Obstacle avoidance method

Abstract

The invention relates to a method for avoiding objects (C1, C2, C3, C4) by a motor vehicle (10), comprising steps of: - detecting objects located in the environment of the motor vehicle, - acquiring data characterizing the position and / or the dynamics of each object then, if several objects have been detected, - verifying whether at least one proximity criterion between at least two of the detected objects is met and, if so, - associating the two objects into a group (G2), - calculating the data characterizing the position and / or the dynamics of said group, and - activating an obstacle avoidance system and / or determining an avoidance trajectory, according to the data characterizing the position and / or the dynamics of said group. Figure for abstract: Fig.10

Description

Title of the invention: Obstacle avoidance method 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 avoiding obstacles.

[0003] The invention also relates to a motor vehicle equipped with a computer adapted to implement this method. State of the art

[0004] In an effort to make motor vehicles safer, they are currently being equipped with driving assistance systems or autonomous driving systems.

[0005] Among these systems, we know in particular the automatic emergency braking systems (better known by the abbreviation AEB, from the English "Automatic Emergency Braking"), designed to avoid any collision with obstacles located in the lane taken by the vehicle, by simply acting on the conventional braking system of the motor vehicle.

[0006] However, there are situations in which these emergency braking systems do not allow a collision to be avoided or are not usable (for example if another vehicle is following the motor vehicle too closely).

[0007] For these situations, automatic avoidance systems (better known by the abbreviation AES, from the English “Advanced Evasive Steering” or “Automatic Emergency Steering”) have been developed which make it possible to avoid the obstacle by diverting the vehicle from its trajectory, by acting on the direction of the vehicle.

[0008] For this AES function to be effective, it is necessary to reliably detect the part of the environment of the motor vehicle which is relevant for the calculation of an optimal avoidance trajectory.

[0009] A parameter generally used for this purpose is called time-to-collision (TTC). A second parameter is formed by the distance required to pass by the detected obstacle without hitting it. Each potential obstacle is thus considered in isolation in order to determine the most dangerous of them and to deduce an optimal avoidance trajectory.

[0010] However, it turns out that by relying on these parameters, the AES obstacle avoidance function is activated in autonomous mode (without driver action) in dangerous situations in which it would be preferable to let the driver carry out the avoidance. It also happens that it is not activated due to too many obstacles to avoid in the same area, which makes the function less available. Presentation of the invention

[0011] In order to overcome the aforementioned drawbacks of the state of the art, the present invention proposes to process the detected objects not independently, but by grouping them together when possible.

[0012] More particularly, the invention proposes a method for avoiding objects by a motor vehicle, which objects are initially considered to be potential obstacles. This method comprises steps of: - detection of objects located in the environment of the motor vehicle, - acquisition of data characterizing the position and / or dynamics of each detected object, then, if several objects have been detected, - verification whether at least one proximity criterion between at least two of the detected objects is met and, if so, - association of the two objects into a group, - calculation of data characterizing the position and / or dynamics of said group, and - activation of an obstacle avoidance system and / or determination of an avoidance trajectory, according to the data characterizing the position and / or dynamics of said group.

[0013] The target grouping method has several advantages.

[0014] The main advantage is that it makes the AES function activatable in a larger number of situations than generally.

[0015] Indeed, in the presence of too large a number of objects to be processed, it is generally planned to leave the AES function inactive whereas here, the objects being grouped, it is possible to process a large number of distinct objects.

[0016] Similarly, when the vehicle is traveling on a two-lane road and one or more objects are located on each lane, it is generally intended not to allow the activation of the AES function. On the contrary, in the present invention, only the distance between the different objects or groups of objects is considered in order to check whether it is possible to activate the AES function.

[0017] Furthermore, it is generally provided to detect on which traffic lane each object is located before activating the AES function. When an object is located between two traffic lanes, the space to pass next to this object is more restricted than if the object were centered on its lane. Therefore, a large safety margin is generally considered before activating the AES function. On the contrary, in the present invention, each group of targets defines an area to be avoided, the position of which does not depend on that of the traffic lanes. It is thus possible to restrict the safety margins to be considered, which makes it possible to activate the AES function in a greater number of situations.

[0018] More generally, this method makes it possible to disregard the concept of traffic lanes in the calculation of the avoidance trajectory, preferring to manage the environment as a single space in which the objects move.

[0019] It will also be noted that the invention makes it possible to simplify the calculations.

[0020] It then makes it possible to manage a greater number of targets at the same time, which is particularly interesting in the case where a peloton of cyclists is detected.

[0021] It should also be noted that it ignores the class of detected objects (cyclists, automobiles, etc.), preferring to group objects of any class according to only one or more proximity criteria.

[0022] Other advantageous and non-limiting characteristics of the method according to the invention, taken individually or in all technically possible combinations, are the following: - the calculation of the data characterizing the position and / or the dynamics of said group is carried out based on the data individually characterizing the objects of the group; - then, these data individually characterizing the objects of the group are no longer considered: the activation of the obstacle avoidance system and / or the determination of the avoidance trajectory is in fact carried out independently of the data individually characterizing the objects of the group; - the proximity criterion relates to the lateral distance between the two objects (along an axis orthogonal to the tangent to the road at the level of the objects); - at the acquisition stage, one of the data is a lateral deviation of trajectory that the motor vehicle must make to avoid each object; - the proximity criterion consists of verifying whether the difference between, on the one hand, the lateral deviation of trajectory to be made to avoid a first of the two objects by the side oriented towards the second object, and on the other hand, the lateral deviation of trajectory to be made to avoid the second object by the side oriented towards the first object, is greater than or equal to a predetermined threshold; - this threshold is greater than or equal to 0; - if at least three objects have been detected, it is planned to order the objects in a succession order going from one edge of the road to the other, then it is checked whether the proximity criterion is met between each pair of successive objects according to said succession order; - it is planned to calculate a relative lateral speed as a function of the difference between, on the one hand, the lateral speed of the motor vehicle in relation to the road on which the vehicle is traveling in a first reference frame oriented along a tangent to the road at the level of the motor vehicle, and, on the other hand, the lateral speed of the object in relation to the road in a second reference frame oriented along a tangent to the road at the level of said object, then each lateral deviation of trajectory is determined as a function of the lateral speed relative ; - the said proximity criterion relates to the longitudinal distance between the two objects; - at the acquisition stage, one of the data relates to the time remaining before the motor vehicle can hit each object; - to verify that the said proximity criterion is met, we check whether the difference between the times before collision of the two objects is less than a threshold; - this threshold is greater than or equal to 0; - at the acquisition stage, one of the data is a lateral trajectory deviation to be made to avoid each object by the same left or right side, and at the calculation stage, one of the data characterizing the group is chosen equal to the largest of the lateral trajectory deviations to be made to avoid the objects of the group on the same side; - in the acquisition step, one of the data characterizing each object is a time remaining before the motor vehicle can hit each object, and in the calculation step, one of the data characterizing the group is chosen to be equal to the smallest of the times remaining before the motor vehicle can hit one of the objects in the group.

[0023] The invention also relates to a motor vehicle comprising at least one steered wheel, a steering system for each steered wheel adapted to be maneuvered by an actuator controlled by a computer, which computer is adapted to implement a triggering method as mentioned above.

[0024] Of course, the various features, variants and embodiments of the invention may be combined with each other in various combinations to the extent that they are not incompatible or mutually exclusive. Detailed description of the invention

[0025] The description which follows with reference to the appended drawings, given as non-limiting examples, will make it clear what the invention consists of and how it can be implemented.

[0026] In the attached drawings:

[0027] [fig. 1] is a schematic view of a motor vehicle according to the invention and two target cars traveling on two separate traffic lanes;

[0028] [fig.2] is a schematic view of the motor vehicle of [fig.l] and one of the two target cars;

[0029] [fig.3] is a view corresponding to that illustrated in [fig.2], illustrating a second step in the process of determining the position of one of the target cars;

[0030] [fig.4] is a view similar to that illustrated in [fig.2], illustrating four reference points used in the context of the invention;

[0031] [fig.5] is a representation of the four markers of [fig.4];

[0032] [fig.6] is a schematic view of one of the target cars of [fig.l] and its track of traffic;

[0033] [fig.7] is a schematic view of an example configuration in which find the motor vehicle conforming to the invention and four cars driving nearby;

[0034] [fig.8] is a view corresponding to that of [fig.7], in which the references of the cars were reordered;

[0035] [fig.9] is a view similar to that of [fig.8], on which the cars have been grouped for the first time;

[0036] [fig.10] is a view similar to that of [fig.8], on which the cars have been grouped a second time;

[0037] [fig.l 1] is a schematic view of another example of configuration in which there are the motor vehicle according to the invention and a group of three cars traveling nearby;

[0038] [fig.12] is a schematic view of another example of configuration in which find the motor vehicle conforming to the invention and three cars driving nearby.

[0039] In [fig.l], a motor vehicle 10 is shown driving on a road on which there are two “objects” forming potential obstacles for the motor vehicle 10. Here, these two objects are formed by cars C1, C2. Alternatively, they could be other types of objects (pedestrians, cyclists, etc.). The objects considered are preferably mobile.

[0040] In the remainder of the description, the motor vehicle 10 is the one which will implement the present invention, and it will be called “EGO vehicle 10”.

[0041] This EGO 10 vehicle conventionally comprises a chassis which delimits a passenger compartment, wheels of which at least two are steered, a powertrain, a braking system and a conventional steering system making it possible to act on the orientation of the steered wheels.

[0042] In the example considered, the steering system is controlled by a power steering actuator which makes it possible to act on the orientation of the steered wheels as a function of the orientation of the steering wheel and / or, depending on the case, as a function of an instruction issued by a CIO computer.

[0043] The CIO calculator comprises at least one processor, at least one memory and different input and output interfaces.

[0044] Thanks to its input interfaces, the CIO calculator is adapted to receive input signals from different sensors.

[0045] Among these sensors, the following are provided, for example: - a device such as a front camera, allowing the position of the EGO vehicle relative to its lane, - a device such as a RADAR or LIDAR remote sensor, making it possible to detect an obstacle located in the path of the EGO 10 vehicle, - at least one lateral device such as a RADAR or LIDAR remote sensor, allowing observation of the environment on the sides of the EGO vehicle.

[0046] The CIO computer thus receives data from several sensors relating to the objects present in the environment of the EGO 10 vehicle. In a conventional manner, this data is combined together so as to provide reliable merged data on each object.

[0047] Thanks to its output interfaces, the CIO calculator is adapted to transmit an instruction to the power steering actuator.

[0048] It thus makes it possible to ensure that the vehicle follows as best as possible, and if the conditions justify it, an obstacle avoidance trajectory.

[0049] Thanks to its memory, the CIO calculator stores data used in the context of the method described below.

[0050] In particular, it stores a computer application, consisting of computer programs comprising instructions whose execution by the processor allows the computer to implement the method described below.

[0051] These programs include in particular an “AES system” which is designed to calculate an obstacle avoidance trajectory and to pilot the EGO 10 vehicle so that it follows this trajectory or to assist the driver in piloting the EGO 10 vehicle so that it follows this trajectory. The AES system has an autonomous mode in which the trajectory tracking is done without the assistance of the driver, and a manual mode in which the AES system helps the driver to avoid the obstacle and in which the driver remains in control of the maneuver. This AES system being well known to those skilled in the art, it will not be described here in detail.

[0052] The computer programs also include AES system activation software, which will make it possible to determine whether the AES system must be activated (taking into account the trajectory of the EGO vehicle and those of the objects present in its environment) and to wait for the best moment to activate it. It is this activation software which is here more precisely the subject of the present invention.

[0053] This software is active as soon as the EGO 10 vehicle is in motion.

[0054] It is implemented in a loop, at regular time steps.

[0055] It comprises a preliminary step of acquiring data relating to the EGO 10 vehicle and its environment, followed by nine main steps. These successive steps can then be described one by one.

[0056] During the preliminary step, the CIO computer receives at least one image acquired by the front camera of the EGO 10 vehicle. It also receives data from the remote detectors. These images and data are then merged.

[0057] At this stage, the CIO computer therefore has an image of the road located in front of the EGO vehicle 10 and merged data which characterizes in particular each object detected and located in the environment of the EGO vehicle 10. This environment is here considered to be the zone located around the EGO vehicle, in which the vehicle's sensors are adapted to acquire data.

[0058] In the example of [fig.l], the vehicle EGO is traveling on a central traffic lane VC, on either side of which there are two other traffic lanes VR, VL.

[0059] The CIO calculator then seeks to determine the positions and shapes of the separation lines NL, L, R, NR of these traffic lanes Vc, VR, VL.

[0060] For this, here, each of these lines is modeled by an equation of polynomial form. Here, the chosen polynomial is of order 3, so that we can write:

[0061] [Math.l] yLine = dx 3 + cx 2 + bx + a

[0062] In this equation: - the term yLine represents the lateral coordinate of the lane separation line considered, - the term x represents the longitudinal coordinate of this line, and - the terms a, b, c and d are the coefficients of the polynomial, determined according to the shape of the line seen by the front camera of the EGO vehicle (or acquired by the CIO calculator in a navigation system comprising a detailed map of the places in which the EGO vehicle moves).

[0063] In practice, these terms are provided by data fusion. They make it possible to model the shape of the lane separation lines up to a distance of around one hundred meters when visibility conditions are good.

[0064] At this stage, it will be noted that in the remainder of this presentation, a term described as “longitudinal” will correspond to the component of a vector along the abscissa of the reference frame considered, and a term described as “lateral” will correspond to the component of a vector along the ordinate of the reference frame considered (the reference frames considered here always being orthonormal).

[0065] The equation [Math. 1] is expressed here in a reference frame (XEG0, YEG0) attached to the EGO 10 vehicle and represented in [fig.l]. This reference frame is oriented so that its abscissa axis extends along the longitudinal axis of the EGO 10 vehicle. It is centered at the level of the front RADAR of the EGO 10 vehicle.

[0066] Alternatively, other simpler or more complex modeling of the geometries of the lane separation lines could be used.

[0067] Once the coefficients a, b, c, d have been determined for each separation line of track, the CIO calculator can implement the nine steps of the process which will make it possible to perceive to what extent the detected objects are dangerous for the EGO vehicle in order to trigger, if necessary, the AES obstacle avoidance system.

[0068] The first step consists of determining the distance which separates the EGO vehicle from the object considered (one of the cars C1, C2).

[0069] The distance calculated here is not a Euclidean distance. Indeed, we want to take into account the shape of the road to determine a distance that the EGO 10 vehicle and the object should travel before colliding.

[0070] The CIO calculator therefore calculates an arc distance LAB here.

[0071] For this, as detailed for example in document FR3077547, the cal culator can use the following equation:

[0072] [Math.2]

[0073] Where: - Lab is the distance of an arc between two points A and B (corresponding to the positions of the EGO vehicle and the object considered), - xA is the longitudinal position of the EGO vehicle (at the RADAR level), and - xB is the longitudinal position of the object considered, in the frame (XEGo, YEG0).

[0074] The second step consists of determining the position of each detected object relative to the traffic lanes of the road, taking into account the equation of each lane separation line and the merged data.

[0075] The CIO calculator knows the coordinates, in the frame (XEG0, YEG0) attached to the vehicle EGO, of a characteristic point of each detected object (hereinafter called “anchor point”). This characteristic point is typically the center of the object seen by the front camera or by the RADAR remote sensor. Here, it will be considered to be the middle of the grille of the car C1, C2

[0076] In the example of [fig.l] where two objects have been detected (the two cars C1, C2), the coordinates of the anchor points are respectively referenced (X_rell, Y_rell) and (X_rel2, Y_rel2).

[0077] In this [fig. 1], the following values ​​have also been represented on the ordinate axis of the reference (XEG0, Yego): - Y_road_NL_l, which is the value of the term yLine of the equation [Math. 1] of the lane separation line NL, at the abscissa point X_rell, - Y_road_NL_2, which is the value of the term yLine of the equation [Math. 1] of the lane separation line NL, at the abscissa point X_rel2, - Y_road_L_l, which is the value of the term yLine of the equation [Math. 1] of the lane separation line L, at the abscissa point X_rell, - Y_road_L_2, which is the value of the term yLine of the equation [Math. 1] of the lane separation line L, at the abscissa point X_rel2, - Y_road_R_l, which is the value of the term yLine of the equation [Math. 1] of the lane separation line R, at the abscissa point X_rell, - Y_road_R_2, which is the value of the term yLine of the equation [Math. 1] of the lane separation line R, at the abscissa point X_rel2, - Y_road_NR_l, which is the value of the term yLine of the equation [Math. 1] of the lane separation line NR, at the abscissa point X_rell, - Y_road_NR_2, which is the value of the term yLine of equation [Math. 1] of the lane separation line NR, at the abscissa point X_rel2.

[0078] Then, by comparing these values ​​with the lateral coordinates Y_rell, Y_rel2 of the cars C1, C2, it is possible to determine on which traffic lane each of these two cars is located.

[0079] For example, the lateral coordinate Y_rell of the car Cl is here between the values ​​Y_road_R_l and Y_road_L_l, which means that this car is located between the lane separation lines L and R.

[0080] At this stage, the CIO calculator can thus know in which traffic lane VL, Vc, VR each detected object is located.

[0081] The third step aims to determine parameters characterizing the kinematics of each object in relation to the lane separation lines.

[0082] In the remainder of the description of this step, we will only be interested in one of these objects (the car C1), for reasons of clarity of the presentation.

[0083] This step includes a first sub-step during which the CIO calculator determines the position of the object relative to one of the lane separation lines. The lane separation line considered is preferably the one which separates the central traffic lane from the traffic lane on which the object considered is located.

[0084] Alternatively, the lane separation line considered could be another line, for example a lane edge line (see Figures 2 and 3), in particular if no line is detected between the traffic lane of the object and that of the EGO 10 vehicle.

[0085] The idea is to discretize an interval of this lane separation line into a finite number N of points, then to select the one which is closest to the object considered. This operation is carried out several times by rediscreting the lane separation line over an interval which is each time reduced and located on either side of the selected point in order, ultimately, to find a good estimate of the point of the traffic lane which is closest to the object considered.

[0086] In practice, as shown in [fig.2], the calculator begins by discretizing the lane separation line R into N coordinate points (Xi, Yi) in the reference frame of the EGO vehicle 10. These points are regularly distributed along this line (in practice, the interval between two consecutive points along the XEGO axis is always the same), a first of the points being located at the height of the EGO vehicle (with a zero abscissa) or at a first predetermined distance from it, and the last of the points being located at a second predetermined distance from the EGO vehicle.

[0087] Then, knowing the coordinates here noted (Xrei;Yrei) of the anchoring point of the car Cl, the calculator can deduce the Euclidean distance BirdDistance between each discretization point of the lane separation line R and the anchoring point of the car Cl, using the equation:

[0088] [Math.3]

[0089] The discretization point for which the Euclidean distance BirdDistance is the smallest is the one which is closest to the car Cl. This point with coordinates (Xs, Ys) is therefore selected.

[0090] Then, as shown in [fig.3], this discretization operation is repeated over a smaller interval and with a finer discretization. The limits of the interval are preferably formed by the points with coordinates (Xs-1, Ys-1) and (Xs+1, Ys+1). The number of discretization points is preferably still equal to N. This new operation then makes it possible to select a new point with coordinates (Xs, Ys).

[0091] After a certain number of loops (for example 10) or when the interval between two discretization points is sufficiently small (for example less than 10 cm), the computer stops repeating these operations in a loop.

[0092] The last point to be selected is called "projection point F". It is considered to be a good approximation of the point on the separation line that is closest to the car CL

[0093] The value of the abscissa Xs of this point is called DistanceXproj.

[0094] The value of the Euclidean distance BirdDistance between the projection point F and the car Cl is called DistTarget2Lane.

[0095] A second sub-step consists for the calculator CIO in determining the speed of the vehicle EGO in a reference frame attached to the road and located at the height of this vehicle EGO 10, and the speed of the car Cl in a reference frame attached to the road and located at the height of this car CL

[0096] In this sub-step, we will formulate the hypothesis that, from the projection point F, the road follows the tangent at this point. We therefore consider it straight from the car CL

[0097] To better understand the calculations, [fig.4] shows four reference points used in the rest of this presentation.

[0098] The first marker is the marker (XEG0, YEG0) already presented, which is attached to the EGO vehicle.

[0099] It will be noted that this reference frame advances at the same time as the EGO 10 vehicle. We have therefore also represented an absolute reference frame (Xabs, Yabs) which merges, at the time of the measurements, with the first reference frame but which is considered to be fixed.

[0100] Another reference point is noted (XiineEG0, YiineEG0); it is attached to the lane separation line R, is oriented so that its abscissa is tangent to this line, and it is centered at the level of the RADAR of the EGO vehicle (the abscissa of this RADAR is zero in the second reference point).

[0101] Yet another reference point is noted (XObj, YObj); it is attached to the car Cl, is oriented so that its abscissa is aligned with the direction of advancement of the car Cl, and it is centered at the level of the anchor point of this car Cl.

[0102] A final reference point is noted (XLineObj, YLineObj); it is attached to the lane separation line R, is oriented so that its abscissa is tangent to this line, and it is centered at the level of the anchor point of the car Cl.

[0103] In [fig.5], the angles separating these reference points are shown: - AngleiineEGO / EGo allows you to move from the (XEG0, YEG0) reference point to the (XiineEG0, YiineEG0) reference point, - AngleiineObj / EGo allows you to move from the (XEG0, YEG0) mark to the (XLineObj, YLineObj) mark, - AngleObj / EGO allows you to move from the (XEG0, YEG0) mark to the (XObj, YObj) mark, - Angleobj / Lineobj allows you to move from the (XLineObj, YLineObj) mark to the (XObj, YObj) mark.

[0104] We will more generally call Angle, i^x / Eco the angle separating the abscissa of the reference frame (X ego, Yego) and the tangent to the lane separation line R, at the abscise point X (expressed in the reference frame (XEG0, YEGO)).

[0105] We can then write:

[0106] [Math.4] Corner LineXIE x O = arctan(d(ydLine^

[0107] With:

[0108] [Math.5] d ( y Line ( x ) ) _ Y 2, o r Y . h

[0109] So, for x = 0, we can write:

[0110] [Math.6] Ançle,. Vl „ = Angle,. = arctan(b\ ° LmeX / Ego ° LineEgoI Ego \ /

[0111] At the abscissa point x = DistanceXproj, we can write:

[0112] [Math.7] Angle .,.. = Angle - Angle,. ° Obj / LmeObj Q Obj / Ego ​​' LineObjfEgo

[0113] The calculator can calculate the longitudinal components VxEGo / LineEGo and lateral Vy EG0 / LineEG0 of the vehicle speed EGO 10 in the reference frame (XLineEG0, YLineEG0) using the following formulas:

[0114]

[0115]

[0116]

[0117]

[0118]

[0119]

[0120]

[0121]

[0122]

[0123]

[0124] [Math. 8] y » Ego! LineEgo [Math.9] Vy Ego / LmeEeo = V r , , *cos {Angle -Angle,. r !v Ego / abs I y / Eei LineEgofEgo \ Ego = V r , , *sin (Angle -Angle,. .. Ego / abs ° ° LmeEgo!Ego \ y Ego In these formulas: - VEG0 / abs is the speed of the vehicle EGO 10 in the absolute reference frame (Xabs, Yabs), measured for example by the sensors located at the vehicle axles; - AngleVEgo / Ego is the angle of the vehicle velocity vector EGO 10 relative to the abscissa of the reference frame (XEG0, YEG0). This angle is assumed to be zero here. The calculator can also calculate the longitudinal VxObj / abs and lateral VyObj / abs components of the speed Vobj / abs of the car Cl in the absolute frame. It uses the following formulas for this: [Math. 10] Vx^,., , = Vx,^, + Vx r , , Obj / abs Obj / Ego Ego / abs [Math. 11] Vy . = Vy + Vy Obj / abs Obj / Ega Ego / abs In these formulas: - VxEG0 / abs and VyEG0 / abs are the components of the speed of the vehicle EGO 10 along the abscissa and ordinate of the absolute reference frame (Xabs, Yabs), and - VxObj / EGO and VyObj / EGo are the components of the speed of the car Cl relative to the vehicle EGO 10 along the abscissa and the ordinate of the reference frame (XEG0, YEG0). We can then write: [Math. 12] V , , = AVx^ .. . 2 + Vv T obj / abs y Obj / abs 'Obj / abs As the following two equations show, we can, from the angles Angle LineObj / Ego ​​AngleEineEgO / EgO, determine the components VxGbj / bneGbj, Vyobj / iineobj of the relative speed "along the lane separation line R" of the car Cl with respect to this line at the projection point F, which allows a better representation of the information. [Math. 13] Vx„, ■ :, • • = V” . .. , *cos (Angle + Angle .. „ - Angle,. r ' Obj / LtneObj obj / abs | ■ y - / Obj ° Obj / Ego ■ LmeObj J Ego

[0125] [Math. 14] Vy / i *sin / Angle +Angle nl . / r -Corner,. nl .. r \ Obj / LineObj oojlabs ° VOhjfObj Obj / Ego ​​° LineObj / Ego}

[0126] In these two equations, AngleVobj / obj is the angle of the velocity vector of the car Cl in the frame attached to this car, and AngleObj / Ego ​​is the heading angle of the car in the frame (XEG0, YEG0).

[0127] In practice, we assume here that the velocity vector of the object is collinear with its heading angle, so that the angle AngleVobj / obj is zero.

[0128] A similar process is applied to determine the relative accelerations “along the lane separation line R” between the vehicle EGO and this line at the point of zero abscissa and between the car Cl and this line at the projection point F.

[0129] Thus, the calculator can calculate the longitudinal components AxEGO / EineEGo and lateral AyEGO / EineEGo of the acceleration of the vehicle EGO 10 in the reference frame (XLineEG0, Y LineEGo) by means of the following formulas:

[0130] [Math. 15] Ax r r = A r . . *cos (Angle -Angle,. „ \ Ego / LtneEgo Ego / abs I ° y / Eso & LineEgo / Ego I \ Ego /

[0131] [Math. 16] A y = 3U , , *sin (Angle -Angle,. r \ r Ego / LineEgo Ego / abs i * VEutgEgo LineEgo / Ego j

[0132] The calculator can also calculate the longitudinal AxObj / LineObj and lateral Ayobj / Lineobj components of the acceleration of the car Cl in the frame (XLineobj, YLineobj) using the following formulas:

[0133] [Math. 17] Ax..,..,. .... =A. .. *cos (Angle + Angle,„.. r -Corner,. ."., r I ObjïLmeObj obj! abs | ° V^JObj ° Obj / hga ° LineObj! Ego}

[0134] [Math.18] ^Obj / UneObj obj!abs LineObj / Egt^

[0135] In these formulas: - AEG0 / abS is the absolute acceleration of the vehicle EGO 10 in the absolute frame; - Aobj / abS is the absolute acceleration of the car Cl in the absolute frame.

[0136] We can then combine the calculated speeds and the calculated accelerations in order to obtain the longitudinal components VRelRouteLongi, ARelRouteLongi and lateral components VRelRouteLat, ARelRouteLat of the relative speed and the relative acceleration of the vehicle EGO and the car Cl, relative to the road taken, using the four equations defined below.

[0137] In practice, we consider that the longitudinal component VRelRouteLongi of the relative speed between the vehicle EGO and the car Cl is equal to the difference between, from a on the one hand, the longitudinal component of the speed of the vehicle EGO expressed in the reference frame (XLineEG0, YLineEG0) attached to the traffic lane at the level of the vehicle EGO, and, on the other hand, the longitudinal component of the speed of the car Cl expressed in the reference frame (XLineObj, YLineObj) attached to the traffic lane at the level of the car Cl.

[0138] In the same way, it is considered that the lateral component VRelRouteLat of the relative speed between the vehicle EGO and the car Cl is equal to the difference between, on the one hand, the lateral component of the speed of the vehicle EGO expressed in the reference frame (X LineEG0, YLineEG0) attached to the traffic lane at the level of the vehicle EGO, and, on the other hand, the lateral component of the speed of the car Cl expressed in the reference frame (X LineObj, YLineObj) attached to the traffic lane at the level of the car Cl.

[0139] We can therefore write:

[0140] [Math. 19] VRelRoute, . - Va' . - Vx„ ... r Longi Ob j / LineObj Ego / Line Ego

[0141] [Math.20] VRelRouteIat=Vy . . - Vy Obj! LineObj Ego! LineEgo

[0142] The components of acceleration can be calculated in a similar way:

[0143] [Math.21] ARelRoute. .= Ax„,.... n..-Axr ,r. r Longi Obj / LmeObj Ego / LineEgo

[0144] [Math.22] ARelRoute, Ay - Ay Obj 1 LineObj Ego! LineEgo

[0145] As will appear in detail later in this presentation, the use of relative speeds makes it possible to provide indications of collision risks that would be difficult to obtain otherwise.

[0146] At this stage, it can be recalled that the CIO calculator knows the value of the distance DistTarget2Lane between the lane separation line (at the projection point F) and the anchor point of the car Cl.

[0147] During a third sub-step, the CIO calculator will determine the distance LaneDY between the projection point F and the point Pprox of the car Cl closest to the lane separation line R (see [fig.6]).

[0148] Here he calculates this distance using the following equation.

[0149] [Math.23] LaneDY = DistTarget^ - sign ( DistTargeALane ). afcfcos (Angle Obj IUrwOh\

[0150] In this equation, the term Width corresponds to the width of the car Cl.

[0151] All of the calculations then make it possible to determine, during a fourth sub- step, a time before collision TTC with the object considered (the car Cl), that is to say the time needed for the EGO vehicle to hit the Cl car if the latter maintain their speed.

[0152] Indeed, at this stage, the calculator knows from equation [Math. 2] the length LAB of the arc separating the vehicle EGO 10 from the car Cl. It also knows from equation [Math. 19] the longitudinal component VRelRouteLongi of the relative speed between the vehicle EGO 10 and the car Cl related to the shape of the road. Finally, it knows from equation [Math. 21] the longitudinal component ARelRouteLongi of the corresponding acceleration.

[0153] The use of these longitudinal components makes it possible, when the road is curved and the vehicles do not have parallel trajectories, to obtain a good approximation of the time before collision TTC.

[0154] Here, the CIO calculator then determines the desired time before collision TTC using the following equation:

[0155] [Math.24] - VRelRoute r . + AVRelRoute T 2 +2 *ARelRoute r *L AB _ Longi V Longi Longi ~ ARelRoute f Longi

[0156] It should be noted that two conditions of validity of this equation must first be verified. These conditions are as follows.

[0157] [Math.25] VRelRoute 1 , + 2*ARelRoutej > 0 and ARelRoute r . 0 Longi Longi Longi

[0158] On the other hand, if the longitudinal component ARelRouteLongi of the relative acceleration is zero, we can write:

[0159] [Math.26] TTC - ^AB 1 VReiRoute r Longi

[0160] Alternatively, the time before collision TTC could have been calculated differently, for example by assuming constant relative speed and / or relative acceleration.

[0161] In summary, at this stage, the computer has, thanks to the exploitation of the merged data, different parameters characterizing the different objects in its environment and which are all potential obstacles in its trajectory. In particular, for each object, it has: - time before collision TTC (equation [Math. 24]), - the position of the object on the road (determined during step 2), and - information confirming the existence of the object (provided during the data merging step).

[0162] Then, during a fourth step, the CIO calculator will carry out an initial filtering of the different detected objects according to the parameters available to it. so as to retain only those which are relevant for the implementation of the AES function (i.e. those which form potential obstacles).

[0163] The filtering operation thus consists of considering that the relevant objects (hereinafter called “targets”) are those whose existence has been validated during the data fusion, whose position is potentially dangerous (in our example, this amounts to verifying that the objects are located on one of the traffic lanes) and for which the time before collision TTC is less than a predetermined threshold.

[0164] If several targets are detected on the same traffic lane, it is also possible to consider only a limited number of them (for example 4), namely those for which the distances to the EGO 10 vehicle are the shortest.

[0165] The fifth step consists, for the CIO calculator, in identifying a lateral trajectory deviation (or, in English, “overlap”) necessary to avoid each target or each group of targets from the right and from the left while avoiding the other objects present on the road.

[0166] This step is implemented in five sub-steps.

[0167] Prior to the first sub-step, the CIO calculator identifies each target by a reference specific to this target.

[0168] In [fig.7], an example of a situation is shown in which four targets are located in the environment of the EGO 10 vehicle, in front of it.

[0169] As an example, each target is identified by the calculator by a reference which is written here in the form Cn, with n a natural integer here equal to 1, 2, 3 or 4.

[0170] The number n of the targets Cn is given here randomly.

[0171] The first sub-step consists of considering each target Cn independently and calculating the deviation ovLn necessary to avoid this target Cn from the right and the deviation ovRn necessary to avoid this target Cn from the left.

[0172] The deviations ovLl, ovRl to be made to avoid the target Cl have been illustrated in [fig.7].

[0173] The determination of these right ovLn and left ovRn deviations is carried out taking into account the trajectory of the EGO 10 vehicle and the fused data. In fact, the data fusion provides kinematic information of the targets relative to the EGO 10 vehicle, which, together with the trajectory of the EGO vehicle, are used to calculate these deviations. These deviations are therefore calculated dynamically based on the dynamics of the EGO 10 vehicle, the dynamics of the targets and the shape of the track

[0174] The example of [fig.7] corresponds to a case in which the EGO 10 vehicle and the targets move in a straight line.

[0175] It can be seen that if a deviation must be made to avoid the target to the right, the value of the deviation ovRn is greater than 0. Otherwise, it is less than or equal to 0. More precisely, if the EGO 10 vehicle does not have to modify its trajectory to avoid the target while passing as close as possible to the latter, the deviation ovRn is equal to 0. On the other hand, if the EGO 10 vehicle does not have to modify its trajectory to avoid the target but would have to modify it if it wanted to pass as close as possible to the latter, the deviation ovRn is strictly less than 0.

[0176] Similarly, if a deviation must be made to avoid the target from the left, the value of the deviation ovLn is greater than 0. Otherwise, it is less than or equal to 0.

[0177] In the case where the road is not straight, we propose to refine the calculation of these deviations by taking into account the following information: - the lateral component VRelRouteLat of the relative speed between the EGO vehicle and the target considered, along the route taken (equation [Math. 20]), - the Angleobj / Lineobj angle, and - time before collision including VAT.

[0178] The first of these pieces of information makes it possible to take into account the actual lateral speed between the EGO 10 vehicle and the target, taking into account the shape of the traffic lane.

[0179] To understand the interest of this parameter, we can consider an example in which the EGO vehicle and the target are traveling on two separate traffic lanes, in opposite directions, following the curves of these two traffic lanes. We then understand that, in theory, the risks of accidents are reduced. In our example, the lateral component VRelRouteLat of the relative speed of the EGO vehicle with respect to the target will be zero, so that the calculated deviation will be less than or equal to zero, which clearly expresses the idea of ​​a theoretically zero risk of collision.

[0180] In other words, the lateral component VRelRouteLat and the time before collision TTC make it possible to weight the influence of the relative lateral and longitudinal speeds on the calculation of the right and left deviations ovLn, ovRn.

[0181] Similarly, the AngleObj / LineObj information makes it possible to determine a more precise value of the impact surface of the target considered, by weighting the length and width of the vehicle.

[0182] The calculation of these right and left deviations Eright, Eieft is carried out here in a manner similar to that described in document FR1907351, with the difference that these calculations make it possible to take into account the three aforementioned pieces of information.

[0183] Thus, the calculation of the half-width of the target will be carried out according to the AngleObj / LineObj- information. This half-width is then used to calculate a preliminary value for each deviation, which does not take into account a safety radius to avoid the target. The lateral component VRelRouteLat is multiplied by the time before collision TTC and then added to this preliminary deviation in order to obtain the desired deviations.

[0184] In other words, if we consider the document FR1907351, to calculate the deviations, it will be necessary to use in the calculation of the lateral coordinate dVy the product of the lateral component VRelRouteLat with the time before collision TTC. The Y coordinate a used in this document (which corresponds here to the Yrei coordinate) will be considered equal to the sum of the coordinate resulting from the data fusion and a term equal to the product of the length of the target by the cosine of the angle AngleObj / LineObj-

[0185] Thus, we can then write:

[0186] [Math.27] E left - Y a + er + Long.cos / Angle \+VRelRoute Lat .TTC obi -, \ LmeOb) /

[0187] In this equation, er is a term to compensate for lateral measurement errors, and Long is the length of the target.

[0188] [Math.28] Y a + er - Long.cos (Angle \ + VRelRoute Lat .TTC \ °b^ LmeObj /

[0189] In the case where data fusion is used, the term er also takes into account the error arising from the data fusion. This term is predetermined and stored in a memory of the computer.

[0190]

[0191] The second sub-step will consist of sorting the targets, in order to order them according to an order which depends on their positions on the road, and more precisely on their distance from one of the edges of the road.

[0192] Here, this operation is performed according to the calculated left deviations ovLn, in descending order. Alternatively, it would of course be possible to apply another sorting method.

[0193] The advantage of the method used here is that the calculation of the deviations takes into account the relative dynamics of the scene, based on the relative lateral speeds of the EGO 10 vehicle and the targets, but also on the calculated pre-collision times TTC which integrate a notion of prediction of the relative positions of the EGO vehicle and the targets.

[0194] Here, as shown in [fig.8], the targets which were previously referenced Cn are now referenced Cn.

[0195] We observe in Figures 7 and 8 that: - target Cl becomes target C2, - target C2 becomes target C4, - target C3 becomes target C3, - target C4 becomes target Cp

[0196] Similarly, the deviations which were referenced ovLn, ovRn are now referenced ovLn, ovRn.

[0197] Thus: - the gaps ovLl, ovRl become ovL2, ovR2, - the gaps ovL2, ovR2 become ovL4, ovR4, - the gaps ovL3, ovR3 become ovL3, ovR3, - the gaps ovL4, ovR4 become ovLi, ovRi.

[0198] This classification of targets allows them to be ordered from left to right in relation to the EGO 10 vehicle.

[0199] In the following, we will note TTCn the time before collision calculated for the target Cn.

[0200] By successively considering each target according to the order thus determined, the CIO calculator can then determine whether it is possible to pass to the right or left of this target.

[0201] To do this, during a third sub-step, the calculator solves the following equations (for all n ranging from 1 to 4 in our example).

[0202] [Math.29] Gap - - (ovL n + ovR n J - dSafe Left^n

[0203] [Math.30] Gap = - (ovR n + ovL n+i ) -dSafe

[0204] In these equations, the parameter dSafe has a strictly positive value and corresponds to a safety distance that one wishes to maintain around the target to avoid passing too close to it. Its value can vary, for example depending on the speeds of the target and the EGO vehicle or the traffic conditions (weather, etc.). It is at least equal to the width of the EGO vehicle.

[0205] The GapLeftn parameter, hereinafter called gap left, corresponds to the width of passage to the left of the target, taking into account other targets.

[0206] The GapRight parameter n, hereinafter called space to the right, corresponds to the width of passage to the right of the target, taking into account the other targets.

[0207] It should be noted here that the calculation of the space to the left GapLeft_n is not necessarily carried out for n equal to 1, since we know that it is possible to pass to the left of the target C i-

[0208] Similarly, the calculation of the space to the left GapRight n is not necessarily carried out for n equal to 4, since we know that it is possible to pass to the right of the target C4.

[0209] At this stage, it can be considered that it is possible for the vehicle EGO 10 to pass to the left of the target Cn considered if and only if the space to the left GapLeft n is greater than or equal to zero.

[0210] In the same way, it can be considered that it is possible for the vehicle EGO 10 to pass to the right of the target Cn considered if and only if the space to the right GapRight n is greater than or equal to zero.

[0211] In the example of [fig.8], we thus obtain: - GapRight l > or = 0, - GapLeft2 > or = 0, ■ GapRight2 < 0, ■ GapLeft 3 < 0, ■ GapRight3 < 0, ■ GapLeft4 < 0.

[0212] In [fig.9], this situation is represented.

[0213] We observe that the values ​​GapRight2, GapLeft_3, GapRight3, GapLeft4 are all strictly less than zero, so that it is a priori not possible to pass between the targets C2, C3 and C4.

[0214] We also observe that the values ​​GapRight 4 and GapLeft_2 are greater than or equal to zero, which means that it is possible to pass from one side of this group of targets C2, C3 and C4 to the other. It is also possible to pass to the right of the target Ci because the term GapRight l is greater than or equal to 0.

[0215] At this stage, the calculator can therefore group the targets when it is not a priori possible to pass between them.

[0216] For this, we use at least one proximity criterion between the targets.

[0217] A first proximity criterion relates to the lateral distance between the targets.

[0218] Thus, to form groups, the CIO calculator identifies whether the GapRight space n or GapLeft_n+i between a target Cn considered and the neighboring target Cn+i (in the order of succession) is strictly less than 0. If this is the case, the two targets are grouped.

[0219] We thus obtain, in the example of [fig.9], a group of three targets, hereinafter called preliminary group Gl.

[0220] The association of targets into groups of several targets could stop there.

[0221] However, it is here planned to consider a second proximity criterion to form the groups. This second proximity criterion relates to the longitudinal distance between the targets.

[0222] The idea is in fact to form a group only if it is not possible for the EGO 10 vehicle to fall back between some of the targets of the preliminary group GL

[0223] For this, the CIO calculator is here programmed to divide the preliminary group(s) Gl when they comprise targets longitudinally distant from each other.

[0224] This sorting of the targets within each group is carried out here by comparing the times before collision TTCn of each target Cn. Alternatively, this sorting could have been carried out using other parameters (notably the arc distances LAb). The advantage of using this parameter TTCn here is that it takes into account the relative longitudinal speeds and accelerations of the targets Cn.

[0225] The CIO calculator operates in the same way for each group of several targets.

[0226] It orders the targets of this group according to their times before collision TTCn, for example in an increasing manner.

[0227] In the example of [fig.10], the targets are then ordered in the following direction: C4, C2, C3.

[0228] Then the calculator calculates the gap between each pair of successive targets (according to the determined order).

[0229] In the example of [fig.10], it therefore calculates: - the gap A4 2 between the time before collision TTC4 associated with the target C4 and the time before collision TTC2 associated with the target C2, - the gap A2 3 between the time before collision TTC2 associated with the target C2 and the time before collision TTC, associated with the target C3.

[0230] Then the calculator compares each of these deviations A4 2, A2.3 with a threshold Sx.

[0231] This threshold could be invariable. However, preferably, it will be chosen as a function at least of the longitudinal speed of the EGO 10 vehicle.

[0232] If the difference between the pre-collision times associated with two successive targets is greater than this threshold, the calculator divides the preliminary group into two groups.

[0233] In the example of [fig.10], at the end of this operation, the targets C4 and C2 then form a first group G2, while the target C3 is isolated.

[0234] At this stage, the calculator will therefore consider each group in the same way as an isolated target. It will therefore associate with it a left deviation ovL;, a right deviation ovR;, and a time before collision TTC; (i being the index of the group considered).

[0235] These parameters are then calculated in the following manner.

[0236] To facilitate understanding, we can rely on the example of [fig. 11] where the group G3 considered comprises three targets Cl, C2, C3.

[0237] The time before collision TTC; of the group will be chosen equal to the smallest time before collision TTCn among the times before collision TTCn of the targets Cb C2, C3 of the group.

[0238] The left deviation ovL; of the group will be chosen equal to the largest left deviation ovLn among the left deviations ovLn of the targets of the group.

[0239] The right deviation ovR; of the group will be chosen equal to the largest right deviation ovRn among the right deviations ovRn of the targets of the group.

[0240]

[0241] In the remainder of this presentation, for simplification, the generic term “target” will be used to designate both a group of several targets and a single target not forming part of any group.

[0242] The sixth step consists, for the calculator, in carrying out a second filtering among the targets, to distinguish those which are critical from the others.

[0243] A target will be classified as critical if it requires the activation of the AES function to be avoided. The most critical target (MCT target) is the one requiring the activation of the AES function as soon as possible.

[0244] A target will be qualified as “moderate risk” if its position is such that it must be taken into account when determining the avoidance trajectory to follow. Thus, a moderate risk target is likely to prevent the activation of the AES function.

[0245] The idea is to successively consider each detected target independently (and therefore independently of its environment).

[0246] To begin with, the CIO calculator considers all targets located on the traffic lane of the EGO 10 vehicle to be critical.

[0247] With regard to targets located on the lanes adjacent to that taken by the EGO 10 vehicle, the computer checks whether they meet additional criteria.

[0248] Here, these criteria are linked to the following parameters: - the arc distance LAB, - the distance LaneDY between the projection point F and the point of the target which is closest to the lane separation line (equation [Math. 23]), and - the lateral component VyObj / iineObj of the target's speed relative to the lane separation line at the projection point F (equation [Math. 14]).

[0249] In [fig.12], three targets C4, C5, C6 are shown, located on two lanes adjacent to that taken by the vehicle EGO 10.

[0250] To determine whether each target is critical or not, the calculator checks whether the following two conditions i) and ii) are met.

[0251] To verify the first condition i), the calculator begins by calculating the distance LaneDY between the target and the lane separation line considered (equation [Math. 23]). This makes it possible to know whether the target is rather close or on the contrary far from the lane separation line considered.

[0252] The CIO calculator deduces a minimum lateral speed threshold, noted VyseuiiMm-

[0253] Then, if the VyObj / iineObj component exceeds the minimum threshold VyseuiiMm, then the first condition (to consider the target as critical) is validated. Otherwise, the target is simply considered as being at moderate risk.

[0254] It should be noted that the threshold used is therefore a variable which is a function of the distance between the target and the lane separation line considered, which makes it possible to take into account the fact that the shorter this distance, the greater the risk of collision.

[0255] The threshold used may also depend on the way in which the target was characterized at the previous time step (critical or non-critical). The idea is that the characterization of the target does not vary at each time step, due to noise in the measurements of the data to be merged. For this, the threshold used for the target to go from non-critical to critical is higher than that used for the target to go from critical to non-critical (like a hysteresis function).

[0256] The second condition ii) makes it possible not to take into account targets resulting from possible perception errors, as well as those having excessively high or aberrant lateral speeds. To do this, the calculator compares the absolute value of the lateral component VyObj / iineObj with a predetermined maximum threshold VyseuiiMax. If this component is greater than the maximum threshold, the second condition is not validated and the target is considered to be at moderate risk. The maximum threshold must be restrictive so as not to consider false detections which sometimes have excessive values.

[0257] This maximum threshold is preferably greater than 2m / s, preferably equal to 3m / s.

[0258]

[0259] Targets that do not meet one and / or the other of conditions i) and ii) are considered moderate risk targets. Other targets are considered critical.

[0260] Classifying perceived targets into critical targets and moderate risk targets reduces calculation times. In addition, this classification of targets simplifies the decision whether or not to activate the AES system, and the decision taken is justifiable to a human driver.

[0261] In [fig. 12], it can be considered that only target C4 fulfills both conditions i) and ii), since none of the other two targets fulfills condition i).

[0262] The seventh step consists, for the CIO calculator, in determining, for each critical target, a critical time Tcrit which integrates two distinct and relevant pieces of information to guarantee obstacle avoidance while minimizing the intrusive nature of the AES function for the driver.

[0263] The interest in calculating this parameter is in particular to make it possible to find which is the most critical target MCT among the critical targets.

[0264] Before describing how this parameter is obtained, it may be recalled that the AES avoidance system can be operated fully automatically (in which case the avoidance trajectory is followed by the power steering actuator, autonomously) or semi-automatically (in which case the avoidance is carried out manually by the driver, the power steering actuator being controlled to help the driver follow an avoidance trajectory once the driver has triggered the avoidance). In the remainder of the description, we will speak of an autonomous mode and a manual mode to designate these two methods respectively.

[0265] Avoidance in manual mode is potentially less effective than avoidance in autonomous mode. Thus, the avoidance trajectory calculated by the AES system will not be the same in manual mode and in autonomous mode. It should be noted that the calculation of this The avoidance trajectory (in the form of a clothoid) is not the subject of the present invention. It can simply be recalled that the shape of this trajectory is calculated according to the dynamic performance of the vehicle and the capabilities of the driver (in manual mode).

[0266] In any case, taking into account the calculated deviations ovln, ovRn, it is possible to construct four avoidance trajectories (four clothoids) to avoid the critical target from the right and from the left, in manual mode and in autonomous mode.

[0267] These clothoids have curvatures which depend on the maximum dynamic capabilities of the vehicle. The shape of these clothoids is then read from a database, depending on the speed of the EGO 10 vehicle.

[0268] Taking into account these deviations and the determined avoidance trajectories, it is possible to deduce four TTS maneuver times (see [fig.7]) which correspond to the time required, in each mode, to carry out the avoidance maneuver from the left or from the right.

[0269] The manner of obtaining these TTS maneuvering times will not be described here since it will depend on the dynamic characteristics of the EGO 10 motor vehicle and the driver's performance. In practice, these TTS maneuvering times could be read from a database established using a battery of tests.

[0270] Knowing the time before collision TTCn and the maneuvering time TTS for each critical target, the calculator can deduce the critical time Tcrit using the formula:

[0271] [Math.31] Tcrit = TTC-TTS

[0272] This critical time therefore becomes zero at the last moment when it is still possible to activate the AES function and avoid collision with the critical target considered, in the manual or automatic mode considered, by avoiding it from the side considered.

[0273] Thus, this critical time Tcrit carries three vital pieces of information about the target considered: the side from which to avoid it, the performance of the system (and the driver) and the time before collision TTCn with the target.

[0274] The eighth step will then consist of using the calculated critical times Tcrit to classify the critical targets and find the most critical target MCT.

[0275] In the case where the driver is in control of the maneuver in manual mode (for example because the autonomous mode could not be activated), the computer selects the critical target for which the critical time is the lowest in the case of avoidance from the right and that for which the critical time is the lowest in the case of avoidance from the left.

[0276] Then, as soon as the CIO computer detects that the driver is initiating an avoidance maneuver, it can decide to activate the AES system to assist the driver in his maneuver according to the two selected critical times.

[0277] To do this, the computer determines the side towards which the driver turns the steering wheel when he begins his avoidance maneuver (to the right or to the left), then it elects, from among the two critical times selected, the one corresponding to the side from which the driver begins his avoidance maneuver.

[0278] The most critical target MCT is the one corresponding to this elected critical time.

[0279] In autonomous mode, the computer proceeds differently. It is in fact necessary to determine the side by which the EGO 10 vehicle must avoid the obstacle(s).

[0280] The idea is then to determine, for each critical target, the best side to avoid it and then to determine the most critical target MCT.

[0281] In practice, the calculator selects, for each critical target, the side for which the critical time Tcrit is the highest.

[0282] Then he elects, among the critical times thus selected, the one which is the weakest.

[0283] Then, the calculator considers that the most critical target MCT is the one whose critical time Tcrit was elected.

[0284] The ninth step consists of the CIO calculator triggering the AES function if necessary and at the optimal time.

[0285] In this case, in autonomous mode, the AES function is triggered as soon as the critical time Tcrit elected falls below a predetermined threshold, for example equal to 0 seconds.

[0286] On the other hand, in manual mode, the triggering of this AES function operates differently.

[0287] The idea is to trigger the AES function and the driver's support in avoidance in a way that adapts to the situation.

[0288] Indeed, if the driver initiates his steering action too early, his maneuver will be considered non-urgent because the situation does not require this type of action from the driver at that time. The AES system will then not be activated.

[0289] Similarly, if a collision is imminent and the autonomous mode has not been able to activate beforehand, then it may be considered too late for the AES system to provide any assistance to the driver. It will then be left to another safety system to minimize the impact.

[0290] Thus, in manual mode, it will be necessary to determine a time interval during which, if an avoidance action carried out by the driver is detected, the AES system will be activated.

[0291] This time interval will be delimited by two limits.

[0292] The first terminal, from which, if the driver initiates an avoidance maneuver, he will be helped, will correspond to a critical time Tcrit strictly greater than 0.

[0293] The second limit, from which it will be considered too late to decide trigger the AES system, corresponds to a critical time Tcrit less than or equal to 0, and preferably strictly less than 0.

[0294] Thus, to activate the AES function, the computer determines whether the critical time Tcrit elected is between these two limits, and it activates the AES function only if this is the case.

[0295] The present invention is in no way limited to the embodiments described and shown, but those skilled in the art will be able to provide any variant in accordance with the invention.

[0296] We could have considered a single proximity criterion which is neither lateral nor longitudinal but which only considers the distance between the targets.

[0297] As a further alternative, one could have started by grouping the vehicles according to their longitudinal distances, then dividing each group thus obtained, if necessary, according to the lateral distances between the vehicles in the group.

Claims

Claims

1. Method for avoiding objects (Cl, C2, C3, C4) by a motor vehicle (10), comprising steps of: - detecting objects (Cl, C2, C3, C4) located in the environment of the motor vehicle (10), - acquiring data characterizing the position and / or dynamics of each detected object (Cl, C2, C3, C4), characterized in that, if several objects (Cl, C2, C3, C4) have been detected, steps are provided of: - verifying whether at least one proximity criterion between at least two of the detected objects (Cl, C2, C3, C4) is fulfilled and, if so, - associating the two objects (Cl, C2, C3, C4) into a group (Gl, G2, G3), - calculating the data characterizing the position and / or dynamics of said group (Gl, G2, G3), and - activating an avoidance system obstacle avoidance (AES) and / or determination of an avoidance trajectory (Tl), according to the data characterizing the position and / or the dynamics of said group (Gl, G2, G3).

2. Avoidance method according to the preceding claim, in which said proximity criterion relates to the lateral distance between the two objects (Cl, C2, C3, C4).

3. Avoidance method according to the preceding claim, in which, in the acquisition step, one of the data is a lateral deviation of trajectory (Eieft, Enght) that the motor vehicle (10) must achieve to avoid each object (Cl, C2, C3, C4), and in which the proximity criterion consists of verifying whether the difference (GapLeft_n, GapRight n) between, on the one hand, the lateral deviation of trajectory (Eieft) to be achieved to avoid a first of the two objects by the side oriented towards the second object, and on the other hand, the lateral deviation of trajectory (Eright) to be achieved to avoid the second object by the side oriented towards the first object, is greater than or equal to a predetermined threshold (dsafe).

4. Avoidance method according to the preceding claim, in which, if at least three objects (Cl, C2, C3, C4) have been detected, it is provided to order the objects (Cl, C2, C3, C4) in an order of succession going from one edge of the road to the other, and in which it is checked whether the proximity criterion is fulfilled between each pair of successive objects according to said order of succession.

5. Avoidance method according to one of the two preceding claims, in which provision is made to calculate a relative lateral speed (VRelRouteLat) as a function of the difference between, on the one hand, the lateral speed (VyEgo / LineEgo) of the motor vehicle (10) relative to the road on which it is traveling in a first reference frame (XLineEgo, YLineEgo) oriented along a tangent to the road at the level of the motor vehicle (10), and, on the other hand, the lateral speed (VyObj / LineObj) of the object (Cl, C2) relative to the road in a second reference frame (XLineObj, YLineObj) oriented along a tangent to the road at the level of said object (Cl, C2), and in which each lateral trajectory deviation (Eieft, Eright) is determined as a function of the relative lateral speed (VRelRouteLat).

6. Avoidance method according to one of the preceding claims, in which said proximity criterion relates to the longitudinal distance between the two objects.

7. Avoidance method according to the preceding claim, in which, in the acquisition step, one of the data relates to the remaining time (TTC) before the motor vehicle (10) can hit each object (Cl, C2, C3, C4), and in which, to verify that said proximity criterion is met, it is checked whether the difference (A2_3, A4 2) between the times before collision (TTC) of the two objects is less than a threshold.

8. Avoidance method according to one of the preceding claims, in which, in the acquisition step, one of the data is a lateral trajectory deviation (Eieft, Eright) to be achieved to avoid each object (Cl, C2, C3, C4) by the same left or right side, and in which, in the calculation step, one of the data characterizing the group is chosen equal to the largest of the lateral trajectory deviations (Eieft, Eright) to be achieved to avoid each object (Cl, C2, C3, C4) of the group (G1, G2, G3) on the same side.

9. Avoidance method according to one of the preceding claims, in which, in the acquisition step, one of the data characterizing each object is a remaining time (TTC) before the motor vehicle (101) can hit each object (Cl, C2, C3, C4), and in which, in the calculation step, one of the data characterizing the group is chosen equal to the smallest of the remaining times (TTC) before the motor vehicle (101) can hit each object (Cl, C2, C3, C4) of the group (G1, G2, G3).

10. Motor vehicle (10) comprising at least one steered wheel, a steering system for each steered wheel adapted to be maneuvered by an actuator controlled by a computer (CIO), characterized in that the computer (CIO) is adapted to implement an avoidance method in accordance with one of the preceding claims.

Images