METHOD FOR CONTROLLING AN AUTONOMOUS VEHICLE

Description

[0001] The invention relates to a method for controlling an autonomous vehicle. The invention further relates to a control device for an autonomous vehicle. The invention also relates to a computer program implementing the aforementioned method. Finally, the invention relates to a recording medium on which such a program is recorded.

[0002] Autonomous vehicle technologies essentially use high-definition maps, in association with speed profiles that satisfy the lateral and longitudinal constraint parameters of the autonomous vehicle, to predefine trajectories corresponding to different autonomous driving scenarios in a given traffic area.

[0003] For example, when an autonomous vehicle approaches an X-shaped intersection, the potential trajectories for approaching the intersection (turn right, turn left, go straight) are already pre-mapped, as well as the associated speed profiles.

[0004] However, these technologies have the disadvantage of being inherently limited to traffic scenarios predefined from maps.

[0005] This is also the case in patent application DE102018113007A1, in which the trajectory of an autonomous vehicle is adjusted according to the environment. The same applies to the paper by David Gonzalez Bautista, "Functional architecture for automated vehicles trajectory planning in complex environments," dated April 3, 2017, which describes an obstacle avoidance algorithm, and the article by Lattarulo Ray et al., "Real-Time trajectory planning method based on n-order curve optimization," October 8, 2020, 24th International Conference on System Theory, Control and Computing, IEE, which deals with the optimization of the kinematic profile of an autonomous vehicle.

[0006] Document US9511767B1 describes a method for controlling an autonomous vehicle that uses data from a perception system to adapt the vehicle's trajectory based on the behavior of surrounding vehicles. However, this solution has drawbacks.

[0007] The aim of the invention is to provide a device and method for controlling an autonomous vehicle that overcomes the above-mentioned drawbacks and improves upon prior art devices and methods for controlling an autonomous vehicle. In particular, the invention aims to provide a device and method that are simple and reliable and that improve autonomous driving.

[0008] To this end the invention relates to a method of controlling an autonomous vehicle according to independent claim 1.

[0009] The determination of a local map can allow the determination, in particular by Bézier curves, of a new curve of the given segment between an entry point and an exit point of the given segment and the new curve can show a continuity G2 with the first trajectory at the entry point and at the exit point.

[0010] The first map optimization step may further include the optimization of the first kinematic profile, the optimization of the first kinematic profile including the determination of a second kinematic profile associated with the given segment, at least one parameter of the second kinematic profile being different from a parameter of the first kinematic profile, the step of controlling the movement of the autonomous vehicle on the given segment further taking into account the second kinematic profile.

[0011] The determination of a second kinematic profile associated with the given segment may include a calculation of a maximum no-collision speed threshold from data from at least one perception means, and the second kinematic profile may include maximum speeds lower than the maximum no-collision speed threshold.

[0012] The maximum no-collision speed threshold can be calculated by solving an equation whereby the maximum no-collision speed threshold is equal to the square root of the product of the braking value required to reach the maximum no-collision speed threshold by twice the maximum detection distance of at least one means of perception.

[0013] The step of detecting a situation to be optimized may include a measurement of a dwell time, during which the autonomous vehicle remains in a situation to be optimized, and a comparison of the dwell time to a trigger threshold associated with said situation, and the optimization step may run when the dwell time is greater than the trigger threshold.

[0014] A set of predefined situations can be stored in an electronic memory, and the step of detecting a situation to be optimized can include a comparison of a perception from at least one means of perception with the set of predefined situations.

[0015] The set of predefined situations may include crossing a roundabout with both lanes empty, and / or entering a roundabout in heavy traffic conditions, and / or merging two traffic lanes into one lane, and / or passing through a congested intersection, and / or overtaking a very slow vehicle on a two-way road with only one lane in each direction.

[0016] The second kinematic profile may include a maximum speed that is at least 20% higher, and / or a maximum acceleration that is at least 200% higher, and / or a jerk that is at least 1000% higher than the first kinematic profile.

[0017] The invention also relates to a control device for an autonomous vehicle according to independent claim 10.

[0018] The invention further relates to an autonomous vehicle according to claim 11.

[0019] The invention also relates to a computer program product according to independent claim 12.

[0020] The invention further relates to a data recording medium according to independent claim 13.

[0021] The attached drawing represents, by way of example, an embodiment of a control device according to the invention and an execution method of a control process according to the invention. [ Fig. 1 ] There figure 1 represents one embodiment of an autonomous vehicle equipped with a control device. Fig. 2 ] There figure 2 illustrates a method for determining a kinematic profile of the autonomous vehicle according to the embodiment of the invention. Fig. 3 ] There figure 3 represents a flowchart of an execution method of a control process according to an embodiment of the invention. Fig. 4 ]. There figure 4 illustrates a first example of implementing the control method according to the embodiment of the invention. Fig. 5 ]. There figure 5 illustrates a second example of implementation of the control method according to the embodiment of the invention.

[0022] The autonomous vehicle 100 according to the embodiment of the invention can be an autonomous vehicle of any type, including a passenger vehicle, a utility vehicle, a truck or even a public transport vehicle such as a bus or a shuttle.

[0023] The autonomous vehicle 100 includes a control system 10 and motion actuators 4.

[0024] The motion actuators 4 are part of the chassis of the autonomous vehicle 100. They include, in particular, a motor torque actuator, a brake actuator, and a steering wheel rotation actuator. The motion actuators 4 receive commands from the control system 10 to implement movement of the autonomous vehicle along a trajectory determined by the control system 10.

[0025] The control system 10 mainly comprises the following elements: a perception system 2, comprising perception means 21 and an electronic horizon processing module 22, an electronic memory 3, and a computing unit 5 comprising a microprocessor 1, a local electronic memory 6 and communication interfaces 7 enabling the microprocessor 1 to communicate with the perception system 2, and the motion actuators 4.

[0026] Perception devices 21 may include all or some of the following equipment: a set of cameras and / or lidars and / or radars. Alternatively or in addition, perception devices 21 could include vehicle-to-vehicle (V2V) communication systems or vehicle-to-everything (V2X) communication systems enabling vehicles to exchange information with each other, with infrastructure, and with pedestrians.

[0027] Advantageously, the perception means 21 allow the environment located 360° around the autonomous vehicle 100 to be perceived.

[0028] In a preferred embodiment, the perception means 21 comprise five cameras and a lidar. The images from the cameras are captured periodically and synchronously and are aggregated with the data from a lidar revolution.

[0029] From the data from the perception means 21, the electronic horizon processing module 22 constructs a structured perception of the driving scene.

[0030] Periodically, particularly upon each data reception, module 22 determines the maximum capabilities of the perception system 2, i.e., a maximum detection distance DLIM for the perception means 21, which varies according to the driving scenario. For example, when the autonomous vehicle 100 is traveling on a highway, the DLIM distance is greater than when it is traveling on a roundabout.

[0031] The DLIM distance also depends on other parameters, such as weather conditions, brightness, the presence of road infrastructure elements (e.g. a tunnel), or surrounding traffic (e.g. a truck).

[0032] In the remainder of this document, the term "roundabout" refers to a traffic circle where vehicles already circulating on the intersection have priority over vehicles entering the intersection. Furthermore, the maximum detection distance of the perception devices is referred to as the "limited perception distance (LPD)."

[0033] The limiting distance of perception (DLIM) can be determined from the perception limit of each perception method. In one embodiment, the limiting distance of perception (DLIM) is the minimum between the perception limit for images from cameras and the perception limit for data from lidar.

[0034] The image perception limit of cameras can be defined as the maximum distance below which the height of any object of interest (vehicle, pedestrian) exceeds a minimum threshold, for example, 25 pixels. The order of magnitude of this initial limit is several tens of meters, for example, around forty meters under normal visibility conditions, particularly in terms of weather.

[0035] The perception limit of data from lidar can be defined as the maximum distance below which the number of laser points reaching an object of interest (vehicle, pedestrian) is greater than a minimum threshold.

[0036] Within a perception zone, delimited by a circle centered on the autonomous vehicle and with a radius equal to the perception limit distance (DLIM), the electronic horizon processing module 22 constructs an occupancy grid of the driving scene. The point clouds from the lidar are discretized into cells. Each cell of the grid is determined to be free, occupied, or undefined. This determination relies in particular on the processing of the cell's point height, such as applying minimum and maximum point height thresholds within the cell.

[0037] The occupancy grid delimits a collision-free zone around the position of the autonomous vehicle 100, as well as a three-dimensional representation of obstacles located in the perception zone.

[0038] The data from the electronic horizon processing module 22 provides a structured perception of the driving scenario over the perception area, including a perception of the traffic surrounding the autonomous vehicle 100, in terms of traffic density and position of subjects of interest (vehicles, pedestrians).

[0039] The control system 10 further includes an electronic memory 3 in which is recorded a set of digital data representing the actual infrastructure of the environment of the autonomous vehicle 100. This digital data relates, in general, to all the elements that can infer a choice in the driving of the autonomous vehicle 100. It relates in particular to the layout and limits of the sections of navigable roads, the number of lanes of the sections of road, the speed limits applicable on the sections of road, the type of road markings and their location, the presence and position of traffic signs, traffic lights and roundabouts.

[0040] These digital data will simply be referred to as "M1 map" thereafter.

[0041] In the embodiment of the invention, the calculator 1 allows the execution of software comprising the following modules: a trajectory determination module 11, which collaborates with perception system 2 and map 3, a situation detection module 12, which collaborates with perception system 2, an optimization module 13, which collaborates with perception system 2, a autonomous vehicle movement control module 14, which collaborates with motion actuators 4.

[0042] One method for implementing the process of controlling an autonomous vehicle is described below with reference to the figure 3 The process comprises four steps, E1 to E4.

[0043] In the first step E1, a planned trajectory T1 is determined between a starting point and an arrival point of the autonomous vehicle 100. For this, we use in particular the map M1 recorded in the electronic memory 3 and the data from the perception system 2.

[0044] In the remainder of this document, the term "trajectory" is used to refer to the temporal evolution of a state vector defining the characteristics of the autonomous vehicle's movement. In a preferred embodiment, the state vector includes a position, specifically x, y coordinates, longitudinal and lateral velocities, and / or longitudinal and lateral accelerations, and / or a yaw rate, and / or a jerk. In the remainder of this document, the term "position" is used to refer to either the x, y coordinates of the state vector or the state vector as a whole.

[0045] The planned trajectory T1 is preferentially divided into trajectory segments, also referred to as "segments" in the rest of the document.

[0046] Advantageously, each segment is defined so that it can lie entirely within the perception zone of the autonomous vehicle 100 when it enters the segment. Specifically, the length of a segment is less than the perception limit distance (DLIM).

[0047] The term "kinematic profile" is used throughout this document to refer to a generic profile of velocity, acceleration, and jerk implemented by the vehicle. The kinematic profile serves as a template for calculating the velocities, accelerations, and jerks involved in trajectory calculations. The template defined by the kinematic profile imposes limits in terms of velocity, acceleration, and jerk.

[0048] The kinematic profile takes into account the constraints imposed by the road infrastructure: road layout, number of lanes, width of each lane, presence of intersections and / or traffic lights, speed limits, etc. It is also determined based on the technical characteristics of the autonomous vehicle. Furthermore, it is defined according to driving comfort criteria, particularly regarding a maximum speed and / or lateral acceleration and / or jerk threshold.

[0049] In one embodiment, the kinematic profile can be defined according to a jerk limitation method (the jerk being the derivative of the acceleration) illustrated by the figure 2 .

[0050] There figure 2 is a graph comprising three curves, J(t), a(t) and v(t), representing respectively the time evolution of the jerk, acceleration and speed of the autonomous vehicle 100.

[0051] The J(t) curve is a step curve: time is divided into fixed intervals, and the jerk value is constant over each interval. The duration of the intervals depends on the vehicle's acceleration limit and the available distance for accelerating or braking.

[0052] In the illustrated example, the jerk can take three discrete values: the zero value, a maximum value Jmax and a minimum value -Jmax.

[0053] The value of Jmax is low in order to promote the comfort of the users of the autonomous vehicle 100. In one embodiment, the value of Jmax is fixed at 1m / s -3< .

[0054] The curve J(t) thus determines the time evolution of the acceleration a(t): the curve a(t) is made up of linear segments whose slope is determined by the value of Jmax and the duration of the time intervals, it evolves between the acceleration values ​​Amax and -Amax.

[0055] The curve a(t) determines the time evolution of the velocity v(t) between a zero velocity and a maximum velocity Vmax, Vmax being able to be determined by the velocity limitation relative to the trajectory segment.

[0056] This method is repeated on each trajectory segment to define a trajectory T1 that conforms to the road infrastructure and is as comfortable as possible without slowing traffic. Thus, for each trajectory segment, the applied speeds, accelerations, and jerks are determined based on a kinematic profile and the M1 map.

[0057] In the rest of the document, we consider that the trajectory T1 is decomposed into N segments S 1 , ..., SN , a first kinematic profile P1 i being associated with each segment S i , each of the first kinematic profiles P1 i being defined as a function of the first map M1 and each of the segments S i being defined as a function of the first map M1 and the first profile associated with it P1 i .

[0058] The term "trajectory segment" therefore refers both to the trajectory curve and to the speed, acceleration and jerk values ​​applied to move the vehicle along the segment curve.

[0059] Thus, given two distinct points C and D, two segments connecting point C to point D are considered different if their respective curves are different, or if the values ​​of velocity, acceleration, or jerk applied to move the vehicle along the curve of each segment are different. The trajectory T1, thus decomposed into trajectory segments, is stored in local memory 6 of the processing unit 5.

[0060] We then move on to step E2. In step E2 we seek to detect a situation to be optimized on a segment S i of the trajectory T1 located in front of the autonomous vehicle 100, the segment S i being entirely within the perception limit DLIM of the perception means 21.

[0061] In other words, step E2 defines a segment Si of the trajectory T1 on which a situation to be optimized is located. Segment Si is situated in front of the vehicle, meaning that segment Si constitutes a portion of the trajectory T1 that the autonomous vehicle has not yet traveled. In the embodiment presented, it is assumed that the situation to be optimized is confined to a single segment. In alternative embodiments, a situation to be optimized could be considered to extend over several consecutive trajectory segments. However, all segments must lie within the perception limit DLIM.

[0062] A situation requiring optimization is a traffic scenario in which the autonomous vehicle's movement is hampered by the initial trajectory T1, particularly by limitations in the first kinematic profile P1 associated with trajectory T1, such as V1max and / or A1max and / or J1max. For example, in heavy traffic, the autonomous vehicle 100 could remain stuck for a very long time at the entrance to a roundabout if the kinematic profile associated with its trajectory does not allow it to accelerate sufficiently to merge into traffic.

[0063] In other words, the P1 kinematic profile implemented by the T1 trajectory was defined according to driving comfort criteria that limit the responsiveness of the autonomous vehicle. The P1 profile allows the autonomous vehicle to handle the vast majority of situations under optimal safety and comfort conditions. However, the P1 profile may prove unsuitable in certain situations. This unsuitability can result in the vehicle slowing down considerably—for example, when the vehicle cannot overtake a truck in front of it—or even keeping it stopped for extended periods, for example, when the vehicle is unable to enter a roundabout due to traffic density.

[0064] Other situations requiring optimization may concern not the first kinematic profile P1, but the first map M1, representing the actual infrastructure of the autonomous vehicle's driving environment. In some cases, the first map M1 induces a particular vehicle trajectory that does not correspond to the usual behavior of a human driver and does not optimize driving. This is the case, for example, on a trajectory segment crossing a two-lane roundabout where no other vehicles are present. The first trajectory T1 determines movement on the outer lane of the roundabout, which requires a significant reduction in the autonomous vehicle's speed before entering the roundabout. The trajectory T1 was thus defined with reference to the map M1 representing a roundabout with two lanes.When a human driver approaches a multi-lane roundabout with no other vehicles, they adapt their vehicle's trajectory to this situation: they anticipate that using both lanes will reduce the curvature of their path, and therefore they enter the roundabout at a significantly higher speed than they would have if they were traveling only in the outer lane. This adaptation by the human driver amounts to virtually modifying the road infrastructure, specifically by considering the roundabout as having only one lane.

[0065] In one execution mode of the process, the data received from the perception system 2, in particular the data from the electronic horizon processing module 22, are compared to a set of predefined situations stored in the local electronic memory 6.

[0066] In the rest of the document, The data from module 22 of electronic horizon processing are called "perception data", the set of predefined situations stored in local electronic memory 6 is called "situation table", the data contained in the situation table and relating to a predefined situation are called "situation parameters".

[0067] The predefined situations correspond to situations that need optimization. They may include: an entry onto a roundabout where both lanes are empty, referred to hereafter as "empty roundabout situation" and / or a prolonged stop at the entrance to a roundabout in heavy traffic conditions, referred to hereafter as "congested roundabout situation" and / or a merging of two traffic lanes into one lane in congested traffic conditions, and / or a prolonged stop at a congested intersection, and / or prolonged holding behind a very slow vehicle on a two-way road with only one lane in each direction.

[0068] Each of the predefined situations can be associated with one or more situation parameters, including road infrastructure elements and / or traffic density and / or a trigger threshold.

[0069] Thus, comparing perception data with the set of predefined situations may include comparing the road infrastructure elements used by trajectory T1 with the road infrastructure elements associated with each of the predefined situations. Road infrastructure elements may include, for example, a multi-lane roundabout, an X-junction, a junction between two lanes, or a two-way road with one lane in each direction.

[0070] Furthermore, the comparison between perception data and predefined situations can also include comparing a traffic density derived from perception data to a density threshold associated with each of the predefined situations. The density threshold associated with a predefined situation can be a minimum density threshold, for example, for the situation of a congested roundabout. The density threshold can also be a maximum threshold, for example, for the situation of an empty roundabout.

[0071] Furthermore, the situation parameters may include a trigger threshold corresponding to a minimum duration for remaining in that situation. This minimum duration varies depending on the predefined situation. For example, it may be zero for an empty roundabout. It may be on the order of one or more minutes for a congested roundabout.

[0072] Other situational parameters not described in this document may be considered.

[0073] A situation to be optimized is detected when all the situation parameters associated with that situation are verified, in particular by comparison with perception data.

[0074] If no situation to optimize is detected, we proceed to step E4 of commanding the movement of the autonomous vehicle 100.

[0075] When a situation requiring optimization is detected, we proceed to step E3 of trajectory optimization.

[0076] The trajectory optimization is carried out through a modification of the first map M1 and / or the first kinematic profile P1, in particular the kinematic profile P1 i associated with the segment S i on which the situation to be optimized is located.

[0077] Advantageously, the situation table associates each predefined situation with at least one predefined optimization method.

[0078] A predefined optimization method could be: a map modification, or a kinematic profile modification, or a map and kinematic profile modification.

[0079] Regarding the situation of the empty roundabout, the situation table can associate it with a map modification consisting of creating a local map M2 that transforms the two-lane roundabout into a single-lane roundabout. Optionally, the situation table could, under certain traffic conditions, associate this situation with a modification of both the map and the kinematic profile.

[0080] In the case of the congested roundabout situation, the situation table can associate it with a modification of the kinematic profile and the calculation of a new trajectory segment S' i, whose curve overlaps with that of the segment S i, and implementing a kinematic profile P2 i with maximum thresholds of speed and / or acceleration and / or jerk greater than those of the kinematic profile P1 i.

[0081] In a preferred embodiment, the determination of a second kinematic profile P2 i includes a calculation of a maximum collision-free velocity threshold VCOL from the data from the perception means 2, and the second kinematic profile P2 i is defined so as to include velocities below the maximum collision-free velocity threshold VCOL.

[0082] The VCOL threshold is the maximum speed at which the autonomous vehicle can move while still having the ability to stop to avoid colliding with obstacles detected by the perception system 2.

[0083] The VCOL threshold is calculated from the driving scene occupancy grid defined by the electronic horizon processing module 21. The driving scene makes it possible to determine which obstacles are closest to the autonomous vehicle 100, in particular subjects of interest such as other vehicles or pedestrians.

[0084] In a simplified embodiment, the VCOL threshold can be calculated by solving the following Math1 equation: VCOL = 2 × A × DLIM Or : A is an acceleration, specifically the braking value needed to reach the VCOL speed, DLIM is the limiting distance of perception.

[0085] To determine the second kinematic profile P2i, the jerk limitation method described earlier can be used. In this case, the increased responsiveness of the autonomous vehicle 100 will be achieved by setting the maximum speed threshold of the second profile P2i to a value Vmax2 that is both higher than the maximum speed threshold Vmax1 of the first profile P1i and lower than the maximum collision-free speed VCOL. Advantageously, the maximum acceleration and / or jerk thresholds Amax2 and Jmax2 of the second profile are respectively higher than the thresholds Amax1 and Jmax1 of the first profile P1i.

[0086] Following the determination of a local map M2 and / or a second kinematic profile P2 i, we loop back to the step E1 of determining a trajectory to calculate a second trajectory segment S' i -or optimized segment S' i - replacing the first trajectory segment S i.

[0087] The optimized segment S' i is determined so as to be compatible with the local map M2 and / or the second kinematic profile P2 i.

[0088] The optimized segment S' i can exhibit the same curve as the first segment S i, particularly when map M1 has not been modified. However, they differ in the speeds and / or accelerations and / or jerks they employ.

[0089] The optimized segment S' i connects the entry and exit points A, B of the given segment S i . The optimized segment S' i is advantageously determined so as to exhibit a continuity G2, that is to say a continuity of curvature, between the different arcs which constitute it, but also a continuity of curvature at point A with the trajectory segment which precedes it, and a continuity of curvature at point B with the trajectory segment which follows it.

[0090] In one embodiment, the optimized segment S' i can be defined by a fifth-degree polynomial, notably using Bézier curves.

[0091] Then we move on to step E4. In the fourth step E4, we command the movement of the autonomous vehicle according to the trajectory determined in the previous steps, this trajectory may include an optimized segment S' i.

[0092] The trajectory is transmitted to the control laws of the autonomous vehicle 100, in order to be transformed into command orders transmitted to the motion actuators 4 of the autonomous vehicle 100.

[0093] Two examples of the implementation of the control process are described by the figures 4 And 5 As a note, the position of the autonomous vehicle 100 on these figures represents the E4 stage of vehicle movement along the optimized trajectory.

[0094] There figure 4 This illustrates a first example of implementing the process. At time T, the autonomous vehicle 100 is traveling on a segment Si-1 of a trajectory T1, the trajectory T1 being calculated based on a map M1 and a kinematic profile P1. The following segment Si lies entirely within the perception zone, defined by the perception limit distance DLIM calculated at time T. The following segment Si involves movement on the outer lane of a roundabout. In the absence of traffic on the roundabout, in step E2, an empty roundabout is detected. In step E3, the map M1 is locally modified by a local map M2 to replace the two-lane roundabout with a single-lane roundabout. Then, a new trajectory S'i is calculated in step E1 using the local map M2, the new trajectory S'i connecting the entry point A of segment Si to the exit point B of segment Si.In step E4, the autonomous vehicle 100 is commanded to move on the segment S'i between points A and B.

[0095] There figure 5This illustrates a second example of implementing the process. In this second example, the following segment Si determines movement along the outer lane of a roundabout. Traffic on the roundabout is heavy, and the autonomous vehicle 100 must stop at the roundabout entrance. After a given period of immobility of the autonomous vehicle at the roundabout entrance, in step E2, a congested roundabout situation is detected. In step E3, a new kinematic profile P2i is defined to facilitate the vehicle's entry onto the roundabout. The kinematic profile P2i is then used in step E1 to determine the new segment S'i connecting points A and B. The segments Si and S'i follow the same curve but do not implement the same speeds and / or accelerations and / or jerks. In step E4, the autonomous vehicle 100 is commanded to move along the segment S'i between points A and B.

[0096] Finally, the control method according to the invention uses perception means to detect a set of traffic scenarios and optimize the trajectory of the autonomous vehicle according to the identified traffic scenario.

[0097] The trajectory optimization according to the invention implements two optimization levers that can be used independently or simultaneously. The first optimization lever consists of locally modifying the map representing the infrastructure of the identified traffic scenario. The second optimization lever consists of temporarily increasing the responsiveness of the autonomous vehicle by raising the maximum speed, acceleration, and / or jerk thresholds applied within its trajectory.

[0098] Optimization levers are determined based on surrounding traffic to avoid collisions. Map adjustments are preferably made when there is no surrounding traffic. Increased vehicle responsiveness occurs in heavy traffic conditions and takes into account a maximum collision-free speed calculated from electronic horizon data. This ensures that trajectory optimization takes place under safe conditions.

[0099] In addition to its reliability in terms of road safety, the invention offers several advantages. First, it reduces the risk of the autonomous vehicle becoming stuck in heavy traffic. By temporarily increasing the autonomous vehicle's responsiveness, it allows it to merge more easily into dense traffic, while returning to an optimal level of driving comfort after the merge maneuver. The invention thus promotes smoother traffic flow and driving comfort by adapting the autonomous vehicle's driving style to the surrounding traffic. More generally, in a range of given situations, the invention allows the autonomous vehicle to replicate the optimization choices made by a human driver facing those same situations.

[0100] Thus, the invention allows the autonomous vehicle to adapt its driving responsiveness based on perception data relating to static (infrastructure) and dynamic (particularly traffic) data in the driving scene. The driving fluidity of the autonomous vehicle is thereby improved, notably avoiding roadblocks, while maintaining driving safety.

Claims

1. Method for controlling an autonomous vehicle (100) equipped with at least one perception means (21) and with a first map (M1) comprising stored digital data representing the actual infrastructure of the environment of the autonomous vehicle (100), the control method comprising: - a step (E1) of defining a first trajectory (T1) of the autonomous vehicle (100) and a first associated kinematic profile (P1); - a step (E2) of detecting, on the basis of data transmitted by the at least one perception means (21), a situation to be optimized on a given segment (S) of the first trajectory (T1), the given segment (S) being located in front of the autonomous vehicle (100), over a distance which is less than a maximum detection distance (DLIM) of the at least one perception means (21), the control method being characterized in that it comprises: - a step (E3) of optimizing the first map (M1), optimizing the first map comprising determining a local map (M2) on the given segment (S), the local map (M2) representing a virtual modification of the actual infrastructure, - a step (E4) of controlling the movement of the autonomous vehicle (100) on the given segment (S), taking into account the local map (M2).

2. Control method according to the preceding claim, characterized in that determining a local map (M2) makes it possible to determine, notably by means of Bezier curves, a new curve of the given segment (S) between an entry point (A) and an exit point (B) of the given segment (S), and in that the new curve has a continuity G2 with the first trajectory (T1) at the entry point (A) and at the exit point (B).

3. Control method according to one of the preceding claims, wherein the step (E3) of optimizing the first map (M1) further comprises optimizing the first kinematic profile (P1), optimizing the first kinematic profile (P1) comprising determining a second kinematic profile (P2) associated with the given segment (S), at least one parameter of the second kinematic profile (P2) being different from a parameter of the first kinematic profile (P1), the step (E4) of controlling the movement of the autonomous vehicle (100) on the given segment (S) furthermore taking into account the second kinematic profile (P2).

4. Control method according to the preceding claim, characterized in that determining a second kinematic profile (P2) associated with the given segment (S) comprises computing a maximum collision-free speed threshold (VCOL) on the basis of the data originating from the at least one perception means (21), and in that the second kinematic profile (P2) comprises maximum speeds below the maximum collision-free speed threshold (VCOL).

5. Control method according to the preceding claim, characterized in that the maximum collision-free speed threshold (VCOL) is computed by solving an equation according to which the maximum collision-free speed threshold (VCOL) is equal to the square root of the product of the braking value required to reach the maximum collision-free speed threshold (VCOL) and twice the maximum detection distance (DLIM) of the at least one perception means (21).

6. Control method according to one of the preceding claims, characterized in that the step (E2) of detecting a situation to be optimized comprises measuring a holding time, during which the autonomous vehicle (100) remains in a situation to be optimized, and comparing the holding time with a triggering threshold associated with said situation, and in that the optimization step (E3) is executed when the holding time is above the triggering threshold.

7. Control method according to one of the preceding claims, characterized in that a set of predefined situations is stored in an electronic memory (6), and in that the step (E2) of detecting a situation to be optimized comprises comparing a perception originating from the at least one perception means (21) with the set of predefined situations.

8. Control method according to the preceding claim, characterized in that the set of predefined situations comprises negotiating a roundabout the two lanes of which are empty, and / or entering a roundabout in dense traffic conditions, and / or two traffic lanes merging into a single lane, and / or crossing a gridlocked junction, and / or overtaking a very slow vehicle on a two-way road comprising a single lane in each direction.

9. Control method according to one of Claims 3 to 5, characterized in that the second kinematic profile (P2) comprises a maximum speed which is at least 20% higher, and / or a maximum acceleration which is at least 200% higher, and / or a jerk which is at least 1000% higher than the first kinematic profile (P1).

10. Device (10) for controlling an autonomous vehicle (100), the autonomous vehicle being equipped with motion actuators (4), with at least one perception means (21) and with a first map (M1) comprising stored digital data representing the actual infrastructure of the environment of the autonomous vehicle (100), the device comprising means configured to implement the method according to one of the preceding claims.

11. Autonomous vehicle (100), characterized in that it comprises a control device (10) according to the preceding claim.

12. Computer program product comprising program code instructions which are stored on a computer-readable medium for implementing the steps of the method according to any one of Claims 1 to 9 when said program is executed by a computer.

13. Computer-readable data storage medium on which a computer program according to the preceding claim is stored.