Automated method for managing the longitudinal speed of a motor vehicle.

Abstract

Method for managing the longitudinal speed of a motor vehicle. Method for managing the longitudinal speed of a motor vehicle comprising a longitudinal speed control module, the method defining a model for managing the longitudinal speed of the motor vehicle, the model being configurable according to a first set of calibration parameters and according to a second set of driving parameters of the motor vehicle, the motor vehicle comprising means for determining the second set of driving parameters, characterized in that it comprises: - an initial training step of the model, comprising the use of different simulated values ​​of the first set of calibration parameters, then - a model update step comprising the use of different observed or measured values ​​of parameters of the first set of calibration parameters and of a second set of driving parameters.Figure for the abridged version: 1.

Description

Title of the invention: Automated method for managing the longitudinal speed of a motor vehicle.

[0001] The invention relates to a method for the automated management of the longitudinal speed of a motor vehicle. The invention further relates to a device for the automated management of the longitudinal speed of a motor 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] The longitudinal speed control of a motor vehicle can be automated, either by a longitudinal speed control module or by an autonomous driving system. Such automatic longitudinal speed management systems make it possible, in particular, to adapt the speed of the equipped vehicle to the presence of other vehicles in its environment. Thus, when active, they act on the vehicle's behavior like a virtual driver taking over part of the driving task.

[0003] Existing automatic longitudinal speed management systems can be configured according to predefined settings, but these do not necessarily correspond to the driver's driving habits.

[0004] Alternatively, some vehicles offer the possibility of configuring the settings of these systems via a human-machine interface. However, the increasing complexity of the systems makes their configuration too complex for users, and they may ultimately turn away from such driver assistance systems.

[0005] The object of the invention is to provide a device and a method for automated longitudinal speed management that overcomes the above drawbacks and improves upon devices and methods for automated longitudinal speed management known in the prior art. In particular, the invention makes it possible to create a device and a method that are simple and reliable and that adapt to a user's driving habits.

[0006] To this end, the invention relates to a method for managing the longitudinal speed of a motor vehicle comprising a longitudinal speed control module, the method defining a model for managing the longitudinal speed of the motor vehicle, the model being configurable according to a first set of calibration parameters and according to a second set of driving parameters of the motor vehicle, the motor vehicle comprising means for determining the second set of driving parameters. The method further comprises: - an initial model training step, including the use of different simulated values ​​of the first set of calibration parameters, then - a model update step including the use of different observed or measured values ​​of parameters from the first set of calibration parameters and a second set of driving parameters.

[0007] In one embodiment, the initial training step includes an iteration over the following substeps, for different values ​​of the first set of calibration parameters: - a sub-step of collecting model training data including (i) assigning a current value to each parameter in the first set of calibration parameters, and (ii) a determination of a current value of each driving parameter of the second set by the determination means, then - a sub-step of integrating into the model a third set of training parameters, the third set comprising the current value of the parameters of the first set of calibration parameters and the current value of the parameters of the second set of driving parameters.

[0008] In one embodiment, the training data collection substep is carried out over a given recording period, - during simulations of a motor vehicle equipped with a longitudinal speed control module, the vehicle moving along a virtual path, and / or - during actual travel of a motor vehicle equipped with a longitudinal speed control module, the vehicle moving on one or more predefined road routes.

[0009] In one embodiment, the integration substep comprises, following the data collection substep, - a segmentation of the given recording duration into a series of sub-periods, then - for each sub-period, a calculation of statistical characteristics relating to the training data recorded during the sub-period, the statistical characteristics including a mean and / or a variance and / or a maximum value, then - an integration into the model of the statistical characteristics relating to the training data of each sub-period.

[0010] In one embodiment, the model update step comprises the following substeps: - a determination, at a given moment of observation, of a common driving mode of the first motor vehicle as being of a first kind if a speed longitudinal of the first motor vehicle is controlled by a human driver, otherwise as being of a second kind, then - if the current driving mode is of the first kind, a sub-step of recording observational data comprising a determination at the time of observation of parameters of the first set of calibration parameters and the second set of driving parameters, from the data from the determination means, or - otherwise, a sub-step at the end of observation including an update of the vehicle's longitudinal speed management model based on observation data collected during the different observation times.

[0011] In one embodiment, the management method includes an iteration of an automatic longitudinal speed management step of the motor vehicle by applying the model as a function of at least one parameter of the second set of driving parameters determined by the determination means, at an instant of the iteration of the automatic longitudinal speed management step of the motor vehicle.

[0012] In one embodiment, the iterations of the model update step and / or the automatic longitudinal speed management step are carried out at a frequency of a computer on which the process is executed, for example at a frequency of 100 Hertz.

[0013] In one embodiment, the management process includes machine learning of the longitudinal speed management model, according to a method chosen from - a method known as the "Vector Support Machine", - a method known as the "K nearest neighbors" method, - a method known as "random forest", and - a method known as "recurrent neural network".

[0014] In one embodiment, the first set of calibration parameters includes - a maximum longitudinal acceleration value applicable when the motor vehicle is preceded by another vehicle, and / or - a maximum longitudinal acceleration value applicable when the motor vehicle is not preceded by any other vehicle, and / or - a setpoint value for the following distance of a vehicle located in front of the motor vehicle, and / or - a maximum value of lateral acceleration in cornering, and / or - a maximum value of longitudinal deceleration.

[0015] In one embodiment, the second set of driving parameters includes at least one parameter taken from - a longitudinal acceleration of the motor vehicle, - a longitudinal speed of the motor vehicle, - a lateral acceleration of the motor vehicle, - a lateral speed of the motor vehicle, - a yaw rate of the motor vehicle, - a distance between the motor vehicle and a vehicle in front of it, - a speed of a vehicle in front of the motor vehicle, - an acceleration of a vehicle in front of the motor vehicle, - a steering angle applied by a driver of the motor vehicle, - an engine torque request applied by the motor vehicle, - a braking torque request applied by the motor vehicle.

[0016] In one embodiment, the process includes a driving phase delimited by - a moment at the start of the driving phase, where a motor vehicle engine is started by means of a key or an ignition button by a driver, and - a moment at the end of the driving phase, where a motor vehicle engine is stopped by means of a key or an ignition button by a driver. Furthermore, between the start time of the driving phase and the end time of the driving phase, a model for managing the longitudinal speed of the motor vehicle is recorded in a volatile memory of the motor vehicle. In addition, at the time of the end of the driving phase, a model for managing the longitudinal speed of the motor vehicle is recorded in a non-volatile memory of the motor vehicle.

[0017] The invention also relates to a longitudinal speed management device for a first motor vehicle, the motor vehicle being equipped with a longitudinal speed control module, an engine and a braking system, the device comprising hardware and / or software elements implementing the method according to the invention.

[0018] The invention also relates to a computer program product comprising program code instructions recorded on a computer-readable medium to implement the steps of the process according to the invention when said program is running on a computer or a computer program product downloadable from a communication network and / or recorded on a computer-readable and / or computer-executable data medium comprising instructions which, when the program is executed by the computer, cause the computer to implement the steps of the process according to the invention.

[0019] The invention also relates to a computer-readable data recording medium on which is recorded a computer program comprising program code instructions for implementing the process according to the invention or a computer-readable recording medium comprising instructions which, when executed by a computer, lead the computer to implement the steps of the process according to the invention.

[0020] The invention also relates to a signal from a data carrier, carrying the computer program product according to the invention

[0021] Fig. 1 represents a motor vehicle equipped with an automated longitudinal speed management device according to the invention.

[0022] Fig. 2 defines longitudinal and lateral speeds of the motor vehicle and of a target vehicle preceding it in its lane of travel.

[0023] The [Fig.3] is a first flowchart of an automated management process according to the invention.

[0024] Figure 4 illustrates a comparative analysis of residual errors relating to three longitudinal speed management models.

[0025] The [Fig.5] is a second flowchart of an automated management process according to the invention.

[0026] An embodiment of a vehicle equipped with a means for implementing an automated longitudinal speed management method is described below with reference to [Fig.1].

[0027] The first motor vehicle 100, or motor vehicle 100, can be a motor vehicle of any type, including a passenger car, a commercial vehicle, a truck, or a public transport vehicle such as a bus or a shuttle. According to the embodiment described, the motor vehicle 100 can be an autonomous vehicle.

[0028] This illustration is not exhaustive. In particular, the motor vehicle could be a non-autonomous vehicle, equipped with a driving assistance system, in particular a driving assistance system corresponding to a level greater than or equal to level 2 of autonomy, i.e. corresponding to a partial autonomy of the vehicle.

[0029] The motor vehicle 100 includes an automated longitudinal speed management system 10 for a motor vehicle, also referred to in the rest of the document as "management system 10".

[0030] The management system 10 can be part of a more global driver assistance system 50, including a longitudinal speed control module 5 capable of transmitting control commands to a motor 6 or to a braking system 7 of the vehicle.

[0031] As a note, the motor vehicle 100 may be equipped with several motors for its movement, for example, a combustion engine and an electric motor. In the remainder of this document, stopping the motor vehicle 100 refers to stopping all the motors used to move the motor vehicle 100.

[0032] With reference to [Fig. 2], longitudinal and lateral speeds of motor vehicle 100 and of a target vehicle 200 preceding it in its lane of travel are defined. It is assumed that motor vehicle 100 is traveling in a lane 40 of a road, and the terminology used in the remainder of this document is defined: - The axis called longitudinal axis 101 of the motor vehicle 100 is defined as an axis of symmetry of the motor vehicle 100 parallel to the axis along which the vehicle moves in a straight line, oriented towards the front of the vehicle. - The axis called lateral axis 102 of the motor vehicle intersects perpendicularly the longitudinal axis 101 at a point located at the center of gravity of the motor vehicle 100, and it is oriented towards the left of the motor vehicle, left and right being defined according to the point of view of the driver. - The velocity vector 103 of the motor vehicle 100 projected onto the longitudinal axis 101 defines the longitudinal component 104 of the velocity vector, called longitudinal velocity. - The velocity vector 103 of the motor vehicle 100 projected onto the lateral axis 102 defines the lateral component 105 of the velocity vector, called lateral velocity. - Similarly, a distance between two vehicles can be projected onto the longitudinal and lateral axes, thus defining a longitudinal distance -or DS tracking distance- and a lateral distance.

[0033] The same terminology is applied to define the position and speed parameters of a second vehicle 200, shown in [Fig. 2]. The second vehicle 200 is characterized by the fact that it travels in the same lane 40 as the motor vehicle 100 and is located directly in front of it. In the remainder of this document, the second vehicle 200 may be referred to interchangeably as the "target vehicle 200".

[0034] A target vehicle 200 can be a motor vehicle of any type, including a passenger vehicle, a utility vehicle, or a motorcycle.

[0035] In the remainder of the document, the term "tracking time" refers to the time it would take, at a given moment, for the motor vehicle 100 to reach the position of a target vehicle 200.

[0036] The longitudinal speed control module 5 is configurable according to a first set ENS1 of calibration parameters, which determine the dynamics of the longitudinal speed control laws of the motor vehicle 100.

[0037] In one embodiment, the first set ENS1 of calibration parameters comprises: - a maximum longitudinal acceleration value applicable when the motor vehicle 100 is preceded by another vehicle, including a target vehicle, and / or - a maximum longitudinal acceleration value applicable when the motor vehicle 100 is not preceded by any vehicle, and / or - a setpoint value for the following distance of a vehicle located in front of the motor vehicle (100), and / or - a maximum value of lateral acceleration in cornering, and / or - a maximum value of longitudinal deceleration.

[0038] In this embodiment, the chosen calibration parameters have a first-order effect on the longitudinal speed control module 5. Other parameters could be considered, for example, maximum jerk values, or minimum values ​​of a final distance between a target and the motor vehicle 100 when stopped.

[0039] The following Table 1 provides an example of values ​​assigned to the first set ENS1 of calibration parameters.

[0040] [Tables 1] Parameters Min Max Maximum longitudinal acceleration 0.5 m / s² 1.5 m / s² Tracking time 1 s 4 s Reference lateral acceleration 0.6 m / s² 1 m / s² Allowed acceleration with target 0.2 m / s² 0.4 m / s² Maximum deceleration -1 m / s² -3.8 m / s² Allowed acceleration without target 0.4 m / s² 1.6 m / s²

[0041] The motor vehicle 100 further includes means for determining 1 a second set ENS2 of driving parameters.

[0042] In the remainder of this document, the term "driving parameters" refers to parameters representative of a driving mode of a user of the motor vehicle 100. The driving parameters are likely to be taken into account in the management of the longitudinal speed of the motor vehicle 100, in particular in the determination of values ​​of the calibration parameters of the longitudinal speed control module 5.

[0043] In one embodiment, the driving parameters include at least one parameter taken from: - a longitudinal acceleration of the motor vehicle 100, - a longitudinal speed of the motor vehicle 100, - a lateral acceleration of the motor vehicle 100, - a lateral speed of the motor vehicle 100, a yaw rate of the motor vehicle 100, - a distance between motor vehicle 100 and a vehicle in front of it, - a speed of a vehicle in front of motor vehicle 100, - an acceleration of a vehicle in front of motor vehicle 100, - a steering angle applied by a driver of motor vehicle 100, - an engine torque request applied by motor vehicle 100, - a braking torque request applied by motor vehicle 100.

[0044] Driving parameters may further include: - weather conditions at a location where the motor vehicle 100 is traveling, for example, rainy weather, - Visibility conditions related to road layout or lighting, - The driver's preferred driving style, for example a sporty driving style, - Traffic conditions, including traffic density, the presence of a target vehicle 200 in front of the motor vehicle 100, the distance or time interval separating the motor vehicle 100 from the target vehicle 200, - road infrastructure, for example the number of lanes of the road, the type of road (departmental road, national road, motorway).

[0045] The means 1 for determining the driving parameters include a means for determining the speed and acceleration of the motor vehicle 100. The means 1 for determining the driving parameters may also include a sensing means, in particular a camera and / or a radar and / or a lidar located at the front of the motor vehicle 100. The determination means may also include a means for geolocating the motor vehicle on a map, enabling, in particular, the determination of an applicable speed limit. Other embodiments of the determination means 1 are conceivable, depending on the driving parameters taken into account.

[0046] During manual operation of a vehicle, a human driver ensures that the driving parameters evolve in a coherent manner with each other, for example, adapting their longitudinal speed to visibility conditions, traffic conditions, road grip conditions, and the desired responsiveness in a given driving situation...

[0047] The invention aims to - on the one hand, the construction of a MOD model representative of how a given user adapts the longitudinal speed of the motor vehicle 100 according to driving parameters, and - on the other hand, the use of the MOD model to determine calibration parameters based on driving parameters, so as to reproduce the driving habits of the given user.

[0048] Initially, prior to the marketing of the motor vehicle 100, particularly during vehicle calibration, the management system 10 implements a learning phase of an initial model MOD0 for longitudinal speed management. The learning of the initial model MOD0 is carried out in the initial training step E100, described later in this document.

[0049] Subsequently, the management system 10 operates alternately between phases of learning a MOD model and phases of operating the MOD model, the MOD model being specific to the driving habits of a given user.

[0050] In other words: - In a first operating mode, M_MANUAL, the longitudinal speed of the motor vehicle 100 can be controlled manually. In this case, the longitudinal speed of the motor vehicle 100 is determined by the driver. In this operating mode, the management system 10 performs a learning process based on the driver's driving habits, focusing on the current values ​​of the driving parameters. The learning process includes the construction of a customized MOD model for managing longitudinal speed. - In a second operating mode, M_AUTO, the longitudinal speed of the motor vehicle 100 is controlled automatically, that is, determined by the management system 10 based on the customized MOD model.Thus, in the second operating mode, the calibration parameters of the longitudinal speed control module 5 are defined so as to apply the driving habits of the driver of the motor vehicle 100.

[0051] In one embodiment, in addition to the determination means 1, the automated longitudinal speed management system 10 of a motor vehicle comprises the following elements: - a human-machine interface 2 allowing the driver to manage the activation and deactivation of the automatic longitudinal speed control, - a user management system 3, - a processing unit 4 comprising a microprocessor 41, a memory 42 and communication interfaces 43.

[0052] The management system 10, and particularly the microprocessor 41, mainly comprises the following modules which cooperate with each other: - a module 410 for initial model training, this module being able to cooperate with memory 42, - a mission start module 411, this module being able to cooperate with the human-machine interface 2 and / or the memory 42, - a model update module 412, this module being able to cooperate with the determination means 1, the human-machine interface 2 and / or the memory 42, - a calibration parameter determination module 413, this module being able to cooperate with the determination means 1, the longitudinal speed control module 5 and / or the memory 42, - a 414 end-of-mission module, this module being able to cooperate with the human-machine interface 2 and / or the memory 42,

[0053] The motor vehicle 100, in particular the management system 10, preferably comprises all the hardware and / or software elements configured to implement the process defined in the object of the invention or the process described below.

[0054] The determination means 1 may include, for example, a radar, and / or a lidar, and / or a camera and / or any other type of sensor suitable for detecting targets in front of the motor vehicle 100.

[0055] The determination means 1 can provide measurements to the microprocessor 41, including: - a longitudinal distance DS between the motor vehicle 100 and the target vehicle 200, - a longitudinal speed 204 of the target vehicle 200, and - a longitudinal acceleration of the target vehicle 200.

[0056] Preferably, the analysis of the images provided by the determination means 1 can also provide data concerning the meteorological conditions that may influence the driving of the motor vehicle 100. In particular, the determination means 1 can detect the presence of rain or snow.

[0057] Thus, in one embodiment, the determination means 1 make it possible to determine at any given moment the meteorological conditions in which the motor vehicle 100 is operating as being rainy, snowy, or otherwise normal conditions. A meteorological criterion can therefore be taken into account in the construction and application of the MOD model. In an alternative embodiment not developed in this document, meteorological data and / or other environmental conditions that may modify visibility could be deduced from, for example, a windshield wiper operation indicator and / or the state of the headlights.

[0058] The determination means 1 include a means for determining the speed and acceleration of the motor vehicle 100, which can be achieved by computers using data relating to the chassis of the motor vehicle 100, data from the wheel rotation speed sensors.

[0059] In one embodiment, a human-machine interface 2 allows the driver to switch between a first-type driving mode, M_MANUAL, in which the driver manually controls the longitudinal speed of the motor vehicle 100, and a second-type driving mode, M_AUTO, in which longitudinal speed control is automatic. Various embodiments are possible for the human-machine interface 2, for example, buttons and / or a touchscreen and / or voice control.

[0060] In one embodiment, the management system 10 can receive information from a user management system 3 for the motor vehicle 100. Thus, the longitudinal speed management system 10 builds a MOD model for each user. In other words, if the motor vehicle 100 is equipped with a user management system, then personalized longitudinal speed management can be implemented for each user of the motor vehicle 100, so as to adapt the vehicle's behavior to each user.

[0061] The MOD model operating module 413 is capable of transmitting, via communication interfaces 43, driving parameter values, for example a tracking time value, to the longitudinal speed control module 5. Module 5 transmits control commands to the motor 6 or the braking system 7 so as to apply the driving parameters determined by the management system 10.

[0062] The memory 42 constitutes a recording medium readable by a computer or calculator comprising instructions which, when executed by the computer or calculator, lead it to implement a management process 10 according to an embodiment of the invention.

[0063] The memory 42 also allows the MOD model to be stored. Advantageously, the memory 42 comprises a volatile memory 421, the contents of which are regularly erased, particularly when the memory is no longer powered, and a non-volatile memory 422, the contents of which persist over time even when it is no longer powered.

[0064] The MOD model will advantageously be built over the course of the missions (or drives) of the motor vehicle 100, a mission being a driving phase delimited by - a start of mission, where the motor vehicle 100 is stopped and the driver starts the vehicle by means of, for example, a key or a contact button, - an end of mission, corresponding to a stop of the vehicle by the driver by means of, for example, a key or a contact button.

[0065] As a note, during the same mission several automatic engine shutdown phases (in particular, those generated by a "stop and start" system) may occur.

[0066] In a preferred embodiment, memory 42 allows recording At the end of the Nth mission of the vehicle 100, a model M0DN is generated, corresponding to an update of an M0DN i model recorded during the previous mission. The update of the MOD model takes into account the learning performed during the Nth mission.

[0067] In a preferred embodiment, the memory 42 comprises volatile memory 421 and non-volatile memory 422. Such an architecture allows, - on the one hand, during the execution of an Nth mission, to record in volatile memory 421 the temporary data necessary for the construction of the M0DN model, and - on the other hand, to record in non-volatile memory 422 the M0DN model generated at the end of an Nth mission, and to retain it until the end of the N+th mission.

[0068] The motor vehicle 100, in particular the management system 10, preferably comprises all the hardware and / or software elements configured so as to implement the method defined in the object of the invention or the method described below.

[0069] The longitudinal speed management method according to the invention comprises - an initial training step E100 of the MOD model, including the use of different simulated values ​​of the first set ENS1 of calibration parameters, then - a step E2 for updating the MOD model, including the use of different observed or measured values ​​of parameters from the first set ENS1 of calibration parameters and the second set ENS2 of driving parameters.

[0070] The initial training step E100 is a preparation phase of the MOD model that takes place before the motor vehicle 100 is used by an end user, i.e., before the vehicle is marketed. During the initial training phase, data collected, in particular by simulation, are fed into the model in order to increase the learning level of the MOD model. The objective of step E100 is to obtain a model suitable for use on data not present in the training data.More specifically, in step E100, the MOD model is trained to be able to determine calibration parameters of the longitudinal speed control module 5 corresponding to driving parameter values ​​not present in the training data.

[0071] Step E2 of the MOD model update is a phase of adapting, or modifying, the MOD model according to the driving habits or behaviors of a user of the motor vehicle 100. In other words, step E2 is a phase of modification of the MOD model from data recorded during so-called "manual" driving phases carried out by the user without use of the longitudinal speed control module 5.

[0072] Data recorded during manual driving phases are injected into the MOD model in order to train the MOD model to predict parameters of the longitudinal speed control module 5 which correspond to a driving dynamic as close as possible to the manual driving dynamic observed in the user of the motor vehicle 100.

[0073] An embodiment of the method for managing the longitudinal speed of a motor vehicle is described below with reference to [Fig. 3]. The method comprises five steps, a step E100 followed by steps E1 to E4.

[0074] Step E100 is executed during the calibration of the motor vehicle 100. At the end of step E100, an initial model MOD0 is obtained.

[0075] In order to construct the initial model MOD0, step E100 includes substeps E101 to E104 which are executed successively.

[0076] In a substep E101 of training data collection, data necessary for the construction of the initial model MOD0 is collected.

[0077] The E101 substep of training data collection is carried out over a given recording period, - during simulations of a motor vehicle equipped with a longitudinal speed control module 5, the vehicle moving on a virtual course, and / or - during real movements of a motor vehicle equipped with a longitudinal speed control module 5, the vehicle moving on one or more predefined road courses.

[0078] The data collected are therefore of two types: - a first type of data relating to data from simulations of a vehicle equipped with a longitudinal speed control module, and - a second type of data relating to data from actual driving of a vehicle equipped with a longitudinal speed control module.

[0079] The first and / or second type training data collection includes (i) assigning a current value to each parameter of the first set ENS1 of control module calibration parameters, and (ii) a determination of a current value of each driving parameter of the second set ENS2 by the determination means 1.

[0080] Data of the first type makes it possible to simulate a wide variety of situations in a limited time, the simulated vehicle driving on a virtual course containing different driving situations, including - driving in curves, - crossing intersections, - situations involving surrounding vehicles, for example vehicles moving at a different speed than the simulated vehicle, or a vehicle merging or stopping in front of the simulated vehicle.

[0081] During the simulations, a set of dynamic quantities implemented by the simulated vehicle are measured. The dynamic quantities are selected so as to be representative of the different dynamic behaviors of the vehicle according to the calibration that has been chosen for the longitudinal speed management system of the simulated vehicle.

[0082] For example, the longitudinal acceleration of the simulated vehicle is influenced by certain parameters of the longitudinal speed management system during straight-line driving phases where the set speed (from the longitudinal speed management system) is suddenly increased by several tens of kilometers per hour. Similarly, the measured distance between the simulated vehicle and a target vehicle is influenced by certain parameters of the longitudinal speed management system during driving phases where a target vehicle pulls in front of the simulated vehicle.

[0083] Advantageously, the data of the first type makes it possible to generate, in a short time, a dataset comprising a wide range of parameters influencing the dynamic behavior of the vehicle via the longitudinal speed management system. The simulations make it possible, for example, to produce recordings where the simulated vehicle applies strong accelerations whenever possible, while maintaining large safety distances. In another example, the simulations make it possible to produce recordings where the simulated vehicle applies more gradual accelerations but is allowed to get closer to the vehicles in front of it.

[0084] Data of the first type are recorded to feed the training database.

[0085] Data of the second type also allows data to be collected according to different configurations of the longitudinal speed management system, but the implementation of actual driving is time-consuming, which limits the number of recordings.

[0086] However, the second type of data offers a significant advantage, since, compared to the first type of data, the measurements taken are more realistic and more representative of the behavior of a real vehicle. In particular, the second type of data allows for learning under more complex conditions, but also conditions closer to those of real-world driving, where the MOD model will be used to predict driving parameters.

[0087] Substep E101 further includes, following the recording of data of the first and second type, a labeling process for relevant data.

[0088] In other words, - Following the collection of data of the first and second type, we have recordings containing a given number N of recorded signals, the recorded signals allowing us to define a calibration of the control module 5 of the longitudinal speed and driving parameters representative of a user's driving style, - Each recorded signal represents a time evolution of a calibration parameter or a driving parameter; for example, a signal represents a time evolution of the speed of the simulated vehicle, or a time evolution of the acceleration of the simulated vehicle, or a time evolution of the following distance applied by the simulated vehicle. - the labeling process associates with each record the relevant Pi calibration parameters for training the MOD0 model.

[0089] An example of driving data measured during a 100-second recording is shown in Table 2. The driving parameters are determined every 10 milliseconds. The different columns of the table contain the following data respectively: - The N_MES column contains the determination or measurement number of the driving parameters. - The ALON column contains the longitudinal acceleration of the motor vehicle 100 in m / s², - The VEGO column contains the longitudinal speed of the motor vehicle 100 in m / s, - the VCIB column contains the longitudinal speed of the target vehicle 200 in m / s, - the DCIB column contains the longitudinal distance measured between the motor vehicle 100 and the target vehicle 200 in m, - the COU column contains the torque controlled by the driver of the motor vehicle, 100 in Nm, - the FR column contains the braking controlled by the driver of the motor vehicle at 100 m / s2. - The ALAT column contains the lateral acceleration of the motor vehicle 100 in m / s², - the VLAC column contains the yaw rate of the motor vehicle 100 in rad / s2,

[0090] [Tables2] N_MES ALON VEGO VCIB DCIB COU FR ALAT VLAC 0 0 25.04 0 0 -117 0 0 0 1 0.43 25.10 0 0 -117 0 0 0 2 0.79 25.14 0 0 -117 0 0 0 3 1.04 25.18 0 0 -117 0 0 0 10992 -0.19 16.27 16.64 43.79 -26.4 0 0.44 -1.54

[0091] Then, starting from each record made in substep E101, we apply, - a segmentation of the given recording duration into a series of sub-periods, then - for each sub-period, a calculation of statistical characteristics relating to the training data recorded during the sub-period, the statistical characteristics including a mean and / or a variance and / or a maximum value, - then an integration into the MOD model of the statistical characteristics relating to the training data of each sub-period.

[0092] In other words, from each recording made in substep E101, in a substep El02 a third set ENS3 of training parameters is integrated into the MOD0 model, the third set ENS3 comprising each current value of the parameters of the first set ENS1 of calibration parameters and each current value of the parameters of the second set ENS2 of driving parameters.

[0093] To this end, firstly, statistical characteristics associated with each of the calibration parameters and each of the driving parameters are extracted. The extraction of statistical characteristics requires segmenting or sampling the temporal signals from the recordings, for example into time intervals of a duration of a few seconds.

[0094] In an example illustrated by Table 3, we thus obtain about a hundred time intervals and we calculate, on each given time interval, the following calibration parameters: - a maximum value of the vehicle's longitudinal acceleration simulated over the given time interval, - an average of the tracking time applied by the simulated vehicle over the given time interval, - a reference lateral acceleration applied over the given time interval, the reference lateral acceleration corresponding to an upper limit of lateral acceleration not to be exceeded to meet a driving comfort criterion when cornering, - a maximum value of the longitudinal acceleration applied by the simulated vehicle when it is preceded by a target vehicle.

[0095] [Tables3] Maximum longitudinal acceleration Tracking time Reference lateral acceleration Maximum acceleration with target. Int_1 0.5 1 1 1 Int_2 1 1 1 1 Int_100 1.5 4 0.8 0.2

[0096] Furthermore, in step E102, data standardization is applied. In other words, a multiplicative factor is applied to the values ​​taken by the various calibration parameters and the various driving parameters so as to bring them into an equivalent range. Thus, regardless of the order of magnitude of the numerical values ​​taken by a parameter, it will have the same weight as the other parameters in the learning process.

[0097] Then we integrate the data thus obtained into the MOD0 model, for the training of the MOD0 model.

[0098] In the described embodiment, a machine learning algorithm is used, in particular of the "vector support" type. The algorithm is based on the implementation of a series of chain regressions, each regression having the objective of determining the value of one of the parameters, from the data of a record being analyzed and a previous parameter having been determined (if it exists).

[0099] Other learning methods could be used, including - a so-called "K nearest neighbors" method, - a so-called "random forest" method, and - a method known as "recurrent neural network".

[0100] Step E102 advantageously includes an analysis of the learning curves of the MOD0 model to evaluate its learning level and thus detect underfitting or overfitting of the model.

[0101] The analysis of the learning curves is carried out by testing the MOD0 model from simulations that have not been used for training the MOD0 model.

[0102] The performance of the MOD0 model can be observed on a histogram of residual errors. The standard residual error measures the accuracy with which the MOD0 model can predict the parameter values ​​with new input data.

[0103] Figure 4 illustrates a comparative analysis of residual errors relating to three MOD, M0D2, M0D3 longitudinal speed management models, the three models being obtained from the same set ENS3 of training parameters: - the first MOD model, corresponding to the graph G1, was carried out according to the "vector support" method, - the second model M0D2, corresponding to graph G2, was created using the so-called "random forest" method, and - the third model M0D3, corresponding to the graph G3, was carried out using the so-called "K nearest neighbors" method.

[0104] The MOD model carried out according to the "vector support" method appears to minimize the value of the errors compared to other methods, the errors measured in graph G1 being more concentrated around the value 0 than in graphs G2 and G3.

[0105] The MOD0 prediction model thus obtained at the end of step E100 is ready to be applied to input data not present in the training data.

[0106] Step E100 is then followed by an iteration over steps E1 to E4, each iteration corresponding to a mission of the motor vehicle 100.

[0107] Stages E1 to E4 take place over the duration of one Nth mission of the motor vehicle 100 - Steps E1 and E4 are executed respectively at the beginning of the Nth mission, and at the end of the Nth mission, and - during the execution of the Nth mission, the process includes an alternation of iterations on one or the other of the steps E2 or E3.

[0108] In a mission start step El, the process for the vehicle's Nth mission is initialized.

[0109] The initialization of the process includes retrieving from non-volatile memory 422 a model M0DN i, which was saved in non-volatile memory 422 at the end of the Nth mission. At the beginning of the Nth mission, the model M0DN is therefore equal to the model M0DN L

[0110] Under certain conditions, the MOD0DN i model may contain the initial MOD0 model. This is the case, for example, - if N=l, that is, if it is the very first mission of motor vehicle 100, or - if the motor vehicle 100 has always been used by applying the automatic mode M_AUTO for longitudinal speed management.

[0111] Furthermore, in step 11, an initial driving mode for the first motor vehicle 100 is determined - as being of a first kind M_MANUAL if the longitudinal speed of the motor vehicle 100 is controlled at the time of initialization by a human driver, - as being of a second kind M_AUTO if the longitudinal speed of the motor vehicle 100 is controlled at the time of initialization by the longitudinal speed control module 5.

[0112] In one embodiment, the initial driving mode of the motor vehicle 100 can be determined by the value of a MODE_V variable stored in memory, in particular in volatile memory 421. The MODE_V variable could take the values ​​M_MANUAL and M_AUTO. For example, the value of the MODE_V variable could be updated in memory based on an action by the driver of the motor vehicle 100 on the human-machine interface 2.

[0113] In one embodiment, the variable MODE_V could further take a third value M_STOP when the driver terminates the current mission.

[0114] If the initial driving mode is determined to be of the first type M_MANUAL, then we proceed to the second learning step E2; if the initial driving mode is determined to be of the second type M_AUTO we proceed to the third step E3 automatic management of the longitudinal speed of the motor vehicle.

[0115] The E2 learning step comprises an iteration, at different observation times T_OBSP, of the following substeps: - a determination, at the observation time T_OBSP, of a current driving mode of the first motor vehicle 100 as being of a first kind M_MANUAL if a longitudinal speed of the first motor vehicle 100 is controlled by a human driver, otherwise as being of a second kind M_AUTO, then - if the current driving mode is of the first kind M_MANUAL, a substep E21 of recording observation data D_OBSP comprising a determination at the observation time T_OBSP of parameters of the first set ENS1 of calibration parameters and of the second set ENS2 of driving parameters from the data from the determination means 1, or - if the current driving mode is the second mode M_AUTO, a sub-step E22 of the end of observation includes an update of a management MOD model of longitudinal velocity from observation data D_OBS collected at different observation times T_OBSP.

[0116] As a note, - for p=l, that is to say during the first iteration of step E2, the determination of a common driving mode of the first motor vehicle 100 is carried out in step El, then - as explained below, for p>l the determination at the observation time T_OBSP of a current driving mode of the first motor vehicle 100 is carried out in a sub-step E22 of the p-th iteration of step E2.

[0117] Thus, at the beginning of an Nth mission of the vehicle, if the initial driving mode calculated in step El is manual, we proceed to step E2 which iterates on sub-step E21 followed by sub-step E22.

[0118] In a first sub-step E21, a current value P;c of the parameters P; of the first set ENS1 of calibration parameters and of the second set ENS2 of driving parameters is determined from the data from the determination means 1.

[0119] In a preferred embodiment, substep E21 begins with a process to verify the conditions required for recording an observation. For example, the required conditions preferably include: - detection of a target vehicle 200, - the stabilization of the longitudinal speed of the motor vehicle 100 and the longitudinal speed of the target vehicle 200, the variations of which must be less than a first given threshold, - the stabilization of a measured longitudinal distance between motor vehicle 100 and target vehicle 200, the variations of which must be less than a second given threshold.

[0120] The first threshold given can correspond to a percentage of the current speed of the motor vehicle, for example 2% of the current speed. In this case, the first threshold is approximately 2 km / h for a current speed of 110 km / h. This percentage can be calibrated to other percentage values.

[0121] The second given threshold may correspond to a percentage of the current distance between motor vehicle 100 and target vehicle 200, for example 1% of the current distance between motor vehicle 100 and target vehicle 200. In one embodiment, the second threshold may be, for example, about 6 meters for a current speed of 110 km / h.

[0122] Preferably, the required conditions must be met for a minimum duration, for example 2 seconds.

[0123] The current values ​​P;c of the parameters can then be determined from the data from the determination means 1.

[0124] Then, in substep E21, we record a pth observation D_OBSPp including the current values ​​P;c of the parameters.

[0125] We then proceed to substep E22 in which we start a timer with a duration equal to one iteration period P_ITER_MAN. In one embodiment, the iteration period of substeps E21 and E22 is determined by the processing frequency of the computer, i.e. 100 Hertz.

[0126] The timer expires at a time T_OBSp+i,=T_OBSp+P_ITER_MAN.

[0127] At time T_OBSp+i, we check - on the one hand, if the vehicle's Nth mission continues and - on the other hand, if the manual driving phase continues.

[0128] To do this, we check that the value of the variable M0DE_V is indeed M_MANUEL. If so, we loop back to the E21 recording step.

[0129] If the value of the variable M0DE_V is different from M_MANUEL, then we detect an end of the learning phase, that is to say the end of a continuous phase of iterations of substeps E21 and E22.

[0130] In this case, substep E22 includes an update of the M0DN model by integrating the observation data D_OBSP acquired during the learning phase which is ending.

[0131] In other words, as expressed by the Mathl formula, the values ​​of the parameters P; are then calculated by performing, for each parameter, a weighted average between the values ​​from the previous learning phases, denoted P;p and the values ​​from the current learning phase, denoted P;c.

[0132] Mathl Vi El, m Pi — has L c.PiC+ aip.PiP djC + CLiP Or - P; is the i-th parameter, - PiC is the current estimate of the ith parameter, - a;c is the number of measurement points for the current estimation of the ith parameter, - Rp is the previous estimate of the i-th parameter - a;p is the number of measurement points for the previous estimation of the ith parameter.

[0133] The resulting M0DN model is stored in volatile memory 421.

[0134] Following sub-step E22, the process can proceed either to an end-of-mission phase implemented in step E4, or to an application phase of the automatic longitudinal speed management model implemented in iterations of step E3: - If the value of the variable M0DE_V is M_STOP, this means that the driver has ended the Nth mission between times T_OBSP and T_OBSp+i. We therefore proceed to step E4, the end of the mission. - If the value of the variable M0DE_V is M_AUT0, this means that the driver has activated automatic longitudinal speed management. In this case, we proceed to step E3 of automatic longitudinal speed management.

[0135] Step E3 includes an iteration at different times T_AUTOq on two substeps E31 and E32.

[0136] In substep E31, the M0DN model available in memory is retrieved and used to calculate estimated values ​​of the calibration parameters which will be automatically applied by the longitudinal speed control module 5 at time T_AUTOq.

[0137] Following substep E31, we proceed to substep E32 in which we start a timer with a duration equal to one iteration period P_ITER_AUTO. In one embodiment, the iteration period of substeps E31 and E32 is determined by the processing frequency of the computer, i.e. 100 Hertz.

[0138] The P_ITER_AUTO period of iteration of substeps E31 and E32 may be different from the P_ITER_MAN period of iteration of substeps E21 and E22.

[0139] The timer expires at a time T_AUTOq +1=T_AUTOq+P_ITER_AUTO.

[0140] At time T_AUTOq+i, we then check if the Nth mission of the vehicle continues and if the automatic driving phase continues.

[0141] To do this, we check that the value of the variable M0DE_V is indeed M_AUT0. If so, we loop back to the application step E31.

[0142] If the value of the variable M0DE_V is different from M_AUT0, the process can then proceed either to an end-of-mission phase implemented in step E4, or to a learning phase implemented in step E2: - If the value of the variable M0DE_V is M_STOP, this means that the driver has ended the Nth mission between times T_AUTOq and T_AUTOq+i. We therefore proceed to step E4, the end of the mission. - If the value of the variable MODE_V is M_MANUEL, this means that the driver has selected the manual driving mode; we then proceed to the E2 learning step.

[0143] In the E4 end-of-mission step, the M0DN model available in volatile memory 421 is retrieved and saved in non-volatile memory 422.

[0144] The data previously recorded in volatile memory are erased. The M0DN model from the Nth mission (stored in non-volatile memory 422) will subsequently serve as the initial input model for step El during the execution of the N+th mission.

[0145] Figure 5 is a second flowchart of an automated management process according to the invention. In a first step 70, the mission of the motor vehicle 100 is started. In a step 71, it is tested whether the mission is completed.

[0146] If the mission is complete, we proceed to mission end step 79, which may include saving data to non-volatile memory. Then, following step 79, we loop back to step 70.

[0147] If the mission continues, we proceed to step 72 of reading from memory a longitudinal speed management model.

[0148] Then, in a step 73, the current driving mode of the motor vehicle 100 is identified.

[0149] Then, in step 74, we check if the current driving mode is manual: - If the current driving mode is manual, we proceed to step 75, which teaches the longitudinal speed management model. Then, in step 77, we test if the mission is complete. If so, we proceed to step 79, which ends the mission. If not, we loop back to step 73, which identifies the current driving mode. - If the current driving mode is automatic, we proceed to step 76, which automatically applies a longitudinal speed management model. Then, in step 78, we check if the mission is complete. If so, we proceed to step 79, which ends the mission. If not, we loop back to step 73, which identifies the current driving mode.

[0150] Finally, the longitudinal speed management method according to the invention has multiple advantages.

[0151] First, it allows the motor vehicle 100 to learn a personalized longitudinal speed management system as the driver performs various tasks. This learning takes into account different criteria likely to influence the driving parameters applied by a human driver. First, the longitudinal speed of the motor vehicle 100 is a key parameter of the model. The conditions under which the vehicle operates are also Factors taken into account, particularly weather conditions, can also be considered. Other conditions likely to influence personalized driving parameters can also be taken into account by the invention, such as ambient light or road surface conditions. In addition, or alternatively, the integration of traffic-related information could allow driving parameters to adapt to traffic density. Indeed, for example, in heavy traffic situations, a driver might tend to reduce their following time to avoid vehicles merging between their vehicle and the target vehicle.

[0152] Thus, thanks to the invention, the driver can experience the same driving sensations in an automatic longitudinal speed management mode as in a manual driving mode. Furthermore, the personalization of driving parameters can advantageously be differentiated according to different users of the motor vehicle 100. By mimicking driver behavior, the invention contributes to driver acceptance of the autonomous driving system and therefore to an increase in its frequency of use. The invention thus improves the safety of the driver and vehicle occupants.

[0153] The longitudinal speed management method according to the invention also makes it possible to simplify the human-machine interface relating to the use of an automatic longitudinal speed management system.

[0154] The longitudinal speed management method according to the invention also reduces the number of parameters requiring predefined settings before the vehicle is put into service. The cost of vehicle development is therefore reduced. Indeed, during the initial stages of the method, an expert driver could train the model by performing several driving tests so that the system can offer a user a set of the most suitable driving parameters.

[0155] Furthermore, the longitudinal speed management method according to the invention can operate solely from data from means on board the motor vehicle 100. In other words, the technical means required for the implementation of the invention are available on any vehicle whose level of autonomy is greater than or equal to 1.

[0156] Furthermore, the longitudinal speed management method according to the invention uses a highly efficient method, known as a "support vector machine," to construct a longitudinal speed management model. In addition, the model update calculations weight each new estimate of a driving parameter based on previous estimates, thus avoiding abrupt model oscillations. This embodiment makes it possible to reduce the impact of a few observations that might deviate from a trend defined by the majority of observations.

[0157] The invention could also make it possible to evaluate the driving of a vehicle user, with the aim of providing him with indicators on a level of safety or a level of energy consumption related to his driving.

Claims

Demands

1. A method for managing the longitudinal speed of a motor vehicle (100) comprising a longitudinal speed control module (5), capable of transmitting control commands to a motor (6) or a braking system (7) of the vehicle, the method defining a model (MOD) for managing the longitudinal speed of the motor vehicle (100), the model being configurable according to a first set (ENS1) of calibration parameters and according to a second set (ENS2) of driving parameters of the motor vehicle (100), the motor vehicle (100) comprising means for determining the second set of driving parameters (ENS2), characterized in that it comprises: • an initial training step (El00) of the model (MOD), comprising the use of different simulated values ​​of the first set (ENS1) of calibration parameters,then • a model update (MOD) step (E2) including the use of different observed or measured values ​​of parameters from the first set (ENS1) of calibration parameters and a second set (ENS2) of driving parameters.

2. A management method according to the preceding claim, characterized in that the initial training step (E100) comprises an iteration on the following substeps, for different values ​​of the first set (ENS1) of calibration parameters: • a substep (E101) of model training data collection (MOD) comprising (i) an assignment of a current value to each parameter of the first set (ENS1) of calibration parameters, and (ii) a determination of a current value of each driving parameter of the second set (ENS2) by the determination means (1), then • a substep (E102) of integration into the model (MOD) of a third set (ENS3) of driving parameters, the third set (ENS3) comprising the current value of the parameters of the first set of calibration parameters and the current value of the parameters of the second set of driving parameters.

3. Management method according to the preceding claim, characterized in that the substep (E101) of collecting training data is carried out over a given recording time, - during simulations of a motor vehicle equipped with a longitudinal speed control module (5), the vehicle moving on a virtual course, and / or - during real movements of a motor vehicle equipped with a longitudinal speed control module (5), the vehicle moving on one or more predefined road courses.

4. A management method according to claim 2 or 3, characterized in that the integration substep (E102) comprises, following the collection substep (E101), - a segmentation of the given recording duration into a series of sub-periods, then - for each sub-period, a calculation of statistical characteristics relating to the training data recorded during the sub-period, the statistical characteristics comprising a mean and / or a variance and / or a maximum value, then - an integration into the model (MOD) of the statistical characteristics relating to the training data of each sub-period.

5. A management method according to any one of the preceding claims, characterized in that the model update step (E2) comprises the following substeps: • a determination, at an observation time (T_OBSP), of a current driving mode of the first motor vehicle (100) as being of a first kind (M_MANUAL) if a longitudinal speed of the first motor vehicle (100) is controlled by a human driver, otherwise as being of a second kind (M_AUTO), then • if the current driving mode is of the first kind (M_MANUAL), a substep (E21) of recording observation data (D_OBSP) comprising a determination at the observation time (T_OBSP) of parameters of the first set (ENS1) of calibration parameters and the second set (ENS2) of driving parameters, from the data from the determination means (1), or • otherwise,a sub-step (E22) at the end of observation comprising an update of the model (MOD) for managing the longitudinal speed of the motor vehicle (100) based on observation data (D_OBSP) collected during the different observation times (T_OBSP).

6. Management method according to the preceding claim, characterized in that it comprises an iteration of a step (E3) of automatic management of the longitudinal speed of the motor vehicle (100) by application of the model (MOD) as a function of at least one parameter of the second set (ENS2) of driving parameters determined by the determination means (1), at an instant (T_AUTOq) of iteration of the step (E3) of automatic management of the longitudinal speed of the motor vehicle (100).

7. A management method according to claim 5 or 6, characterized in that the iterations of step (E2) of updating the model and / or step (E3) of automatically managing the longitudinal speed are carried out at a frequency of a computer on which the process is executed, for example at a frequency of 100 Hertz.

8. Management method according to any one of the preceding claims, characterized in that it comprises automatic learning of the longitudinal speed management model (MOD), according to a method chosen from - a so-called "Vector Support Machine" method, - a so-called "K nearest neighbors" method, - a so-called "random forest" method, and - a so-called "recurrent neural network" method.

9. Management method according to any one of the preceding claims, characterized in that the first set (ENS1) of calibration parameters comprises - a maximum value of longitudinal acceleration applicable when the motor vehicle (100) is preceded by another vehicle, and / or - a maximum value of longitudinal acceleration applicable when the motor vehicle (100) is not preceded by any vehicle, and / or - a setpoint value for following distance of a vehicle located in front of the motor vehicle (100), and / or - a maximum value of lateral acceleration in cornering, and / or - a maximum value of longitudinal deceleration.

10. A management method according to any one of the preceding claims, characterized in that the second set (ENS2) of driving parameters comprises at least one parameter taken from: - a longitudinal acceleration of the motor vehicle (100), - a longitudinal speed of the motor vehicle (100), - a lateral acceleration of the motor vehicle (100), - a lateral speed of the motor vehicle (100), - a yaw rate of the motor vehicle (100), - a distance between the motor vehicle (100) and a vehicle in front of it, - a speed of a vehicle in front of the motor vehicle (100), - an acceleration of a vehicle in front of the motor vehicle (100), - a steering angle applied by a driver of the motor vehicle (100), - a motor torque request applied by the motor vehicle (100), - a braking torque request applied by the motor vehicle (100).

11. A management method according to any one of the preceding claims, characterized in that the method comprises a driving phase delimited by - a start time of the driving phase, where an engine of the motor vehicle (100) is started by means of a key or an ignition button by a driver, and - a stop time of the driving phase, where an engine of the motor vehicle (100) is stopped by means of a key or an ignition button by a driver, in that, between the start time of the driving phase and the stop time of the driving phase, a model for managing the longitudinal speed of the motor vehicle is recorded in a volatile memory (421) of the motor vehicle, and in that at the stop time of the driving phase, a model for managing the longitudinal speed of the motor vehicle is recorded in a non-volatile memory (422) of the motor vehicle.

12. Device (10) for managing the longitudinal speed of a first motor vehicle (100), the motor vehicle being equipped with - a processing unit (4) comprising a microprocessor (41), a memory (42) and interfaces (43), - a longitudinal speed control module (5), - a motor (6) and - a braking system (7), the device comprising hardware and / or software elements (1, 2, 3, 4, 41, 42, 43, 411, 412, 413, 414, 421, 422) implementing the method according to any one of claims 1 to 11.

13. Product computer program comprising program code instructions recorded on a computer-readable medium to implement the steps of the process according to any one of claims 1 to 11 when said program is run on a computer.

14. A computer-readable data storage medium on which a computer program comprising program code instructions for implementing the process according to any one of claims 1 to 11.

Images