Device for planning the path and / or trajectory of a motor vehicle

Abstract

The invention relates to a device (2) for planning a path and / or trajectory of a motor vehicle (4), the device comprising a module (6) for receiving a sequence of input variables, and hardware and software means for determining a control law corresponding to the path and / or the trajectory from the received sequence of input variables. The device (2) comprises a temporal classification means (8) in the form of a compact representation set.

Description

Technical Field

[0001] The application relates to the field of an apparatus and a method for planning a path and / or trajectory of a motor vehicle, and a computer program for implementing such a method. Background Art

[0002] At present, motor vehicles are equipped with advanced driver assistance devices with increasing performance levels. Advanced driver assistance devices are intended to allow autonomous driving of motor vehicles, that is, driving without the intervention of the driver. In particular, advanced driver assistance devices can be used to plan the path and / or trajectory of a motor vehicle. In this application, the term "planning" and its derivatives will be interpreted according to their usual definition in the field of advanced driver assistance devices, that is, the planning of a path or trajectory means that a series of states (position, speed, acceleration) of the motor vehicle in the future corresponds to a path or trajectory from one state to another desired state. In this application, the path of a vehicle will be understood as a geometric shape corresponding to the travel of the vehicle between a starting point and an arrival point. The trajectory of a vehicle will be understood as the change in the position of the vehicle between a starting point and an arrival point over time.

[0003] Known examples of methods for planning a path and / or trajectory of a motor vehicle include in particular methods for planning by solving an optimal control problem modeled on the basis of first-order Lagrangian mechanics equations. Methods for generating a path and / or trajectory generally aim to find a good compromise between comfort and safety of the vehicle occupants.

[0004] An example of a method for generating trajectories can be found in the paper “Optimal trajectory generation for dynamic street scenarios in a Frenet frame, IEEE International Conference on Robotics and Automation” by M. Werling, J. Ziegler, S. Kammel and S. Thrun [1] (hereinafter referred to as the “Werling method”). The Werling method relates to an indirect method for solving an optimal control problem. According to this method, some input parameters are selected, the optimal control problem is partially solved by considering only the selected input parameters, and the solution is verified a posteriori using the unselected input parameters and the selected input parameters. However, the use of an indirect method lacks robustness due to the large number of input parameters that have to be considered. Moreover, the fact that the collision avoidance constraints are verified a posteriori may result in the planning of a large number of paths and / or trajectories being ultimately rejected due to the risk of collision, so that a lot of resources are used for the large number of rejected calculations.

[0005] Another example of a trajectory planning method (hereinafter referred to as "Mercy method") is described in the paper "Time-optimal motion planning in the presence of moving obstacles" by T. Mercy, W. Van Loock, G. Pipeleers and J. Swevers [2]. This example uses a direct method to solve the optimal control problem. According to the direct method, all constraints are taken into account when solving the problem. Therefore, the present invention is closer to Mercy's method than to Werling's method. However, the complexity of the algorithm associated with Mercy's method makes it impossible to incorporate it into the on-board equipment of a motor vehicle.

[0006] Other examples of path and / or trajectory planning methods include potential field methods, in which each obstacle produces an anti-gravity field that tends to prevent the vehicle from approaching it. However, such methods run the risk of getting stuck in a local minimum and therefore may not produce an optimal solution to the path and / or trajectory planning problem. Summary of the invention

[0007] In view of the above, it is an object of the present invention to enable planning of a path and / or trajectory of a motor vehicle while overcoming the above mentioned disadvantages.

[0008] More specifically, the present invention proposes to enable a sufficiently robust path and / or trajectory generation while reducing the computational complexity.

[0009] For this purpose, a device for planning a path and / or trajectory of a motor vehicle is proposed, which device comprises a module for receiving a series of input variables, and hardware and software means for determining a control law corresponding to the path and / or the trajectory based on the received series of input variables.

[0010] According to one of its general characteristics, the device comprises temporal classification means in the form of a set of compact representations.

[0011] By using a temporal classifier, paths and / or trajectories can be planned based on the historical records. Like the Mercy method, this approach enables all input parameters to be taken into account, while having a relatively low complexity. In fact, the use of a temporal classifier requires a lot of work to design and tune the algorithm, but a small amount of work for the inference phase of planning. Therefore, the computational resource requirements are shifted from the online inference case to the offline training phase.

[0012] In this application, the expression "control law" will be interpreted in the following sense: a control law corresponding to a path or trajectory comprises the set of commands required to operate the actuators in order to make the vehicle follow this path or trajectory while modeling the physical constraints of the chassis.

[0013] The expression "temporal classification" denotes the act of classifying a plurality of input sequences as a function of time.

[0014] The expression "compact representation set" denotes the set formed by input sequences that have undergone operations to become compact, ie, for example, dropping the notion of time in the input sequence, or deleting certain points in the input sequence in order to reduce their memory footprint.

[0015] According to one embodiment, the temporal classification device is configured to define a surjection of an input set consisting of different input variable sequences of variable length to an output set consisting of different output variable sequences of variable length.

[0016] In particular, this characteristic is made possible by the use of a temporal classification device. Thus, the temporal classification device does not select a sequence of input variables of a predefined length, but continues to receive input variables until it has enough data to determine a satisfactory solution, thereby increasing the length of the sequence. The surjection between the input set and the output set is the application of a surjection of the input trajectory set corresponding to the stored trajectories of different categories to the output trajectory set. For example, the application may belong to a grouping in which the system used is also able to interpolate or extrapolate trajectories not seen during learning. In the present application, the length of the sequence can be temporal, i.e. corresponding to the sampling period of the sequence, or metric, i.e. corresponding to the sampling distance of the sequence. Similarly, the temporal classification device generates a variable sequence at its output that can take any length.

[0017] Preferably, the input set includes a temporal input variable sequence and / or a metric input variable sequence.

[0018] According to this variant, trajectories of various classes are stored in the input set in memory by temporal classification, but the trajectories have different lengths in order to better represent possible situations.

[0019] Advantageously, the input set comprises at least one sequence of input variables selected from: a sequence of raw variables obtained from sensors, a sequence of variables for incorporating a representation of the vehicle's external environment, a sequence of variables for representing internal parameters of the vehicle, a sequence of variables for representing the agent's behavior, and a sequence of variables for representing the agent's intention.

[0020] In this application, the term "representation" is used to refer to a data set representing an event, parameter, state, etc. A "merged representation" is a data set obtained by merging data obtained from different sensors and / or estimators.

[0021] In one embodiment, the output set includes a temporal output variable sequence and / or a metric output variable sequence.

[0022] Preferably, the output set comprises at least an output sequence selected from: a sequence of coordinate pairs of route points, a sequence of abscissas of route points, and a sequence of ordinates of route points.

[0023] In another embodiment, the temporal classification means is configured to use a connectionist temporal classification function.

[0024] The temporal classification function may be of various types, such as: Hidden Markov Field, RNN-LSTM, RNN-LSTM-CTC (Connectionism). By using such a function, efficient training of the temporal classification device may be implemented.

[0025] According to yet another embodiment, the temporal classification means comprises a hidden Markov chain model and / or a recurrent neural network, preferably a recurrent neural network with long short-term memory.

[0026] Such an example is particularly suitable for forming temporal classification devices, as these devices can be easily trained during an offline training phase.

[0027] According to another aspect, a method for planning a path and / or trajectory of a motor vehicle is proposed, the method comprising receiving a sequence of input variables and determining a control law corresponding to the path and / or trajectory of the motor vehicle based on the sequence of input variables.

[0028] According to one of its general characteristics, the method consists in using a temporal classification means in the form of a set of compact representations.

[0029] According to one embodiment, the method comprises an offline training phase of the temporal classification device, in which a set of output variable sequences each corresponding to the same manipulation is grouped into a first level grouping.

[0030] This grouping is particularly suitable for the purpose of preparing the temporal classification device for its correct operation during the inference phase.

[0031] In this application, the expression "offline" is used to mean that the off-line phase occurs outside the time period when the vehicle is controlled by the control law, for example when the vehicle is not moving.

[0032] According to one embodiment, during the offline training phase, for each first level grouping, sets of variable sequences each corresponding to manipulations having substantially equal implementation periods and / or substantially equal implementation distances are grouped into second level groups.

[0033] Therefore, the temporal classification means can be finely tuned, thereby improving the robustness of the planning without thereby increasing the computational complexity.

[0034] In one embodiment, the method comprises: transmitting an index and at least one time modulation factor by the time classification device, reading a path associated with the transmitted index in a lookup table, and time modulating the path read by the transmitted time modulation factor.

[0035] The method according to the invention therefore consists in storing a plurality of path patterns with a certain degree of freedom with respect to time, which degree of freedom becomes a parameter. This limits the computational complexity still further.

[0036] Advantageously, the temporal classification means transmits the temporal modulation factor in the longitudinal direction and / or transmits the temporal modulation factor in the transverse direction.

[0037] According to yet another aspect, a computer program is proposed, comprising code configured for implementing the method as defined above when executed by a processor or an electronic control unit. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Other objects, characteristics and advantages of the present invention will become apparent from a careful reading of the following description, which is given by way of non-limiting example only and is made with reference to the accompanying drawings, in which:

[0039] - Figure 1 is a schematic representation of a planning device according to an embodiment of the present invention,

[0040] - Figure 2 is a schematic representation of a planning method according to an embodiment of the present invention,

[0041] - Figure 3 Shows Figure 2 The first stage of the method,

[0042] - Figure 4 Shows Figure 2 The second stage of the method, and

[0043] - Figure 5 and Figure 6 The route selection of the vehicle in two corresponding operating situations is schematically shown. DETAILED DESCRIPTION

[0044] refer to Figure 1, schematically representing a planning device 2. The device 2 is intended to plan a path for a motor vehicle 4 incorporating it. Obviously, other configurations of the device 2 for planning the trajectory of the vehicle 4 can be envisaged without departing from the scope of the invention. In the example shown, the vehicle 4 is autonomously driven. Without departing from the scope of the invention, it can be envisaged that the vehicle incorporating the device 2 is not autonomously driven; that is, the vehicle is driven by a driver. In this case, the device 2 is intended only to provide assistance to the driver (for example in dangerous situations).

[0045] In the example shown, the device 2 forms part of an autonomous driving framework (not shown). More specifically, the device 2 is data linked to a prediction module (not shown), a sensor merging module (not shown) and a positioning module (not shown) belonging to the autonomous driving framework.

[0046] The device 2 comprises a receiving module 6 that is data linked with the other modules of the autonomous driving architecture and with the onboard computer of the vehicle. More precisely, the module 6 is capable of receiving a plurality of sequences of input variables. A sequence is a collection of values ​​taken by a variable. A time series is a collection of values ​​taken by a variable at a plurality of different moments. A measurement series is a collection of values ​​taken by a variable at a plurality of different locations. In the example shown, the sequence is temporal and consists of the values ​​of the variable taken at a plurality of moments separated by an interval of 40 ms.

[0047] In the example shown, the module 6 receives a sequence of input vectors whose scalars are input variables that can be internal constraints and external constraints. Internal constraints relate in particular to the comfort and safety of the passengers in the vehicle 4 and include the driver's intention, predictions of future maneuvers, the speed of the vehicle 4, or the input of the passengers to the driving mode of the vehicle 4. External constraints relate in particular to environmental conditions and traffic and include the position, speed and acceleration of other vehicles, the curvature of the road, meteorological conditions, the type of road surface, day or night conditions, or information related to visibility.

[0048] The device 2 includes a recurrent neural network 8, which is generally referred to by the English term "Recurrent Neural Network" or the corresponding acronym "RNN". Reference may be made to the document "Trajectory Generation Using RNN with Context Information for Mobile Robots" by Y.-M. Lee and J.-H. Kim in the journal Robot Intelligence Technology and Applications [3], which mentions an example of a recurrent neural network. The recurrent neural network 8 is data-linked to the receiving module 6.

[0049] In the example shown, the recurrent neural network 8 is of the long short-term memory type. This property is usually referred to by the English term “Long Short Term Memory” or the corresponding acronym “LSTM”.

[0050] The recurrent neural network 8 is configured to use a connectionist temporal classification function. This function is usually referred to by the English term "Connectionist Temporal Classification" or the corresponding acronym "CTC". The recurrent neural network 8 thus forms a temporal classification means of the device 2. In addition, the use of a recurrent neural network makes it possible to generate a variable length sequence of output variables by collecting a variable length sequence of input variables.

[0051] The recurrent neural network 8 is configured to determine and transmit an index ind, a longitudinal modulation factor lon and a lateral modulation factor lat based on the input variables received by the receiving module 6. More specifically, the variables ind, lon and lat are determined so that the sum of the lateral accelerations experienced by the vehicle 4 during the maneuver is minimized while ensuring a degree of safety, avoiding any obstacles and respecting the maximum maneuvering length.

[0052] Obviously, the invention is not limited to the use of recurrent neural networks 8. In particular, Hidden Markov Models (also known by the English term “Hidden Markov Model” or the corresponding acronym “HMM”) may be used.

[0053] The device 2 has a lookup table 10. The table 10 may be referred to by the English name "Look-Up Table" or the corresponding acronym "LUT". The table 10 is data-linked with the recurrent neural network 8. The table 10 contains a plurality of path shapes based on the index ind. The path shapes stored in the table 10 correspond to possible paths of the vehicle 4.

[0054] The device 2 comprises a temporal modulator 12. The modulator 12 is data linked to the recurrent neural network 8 and to the table 10. More precisely, the modulator 12 is capable of modulating the longitudinal position and the transverse position of the points forming the path delivered by the table 10, respectively, with respect to time, based on the longitudinal modulation factor and the transverse modulation factor delivered by the recurrent neural network 8. In other words, the modulator 12 is capable of expanding or contracting the path delivered by the table 10. The path expanded or contracted by the modulator 12 is associated with a speed profile that determines the time sequence of the route points. This time sequence is supplied to a control law device (not shown) of the vehicle 4.

[0055] refer to Figure 2 , which shows the main stages of a method for planning a path for a motor vehicle 4 , which can be implemented by the device 2 .

[0056] The method comprises a first phase P01 of training a recurrent neural network 8. Phase P01 is implemented when the onboard computer (not shown) of the vehicle 4 is offline, ie before transport of the vehicle 4, during maintenance of the vehicle 4, or during long-term parking of the vehicle 4.

[0057] refer to Figure 3 , phase P01 comprises a first step E01 of recording actual driver behavior. For example, the behavior of a test driver on a vehicle of the same model as vehicle 4 may be recorded. The type of maneuver may be identified and recorded while identifying and recording the behavior. For example, the action of leaving the center of a lane to reach the center of a lane (which may be the same lane or a different lane) may be identified and recorded.

[0058] Phase P01 comprises a second step E02 of grouping the recorded data by maneuver type. More precisely, in step E02, a plurality of first-level groups are created, each group corresponding to a respective maneuver recorded in step E01. For example, if during step E01 data have been recorded respectively for a maneuver of changing lanes, a maneuver of re-centering within the same lane, an emergency braking maneuver and an emergency evasive maneuver of the vehicle, four first-level groups are respectively associated with the aforementioned four maneuvers.

[0059] The method comprises a third step E03 of grouping the recorded data by maneuver length. In step E03, a plurality of second-level groups are created in each first-level group, each second-level group corresponding to a respective maneuver length recorded in step E01. For example, if data have been recorded for an emergency braking maneuver for a distance of 20 meters (+ / -10%), for an emergency braking maneuver for a distance of 30 meters (+ / -10%) and for an emergency braking maneuver for a distance of 50 meters (+ / -10%), respectively, the first-level group associated with the emergency braking maneuver is subdivided into three second-level groups respectively associated with the three aforementioned maneuver lengths.

[0060] Phase P01 comprises a fourth step E04 of enriching table 10. In step E04, the trajectories stored in table 10 are refined using the trajectories received in step E01.

[0061] Phase P01 comprises a step E05 of separating the distinguishing characteristics associated with the different groups defined in steps E02 and E03. In particular, in step E05, the distinguishing characteristics may be:

[0062] - information obtained from the merging module, such as: status information about the vehicle 4; status information about surrounding objects; a map representing, for each neighboring vehicle, the position, speed and acceleration of the neighboring vehicle and the free space or risk-free space around the neighboring vehicle; and the boundary lines of the lanes,

[0063] - context-related information, such as road surface, weather forecast, day or night conditions or visibility information,

[0064] - information about the vehicle 4, such as the selection of a driving mode,

[0065] - Information about intentions, such as actual or simulated operation of indicator lights, sudden changes in direction angle, sudden braking (thus detecting the intention of emergency braking).

[0066] Phase P01 comprises a step E06 of defining a temporal classification means. In this case, a recurrent neural network 8 of the RNN-LSTM type using a CTC function is defined in step E06.

[0067] Phase P01 comprises a step E07 of training the recurrent neural network 8. Step E07 is implemented by using as input the set of distinctive features separated in step E05 and by using as output the possible trajectories recorded and grouped in steps E01, E02 and E03.

[0068] The invention is not limited to the steps described above with reference to the examples presented. In particular, instead of using the grouping technique implemented in steps E02 and E03, a competitive learning strategy can be used. Other variants that can be used are described in the publication by C. Sung, D. Feldman and D. Rus “Trajectory clustering for motion prediction, International Conference on Intelligent Robots and Systems” [4].

[0069] refer to Figure 2 , the method comprises a second phase P02 of inference. Phase P02 is performed online; that is, it is continuously implemented while the vehicle 4 is moving. Although phase P01 is represented above and has been described before phase P02, phase P01 may be after and / or before phase P02.

[0070] refer to Figure 4 Phase P02 comprises a first step F01 of receiving a sequence of input variables. The received sequence of input variables consists of n vectors, each of which is generated by module 10 at time t i The received sequence is composed of different input variables, wherein i is in the range of 1 to n. The value of the number n is selected so that the length of the sequence of input variables is sufficient to determine the relevant values ​​ind, lon and lat. The input variables are substantially the same as the distinguishing characteristics separated in step E05 of phase P01, usually data obtained from a sensor merging module and other information from certain specific sensors (such as road keeping and visibility conditions) or information about the intention from the behavior planner. In the example of this embodiment, the received sequence is temporal. Without departing from the scope of the invention, another type of sequence, in particular a metric sequence, can be envisaged.

[0071] Phase P02 comprises a second step F02 of inputting the sequence of input variables into the recurrent neural network 8. In step F02, the index ind, the longitudinal modulation factor lon and the transverse modulation factor lat delivered by the recurrent neural network 8 are collected.

[0072] Phase P02 comprises a third step F03 of entering the index ind into table 10. In step F03, the path shapes delivered by table 10 are collected.

[0073] Phase P02 comprises a fourth step of temporal modulation of the path shape collected in step F03. In step F04, a longitudinal expansion or contraction of the path shape collected in step F03 by a factor lon and a transverse expansion or contraction of the path shape collected in step F03 by a factor lat are implemented.

[0074] Phase P02 comprises a step F05 of checking for collision avoidance. In step F05, it is checked whether the modulated path obtained at the end of step F04 does not present a risk of collision of the vehicle 4 with another object. If there is a risk of collision with another object, an error signal is transmitted so that phase P02 is repeated by selecting a different path.

[0075] Figure 5 and Figure 6 An example of a path that may be planned by the device 2 for the vehicle 4 is shown. Figure 5 A first operating condition of the vehicle 4 is shown, and Figure 6 A second operating situation of vehicle 4 is shown.

[0076] Figure 5 and Figure 6 A road 14 is schematically shown, which includes a plurality of lanes delimited transversely by a first boundary line 16, a second boundary line 18 and a third boundary line 20. Also schematically shown are the center lines 22 of the lanes between lines 16 and 18 and the center lines 24 of the lanes between lines 18 and 20. In both operating situations, a path is planned that enables the vehicle 4 to change lanes. In the initial state and in the final state, the lateral velocity is zero. The longitudinal velocity in the initial state and in the final state is the same.

[0077] refer to Figure 5 , in the recurrent neural network 8, the corresponding index ind is transmitted 56 A path 26 schematically represented by a series of circles is then transmitted by table 10. Path 26 corresponds to the shape of the path collected in step F03 of the method according to the invention. The recurrent neural network 8 transmits the factor lon 5 , which is chosen to minimize the sum of the lateral accelerations during the entire maneuver. Figure 5 and Figure 6 In the example, the factor lat passed by the recurrent neural network 56 = 1. The path 28 schematically represented by a series of crosses corresponds to the factor lon transmitted by the modulator 12 in the recurrent neural network 8. 5 Path 28 corresponds to the modulated path obtained at the end of step F04 of the method according to the invention. Under these conditions, vehicle 4 will follow path 28.

[0078] refer to Figure 6 , the first obstacle 28 and the second obstacle 30 are located on the road 14. The index ind transmitted by the recurrent neural network 8 56 and Figure 5 Therefore, we refer again to Figure 5 However, the recurrent neural network 8 takes into account the existence of obstacles 28 and 30 and transfers the factor lon 6 so that obstacles 28 and 30 can be avoided. The path 32 schematically represented by a series of squares corresponds to the path 32 generated by the modulator 12 at a factor of lon. 6 Path 32 therefore corresponds to the modulated path obtained at the end of step F04 and will be the path followed by vehicle 4 in this operating mode.

[0079] In view of the above, the present invention makes it possible to provide an optimal path and / or trajectory while allowing all internal and external constraints to be taken into account, while using very few computing resources in the inference phase. In particular, by using a temporal classification device, the demand for computing resources is shifted to an offline training phase, making it possible to obtain a robust plan while limiting its complexity.

[0080] List of cited public documents:

[0081] [1]Moritz Werling, Julius Ziegler, Kammel et al., Optimal trajectory generation for dynamic street scenarios in a frenet frame. In Robotics and Automation (ICRA), 2010 IEEE International Conference on. IEEE, 2010, pp. 987–993 (ISBN 978-1-4244-5040-4).

[0082] [2] Tim Mercy, Wannes Van Loock, Goele Pipeleers, et al., Time-optimal motion planning in the presence of moving obstacles, 2015.

[0083] [3] You-Min Lee and Jong-Hwan Kim, Trajectory Generation Using RNN withContext Information for Mobile Robots. Robot Intelligence Technology and Applications, 4. Springer, Cham, 2017, pp. 21–29 (ISBN: 978-3-319-31291-0).

[0084] [4] Cynthia Sung, Dan Feldman, and Daniela Rus. Trajectory clustering for motion prediction. Intelligent Robots and Systems (IROS), 2012 IEEE / RSJ International Conference on. IEEE, 2012, pp. 1547–1552.

Claims

1. A device (2) for planning a path and / or trajectory of a motor vehicle (4), the device comprising a module (6) for receiving a sequence of input variables, and hardware and software means for determining a control law corresponding to the path and / or trajectory based on the received sequence of input variables, wherein: A control law corresponding to a path and / or trajectory comprises a set of commands required to operate actuators so that the vehicle follows the path or trajectory while modeling the physical constraints of the chassis; characterized in that the device comprises a temporal classification device in the form of a set of compact representations; wherein the temporal classification represents an action of classifying a plurality of input sequences as a function of time, and the set of compact representations represents a set formed by input sequences that have undergone an operation of becoming a compact state; and wherein the temporal classification device is configured to use a connectionist temporal classification function.

2. The device (2) as claimed in claim 1, wherein: The temporal classification device is configured to define a surjection of an input set consisting of different input variable sequences of variable length to an output set consisting of different output variable sequences of variable length.

3. The device (2) as claimed in claim 2, wherein: The input set includes a time input variable sequence and / or a measurement input variable sequence; wherein the measurement input variable sequence is a set of values ​​taken by the input variable at multiple different positions.

4. The device (2) as claimed in claim 2 or 3, wherein: The input set includes at least one input variable sequence selected from the following: a raw variable sequence obtained from a sensor, a variable sequence for incorporating an external environment representing the vehicle, a variable sequence for representing internal parameters of the vehicle, a variable sequence for representing the behavior of an agent, and a variable sequence for representing the intention of an agent.

5. The device (2) as claimed in any one of claims 2 to 4, wherein: The output set includes a temporal output variable sequence and / or a metric output variable sequence.

6. The device (2) as claimed in any one of claims 2 to 5, wherein: The output set comprises at least an output sequence selected from: a sequence of coordinate pairs of route points, a sequence of abscissas of route points, and a sequence of ordinates of route points.

7. The device (2) as claimed in any one of claims 1 to 6, wherein: The time classification device includes a hidden Markov chain model and / or a recurrent neural network.

8. The device (2) as claimed in any one of claims 1 to 6, wherein: The time classification device comprises a long short-term memory recurrent neural network (8).

9. A method for planning a path and / or trajectory of a motor vehicle (4), the method comprising receiving (F01) a sequence of input variables and determining a control law corresponding to the path and / or trajectory of the motor vehicle based on the sequence of input variables, characterized in that The method includes using (F02) a temporal classification device in the form of a compact representation set; wherein the control law corresponding to a path and / or trajectory includes a set of commands required to operate actuators to cause the vehicle to follow the path or trajectory while modeling the physical constraints of the chassis; wherein the temporal classification represents the action of classifying multiple input sequences as a function of time, and the compact representation set represents a set formed by input sequences that have undergone operations to become compact states; and wherein the temporal classification device is configured to use a connectionist temporal classification function.

10. The method of claim 9, comprising an off-line training phase (P01) of the temporal classification device, in which sets of output variable sequences each corresponding to the same manipulation are grouped into first-level groups.

11. The method of claim 10, wherein: During this offline training phase (P01), for each first-level grouping, sets of variable sequences each corresponding to manipulations having substantially equal implementation periods and / or substantially equal implementation distances are grouped into second-level groups.

12. The method according to any one of claims 9 to 11, comprising: The time classification device transmits (F02) an index (ind, ind 56 ) and at least one temporal modulation factor (lon, lon5, lon6, lat, lat 56 ), read (F03) in the lookup table (10) and the transmitted index (ind, ind 56 ) associated with the path, and the time modulation factors (lon, lon5, lon6, lat, lat) transmitted by 56 )The path read is time modulated (F04).

13. The method according to any one of claims 9 to 12, wherein: The time classification device transmits (F02) a time modulation factor (lon, lon5, lon6) in the longitudinal direction and / or transmits a time modulation factor (lat, lat 56 ).

14. A computer program product comprising computer instructions configured to implement the method according to any one of claims 9 to 13 when executed by a processor or an electronic control unit.

Images