the virtual trailer model.

[0021] Preferably, before parking the virtual model in said parking space, said natural person indicates the position of the parking space.

[0022] The invention also relates to a reversing assistance device for a motor vehicle to the rear of which a trailer is attached, which includes a reversing camera and a computer which is programmed to implement a driving assistance method as described above and which is parameterized according to the aforementioned parameterization method.

[0023] Of course, the various features, variants, and embodiments of the invention can be combined with one another in various ways, provided they are not incompatible or mutually exclusive. Detailed description of the invention

[0024] The following description with regard to the attached drawings, given by way of non-limiting examples, will make it clear what the invention consists of and how it can be carried out.

[0025] On the attached drawings:

[0026] [fig. 1] is a schematic view of a motor vehicle equipped with a reversing driving assistance device according to the invention;

[0027] [fig.2] is a schematic top view of a trailer and motor vehicle from figure 1, entering a parking space;

[0028] [fig.3] is a view taken by a reversing camera of the motor vehicle in figure 1.

[0029] In the remainder of this exposition, a motor vehicle and a trailer will be discussed, the whole being called "actual coupling 1".

[0030] A virtual model of this motor vehicle and trailer will also be discussed, the whole being referred to this time as "virtual coupling 1'".

[0031] Figure 2 shows the virtual model of the motor vehicle 10' and its trailer 11', while Figure 3 shows a part of the real motor vehicle 10 and its trailer 11.

[0032] The motor vehicle 10 could be of any type. It could, for example, be of a truck. In the example illustrated in figure 1, it is rather a car.

[0033] This car conventionally comprises a chassis, four wheels 12 of which at least two are steerable, a steering system 16, a powertrain 17 and braking means 18.

[0034] The steering system 16 includes a steering wheel for controlling the steering of the steerable wheels. The angle of the steering wheel relative to its midpoint will hereafter be referred to as the "steering angle". The steering system may also optionally include an actuator 16A for varying the steering angle of the steerable wheels when a control unit 13 requires it, without the driver having to exert any effort on the steering wheel.

[0035] The powertrain 17 comprises an engine (electric, internal combustion or hybrid) and a system for transmitting power from the engine to the drive wheels 12. It is controlled by means of an accelerator pedal available to the driver 20. It may also optionally be controlled by an actuator 17A under the control of the computer 13.

[0036] The braking means 18 are controlled by means of a brake pedal available to the driver. They may also optionally be controlled by an actuator 18A under the control of the computer 13.

[0037] The motor vehicle 10 includes at least one camera 14, referred to as a reversing camera. This camera 14 is oriented towards the rear of the vehicle, so as to acquire images of the environment located at the rear of the vehicle, such as that illustrated in Figure 3. The optical axis of this camera 14 is thus located in the vertical median plane of the motor vehicle.

[0038] This camera 14 has a wide field of view, extending horizontally over approximately 180 degrees and vertically over approximately 140 degrees. This fisheye camera is subject to distortion problems. A distortion correction map can be used, if necessary, to process the acquired images.

[0039] The vehicle includes a Human-Machine Interface allowing the computer 13 to communicate information to the driver of the motor vehicle, and vice versa. Here, this interface is in the form of a touch screen 15. Alternatively, it could be in another form (speakers, control buttons, etc.).

[0040] The calculator 13 is specifically designed to control the aforementioned actuators 16A, 17A, 18A and / or the touch screen 15.

[0041] This computer 13 includes a processor, a memory and a data exchange interface, connected for example to a CAN network of the vehicle.

[0042] Thanks to this data exchange interface, the computer 13 is adapted to receive various information, for example the steering wheel angle, information entered by the driver via the touch screen 15, and images acquired by the 14-inch reversing camera.

[0043] Thanks to its memory, the computer 13 stores a computer application, consisting of computer programs including instructions whose execution by the processor allows the implementation by the computer of the process described at the end of this presentation.

[0044] Here, the vehicle 10 is equipped at the rear with a tow ball (not visible in figure 1) allowing the trailer 11 to be hitched.

[0045] The present invention relates to a way of assisting the driver to reverse when the trailer 11 is coupled to the motor vehicle 10. This assistance is provided in the form of "instructions", which may consist of information to be provided to the driver, via the touch screen 15, or of control signals from the actuators 16A, 17A, 18A allowing control of the direction and even the speed of the vehicle during the reversing maneuver.

[0046] Here, a "maneuver" is defined as the set of operations enabling the vehicle to reach a parking space from an initial position.

[0047] According to a particularly advantageous feature of the invention, the instructions are calculated by the computer 13, by means of an artificial neural network which is recorded therein.

[0048] To configure the neural network so that it provides satisfactory instructions, three successive steps are planned: - a step in the development of a database, - a training step for the neural network using this database, and - a backup of the neural network in the computer.

[0049] In the remainder of this presentation, we will first describe how the database will be developed, then how the neural network will be trained, and finally how the computer will be used in practice to help the driver of the vehicle to perform a reversing maneuver with a trailer.

[0050] The first step involves firstly a simulation operation which consists, for a natural person hereinafter referred to as "operator", in carrying out a large number of reverse maneuvers, then a second operation of storing in the database the data collected during the first operation.

[0051] The maneuvers could be carried out on a real motor vehicle to which a trailer would be attached and which would have sensors to determine, during each maneuver, the position of the motor vehicle in relation to a target area (in our example a parking space), as well as the steering angle and the speed of the motor vehicle.

[0052] However, for ecological, cost and precision reasons, these maneuvers are preferably carried out virtually.

[0053] For this, a digital model of the motor vehicle 10' and trailer 11' (the virtual coupling 1') is used.

[0054] Here, this virtual coupling 1' is programmed to be piloted by means of a simulator comprising a steering wheel, a pedal set, and a screen simulating the view that a driver would have sitting in the passenger compartment of the motor vehicle 10.

[0055] The virtual coupling 1' allows the way in which the motor vehicle 10 and its trailer 11 move, both forward and backward, when it is given instructions for steering angle and movement of the accelerator and brake pedals.

[0056] This digital model naturally depends on the characteristics of the motor vehicle 10 and its trailer 11. Indeed, the trajectory taken in reverse by the actual coupling 1 depends on the distance between the two axles of the vehicle, the distance between the axle of the trailer and the coupling ball of the vehicle...

[0057] In other words, the numerical simulations of maneuvers that will be carried out with this virtual coupling 1' will be specific to the model of the motor vehicle 10 and to a particular type of trailer 11.

[0058] As shown in Figure 2, for the simulation of maneuvers, the virtual coupling 1' is placed in a virtual environment 2' representing at least one road and parking spaces 20. In practice, the virtual environment represents a city with its buildings, roads, cars and other vehicles using the roads...

[0059] Here, the virtual environment simulation software 2' used is that developed by Scaner Studio®. The virtual model parameterizing the movement of the motor vehicle 10' and the trailer 11' is programmed using Simulink®.

[0060] In this simulation, the motor vehicle 10' is equipped with at least one virtual image sensor, whose images are transmitted to the screen of the simulator in which the operator is located. These images are therefore used by the operator to perform reversing maneuvers.

[0061] This virtual image sensor is preferably a virtual reversing camera having the same characteristics as those of the reversing camera 14 of the real motor vehicle 10.

[0062] As Figure 3 clearly shows, the disadvantage of such a reversing camera 14 is that the trailer 11 hides a large part of the environment, making it difficult for the operator to perform precise maneuvers.

[0063] This is why the virtual model of motor vehicle 10' is equipped with at least one other virtual image sensor, which this time does not correspond to any sensor provided on the real motor vehicle 10.

[0064] Here, this second virtual sensor is a camera allowing a top view of the virtual coupling 1' and its environment 2'. This view is shown in Figure 2. More precisely, it is a distortion-free camera positioned vertically above the center of inertia of the motor vehicle 10', at a height of 5m, and oriented vertically towards the ground.

[0065] The images acquired by the two virtual image sensors are therefore transmitted onto the screen of the simulator.

[0066] To perform reverse maneuvers in this simulator, the operator always operates in the same way.

[0067] It starts by driving on a road in the virtual environment 2' until it detects a parking space.

[0068] This detection can only be carried out by the operator, who uses the interfaces at his disposal to point to the parking space and its outline on the images he has available.

[0069] Alternatively, this detection can be carried out automatically, for example in the manner presented in document FRI853260.

[0070] Then, the operator moves forward so as to place the virtual coupling 1' in a position, called the initial position, which he judges to be a promising starting point for easily parking the motor vehicle 10' and its trailer 11' in the parking space 20. The operator will ensure, at each maneuver, that he varies his starting position so as to modify the lateral gap between the virtual coupling 1' and the parking space 20 as well as the inclination of this coupling relative to this space.

[0071] In the initial position, the trailer 11' is placed in the axis of the motor vehicle 10'.

[0072] Once parking space 20 is selected and the virtual coupling 1' is in the starting position, the relative position of parking space 20 with respect to the motor vehicle 10' is acquired.

[0073] For this purpose, the coordinates of the parking space 20 on one of the images acquired by the virtual reversing camera or by the second camera, in pixels, are transformed into three-dimensional coordinates in the frame of the virtual coupling 1'. Preferably, the center of the frame is positioned in the middle of the front axle of the motor vehicle 10', the abscissa axis x is oriented towards the rear of the vehicle, the ordinate axis y is oriented towards the right of the vehicle, and the third axis z is oriented vertically upwards.

[0074] Then, the operator engages reverse gear and parks the virtual coupling 1', by varying the steering angle of the steering wheel and adjusting the speed of this coupling.

[0075] During the entire maneuver, trajectory data are recorded.

[0076] In practice, this data includes: - the steering angle, - the speed of the vehicle, - the coordinates of parking space 20 in the virtual coupling reference 1'.

[0077] This data triplet is recorded at regular time steps, for example every hundredth of a second.

[0078] The coordinates of parking space 20 can be expressed in different ways. For example, they can be the coordinates of the four corners of parking space 20, which allows us to deduce the position and orientation of the virtual coupling 1' relative to this space.

[0079] Preferably, the speed imposed on the virtual model during the maneuver is between a minimum threshold (for example, 1 km / h) and a maximum threshold (for example, 8 km / h). Preferably, this speed is kept substantially constant by the operator throughout the entire maneuver (except at the beginning and end of the maneuver).

[0080] This type of virtual maneuver is repeated a large number of times in different parking spaces. The number of virtual maneuvers will depend on the variety of situations one wishes to anticipate, as well as other parameters. In any case, it will be at least several hundred.

[0081] The virtual maneuvers are repeated on all types of parking spaces (angled, perpendicular, parallel). They are repeated on roads of different widths, presenting various obstacles (posts, another car parked next to it...).

[0082] For each virtual maneuver performed, a set of data triplets is obtained.

[0083] All these sets of triples are then recorded in a database.

[0084] The database architecture may be as follows. It will include as many tables as there are virtual maneuvers performed. Each table, associated with one of these maneuvers, will include a "steering angle" field, a "speed" field, and one or more "position" fields storing the coordinates of the parking space in the virtual coupling's reference frame.

[0085] In our example where the three-dimensional coordinates of the four corners of the parking space are acquired, twelve "position" fields will be provided (corresponding respectively to the 3 coordinates x, y, z of the four corners).

[0086] For each time step, the table will contain a new record. The first of these records will therefore store the initial value of the steering angle, the initial value of the speed and the initial position of the parking space relative to the virtual hitch.

[0087] It is of course possible to enrich the database with other elements, for example the shape of the parking space, the position of obstacles around this parking space...

[0088] Once the database has been obtained, the second learning step consists of providing the database to the neural network so that it can train itself to determine optimal steering angle and speed instructions, taking into account the position of the coupling in its environment.

[0089] The neural network used here is a model of the MTSF - LSTM type (from the English "Multivariate Time Serial Forecasting - Long Short-Term Memory").

[0090] The input layer of the neural network has as many artificial neurons as each table in the database has "position" fields. In our example, it therefore has twelve.

[0091] The output layer here comprises two artificial neurons which respectively provide a steering angle command and a speed command.

[0092] The neural network training step then allows the different artificial neurons to be parameterized so that the neural network can determine the variations over time of the steering angle to be applied to the steering wheel of the motor vehicle 10 to reach any parking space as a function of the initial posture and then the instantaneous posture of the motor vehicle 10 in relation to the parking space 20.

[0093] Once the neural network is properly parameterized, this neural network is implemented in the computer of the real motor vehicle 10.

[0094] It may be noted that several neural networks, parameterized differently and each associated with a distinct type of trailer, can be recorded in the computer.

[0095] So, when a driver is at the wheel of the motor vehicle 10 and wishes to reverse into a parking space 20, he can be assisted by the computer 13. We can now describe how this driving assistance method is implemented.

[0096] The first step for the computer 13 is to acquire the dimensional characteristics of the trailer 11.

[0097] This step can be performed manually by the driver, who will be asked to enter information such as the length and width of the trailer, the number of axle(s), the distance between the coupling ball and each axle...

[0098] Alternatively, this step could be carried out differently, for example by analyzing the images acquired by the reversing camera or by determining how the trailer behaves when reversing.

[0099] At the end of this step, the computer 13 is able to determine which network of neurons it must use from among the set of recorded neural networks.

[0100] The second step consists, for the driver 20, in positioning his vehicle in a position suitable for starting the parking maneuver. This position is such that the parking space 20 is visible in the images acquired by the reversing camera (see figure 3) and that the trailer 11 is positioned in line with the motor vehicle 10.

[0101] The third step consists, for the driver 20, of starting the parking assistance function.

[0102] For this purpose, the user can, for example, indicate the position of the parking space 20 in which they wish to park using the touchscreen 15. Alternatively, the parking space can be located in another way, as described, for example, in document FR1853260.

[0103] In this way, the computer 13 is able to determine the position of this parking space in the frame of the actual coupling 1, in the same way as above, for example by determining the three-dimensional coordinates of the four corners of the parking space 20.

[0104] Based on this single position data point, the neural network will be able to provide the steering angle and speed instructions most suitable for parking the vehicle. These instructions can vary at every moment of the maneuver, for example every hundredth of a second, so that the actual coupling 1 can perform complex parking maneuvers.

[0105] In practice, these steering angle and speed instructions can be used in various ways, depending on whether the motor vehicle 10 has actuators 16A, 17A, 18A allowing the vehicle to be parked automatically and depending on the wish expressed by the driver 20.

[0106] Thus, in a first example, the instructions can simply be displayed on the touch screen 15, so that the driver 20 knows at all times what steering angle to apply and what speed to adopt.

[0107] In a second example, the driver 20 can simply control the speed of the vehicle, using the speed setting displayed on the touch screen 15, while the steering wheel orientation is automatically controlled by the actuator 16A according to the calculated steering angle setting.

[0108] In a third example, the vehicle actuators 16A, 17A, 18A can be controlled by the computer 13 so as to park the actual coupling 1 without the help of the driver.

[0109] The present invention is in no way limited to the embodiment described and represented, but a person skilled in the art will be able to make any variation in accordance with the invention.

[0110] Thus, by way of example, one could consider that the neural network does not does not provide a speed instruction, the latter being then chosen by the driver or by the computer, depending on the situation perceived.

Claims

Claims

1. Method for assisting in reversing a motor vehicle (10) to the rear of which a trailer (11) is coupled, the method comprising: - a step of acquiring the relative posture of the motor vehicle (10) with respect to a target zone (20), and - a step of determining a steering angle setpoint enabling the motor vehicle (10) to reverse to said target zone (20), by means of a computer (13) in which an artificial neural network is recorded, the relative position of the target zone (20) with respect to the motor vehicle (10) being provided as input to said artificial neural network and said steering angle setpoint being generated as output from said artificial neural network.

2. Driving assistance method according to claim 1, in which the target zone (20) is a parking space selected in advance by the computer (13) or by a driver of the motor vehicle (10).

3. Driving assistance method according to one of claims 1 and 2, in which, in the determination step, the artificial neural network generates as output a speed instruction allowing the motor vehicle (10) to reverse to said target zone (20).

4. Driving assistance method according to one of claims 1 to 3, in which, after the determination step, the steering angle instruction is displayed or transmitted vocally to the driver of the motor vehicle (10).

5. Driving assistance method according to one of claims 1 to 3, in which, after the determination step, the steering angle setpoint is transmitted to an actuator controlling the orientation of the steered wheels of the motor vehicle.

6. Method for configuring a calculator (13) for assisting with driving in reverse of a motor vehicle (10) to the rear of which a trailer (11) is coupled, said method comprising: - a step of developing a database associating with each initial posture of a reference vehicle (10') relative to a target zone (20), trajectory data enabling the reference vehicle (10') to reverse to said target zone (20), and - a step of learning a neural network, which makes it possible to de- terminating the variations over time of a steering angle to be applied to the steering wheel of the motor vehicle (10) to reach any target zone as a function of an initial posture of the motor vehicle (10) relative to the target zone (20), said neural network receiving said database as training data, and - a saving of the neural network in the computer.

7. Parameterization method according to the preceding claim, in which the database is developed using as reference vehicle (10') a virtual model of the motor vehicle (10) equipped with a trailer (11') in a virtual environment (2') comprising a plurality of parking spaces, a physical person driving the virtual model so as to park it in reverse in said parking spaces.

8. A parameterization method according to the preceding claim, wherein, to drive the virtual model, the driver uses a view taken from the virtual motor vehicle model and facing towards the rear of the virtual vehicle model, and a top view of the virtual model.

9. Parameterization method according to one of the two preceding claims, in which, before parking the virtual model in said parking space, said natural person indicates the position of the parking space.

10. Device for assisting driving in reverse of a motor vehicle (10) to the rear of which a trailer (11) is coupled, characterized in that it comprises a reversing camera and a computer which is programmed to implement a driving assistance method according to one of claims 1 to 5 and which is parameterized according to the parameterization method according to one of claims 6 to 9.

Images