Method for adjusting an on-board vehicle system based on a prediction of an adjustment action on the on-board system.

Abstract

The present invention relates to a method and device for adjusting a vehicle's embedded system based on a prediction of an adjustment action on the embedded system. The method comprises the following steps: - obtaining (32) the prediction of the adjustment action on the embedded system from a prediction model fed as input a first data point representing an interior and / or exterior context of a vehicle's passenger compartment; - obtaining (34) a second data point representing a vehicle occupant's reaction to the presentation of information representing the prediction of the adjustment action on the embedded system; - adjusting (35) the prediction model based on the prediction of the adjustment action and the second data point; and - adjusting (36) the embedded system based on the prediction of the adjustment action and the second data point. Figure for the abstract: Figure 3

Description

Title of the invention: Method for adjusting an embedded vehicle system based on a prediction of an adjustment action on the embedded system. technical field

[0001] The present invention relates to the adjustment of an embedded vehicle system based on a prediction of an adjustment action on the embedded system. Technological background

[0002] Contemporary vehicles are equipped with driver assistance systems, known as AD AS (from the English "Advanced Driver-Assistance System" or in French "Système d'aide à la conduite avancé").

[0003] Some of these ADAS systems assist the driver by automatically controlling, for example, the vehicle's speed or trajectory. Speed ​​is controlled via a cruise control system, for example, an ACC (Adaptive Cruise Control) system, based on a set speed and a set following distance. The vehicle's trajectory is controlled, for example, via a lane-keeping system or a semi-automatic lane-change assist system.Maintaining a vehicle in its lane is achieved through a lane keeping assist system (also called a lane position assist system), known for example as LPA (Lane Position Assist), LKA (Lane Keeping Assist), or LCA (Lane Centering Assist). Such a system keeps the vehicle within its lane, for example, in the center of the lane, by recognizing the lane markings and setting a torque threshold that the driver must exceed to change the vehicle's trajectory and override the LKA system.Semi-automatic lane change assistance is provided by a semi-automatic lane change system, known as SALC (Semi-Automatic Lane Change). The primary function of a SALC system is to assist the driver when they wish to change lanes. It detects the activation of the turn signals on one side of the vehicle to indicate the driver's intention to change lanes from the current lane to the target lane on the side where the turn signals are activated. Once activated by the driver, the SALC system initiates the lane change after performing a few checks. There are also ADAS systems that detect and recognize traffic signs, known as TSR (Traffic Sign Recognition) systems. These systems use onboard cameras to capture images of the front of the vehicle. These images are analyzed by an onboard computer to detect pictograms corresponding to signs defined by the highway code. The TSR system can then transmit data that can influence the operation of other ADAS systems. For example, if a speed limit sign or a stop sign is detected, the data transmitted by the TSR system can instruct the ACC system to reduce the set speed. There are also ADAS systems that detect whether the driver's hands are on the steering wheel.These hand-on-steering-wheel detection systems are called HOD systems (Head-Off Detection Sensing System). There are also driver behavior monitoring systems such as attention or facial emotion prediction systems that predict driver behavior, for example, by analyzing images of the driver's face taken from cameras positioned inside the vehicle.

[0004] Contemporary vehicles also include so-called comfort systems such as road navigation systems, ventilation control, vehicle cabin air conditioning or infotainment.

[0005] ADAS and comfort systems include means for interacting with the interior and / or exterior environment of the vehicle's passenger compartment. For example, modern vehicles are equipped with sensors to provide ADAS systems with information about the exterior environment of the vehicle's passenger compartment, such as cameras, radar, or microphones, etc. They may also be equipped with sensors to provide comfort systems with information about the interior environment of the vehicle's passenger compartment, such as temperature / humidity sensors or sensors to determine the number of occupants in the vehicle, etc.

[0006] Contemporary vehicles also include means of restitution so that a vehicle occupant can become aware of information provided by an AD AS and / or comfort system.

[0007] Contemporary vehicles also include means of interfacing with embedded systems so that a vehicle occupant can provide information such as choices and validations of parameter settings of these embedded systems.

[0008] The interface means may take the form of physical buttons and / or touch screens which are configured to display, via a graphical Human-Machine Interface (HMI), the operating status of the systems embedded in the vehicle, possible choices of embedded system parameters or predictions of embedded system adjustment actions.

[0009] However, the embedded systems of contemporary vehicles are not automated and require the driver of the vehicle (or one of the occupants of the vehicle) to initiate an action from the human-machine interface means in order for these embedded systems to start up and for an action to adjust these embedded systems to be validated. Summary of the present invention

[0010] One object of the present invention is to solve at least one of the problems of the technological background described above.

[0011] Another object of the present invention is to automate vehicle embedded systems.

[0012] According to a first aspect, the present invention relates to a method for adjusting an embedded vehicle system based on a prediction of an adjustment action on the embedded system, the method being implemented by at least one computer embedded in the vehicle, the method comprising the following steps: - obtaining initial data representative of an internal and / or external context of a vehicle's passenger compartment; - obtaining the prediction of the adjustment action on the embedded system from a prediction model fed as input by the first data; - control of a means of providing information on board the vehicle to provide in the vehicle's passenger compartment information representative of the prediction of the adjustment action on the onboard system; - obtaining a second data point representing a reaction from a vehicle occupant to the rendering of information representing the prediction of the adjustment action on the on-board system; - adjusting the prediction model based on the prediction of the adjustment action and the second data point; and - setting up the embedded system based on the prediction of the setting action and the second piece of data.

[0013] Such a method allows an action to adjust an on-board system (ADAS or comfort) to be predicted according to the internal and / or external context of the vehicle's passenger compartment and offered to the vehicle's occupants even if this on-board system has not been initiated (launched / validated) beforehand by a vehicle occupant.

[0014] The process makes it possible to adapt the comfort of the vehicle's occupants according to the evolution of the interior and / or exterior conditions of the vehicle's passenger compartment during a journey.

[0015] The method also makes it possible to personalize the predictions of adjustment actions based on the decisions made by the vehicle occupants during past journeys in order to predict future on-board system adjustment actions that are better suited to the vehicle occupants during future journeys.

[0016] According to a particular and non-limiting embodiment of the present invention, the internal context is defined by at least one data point obtained from at least one sensor on board the vehicle.

[0017] This embodiment is advantageous because it allows for the prediction and proposal to the vehicle occupants of on-board system adjustment actions based on changes in driving scenarios and / or road traffic or weather conditions, for example.

[0018] According to a particular and non-limiting embodiment of the present invention, the internal context is defined by a data representation of information obtained from a driver profile of the vehicle recorded in a memory of the vehicle.

[0019] This embodiment is advantageous because the method can predict on-board system adjustment actions that are customized according to the vehicle driver's adjustment preferences.

[0020] According to a particular and non-limiting embodiment of the present invention, the external context is defined by at least one piece of data obtained by at least one sensor on board the vehicle.

[0021] According to a particular and non-limiting embodiment of the present invention, the method further comprises a step of recording in a memory dedicated to memorizing a driver profile, a third data representative of the adjustment of the prediction model from the prediction of the adjustment action and the second data.

[0022] This embodiment allows the prediction model to be adjusted locally, i.e. from on-board resources of the vehicle, thus avoiding any transmission of personal data of the driver.

[0023] According to a particular and non-limiting embodiment of the present invention, the model for predicting the adjustment action of the vehicle's on-board system comprises a pre-trained neural network, and the adjustment of the prediction model from the prediction of the adjustment action and the second data point is implemented by reinforcement learning.

[0024] This example is advantageous because it allows for adjustment of the prediction model based on the driver's decisions regarding the proposed adjustment action predictions, using a reinforcement learning method whose implementation Implementation requires computing and memory resources that correspond to those that a contemporary vehicle can provide.

[0025] According to a second aspect, the present invention relates to a device for adjusting an embedded vehicle system based on a prediction of an adjustment action on the embedded system, the device comprising a memory associated with a processor configured for implementing the steps of the process according to the first aspect of the present invention.

[0026] According to a third aspect, the present invention relates to a vehicle, for example of the automobile type, comprising a device as described above according to the second aspect of the present invention.

[0027] According to a fourth aspect, the present invention relates to a computer program which includes instructions adapted for carrying out the steps of the process according to the first aspect of the present invention, in particular when the computer program is executed by at least one processor.

[0028] Such a computer program may use any programming language, and be in the form of source code, object code, or an intermediate form between source code and object code, such as in a partially compiled form, or in any other desirable form.

[0029] According to a fifth aspect, the present invention relates to a computer-readable recording medium on which is recorded a computer program comprising instructions for carrying out the steps of the process according to the first aspect of the present invention.

[0030] On the one hand, the recording medium can be any entity or device capable of storing the program. For example, the medium can include a storage means, such as a ROM, RAM, CD-ROM or a microelectronic circuit-type ROM, or a magnetic recording means or a hard disk drive.

[0031] On the other hand, this recording medium can also be a transmissible medium such as an electrical or optical signal, such a signal being able to be transmitted via an electrical or optical cable, by conventional or radio frequency, by self-directing laser beam, or by other means. The computer program according to the present invention can, in particular, be downloaded from an Internet-type network.

[0032] Alternatively, the recording medium may be an integrated circuit in which the computer program is incorporated, the integrated circuit being adapted to execute or to be used in the execution of the process in question. Brief description of the figures

[0033] Other features and advantages of the present invention will become apparent from the description of the particular and non-limiting embodiments of the present invention below, with reference to the attached Figures 1 to 5, in which:

[0034] [Fig-1] schematically illustrates part of a vehicle's passenger compartment, according to a a particular embodiment of the present invention.

[0035] [Fig.2] schematically illustrates a field of vision associated with a driving position of the vehicle of [Fig.1], according to a particular and non-limiting embodiment of the present invention.

[0036] [Fig.3] illustrates a flowchart of the different steps of a method for adjusting an embedded vehicle system from a prediction of an adjustment action on the embedded system, according to a particular and non-limiting embodiment of the present invention.

[0037] [Fig.4] schematically illustrates a principle of reinforcement learning of a neural network.

[0038] [Fig.5] schematically illustrates a device configured to adjust an on-board vehicle system based on a prediction of an adjustment action on the on-board system, according to a particular and non-limiting embodiment of the present invention. Description of examples of achievements

[0039] A method and a device for adjusting an on-board vehicle system based on a prediction of an adjustment action on the on-board system will now be described in what follows with joint reference to Figures 1 to 5. The same elements are identified with the same reference signs throughout the following description.

[0040] The terms "first," "second" (or "firsts," "seconds"), etc., are used in this document by arbitrary convention to allow for the identification and distinction of different elements (such as operations, means, etc.) implemented in the embodiments described below. Such elements may be distinct or correspond to a single element, depending on the embodiment.

[0041] Fig. 1 presents a view of the interior of the front part of the passenger compartment of a vehicle 10 including in particular a steering wheel, a dashboard, and a windscreen 100.

[0042] Vehicle 10 corresponds, for example, to a vehicle with an internal combustion engine, with electric motor(s), or to a hybrid vehicle with an internal combustion engine and one or more electric motors. Vehicle 1 thus corresponds, for example, to a land vehicle, such as a car, a truck, a bus, a tram, a train, etc.

[0043] The vehicle 10 includes means for implementing embedded systems such as AD AS systems and / or comfort systems.

[0044] These embedded systems include or are associated with vehicle-mounted sensors. This sensor or these sensors are configured to obtain data that represent an interior and / or exterior context of the vehicle's passenger compartment 10 and to provide this data to the vehicle's embedded systems 10 if necessary.

[0045] This sensor or these sensors correspond, for example, to one or more of the following sensors: - one or more millimeter-wave radars arranged on the vehicle 10, for example at the front, at the rear, on each front / rear corner of the vehicle; each radar is adapted to emit electromagnetic waves and to receive the echoes of these waves reflected by one or more objects, in order to detect obstacles and their distances from the vehicle; and / or - one or more LIDAR(s) (Light Detection and Ranging), a LIDAR sensor corresponding to an optoelectronic system composed of a laser emitter, a receiver including a light collector (to collect the portion of the light emitted by the emitter and reflected by any object located in the path of the light rays emitted by the emitter) and a photodetector that transforms the collected light into an electrical signal; a LIDAR sensor thus makes it possible to detect the presence of objects located in the emitted light beam and to measure the distance between the sensor and each detected object; and / or - one or more cameras (with or without a depth sensor) for acquiring one or more images of the environment around the vehicle within the field of vision of the camera(s); - one or more temperature and / or humidity sensors that provide temperature and / or humidity data inside or outside the vehicle's passenger compartment 10;

[0046] - sensors that allow the number of occupants of the vehicle to be obtained; - speed sensors, vehicle acceleration 10; - sensors that are configured to obtain data representative of a driving scenario (urban, semi-urban, etc.); - sensors that are configured to obtain data representative of road traffic and environmental conditions (weather forecasts and / or current weather conditions around the vehicle 10 for example); - sensors for measuring the perspiration of the driver of vehicle 10 which can be based on image analysis of the driver of vehicle 10; - haptic actuators positioned for example at the steering wheel or driver's seat. - a sensor for recognizing facial emotions of the driver;

[0047] The vehicle 10 may also be equipped with a navigation system, which uses, for example, geolocation information provided by a satellite positioning system such as the GPS (Global Positioning System) or the Galileo system. This geolocation information is combined with mapping data, particularly road data, to display the route on a map generated from the mapping data. The vehicle 10 may then be equipped with road surface condition sensors that are linked to this navigation system to provide data representative of the condition of the road surface on which the vehicle 10 is traveling or will soon be traveling.

[0048] The vehicle 10 may also have an on-board voice recognition sensor which includes at least one microphone positioned in the vehicle's passenger compartment and a computing unit configured to analyze the human voice captured by said at least one microphone.

[0049] The data obtained by the vehicle's on-board sensors 10 can be fused together by an on-board data fusion system of the vehicle 10 so that this fused data represents a current state of the interior and / or exterior context of the vehicle's passenger compartment.

[0050] The present invention is not limited to a particular type of embedded system or to a particular type of sensor but extends to any embedded and adjustable system of the vehicle 10 from an adjustment action defined from data which define an interior and / or exterior context of the passenger compartment of the vehicle 10.

[0051] For example, if the present invention is implemented to adjust speed setpoint settings for an ACC system, the fusion system may require data from speed sensors and chase state sensors.

[0052] According to another example, if the present invention is implemented to adjust temperature setpoint settings for an air conditioning system, the fusion system may require data from temperature and / or humidity sensors located inside the vehicle's passenger compartment 10 but also those obtained from environmental condition sensors and driver perspiration measurement sensors of the vehicle.

[0053] The vehicle 10 also includes means of restitution so that an occupant of the vehicle 10 can become aware of information provided by an AD AS and / or comfort system.

[0054] These means of reproduction can be loudspeakers for reproducing audio feedback, haptic actuators, or even one or more screens. Among these screens, it is known to incorporate one or more LCD screens (from the English "Liquid Crystal Display"), for example of the TFT type (from the English "Thin-Film Transistor" or in French "Transistor en film mince"), or OLED (from the English "Organic Light-Emitting Diode" or in French "Diode électroluminescente organique").

[0055] These means of displaying information can also take the form of a head-up display that allows useful driving information to be displayed at the driver's eye level, for example on a transparent or semi-transparent strip mounted on the dashboard behind the steering wheel. Such a head-up display is described, for example, in document WO2013 / 189808 A published on December 27, 2013.

[0056] Contemporary vehicles also include means of interfacing with embedded systems so that a vehicle occupant can provide information such as choices and validations of parameter settings of these embedded systems.

[0057] The interfacing means may take the form of a speech recognition system coupled with a microphone.

[0058] The interface means can take the form of physical buttons and / or touch screens which are configured to display, via a graphical Human-Machine Interface (HMI), the operating status of the systems embedded in the vehicle, possible choices of embedded system parameters or predictions of action for setting embedded systems.

[0059] The interfacing means may also take the form of a display system including a control device for the head-up display device.

[0060] The head-up display device is configured to display floating images, each comprising one or more graphic objects, by projection, onto a surface 101 of the windshield 100, of at least two light beams, each carrying data representative of a floating image.

[0061] At least one floating image advantageously comprises a set of graphic objects, each representing a parameter or operating state of the vehicle 10. By way of example, graphic objects representing parameters or operating states of the vehicle 10 correspond to all or part of the following information, in any possible combination: - information representing instantaneous speed; and / or - information representing the rotation of a drive shaft (also called tachometer information); and / or - information representing the distance traveled; and / or - information representing the fuel level or the battery charge level in the case of an electric vehicle; and / or - information representative of the temperature of an engine coolant; and / or - at least one piece of information representative of a warning indicator (for example, battery charge indicator, engine oil pressure indicator, engine oil or coolant temperature indicator, brake failure indicator, etc.), corresponding, for example, to a pictogram that appears or changes color in the event of a warning; and / or - at least one piece of information representative of a warning indicator (for example, engine oil level warning light, airbag warning light (also called inflatable cushion), brake pad wear indicator, etc.), corresponding for example to a pictogram that appears or changes color when a warning is issued; and / or - at least one piece of information representative of an indicator of operation of an on-board system (positioning light indicator, dipped or main beam indicator, hazard warning light indicator, rear window demister indicator, etc.), corresponding for example to a pictogram that appears or takes on a specific color when the on-board system is put into operation.

[0062] Of course, the above list of information displayed via a floating image is given by way of example and is not limited to the examples above. Furthermore, the information displayed corresponds, for example, to the information usually displayed on the dashboard (or instrument cluster).

[0063] At least one other floating image advantageously comprises a set of graphic objects, each representative of the external context of the vehicle 10. By way of example, contextual or environmental information of the vehicle 10 corresponds to all or part of the following information, according to all possible combinations: - representative mapping information of the vehicle's environment 10, for example obtained from the vehicle's on-board navigation system or from a navigation system installed on a mobile communication device (for example, a smartphone connected wirelessly to the vehicle 10; and / or - representative vehicle navigation information, for example a route to be followed by the vehicle, the vehicle's current position, the vehicle's instantaneous speed, the speed limit applicable to the section of road on which the vehicle is traveling 10; this information is obtained, for example, from the vehicle's on-board navigation system or from the navigation system installed on a mobile communication device; and / or - information representative of an object detected in the environment, for example the presence of a vehicle preceding vehicle 10, the presence of a pedestrian or an animal on the road in front of vehicle 10, the presence of a road sign, the presence of a stationary object on the road, the presence of a tunnel entrance, etc.; this information is obtained, for example, from one or more object detection sensors on board vehicle 10 and, for example, associated with one or more AD AS systems of vehicle 10;according to one variant, this information is received from another vehicle or infrastructure connected to vehicle 10 via wireless communication using a vehicle-to-everything (V2X) communication mode, including vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I), and / or vehicle-to-pedestrian (V2P) communication modes; and / or - information representative of an event detected in the environment, for example, information on a disruption on the road, such as an accident, a traffic jam, information on particular weather conditions that may disrupt traffic (snow, fog, rain, ice);This information is, for example, obtained from one or more object detection sensors on board vehicle 10 and, for example, associated with one or more AD AS systems of vehicle 10; according to one variant, this information is received from another vehicle or from the infrastructure connected to vehicle 10 via wireless communication using a vehicle-to-everything (V2X) communication mode, including vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I), and / or vehicle-to-pedestrian (V2P) communication modes; according to another variant, this information is received from one or more servers in the cloud via a 4G or 5G cellular wireless network.

[0064] Figure 2 schematically illustrates a field of vision associated with a position of driving vehicle 10 of [Fig.1], according to a particular and non-limiting embodiment of the present invention.

[0065] Figure 2 illustrates the front part of vehicle 10 from a top view. The field The vision point 210 is associated with a viewpoint 21 which corresponds to the viewpoint from which a driver is supposed to look at the road ahead when driving the vehicle 10. This viewpoint 21 is, for example, positioned in the middle of the driver's seat (front left in the direction of travel of the vehicle 10 according to the particular example in [Fig. 2]), at the location of the driver's head if the driver were seated in the seat driver. Viewpoint 21 is, for example, positioned in the center of a headrest resting on the driver's seat when the seat is equipped with such a headrest.

[0066] The field of vision 210 extends for example around a main vision axis 211, and covers an area between an upper vertical limit (forming for example an angle of 15° with the axis 211 from the viewpoint 21), a lower vertical limit (forming for example an angle of 15° with the axis 211 from the viewpoint 21), a right lateral limit (forming for example an angle of 20 or 25° with the axis 211 from the viewpoint 21) and a left lateral limit (forming for example an angle of 25 or 30° with the axis 211 from the viewpoint 21).

[0067] A graphic object 2100 of a floating image is advantageously projected or displayed within the field of vision 210, to ensure that the driver is looking at the road or the environment in front of the vehicle 10 according to this field of vision 210.

[0068] A floating image comprising the graphic object is advantageously collimated on an image plane 212 perpendicular to the axis 211 and positioned at a particular distance from the viewpoint 21 along the axis 211, such a distance being advantageously greater than or equal to 3.5 m, for example equal to 3.5, 4 or 4.5 m.

[0069] The graphic object 2100 depends on the type of information projected onto the surface 101. When the purpose of displaying a floating image is to provide a prediction of an action to adjust an on-board system to the driver of the vehicle 10, the graphic object can be displayed centered with respect to the axis 211. The graphic object 2100 corresponds to a virtual element added by embedding or superimposing this virtual element onto an element of the real world.

[0070] Figure 3 illustrates a flowchart of the different steps of a process for adjusting an on-board vehicle system based on a prediction of an adjustment action on the on-board system, according to a particular and non-limiting embodiment of the present invention.

[0071] In a step 31, a first data point 311 is obtained. The first data point 301 is representative of an interior context and / or an exterior context of a passenger compartment of the vehicle 10.

[0072] According to a particular and non-limiting embodiment of the present invention, the internal context is defined by a data representation of information obtained from a driver profile of the vehicle recorded in a memory of the vehicle 10.

[0073] This embodiment is particularly advantageous in the case of infotainment systems. Indeed, as soon as a driver of vehicle 10 enters vehicle 10, the first data point 311 can be representative of that driver's identity. The driver's identity can, for example, be obtained from a facial recognition sensor that obtains the driver's identity by The driver's face in vehicle 10 is compared from an image acquired by an onboard camera in vehicle 10 with a set of pre-recorded faces corresponding to pre-recorded driver profiles in vehicle 10. If the driver's identity in vehicle 10 matches one of the pre-recorded identities, then parameters of one or more onboard systems in vehicle 10 can be adjusted according to information from the associated driver profile. For example, the driver's seat can be adjusted by a driver's seat adjustment system according to a pre-recorded driver's seat position in the driver profile. A sound level or playlist can be retrieved from the driver profile and played back according to the sound level, and so on.

[0074] According to a particular and non-limiting embodiment of the present invention, the internal context is defined by at least one data point obtained from at least one sensor on board the vehicle 10.

[0075] According to a particular and non-limiting embodiment of the present invention, the external context is defined by at least one data point obtained from at least one sensor on board the vehicle 10.

[0076] The first data 311 representing the internal and / or external context can be obtained from a data or even a combination of data obtained from on-board sensors of the vehicle 10. For example, the first data 311 representing an external context of the vehicle 10 defined by data obtained from road condition sensors configured to determine the condition of the road on which the vehicle 10 is traveling. The first data 311 can also be representative of weather conditions or even representative of a driving scenario, that is to say representative of the activation state of ADAS or comfort systems and / or their current parameters usually used by a driver of the vehicle 10.According to another example, the first data point 311 could be representative of the driver's perspiration level from a perspiration sensor based on images of the driver captured by cameras, or it could be representative of a facial emotion of the driver obtained from a driver facial emotion recognition sensor.

[0077] In a step 32, a prediction of an adjustment action on an on-board system of the vehicle 10 is obtained from a prediction model fed as input by the first data 311.

[0078] According to one example, the prediction of the adjustment action of the embedded system can be obtained by an embedded system of the vehicle 10 implementing the prediction model.

[0079] According to another example, the prediction of the adjustment action of the embedded system can be obtained by a system dismounted from the vehicle 10 and received by the vehicle 10 via an embedded communication system of the vehicle 10.

[0080] According to a particular and non-limiting embodiment of the present invention, the prediction model is a neural network trained from a learning phase.

[0081] The neural network comprises a set of layers of artificial neurons.

[0082] For example, each artificial neuron is a perceptron, that is to say a linear classifier comprising generally several inputs and a single output and characterized by an activation function, weights (or synaptic coefficients) and a bias (or threshold).

[0083] For example, a perceptron with n inputs (¾ ..., x„) and a single output o is defined by n weights ( Wj, ..., M'„) and a bias (or threshold) 0:

[0084] , h O = / (z) = 1 1 0 otherwise

[0085] The output o then results from applying the Heaviside function to the postsynaptic potential z given by:

[0086]

[0087]

[0088] with a non-linear activation function H(x) given for example by: r» rrz a f0 « X< 0 V xe P, € € H ( x ) = _. . , It if x > 1

[0089] The internal parameters of the neural network are then these weights and biases for the set of perceptrons.

[0090] The present invention is not limited to this definition of perceptron or to the use of other basic elements forming a layer of the neural network. It is also not limited to the number of perceptrons (or other basic elements) used per layer or to the number of layers. The internal parameters of the neural network are usually weights and biases, regardless of the basic elements of the neural network layers.

[0091] The primary function of the neural network is to calculate the postsynaptic potential z, which is usually defined by a vector in a feature space (called the latent space). The vector z is formed of neuronal activations, which we will subsequently call features. The feature vector z is then used to associate a traffic situation class identifier with an input image presented as input to the trained neural network. The feature space has one dimension less than that of the input image space. For this reason, the neural network can be seen as comprising an encoder and a decoder. The input image presented as input to the neural network is then the input image presented as input to the encoder. The output of the encoder is then a feature vector obtained from the input image by applying operations defined by the perceptrons (or other basic elements) of the different layers of the neural network.

[0092] The neural network is pre-trained from a set of road situations in order to classify an input data (representative of an interior and / or exterior context of the passenger compartment of a vehicle) into a class of action of setting of vehicle embedded system, that is to say to associate an identifier of class of action of setting of vehicle embedded system to the input data.

[0093] For example, if several vehicle embedded system tuning action classes correspond to different input data, the neural network can provide different vehicle embedded system tuning action class identifiers, i.e. predict different tuning actions on the same embedded system or tuning actions on different embedded systems.

[0094] The decoder makes it possible to obtain an input data reconstructed from the feature vector. The decoder applies to the feature vector dual operations to those of the encoder which are implemented to obtain the feature vector from the input data.

[0095] Feedback connection weights and other biases can be determined during the training of the neural network and be part of the internal parameters of the neural network.

[0096] The neural network is also configured to provide a vehicle embedded system adjustment action class identifier to an input data reconstructed at the output of the decoder.

[0097] In a step 33, a vehicle 10 restitution means is controlled to render information representative of the prediction of the adjustment action on the system on board inside the passenger compartment of the vehicle 10.

[0098] For example, the means of displaying the information may be a head-up display device which can then be controlled to display a floating image comprising at least one graphic element representing the information to be displayed. This at least one graphic element may take the form of an icon, text, an image, etc.

[0099] According to another example, the means of rendering can be a sound rendering system which can then be controlled to emit via loudspeakers positioned inside the passenger compartment of the vehicle 10 a sound signal representative of the information to be rendered.

[0100] In a step 34, a second data 312 is obtained, the second data 312 is representative of a reaction of an occupant of the vehicle 10 to the rendering of the information representative of the prediction of the adjustment action on the on-board system.

[0101] The reaction of the occupant of the vehicle 10 is obtained from an interface means on board the vehicle 10 such as the voice recognition system coupled with a microphone positioned in the passenger compartment of the vehicle 10 or a press on a physical button or on an area of ​​a touch screen to for example choose and / or validate a proposed adjustment action prediction.

[0102] When the prediction model is implemented by a pre-trained neural network, the adjustment can be implemented by a learning method known from the prior art which would be a function of the reaction of the vehicle driver (second data 312) to the rendering of the information representative of the action prediction at the output of the prediction model.

[0103] Fundamentally, reinforcement learning is an important part of artificial intelligence, which deals with the design of agents that interact with complex environments and learn to do so without explicit human supervision. In reinforcement learning, a neural network is trained to maximize an assumed available reward function, rather than adjusting internal parameters of that neural network so that the output data of the neural network is as close as possible to ground truth data. Reinforcement learning has been used, in particular, in autonomous driving systems to predict and plan vehicle trajectories, but also to make Large Language Models (LLMs) more efficient in their interactions with users of those models.

[0104] Reinforcement learning (RL) involves learning by interacting with an environment 410 ([Fig. 4]). An agent 400 learns from the consequences of its actions 401, rather than receiving explicit instruction. The agent 400 selects its actions 401 based on past experiences (exploitation) and new choices (exploration), which essentially amounts to trial-and-error learning. The reinforcement signal that the agent 400 receives is a numerical reward signal 411, which encodes the success of the outcome of an action 401, and a model state 412, and the agent 400 seeks to learn to select actions that maximize the reward accumulated over time.

[0105] The present invention relates to a method and device for adjusting a vehicle's embedded system based on a prediction of an adjustment action on the embedded system. The method trains a pre-trained neural network that learns to predict parameter adjustments of these embedded systems based on vehicle occupants' reactions to predictions of adjustment actions of embedded systems in order to personalize the settings of the ADAS and comfort systems. vehicle 10 during future journeys of vehicle 10 if interior and / or exterior contexts of the passenger compartment of vehicle 10 are reproduced during these future journeys.

[0106] In a step 36, the embedded system is tuned from the prediction of the tuning action and the second data 312.

[0107] For example, if the second data 312 validates a prediction of an action setting of an on-board system, such as for example a prediction of temperature setpoint, then the air conditioning system is set so that the temperature inside the vehicle 10 is regulated on the basis of this temperature setpoint.

[0108] According to another example, the second data can be representative of a value chosen from a set of values ​​proposed by the rendering of the prediction of adjustment action on an embedded system.

[0109] For example, the rendering may offer several electric power charging station destinations along the route of vehicle 10. The second data may then represent one of these destinations and the navigation system then recalculates the new destination.

[0110] According to one variant, in an optional step 37, a third data representative of the adjustment of the prediction model from the prediction of the adjustment action and the second data 312 is stored in a memory dedicated to storing a conductive profile.

[0111] Using the example of choosing the charging station destination above, the third piece of data can be represented by this destination choice combined with the internal and / or external context of vehicle 10 when the prediction provided by the prediction model is rendered. This third piece of data can then be used for reinforcement learning of the neural network if the prediction model is a pre-trained neural network. The prediction model is then trained based on the driver's choices, and future predictions of onboard system adjustment actions for vehicle 10 will better reflect the driver's preferences.

[0112] Examples of scenarios will now be given for illustrative purposes to indicate to the reader that the above process can be applied to many vehicle embedded systems 10.

[0113] According to a first scenario, suppose that the driver of vehicle 10 is driving on a deformed road surface while having set a speed (from the speed control system) that is excessive relative to the road surface quality. The navigation system can then provide a current geographical position of vehicle 10, and an on-board road surface quality detection system can detect the beginning of a section of deformed road surface at this geographical position (first given 311). The process can then predict a speed setting appropriate to the road surface quality. The process monitors the road surface and may display information on the head-up display indicating road surface deformation, the predicted speed setting, and an icon confirming this new speed setting. The process can then obtain the second data point (312), for example, driver acceptance of the new speed setting, and then control the cruise control system to replace the current speed setting with the adjusted speed setting. The prediction model, in this case the pre-trained neural network, can then undergo a reinforcement learning phase to adapt to interactions with the vehicle's driver.

[0114] According to another scenario, the first data point 311 can represent vehicle onboard system settings preferences specific to the driver of vehicle 10. For example, these settings preferences can be part of a user profile stored in the vehicle 10's memory. For example, this user profile can include contacts and favorite refueling stations. This information can be used during a reinforcement learning phase of the neural network (prediction model) to adapt the predictions provided by this prediction model to the driver's preferences.

[0115] Figure 5 schematically illustrates a device 5 configured for, according to a particular and non-limiting embodiment of the present invention. The device 5 corresponds, for example, to a device embedded in the vehicle 10, for example a computer.

[0116] According to a particular embodiment, device 5 corresponds to a mobile communication device such as device 100.

[0117] The device 5 is, for example, configured to carry out the operations described opposite Figures 1, 2, and 4 and / or the steps of the process described opposite [Fig. 3]. Examples of such a device 5 include, but are not limited to, embedded electronic equipment such as a vehicle's on-board computer, an electronic control unit such as an ECU (Electronic Control Unit), a smartphone, a tablet, or a laptop computer. The elements of the device 5, individually or in combination, may be integrated into a single integrated circuit, into several integrated circuits, and / or into discrete components. The device 5 may be implemented in the form of electronic circuits or software (or computer) modules, or a combination of electronic circuits and software modules.

[0118] The device 5 includes one (or more) processor(s) 50 configured to execute instructions for carrying out the steps of the process and / or for executing instructions from the software embedded in the device 5. The processor 50 may include integrated memory, an input / output interface, and various circuits known to a person skilled in the art. The device 5 further includes at least one memory 51 corresponding for example to volatile and / or non-volatile memory and / or includes a memory storage device which may include volatile and / or non-volatile memory, such as EEPROM, ROM, PROM, RAM, DRAM, SRAM, flash, magnetic or optical disk.

[0119] The computer code of the embedded software(s) including the instructions to be loaded and executed by the processor is for example stored on memory 51.

[0120] According to various particular and non-limiting embodiments, the device 5 is coupled in communication with other similar devices or systems and / or with communication devices, for example a TCU (Telematic Control Unit), for example via a communication bus or through dedicated input / output ports.

[0121] According to a particular and non-limiting embodiment, the device 5 includes a block 52 of interface elements for communicating with external devices, for example a remote server or the cloud, other nodes of the ad hoc network. The interface elements of the block 52 include one or more of the following interfaces: - radio frequency RF interface, for example of the Wi-Fi® type (according to IEEE 802.11), for example in the 2.4 or 5 GHz frequency bands, or of the Bluetooth® type (according to IEEE 802.15.1), in the 2.4 GHz frequency band, or of the Sigfox type using UBN (Ultra Narrow Band) radio technology, or LoRa in the 868 MHz frequency band, LTE (Long-Term Evolution), LTE-Advanced; - USB interface (from the English "Universal Serial Bus" or "Universal Serial Bus" in French); - HDMI interface (from the English "High Definition Multimedia Interface", or "High Definition Multimedia Interface" in French); - LIN interface (from the English "Local Interconnect Network", or in French "Réseau interconnecté local").

[0122] According to another particular and non-limiting embodiment, the device 5 includes a communication interface 53 which enables communication with other devices (such as other computers in the embedded system) via a communication channel 54. The communication interface 53 corresponds, for example, to a transmitter configured to transmit and receive information and / or data via the communication channel 54. The communication interface 53 corresponds, for example, to a CAN (Controller, Access, and Communication) wired network. Area Network » or in French « Controller Network">), CAN FD (from the English « Controller Area Network Flexible Data-Rate» or in French « Controller Area Network Flexible Data Rate">), FlexRay (standardized by ISO 17458) or Ethernet (standardized by ISO / IEC 802-3).

[0123] According to a particular and non-limiting embodiment, the device 5 can provide output signals to one or more external devices, such as a display screen 55, touch or not, one or more speakers 56 and / or other peripherals 57 (projection system, microphones, etc.) via output interfaces 58, 59, 60 respectively. According to a variant, one or more of the external devices is integrated into the device 5.

[0124] Of course, the present invention is not limited to the embodiments described above but extends to a method for adjusting an on-board vehicle system based on a prediction of an adjustment action on the on-board system, which would include secondary steps without falling outside the scope of the present invention. The same would apply to a device configured for implementing such a method.

[0125] The present invention also relates to a vehicle, for example an automobile or more generally an autonomous land-powered vehicle, comprising the device 5 of [Fig.5].

Claims

Demands

1. A method for adjusting a vehicle's embedded system based on a prediction of an adjustment action on the embedded system, the method being implemented by at least one computer embedded in the vehicle, the method comprising the following steps: - obtaining (31) a first data item representing an interior and / or exterior context of a vehicle's passenger compartment; - obtaining (32) the prediction of the adjustment action on the embedded system from a prediction model fed as input by the first data item; - controlling (33) a means for displaying information embedded in the vehicle to provide information in the vehicle's passenger compartment that represents the prediction of the adjustment action on the embedded system; - obtaining (34) a second data item representing a reaction of a vehicle occupant to the display of information representing the prediction of the adjustment action on the embedded system;- adjustment (35) of the prediction model from the prediction of the adjustment action and the second data point; and - adjustment (36) of the embedded system from the prediction of the adjustment action and the second data point.

2. A method according to claim 1, wherein the internal context is defined by at least one data point obtained from at least one sensor on board the vehicle.

3. Method according to claim 1, wherein the internal context is defined by a data representation of information obtained from a vehicle driver profile stored in a vehicle memory.

4. Method according to claim 1, wherein the external context is defined by at least one data point obtained by at least one sensor on board the vehicle.

5. A method according to claim 1, further comprising a step (37) of recording in a memory dedicated to storing a driver profile, a third data item representing the adjustment of the prediction model from the prediction of the adjustment action and the second data item.

6. A method according to claim 1, wherein the model for predicting the adjustment action of the vehicle's on-board system comprises a pre-trained neural network, and the adjustment of the prediction model from the prediction of the adjustment action and the second data point is implemented by reinforcement learning.

7. Device (3) for adjusting an on-board vehicle system based on a prediction of an adjustment action on the on-board system, said device (3) comprising a memory (31) associated with at least one processor (30) configured for carrying out the steps of the method according to any one of claims 1 to 6.

8. Computer program comprising instructions for carrying out the method according to any one of claims 1 to 6, when such instructions are executed by a processor.

9. Computer-readable recording medium on which is recorded a computer program comprising instructions for carrying out the steps of the process according to any one of claims 1 to 6.

10. Vehicle comprising a device according to claim 7.

Images