Method for controlling a motor vehicle

Abstract

The invention relates to a method for controlling a motor vehicle (10) driving in a lane (31) on a road, the vehicle comprising a computer programmed so as to perform: i) a line tracking function which comprises determining a lane trajectory (40) and controlling the motor vehicle so as to follow this trajectory; ii) a target tracking function which comprises determining a target trajectory (45) and controlling the motor vehicle so as to follow this trajectory. According to the invention, the method comprises a step of calculating, on the basis of the lane trajectory and of the target trajectory, a spatial offset (D1, D2, D3) representative of a distance between these trajectories, when the tracking function is active, a step of controlling the motor vehicle according to this spatial offset.

Description

Description Title of the invention: Method for piloting a motor vehicle 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 piloting a motor vehicle travelling on a traffic lane of a road, the vehicle comprising a computer programmed to implement: i) a lane following function comprising, in the active state, the determination of a traffic lane trajectory on the basis of data relating to edge lines delimiting the traffic lane and the piloting of the motor vehicle so as to follow the traffic lane trajectory; ii) a target following function comprising, in the active state, the determination of a target trajectory on the basis of data relating to a third vehicle preceding the motor vehicle on the traffic lane and the piloting of the motor vehicle so as to follow the target trajectory.

[0003] The invention also relates to a motor vehicle adapted to implement such a method. It applies more particularly to cars and other motorized vehicles traveling on roads. State of the art

[0004] In an effort to improve the safety of motor vehicles, they are currently being equipped with driver assistance systems and even highly automated driving systems.

[0005] Driver assistance systems typically include a lane-keeping function that automatically guides the vehicle within its lane based on lane markings. This lane-keeping function is generally a lane-centering feature, also known as lane centering assistance (LCA), designed to keep the vehicle centered in the lane.

[0006] The centering function needs to know the position of the edges of the lane the vehicle is traveling in to operate. Currently, this is achieved using a sensor, such as a camera, equipped with image processing capabilities to determine the position of each lane marking. The vehicle's onboard computer can then deduce the position of the center line of the lane, which allows it to subsequently... to automatically steer the vehicle so that it follows this centerline.

[0007] Unfortunately, this detection method is not always entirely satisfactory. This is particularly true when the camera fails to detect the lane edges (for example, because the lane markings are faded, the lighting is poor, or the markings are obscured by a truck in front of the vehicle) or when the markings are absent (for example, because the lane splits into two lanes or because two lanes merge into one). The centering function can then no longer operate effectively and must be deactivated, which is inconvenient for the driver.

[0008] To address this shortcoming, driver assistance systems often include a target-following function (commonly known as "Autosteer"), which detects and follows the vehicle in front of the one being monitored. This function allows, for example, a truck driver to safely follow another truck. By mimicking the target's trajectory, the vehicle theoretically remains correctly positioned within its lane.

[0009] In practice, the vehicle's steering switches from the line-following function (which remains the default operating mode when lines are detected) to the target-following function when the perception module is no longer able to detect lines.

[0010] However, this switching between functions can lead to discontinuities in vehicle control. For example, when the target is not centered on the lane and the vehicle is, the vehicle may abruptly veer off course when the target tracking function activates. In other words, the vehicle is prone to deviations during the switch between lane-following and target tracking functions. Presentation of the invention

[0011] In this context, the present invention proposes an improved transmission between these two tracking functions.

[0012] More specifically, the invention proposes a method for controlling a motor vehicle using a computer as defined in the introduction. The method comprises, when the lane-following function is activated and the target-following function is deactivated, and then when the target-following function is to be activated and the lane-following function is to be deactivated, a calculation step by the computer, based on the lane trajectory and the target trajectory, of a spatial offset representing a distance between the lane trajectory and the target trajectory. Then, when the target-following function is active, a step of piloting the motor vehicle according to said spatial offset.

[0013] Thus, thanks to the invention, the transition between lane-following and target-following control is seamless. Indeed, the difference in trajectory is anticipated while the lane-following function is still active by analyzing the position of the other vehicle in the lane. Therefore, by calculating the spatial offset, the computer can compensate for any discontinuity between the guidance provided by the lane-following function and that provided by the target-following function.

[0014] Thus, for example, when switching from lane-following to target-following, the control unit can follow the target trajectory, compensated by the spatial offset, to maintain continuity with the lane-following trajectory. In other words, by applying the spatial offset to the target trajectory, the control unit prevents the vehicle from deviating during the switch between lane-following and target-following.

[0015] Additionally, when switching from target tracking to line tracking, the calculator also takes into account the spatial offset to avoid or at least minimize deviations.

[0016] Furthermore, the method according to the invention makes it possible to improve the target tracking function by taking into account, throughout the tracking of the third vehicle, the spatial offset.

[0017] For example, by analyzing successive positions of the other vehicle on the road, the computer can detect an oscillating trajectory. By calculating the spatial offset, the computer can compensate for these oscillations to make the vehicle's steering more linear than the target trajectory when the target tracking function is active.

[0018] In essence, the method according to the invention therefore makes it possible to maintain consistent trajectories between the traffic lane following trajectory and the target following trajectory.

[0019] Other advantageous and non-limiting features of the process according to the invention, taken individually or in all technically possible combinations, are as follows: - when the tracking function is active, the steering of the motor vehicle according to said spatial offset is carried out by means of a neural network; - the neural network is adapted to detect oscillations in the target trajectory or a difference between a curvature of the target trajectory and a curvature of the traffic lane trajectory and, when the target tracking function is active, to determine a suitable trajectory attenuating the oscillations of the target trajectory or modifying the curvature of the target trajectory; - the neural network is a forward-propagating neural network or a network of recurrent neurons; - the neural network is adapted to determine variable lateral compensation based on spatial offset, and to correct or adapt the target trajectory based on the variable lateral compensation; - a preliminary computer configuration operation is planned, including: a step of developing a database associating typical trajectories that a vehicle is likely to follow and typical profiles that the edge lines are likely to follow with theoretical values ​​of the variable lateral compensation; a step of training the neural network using the database; and a step of saving the neural network in the computer; - the calculation of variable lateral compensation also takes into account at least one of the following parameters: spatial offset; an instantaneous movement of the motor vehicle; an instantaneous movement of the third vehicle; data relating to the edge lines; information from a motor vehicle navigation system; contextual information on a state of the traffic lane; - the target trajectory is determined by fitting a polynomial function to a plurality of positions of the third-party vehicle; - the spatial offset is calculated according to a lateral direction inclined relative to the line following trajectory.

[0020] The invention also proposes a motor vehicle traveling on a traffic lane of a road, the vehicle comprising a computer programmed to implement the aforementioned tracking functions and to implement a control method as described above.

[0021] Of course, the different features, variants and embodiments of the invention can be combined with each other in various ways as long as they are not incompatible or mutually exclusive. Detailed description of the invention

[0022] The description that follows, 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.

[0023] Regarding the attached drawings:

[0024] [Fig.1] is a schematic view of a motor vehicle adapted to implement a process according to the invention;

[0025] [Fig.2] is a diagram illustrating algorithmic blocks of a computer in the motor vehicle of [Fig.1];

[0026] [Fig. 3] is a diagram representing, at three successive moments, the motor vehicle of [Fig. 1], seen from above, and a vehicle preceding it following a trajectory straight line, off-center to the left;

[0027] [Fig.4] is a diagram illustrating algorithmic blocks of a neural network from the computer in [Fig.2];

[0028] [Fig.5] is a diagram representing at three successive moments the motor vehicle of [Fig.1], seen from above, and a vehicle preceding it adopting an oscillating trajectory off-center to the left;

[0029] [Fig.6] is a diagram representing at two successive moments the motor vehicle of [Fig.1], seen from above, and a vehicle which precedes it during a turn.

[0030] Figure 1 shows a motor vehicle 10 adapted to implement the invention. This is a car. Alternatively, it could be another type of vehicle (truck, motorcycle, etc.).

[0031] In this figure, the motor vehicle 10 is travelling on a traffic lane 31 of a road 30. We observe that the road 30 has two lateral marking lines 34, 35 (which delimit it) and a central marking line 33 delimiting two traffic lanes 31, 32 of the road 30.

[0032] A traffic lane is defined here as the portion of a road on which a motor vehicle is permitted to travel. Such a traffic lane is generally delimited by edge lines. In [Fig. 1], the edge lines include the right-hand side marking line 35 (relative to the direction of travel of the vehicle) and the center marking line 33.

[0033] Alternatively, a traffic lane can also be delimited, for example on one side only, by a structure such as a safety barrier, a fence, bollards, or a feature of the terrain such as a shoulder. In the context of this invention, this delimitation, provided it extends continuously, also defines an edge line.

[0034] A road (or highway) is defined as a set of traffic lanes. In the illustrative example in [Fig. 1], road 30 therefore has two traffic lanes 31, 32 on which vehicles can travel in the same direction.

[0035] Figure 3 shows motor vehicle 10 and a third vehicle preceding it in the same lane 31, across three sections P1, P2, and P3 of the road 30, corresponding to three successive moments. In the first section P1, the edge lines are formed by the two lateral markings 34 and 35. The second section P2 represents a widening of the lane, with the lateral markings 34 and 35 moving apart. In the third section P3, the lane is split, and the edge lines are formed by one lateral marking 35 and the center line 33.

[0036] The motor vehicle 10, which is the subject of the present invention, is therefore the one that follows the other. It will hereafter be referred to as vehicle-ego 10. The third vehicle will be called target vehicle 20.

[0037] The target vehicle 20 could be of any type. We will only consider that it is in motion and traveling in the same traffic lane 31 as the ego vehicle 10, in front of it.

[0038] As shown in [Fig.1], the ego-vehicle 10 classically includes a passenger compartment in which there is, among other things, a seat for the driver 20 of the vehicle and a steering wheel 12.

[0039] This ego 10 vehicle typically includes a powertrain, a braking system, and a steering system for turning the vehicle. Typically, the steering system includes an electronically controlled power steering actuator, the powertrain includes an electronically controlled engine control actuator, and the braking system includes an electronically controlled brake actuator.

[0040] The ego 10 vehicle also includes an electronic and / or computer processing unit, hereinafter referred to as the computer 11, comprising at least one microprocessor, at least one memory, and input and output interfaces. The computer is embedded in the ego 10 vehicle.

[0041] Thanks to its input interfaces, the calculator 11 is adapted to receive various input data from third-party sensors or calculators.

[0042] Among these sensors, for example, a front camera is planned to detect the edge lines of the traffic lane 31 being used, and a remote detector (RADAR and / or LIDAR) to detect objects in the environment of the ego vehicle 10 (in particular the target vehicle 20).

[0043] Thanks to its output interfaces, the computer is suitable for controlling the power steering actuator, the engine control actuator, and the braking actuator.

[0044] Thanks to its memory, the computer stores a computer application, consisting of computer programs including instructions whose execution by the computer allows the implementation of a line tracking function and a target tracking function.

[0045] Here, the line-following function is a centering function, hereafter referred to as function LC A, designed to keep the ego-vehicle 10 in the center of its traffic lane 31.

[0046] Alternatively, the lane-following function could be a single edge-line following function, i.e., a single lane boundary, designed to keep the ego-vehicle at a given distance from the edge line.

[0047] The target tracking function, hereafter referred to as the AUS function, is designed to ensure the tracking of the target vehicle 20.

[0048] The execution of instructions by the calculator 11 more generally allows the implementation implementation of the process described below.

[0049] In summary, the LCA function will be preferred for laterally steering the ego 10 vehicle in order to center it in its lane 31. Lateral direction is defined here as the left-right direction of the ego 10 vehicle which is orthogonal to the front-back and up-down directions of the ego 10 vehicle. The AUS function will be used when the edge lines are not visible or usable, so the LCA function must be interrupted.

[0050] Figure 2 shows a schematic representation of part of the software architecture of calculator 11, which allows choosing which function to activate, how to implement it and how to proceed with transitions between functions.

[0051] A first block, B1, is designed, in the initial stage, to receive data from the sensors. This data could already be processed by processors embedded in the sensors. Typically, the camera could incorporate a processor adapted to establish the mathematical equations for the track edge lines.

[0052] However, here we will consider this to be raw, unprocessed data. When it receives this data, the computer 11 fuses it, that is, it processes it in a combined manner in order to deduce information about the dashboard lines (typically their equations) and information about the objects present in the environment of the ego vehicle 10 (typically the position of the target vehicle 20). The fusion operation typically consists of comparing the data from the camera and the remote sensor in order to obtain new, reliable, and precise data.

[0053] A second block B2 is adapted to determine, in a second step, a traffic lane trajectory 40 on the basis of first data, provided by the first block Bl, relating to the edge lines delimiting the traffic lane 31.

[0054] A third block B3 is adapted to determine, in a third step, a target trajectory 45 on the basis of second data, provided by the first block Bl, relating to the target vehicle 20.

[0055] A fourth block B4 is adapted to determine, on the basis of the traffic lane trajectory 40 and the target trajectory 45, a spatial offset D representative of a lateral distance between the traffic lane trajectory 40 and the target trajectory 45.

[0056] The fourth block B4 is also designed to choose which function to activate from among the aforementioned LCA and AUS functions.

[0057] Regardless of the tracking function selected, the fourth block B4 is suitable for determining steering quantities relating to the ego-vehicle 10 including for example: a deviation from the midline, a heading angle, a curvature, a curvature derivative.

[0058] When the LCA function is active, the fourth block B4 primarily uses the Data from the second block B2 is used to determine the control parameters. When the AUS function is active, the fourth block B4 uses both the data from the second block B2 and the third block B3 to determine the control parameters.

[0059] Typically, the fourth block B4 favours the LCA function as long as the information on the lane edge lines is deemed reliable, and the AUS function otherwise (provided that a target vehicle 20 precedes the ego vehicle 10).

[0060] Thus, in general, the switch between the two functions can be made based on the following information: - high-quality edge line marking, - a criterion of visibility of road 30 by the camera, - a geometric characteristic of the traffic lane 31, such as its width, - a dynamic characteristic of the motor vehicle 10 such as its speed, - a visibility criterion of the target vehicle 20 by the remote detector.

[0061] In practice, the LCA function remains active whenever conditions allow, with the AUS function taking over when conditions are no longer favorable. Block B4 is specifically responsible for determining when to use the AUS function rather than the LCA function. In this case, the AUS function can only be activated if there is a third vehicle preceding the ego 10 vehicle in the same lane.

[0062] The fourth block, B4, sends information about the tracking function to be used, as well as the control parameters, to a fifth block, B5, which is responsible for transmitting a control command to the power steering actuator. This control command is presented here as an angle to be applied to the steering wheel or the wheels.

[0063] We can then describe in more detail the process implemented by the computer 11 to automatically control the ego vehicle 10 in its lane 31 using the LCA and AUS functions. This process will be implemented in a loop, at regular time steps, i.e. with a constant sampling step.

[0064] To determine the trajectory of traffic lane 40, the second block B2 receives as input the first data relating to the edge lines delimiting the traffic lane 31, which include in particular the position of the edge lines in an OXY frame attached to the ego-vehicle 10 (see [Fig.3]).

[0065] This OXY coordinate system is attached to the ego-vehicle 10 in that the center O is located at the level of the ego-vehicle 10, for example on its front face, as shown in Figures 1 and 3, or at its center of gravity or its rear axle. The x-axis is oriented forward, along the longitudinal axis of the ego-vehicle 10. The y-axis is oriented laterally, to the left, at a right angle to the x-axis.

[0066] The computer 11 then determines the equations of the edge lines bordering its traffic lane 31. From this, it can deduce the equation of the centerline of its traffic lane 31. The centerline is defined here as the line equidistant from the two edge lines, i.e., the left and right lines, delimiting the traffic lane 31. Since the LC A function ensures the centering of the vehicle 10 within its traffic lane 31, this equation of the centerline thus characterizes the trajectory of the traffic lane 40. In other words, the trajectory of the traffic lane 40 is superimposed on the centerline. Hereafter, only the centerline will be used interchangeably to refer to both the centerline and the trajectory of the traffic lane 40.

[0067] As is well known, the control unit 11 stores a controller which, given this midline equation, determines a control command for the power steering actuator within the framework of the LCA function. This controller is a mathematical operator that receives as input various variables, at least one of which is a function of the midline, and provides as output the control command to follow this midline.

[0068] Thus, in its active state, the LCA function allows the ego vehicle 10 to be steered to follow the center line. The control unit 11 then transmits the appropriate steering command. In its inactive state, the LCA function does not determine the center line (for example, if one of the edge lines is not visible), and the control unit 11 therefore does not transmit any steering command to keep the ego vehicle 10 in the center of the traffic lane 31.

[0069] In its active state, the AUS function allows the ego vehicle 10 to be guided to follow the target vehicle 20. As explained later, the control unit 11 then transmits the appropriate guidance command, also taking into account the spatial offset D. In its inactive state, the AUS function does not guide the ego vehicle 10 to follow the target vehicle 20. Thus, the control unit 11 does not transmit any guidance command to follow the target vehicle 20. However, even in its inactive state, the AUS function allows the trajectory of the target 45 to be determined, which in turn allows the spatial offset D to be determined.

[0070] To determine the trajectory of target 45, the third block B3 receives as input the second data relating to the target vehicle 20 which includes in particular the position of the target vehicle 20 in the OXY frame.

[0071] It should be noted that this position of the target vehicle 20 will be recorded in the form of two coordinates (x, y) expressed in the OXY frame.

[0072] Each position of the target vehicle 20 is stored in the memory of the computer 11. It should be noted here that only the last N positions of the target vehicle 20 will remain recorded in this memory (with N a predetermined constant, preferably- (actually greater than 10). The oldest position of the target vehicle 20 will therefore be deleted each time a new position is recorded. In practice, the part of the computer's memory 11 associated with storing these positions, hereinafter referred to as the register, will have a fixed and predefined size.

[0073] This register has two fields, one for the axial coordinate and the other for the lateral coordinate of the position of the target vehicle 20. It also contains N records corresponding to the coordinates of the target vehicle 20 at the N previous time steps.

[0074] Given the positions stored in the register, the third block B3 of the computer 11 is able to calculate the coefficients of a function illustrating the trajectory of the target vehicle 20.

[0075] Here, the function considered is a polynomial function. It is preferably of order 3. Thus, the equation of the trajectory of the target vehicle 20 is defined as follows:

[0076] [Math.l] y = c0 + cc + c2.x 2 + CX 3

[0077] The objective is therefore to determine the values ​​of the coefficients c0, c bc2, c3 are such that the polynomial curve passes as close as possible to the N points recorded in the register. Thus, the trajectory of target 45 is determined by fitting a polynomial function, that is, by parameterizing the coefficients that define it.

[0078] For this, we could use a least squares minimization method.

[0079] However, the solution used here is different.

[0080] We can first write, for the N points with coordinates (x, y,) corresponding to the last N positions of the target vehicle 20 in the frame attached to the ego vehicle 10:

[0082] The variable xi corresponds to the longitudinal coordinate measured along the X-axis of the OXY coordinate system. The variable 3^ corresponds to the estimated coordinate of point i. This estimated coordinate is calculated at each time step and is recorded in a third field of the register, based on the previous lateral coordinates.

[0083] We can also introduce the variable T, which corresponds to the measured lateral coordinate of point i, and the coefficients CQ. C 1> C 2, C 3 (j=0,..,3) of the trajectory which are also estimated at each time step so as to correspond precisely to the trajectory of target 45.

[0084] The CQ coefficients, sont more specifically determined by the following equation:

[0088] At this stage, the fourth block B4 will be able to calculate the spatial shift D.

[0089] The spatial offset D is a distance defined in the lateral direction, which here also corresponds to the direction perpendicular to the median line (in a plane parallel to the road). When the LC A function is active and the ego vehicle 10 moves along the median line, the lateral direction can therefore be approximated to the Y-axis at the center O of the OXY coordinate system.

[0090] Here, the spatial offset D is defined, for example, as a distance, perpendicular to the midline, between the midline and the target trajectory 45. Alternatively, the spatial offset D can be defined along another direction making a non-zero angle with the median line, i.e. a direction distinct from the longitudinal direction. In all cases, the spatial offset D is a lateral offset in the sense that it represents a deviation between the traffic lane trajectory 40 and the target trajectory 45 along a direction inclined with respect to the longitudinal axis.

[0091] As shown in [Fig. 3], the spatial offset D can be calculated at a point along the centerline (and therefore along the superimposed trajectory of the traffic lane 40). The spatial offset can, for example, be calculated at the center of gravity of the ego vehicle 10, in which case it is referenced as DI in [Fig. 3]. It can also be calculated at the center of gravity of the target vehicle 20, in which case it is referenced as D3 in [Fig. 3]. Furthermore, it can be calculated at a predefined or variable aiming distance L in front of the ego vehicle 10 and behind the target vehicle 20, in which case it is referenced as D2 in [Fig. 3]. The aiming distance L varies, for example, depending on the speed of the ego vehicle 10.

[0092] The spatial offset D is, for example, calculated as the difference in ordinate between a point on the midline and a point on the target trajectory 45 that shares the same abscissa as said midline point. The spatial offset D is therefore a distance value.

[0093] Preferably, the spatial offset D is calculated at a plurality of points along the midline over time. More specifically, it is calculated once at each time step. The spatial offset D then comprises a set of distance values, each value corresponding to a difference in ordinates and being bijectively associated with an abscissa in the OXY coordinate system. In practice, the spatial offset D is calculated here for the N recorded positions of the target vehicle 20.

[0094] The distance values ​​included in the spatial offset D are positive when the trajectory of target 45 is located to the left of the midline, as shown in [Fig. 3], and negative when the trajectory of target 45 is located to the right of the midline. This convention depends on the orientation chosen for the OXY coordinate system.

[0095] In a first embodiment, when the AUS function is activated, the steering of the ego vehicle 10 according to the spatial offset D involves tracking the trajectory of target 45 minus the spatial offset D. More specifically, the value or one of the values ​​included in the spatial offset D is subtracted. When the target vehicle 20 is off-center to the left, the spatial offset D is positive, and subtracting it induces an offset to the right. Conversely, when the target vehicle is off-center to the right, the spatial offset D is negative, and subtracting it induces an offset to the left.

[0096] The idea, in this first embodiment, is thus to apply a lateral compensation of fixed value to the trajectory of target 45.

[0097] This first embodiment is illustrated more specifically in [Fig. 3]. It works particularly effectively when the target vehicle 20 follows a trajectory parallel to the traffic lane 31. In this example, the target vehicle 20 remains off-center to the left.

[0098] In [Fig. 3], on the first PI segment of route 30, therefore during an initial period, the LC A function is activated and the AUS function is deactivated. On this first PI segment, the spatial offset is calculated, for example, the one referenced D3 at the level of the target vehicle 20.

[0099] On a second section P2 of the road, the traffic lane 31 of the ego-10 vehicle widens. The control of the ego-10 vehicle then switches from the LCA function to the AUS function: the LCA function must be deactivated and the AUS function must be activated.

[0100] As illustrated on this second section P2, when the AUS function is active, the piloting of the ego-vehicle 10 along the target trajectory 45 is carried out taking into account the spatial offset D.

[0101] We can clearly observe, during the transition from the first section PI to the second section P2, that the vehicle-ego 10 adopts a consistent trajectory, i.e., without any change in slope. In other words, the vehicle-ego 10 maintains a straight trajectory during the transition between the LCA function and the AUS function.

[0102] This figure clearly shows that without taking into account the spatial shift, the ego vehicle 10 would have made a potentially dangerous deviation to start following the target vehicle 20 during the passage from the first section PI to the second section P2.

[0103] Advantageously, the spatial offset D is taken into account throughout the period when the AUS function is active (which corresponds to the second segment P2). On the second segment P2, the ego vehicle 10 then follows a straight, adapted trajectory 41 in line with the midline of the first segment PL. To achieve this, schematically, the computer 11 applies a fixed lateral compensation value VI, which here is equal to the spatial offset value D3, to the target trajectory 45.

[0104] Finally, the transition from the second section P2 to the third section P3 illustrates the switch from the AUS function to the LCA function. Indeed, as seen in [Fig. 3], a new edge line, here the central marking line 33, appears once the lane split is complete. The LCA function can therefore operate again, and the ego vehicle follows the center line.

[0105] In a second embodiment, when the LCA function is activated, the steering of the ego-vehicle 10 according to the spatial offset D includes the use of a variable lateral compensation determined on the basis of the spatial offset D. The term "variable" means that the variable lateral compensation varies over time or along the X-axis, i.e., with the forward movement of the ego-vehicle 10. The tra- target jection 45 is then more finely adapted than with a fixed value lateral compensation.

[0106] Preferably, in this second embodiment, the spatial offset D comprises a plurality of distance values ​​which follow each other temporally and which are taken into account by the calculator 11.

[0107] The idea of ​​this second embodiment is to determine the trajectory of target 45, while the LC A function is still active, in order to take into account the road context and the type of driving of the target vehicle 20. For this, the fourth block B4, an example of whose architecture is shown schematically in [Fig.4], includes here a neural network.

[0108] The neural network is represented schematically by a first sub-block B41. This first sub-block B41 is adapted to determine variable lateral compensation. It transmits the variable lateral compensation as output to a second sub-block B42. Sub-block B41 here has several layers of neurons, for example, from 4 to 6 layers. The neural network is, for example, a forward-propagating network or a recurrent neural network.

[0109] The neural network receives, among other things, the spatial offset D as input. The neural network also receives the position of the target vehicle 20, the speed of the target vehicle 20, and the equations of the edge lines, expressed in the coordinate system attached to the ego vehicle. The first sub-block B41 also receives as input data representing the instantaneous displacement of the ego vehicle 10 or information from a navigation system of the ego vehicle 10.

[0110] The neural network (first sub-block B41) also receives contextual information as input about the state of lane 31 or the behavior of the target vehicle 20. This contextual information includes, for example, the width of lane 31, the presence of roadworks on lane 31, information detected on road signs, or confirmation of a lane change by the target vehicle 20. Thus, for example, when the contextual information indicates a temporary narrowing of the lane, the neural network can determine the variable lateral compensation in order to reduce the trajectory correction. Indeed, applying too much lateral compensation could cause the target vehicle 10 to leave lane 31.

[0111] This second sub-block B42 allows consolidation by comparing the variable lateral compensation to the position of the target vehicle 20 or to the target trajectory 45.

[0112] The fourth block B4 then includes a third sub-block B43 adapted to take into account the variable lateral compensation when the AUS function.

[0113] At the output, the third sub-block B43 sends information on the control parameters to the fifth block B5, which then transmits a control instruction to 1. Power steering actuator.

[0114] Here, the neural network is trained before being recorded on computer 11. Alternatively, the training could be performed in real time, particularly during the driving of the ego vehicle, allowing the neural network to adapt to the user's preferences. However, in this case, training beforehand, compared to real-time training, reduces the required computing power and simplifies the neural network architecture.

[0115] Here, the process includes a step of developing a database of typical trajectories that the target vehicle 20 is likely to follow and typical profiles that the edge lines are likely to follow.

[0116] The typical trajectories that the target vehicle 20 is likely to follow are representative of driving behaviors. They include, for example, trajectories aligned with the centerline, trajectories offset from the centerline, oscillating trajectories (for example, around the centerline), centered trajectories in turns, tight trajectories in turns, etc.

[0117] The typical profiles that the edge lines are likely to follow are representative of road contexts, particularly those relating to traffic lane 31. They include, for example, edge line profiles associated with straight sections of traffic lane 31, curved sections of traffic lane 31, lane splits or mergers of traffic lane 31, etc.

[0118] Once the database of typical trajectories for the target vehicle (20) and typical profiles of the vehicle's edge lines has been developed—which amounts to determining typical lateral offsets—a theoretical variable lateral compensation is associated with each group consisting of a typical trajectory and a typical profile. The theoretical variable lateral compensations are determined, for example, to ensure safe driving. The process then includes a step of training the neural network using the database and the theoretical variable lateral compensations. In other words, the typical trajectories and profiles are the inputs, and the theoretical variable lateral compensations are the outputs on which the neural network is trained. This supervised learning allows for an effective response to the various road conditions that may arise.

[0119] Finally, the process includes a step of saving the neural network in the memory of computer 11.

[0120] Thanks to the neural network, the computer 11 can analyze the trajectory of target 45 by comparing it to the centerline and thus learn to correct or adapt the trajectory of target 45 when the AUS function is activated and the LCA function is deactivated. Correcting or adapting the trajectory of target 45 when the AUS function is active makes it possible, for example, to make piloting safer or to make the driving more in line with the preferences of the vehicle driver - ego 10.

[0121] In this second embodiment, in addition to preventing discrepancies during transitions between the LCA and AUS functions (as in the first embodiment), the computer 11 can therefore adapt the tracking of the target vehicle 20 based on the learning performed while the LCA function is active. Advantageously, the spatial offset D is thus taken into account for the entire period during which the AUS function is active. This improves the safety and comfort of the passengers of the ego vehicle 10.

[0122] The correction, or adaptation, by computer 11 of the trajectory of target 45 on the basis of variable lateral compensation, when the AUS function is active, is illustrated by two examples in figures 5 and 6.

[0123] In [Fig. 5], the target vehicle 20 exhibits an oscillating trajectory. In the first segment P1, the control unit 11 is adapted to detect these oscillations by comparing the trajectory of the target 45 to the centerline. The control unit 11 is then programmed to attenuate the oscillations of the target vehicle 20 in the second segment P2, i.e., when the vehicle is being controlled by the AUS function.

[0124] On [Fig.5], four values ​​DI', D2', D3', D4' of spatial offset are illustrated on the first portion PL. It is understood that the computer 11, on the basis of these values, can analyze the oscillations of the target trajectory 45, for example an average period and an average amplitude of the oscillations.

[0125] Thus, as shown in [Fig. 5], thanks to the neural network, the ego vehicle 10 follows a straight, adapted trajectory 41 on the second segment P2, even though the AUS function is active and the target vehicle 20 adopts an oscillating trajectory. To achieve this, different variable lateral compensation values ​​are applied at different points along the target trajectory 45. Here, schematically, five lateral compensation values ​​VI', V2', V3', V4', V5' are represented, for example, at five points along the target trajectory 45.

[0126] Furthermore, the neural network can determine what value of the spatial offset should be taken into account when switching from the LCA function to the AUS function, for example as a function of the average period of the oscillations, to prevent a deviation of the ego-vehicle 10.

[0127] Finally, the transition from the second section P2 to the third section P3 illustrates the switch from the AUS function to the LCA function. Indeed, as seen in [Fig. 3], a new edge line, here the central marking line 33, appears once the lane split is complete. The LCA function can therefore operate again, and the ego vehicle 10 follows the center line.

[0128] In [Fig. 6], the target vehicle 20 exhibits a curved trajectory with a greater curvature than that of the traffic lane 31. In the first segment PI, the Calculator 11 is adapted to determine the difference between a curvature of the target trajectory 45 and a curvature of the midline. In [Fig. 6], two values ​​DI”, D2” of spatial offset are illustrated.

[0129] The computer 11 can then determine a transform designed to adapt the curvature of the target trajectory 45 on the second segment P2, that is, when steering is performed by the AUS function, for example, so that the ego vehicle 10 takes a less sharp turn. Thus, as shown in [Fig. 6], thanks to the neural network, the ego vehicle 10 follows an adapted trajectory 41 that is generally parallel to the edge lines, even though the AUS function is active and the target vehicle 20 passes close to the inside of the turn. To achieve this, different lateral compensation values ​​are applied at different points on the target trajectory 45. Here, schematically, two variable lateral compensation values ​​VI”, V2”, are shown, for example, at two points on the target trajectory 45.

[0130] The neural network can take into account the driver's preferences and, for example, learn that the driver prefers a sporty driving style to determine variable lateral compensation. When the AUS function is activated, the computer 11 can then steer the ego vehicle 10 to the inside of the curve even though the target vehicle 20 remains centered between the edge lines.

[0131] The present invention is in no way limited to the embodiments described and represented, but a person skilled in the art will be able to make any variation in accordance with the invention.

Claims

Claims

1. A method for controlling a motor vehicle (10) traveling on a traffic lane (31) of a road, the vehicle comprising a computer (11) programmed to implement: i) a line-following function comprising, in the active state, determining a traffic lane trajectory (40) on the basis of data relating to edge lines delimiting the traffic lane (31) and controlling the motor vehicle (10) so as to follow the traffic lane trajectory (40); ii) a target-following function comprising, in the active state, determining a target trajectory (45) on the basis of data relating to a third-party vehicle (20) preceding the motor vehicle (10) on the traffic lane (31) and controlling the motor vehicle (10) so as to follow the target trajectory (45);characterized in that the method comprises, when the line tracking function is activated and the target tracking function is deactivated and then the target tracking function must be activated and the line tracking function must be deactivated, a step of calculating, by the calculator (11), on the basis of the traffic lane trajectory (40) and the target trajectory (45), a spatial offset (D, D1, D2, D3, D1', D2', D3', D4') representative of a distance between the traffic lane trajectory (40) and the target trajectory (45) then, when the target tracking function is active, a step of piloting the motor vehicle (10) as a function of said spatial offset (D, D1, D2, D3, D1', D2', D3', D4').;

2. Method according to claim 1, wherein, when the tracking function is active, the steering of the motor vehicle (10) as a function of said spatial offset (D, D1, D2, D3, D1', D2', D3', D4') is carried out by means of a neural network.

3. Method according to claim 2, wherein the neural network is adapted to detect oscillations of the target trajectory (45) or a difference between a curvature of the target trajectory (45) and a curvature of the taxiway trajectory (40) and, when the target tracking function is active, to determine an adapted trajectory (41) attenuating the oscillations of the target trajectory (45) or modifying the curvature of the target trajectory (45).

4. The method of claim 2 or 3, wherein the neural network is a forward propagation neural network or a recurrent neural network.

5. Method according to one of claims 2 to 4, wherein the neural network is adapted to determine a variable lateral compensation (VI', V2', V3', V4', V5', VI”, V2”) on the basis of the spatial offset (D, Dl, D2, D3, Dl', D2', D3', D4'), and to correct or adapt the target trajectory (45) on the basis of the variable lateral compensation (VI', V2', V3', V4', V5', VI”, V2”).

6. Control method according to claim 5, in which a prior operation of parameterizing the computer (11) is provided, comprising: - a step of developing a database associating typical trajectories that a vehicle is likely to follow and typical profiles that the edge lines are likely to follow with theoretical values ​​of variable lateral compensation (VI', V2', V3', V4', V5', VI”, V2”); - a step of learning the neural network using the database; and - a step of saving the neural network in the calculator (11).

7. Method according to claim 5 or 6, wherein the calculation of the variable lateral compensation (VI', V2', V3', V4', V5', VI”, V2”) also takes into account at least one of the following parameters: - the spatial shift (D, Dl, D2, D3, Dl', D2', D3', D4'); - instantaneous movement of the motor vehicle (10); - instantaneous movement of the third-party vehicle (20); - data relating to edge lines; - information from a motor vehicle navigation system (10); - contextual information on the state of the traffic lane (31).

8. A method according to one of claims 1 to 7, wherein the target trajectory (45) is determined by fitting a polynomial function to a plurality of positions of the third-party vehicle (20).

9. Method according to one of claims 1 to 8, wherein the spatial offset (D, D1, D2, D3, DI', D2', D3', D4') is calculated in a lateral direction inclined relative to the line tracking trajectory.

10. Motor vehicle (10) traveling on a traffic lane (31) of a road, the vehicle comprising a programmed computer (11) to implement: i) a line tracking function comprising, in the active state, determining a traffic lane trajectory (40) on the basis of data relating to edge lines delimiting the traffic lane (31) and steering the motor vehicle (10) so as to follow the traffic lane trajectory (40); ii) a target tracking function comprising, in the active state, determining a target trajectory (45) on the basis of data relating to a third-party vehicle (20) preceding the motor vehicle (10) on the traffic lane (31) and steering the motor vehicle (10) so as to follow the target trajectory (45); characterized in that the computer (11) is programmed to implement a steering method according to one of claims 1 to 9.

Images