PROCEDURE FOR CREATING ROAD GUIDANCE INSTRUCTIONS

Description

[0001] The present invention relates generally to guidance systems, and in particular (but not only) to guidance systems integrated into motor vehicles.

[0002] It relates more particularly to a method for developing voice guidance instructions for an individual, comprising the steps of: determination by a computer of a route to be taken, acquisition of at least one image of the individual's environment, processing of said image to detect at least one object present in the environment and to characterize said object, and establishment and transmission of a voice guidance instruction providing the individual with information to carry out a maneuver in order to follow said route.

[0003] It also relates to a device for generating voice guidance instructions.

[0004] It applies preferentially to cases where the individual is the driver of a motor vehicle. STATE OF THE ART

[0005] Satellite navigation and positioning systems are particularly effective and appreciated for determining a route to be taken, especially for a vehicle driver.

[0006] The advantage over a paper map is considerable, especially because the individual's position is updated in real time. Most navigation systems display directions on maps that are schematic representations of the vehicle's surroundings.

[0007] Typically, such a navigation system is designed to provide the driver with route guidance instructions in two ways: visually through a display screen and vocally through the vehicle's speakers.

[0008] The visual representation of the maneuver to be performed on the display screen is not always clear because the maps alter the distances and angles between road axes. It is sometimes difficult for the driver to fully understand the maneuver to be performed.

[0009] The instructions of navigation systems through their visual representation and the associated voice instructions are not always consistent and give rise to misunderstandings on the part of the driver with regard to the maneuver to be carried out and the perception of the road infrastructure, its configuration and the driving environment.

[0010] In particular, we observe that currently, voice guidance instructions are generated systematically, like a road map, based on the distance between the vehicle and the next maneuver to be carried out.

[0011] It therefore often happens that these instructions are poorly adapted to the environment, repetitive and predictable, or even out of step with what a driver would expect.

[0012] Another disadvantage of known navigation systems is that instructions are generated independently of each other. The instructions are independent of each other. The content of each instruction does not take into account the information given in previous instructions and the next instructions to come depending on the maneuvers that the driver must perform.

[0013] These drawbacks lead, in practice, a majority of users to deactivate the voice function of the navigation system.

[0014] Document FR3078565 relates to a method for developing route guidance instructions adaptive to the environment, comprising steps of: - determining a route between a starting position and a desired destination, from a cartographic database; - acquiring an image in the individual's direction of vision; - searching for at least one point of interest visible on the acquired image; - evaluating the visibility of the point of interest on the acquired image by means of a visibility level; - generating a guidance instruction highlighting a direction to take in relation to the point of interest; and - transmitting the guidance instruction to the individual. PRESENTATION OF THE INVENTION

[0015] The solution is given by the features of the independent claim. Variations are described by the features of the dependent claims.

[0016] In order to overcome the aforementioned drawbacks of the state of the art, the present invention proposes a method for developing voice guidance instructions as defined in the introduction, in which a step of determining a level of complexity of said maneuver is provided, and in which, in the establishment step, said voice guidance instruction is formulated using an index which is deduced from the characterization of said object, only if the level of complexity of said maneuver is greater than a first threshold.

[0017] In other words, the invention proposes to generate more or less elaborate voice guidance instructions, depending on the complexity of the situation encountered.

[0018] For example, if the driver is faced with a simple situation such as a crossroads, and there is no object obstructing the driver's view of the intersection, the voice guidance instruction is developed without adding any particular detail.

[0019] On the other hand, in a more complex situation, for example because an object (such as a truck) hides part of the intersection, or because the intersection itself is more complex, the voice guidance instruction is elaborated by adding clues allowing the driver to orient himself. These can be visual clues (for example the truck, a sign, a road infrastructure) or other types of clues, such as for example contextual clues (in order for example to announce to the driver when he must turn).

[0020] The invention thus proposes to generate voice guidance instructions as close as possible to the way a human co-pilot would use them. In this way, the ease of understanding of these instructions is increased and the frequency of these instructions is adapted, which makes it possible to reduce the cognitive load required of the driver, and which reduces errors and hesitations by the driver.

[0021] Other advantageous and non-limiting characteristics of the method according to the invention, taken individually or in all technically possible combinations, are the following: said index is a visual index immediately or shortly visible to the individual, or a contextual index informing the individual about when he must carry out said maneuver; said voice guidance instruction being structured by affixing at least one action and said index; the level of complexity is calculated according to at least one of the following criteria: the position of the vehicle relative to the area where said maneuver must be carried out, the type of road infrastructure present in the area where said maneuver must be carried out, the tension felt by the individual, the amount of road traffic, the number of objects detected on at least part of the acquired image, the individual's knowledge of the road infrastructure, the phase in which the vehicle is located;if the road infrastructure is a roundabout, the level of complexity is calculated based on at least one of the following criteria: the angle between an axis of arrival of the individual on the roundabout and an axis of exit of the individual from the roundabout, the angle between at least two roads leading to the roundabout, the number of roads leading to the roundabout, the difference between the number of entries to the roundabout and the number of exits from the roundabout, the number of lanes present on the roundabout, and the external diameter of the roundabout; said index being a visual index visible to the individual, if, between the moment at which the voice guidance instruction was issued vocally and the moment at which the maneuver must be carried out, the visual index disappears from the environment visible to the individual, a step of generating a new voice guidance instruction not using said visual index is provided;said index being a visual index visible to the individual, if, between the time at which the voice guidance instruction was issued vocally and the time at which the maneuver must be carried out, another visual index indistinguishable from the visual index used in the voice guidance instruction appears in the environment visible to the individual, a step is provided for generating a new voice guidance instruction different from the voice guidance instruction which was issued vocally according to the invention, if the level of complexity of said maneuver is greater than a second threshold strictly greater than said first threshold, said voice guidance instruction is generated using two indices; it is provided to determine a cost relating to an additional duration or distance that would be caused by an error in following said path, and the level of complexity is determined as a function of said cost;it is intended to distinguish at least two distinct and successive phases prior to carrying out said maneuver, including a phase of anticipation of the maneuver during which a voice guidance instruction can be issued to indicate to the individual how to position himself to approach the maneuver and a description phase during which a voice guidance instruction is issued to indicate to the individual how to carry out the maneuver, and in which the voice guidance instructions issued during these phases use distinct indices; during the anticipation phase, it is intended to state a voice guidance instruction if the road taken or to be taken contains several lanes; the index used during the anticipation phase in the voice guidance instruction refers to signs only if the individual is outside a built-up area;at least one other phase is further provided which follows the description phase and during which a voice guidance instruction can be developed, said other phase being an accompaniment phase simultaneous with said maneuver, or a reassurance phase subsequent to said maneuver and making it possible to indicate to the individual whether he has correctly carried out the maneuver, or a continuity phase making it possible to inform the individual about the active nature of said development method; if the voice guidance instruction issued during the description phase did not refer to any visual cue and the vehicle slows down, provision is made during the accompaniment phase to issue a voice guidance instruction using an index; during the reassurance phase, provision is made to issue a voice guidance instruction using a visual cue distinct from the visual cue(s) used in the voice guidance instruction(s) issued during the preceding phases;and during the continuity phase, it is planned to develop a different voice guidance instruction depending on whether the radius of curvature of the route taken is less than or greater than a threshold. ;

[0022] According to one aspect of the invention, the step of determining by a computer a route to be taken comprises retrieving from a cartographic database static visual markers located on said route and attributes associated with said static visual markers and determining an attribute common to several static visual markers. The voice guidance instruction is formulated in the establishment step using an index comprising the common attribute when the object is characterized as comprising the common attribute in the image processing step.

[0023] According to one aspect of the invention, the visual static markers are road signs, the common attribute comprises a name that is inscribed on a road sign or a background color of a road sign, the object is a road sign.

[0024] According to one aspect of the invention, the step of processing said image comprises the detection of several objects, the development method further comprises a step of classifying the detected objects, an object characterized as having the common attribute being classified as a priority, the object selected for the establishment step being the object classified as a priority.

[0025] The invention also proposes a device for generating road guidance instructions adapted to the environment, comprising: a means for acquiring images of the environment, a cartographic database, a navigation and geolocation system, a processing module which is adapted to receive and combine elements coming from the acquisition means, the cartographic database and the navigation and geolocation system, and which is programmed to implement a method as mentioned above, and a voice transmitter adapted to transmit an indication to the individual.

[0026] Of course, the various features, variants and embodiments of the invention may be combined with each other in various combinations to the extent that they are not incompatible or mutually exclusive. DETAILED DESCRIPTION OF THE INVENTION

[0027] The description which follows with reference to the appended drawings, given as non-limiting examples, will make it clear what the invention consists of and how it can be implemented.

[0028] On the attached drawings: [ Fig. 1 ] is a schematic view of a front part of a motor vehicle equipped with a device for generating route guidance instructions according to the invention.

[0029] The present invention may preferably be used in any motor vehicle (car, truck, motorcycle, etc.).

[0030] On the figure 1 , the passenger compartment of a motor vehicle 1 equipped with a device for generating road guidance instructions making it possible to implement the invention is schematically represented.

[0031] This device comprises means for acquiring images of the vehicle's environment, a guidance unit 5 and an interface 11 adapted to transmit guidance instructions to the driver of the motor vehicle 1.

[0032] The image acquisition means comprise, for example, a camera 10 facing the front of the vehicle, which makes it possible to acquire images of the environment seen by the driver when the latter has his gaze turned towards the road.

[0033] The interface 11 comprises at least one system for transmitting an audio signal 12. In addition, it here comprises a touch screen 14. Alternatively, it could be without a screen.

[0034] The audio signal transmission system 12 is designed to transmit voice instructions to the individual. It is arranged in the passenger compartment, at a distance from the individual such that the latter can clearly hear the communicated voice instructions. For example, these are the speakers (or "loudspeakers") installed in the vehicle for listening to, for example, the radio or the loudspeaker of a portable system such as a telephone or an external navigation device.

[0035] The guidance unit 5 comprises a cartographic database 20, a navigation and geolocation system 22, a computer memory (hereinafter called “memory 24”) and a computer (hereinafter called “processing module 25”).

[0036] The cartographic database 20 includes two-dimensional or even three-dimensional map data, and in particular the locations (longitude and latitude) of the road axes in a given space (for example in Europe). The cartographic database 20 also includes additional information such as the shape of buildings, the location of service stations and characteristic places in the environment (for example historical monuments or rivers), the road surface (paved, earth, tarred, etc.). The cartographic data is designed to be able to be read by the processing module 25.

[0037] Advantageously, the cartographic database 20 includes static visual markers such as road signs and attributes associated with each road sign such as the background color of the sign, the direction or the name written thereon.

[0038] The navigation and geolocation system 22 is adapted to determine the current position of the motor vehicle in space and to determine a route between the current position (or starting position) of the motor vehicle 1 and the desired destination, taking into account the data stored in the cartographic database 20. The navigation and geolocation system 22 may for example be a satellite navigation and positioning system (or “GPS” according to the commonly used acronym). The current position of the vehicle and the route are designed to be able to be read by the processing module 25.

[0039] The route determined by the navigation and geolocation system 22 is stored in the memory 24. The memory 24 also stores parameters such as threshold values ​​useful for implementing the invention. All the elements stored in the memory 24 are capable of being read by the processing module 25. These elements will be detailed later in this presentation.

[0040] The processing module 25 is programmed to receive and combine the various elements coming from the camera 10, the map database 20, the navigation and geolocation system 22 and the memory 24. The processing module 25 is also programmed to transmit data to the speakers 12 and to the screen 14 in order to communicate guidance instructions to the individual. In practice, this processing module 25 is formed by a microcontroller.

[0041] The guidance instruction generation device described above makes it possible to implement the method described below. Step S1: Determining the route

[0042] This process begins when the motor vehicle 1 is started. Then, the processing module 25 determines, during a first step, the route that the motor vehicle 1 must take.

[0043] For this purpose, the current (starting) position is determined using the navigation and geolocation system 22. Alternatively, the individual can specify a starting position different from the position at which he is currently located by manually entering the target starting position. For example, the individual can record the target starting position via the screen 14.

[0044] The individual then specifies a desired destination. For example, the desired destination can be entered on screen 14.

[0045] The processing module 25 combines the starting position, the desired destination and the map data corresponding to the geographical area concerned. The processing module 25 transmits this information to the navigation and geolocation system 22 which then generates a route.

[0046] The determined path is stored in memory 24.

[0047] Advantageously, the processing module 25 retrieves from the cartographic database 20, road signs which are located on said route, as well as attributes associated with said signs. As will be better explained below, the processing module 25 is able to determine whether several signs have a common attribute, identify a common attribute and the signs comprising said common attribute.

[0048] The route is made up of road sections. A road section is, for example, defined as a part of a road axis extending between two successive intersections.

[0049] Each section of road can then be characterized by multiple parameters, such as for example the number of lanes it has, the authorized direction(s) of traffic, the type of intersection, the number of incoming and outgoing lanes at each intersection, etc.

[0050] The route passes through a large number of waypoints such as: road infrastructure (an interchange, a roundabout, a crossing, etc.), municipalities or localities, significant places (pharmacy, bakery, shopping center, fountain, bus shelter), remarkable places (museum, monument, cathedral, train station, airport, etc.).

[0051] At this stage, before describing the rest of the process, we can define other concepts which will be useful for understanding this presentation.

[0052] The first concept is that of "action." This is a unitary movement to be carried out by the vehicle (for example, "move into the right lane," "follow the red car").

[0053] The concept of "target" refers to the area where the motor vehicle must be positioned. This may include road infrastructure elements (traffic lights, stop signs, direction signs, road axes, etc.).

[0054] The concept of "maneuver" refers to a sequence of one or more actions to achieve a "purpose." A maneuver is performed to adapt the position of the motor vehicle in relation to the road infrastructure and to achieve the purpose. For the same maneuver, several pairs of an action and a target can be concatenated. For example, a maneuver can be stated as follows: "Move into the left lane, then turn left onto highway X."

[0055] The concept of "clue" covers any element of language that can be used in a voice guidance instruction, and which, when used, is defined on the basis of the images acquired by the camera in order to facilitate the understanding of the situation by the driver of the vehicle.

[0056] The concept of “visual index” (or “visual reference”) designates static elements (road axes, signs, road infrastructures, and more generally any element listed in the cartographic database 20) or dynamic elements (vehicles, pedestrians or more generally any element detectable by the processing module 25 on the images acquired by the camera 10), which can be used to describe an action to be carried out or when to carry it out. These elements make it possible to enrich a guidance instruction either by adding a detail (for example “Turn right after the traffic light”) or by describing it in relation to this element (for example “Go along the blue building”).

[0057] The concept of a "contextual cue" (or "temporal cue") refers to when the maneuver should take place. This information indicates the time the driver has to complete the maneuver. For example, such a cue could be stated as follows: "now," "right away," "as soon as possible," "at the roundabout," "at the traffic light."

[0058] The concept of "attribute", for a visual cue or target, designates a characteristic of this visual cue or target, allowing it to be differentiated among other markers. This can be a color, a text, a direction, a type of road axis, a lane width, a type of surface, etc.

[0059] We can consider two categories of attributes: egocentric attributes, i.e. relative to the position of the motor vehicle 1 (location of the target as a function of the position of the vehicle, positioning of the different elements surrounding the vehicle, number of the exit of a roundabout taking into account the position of the vehicle, positioning to adopt on the road, etc.), exocentric attributes, i.e. positioned absolutely (location of a visual index as a function of one or more fixed reference points in the environment, which therefore does not vary with the position of the vehicle, for example a number of a road on a multi-lane road, etc.).

[0060] A target will generally be formed by an egocentric attribute, while a visual cue will generally be formed by an exocentric attribute.

[0061] The concept of "familiarity criterion" refers to language elements that reinforce the human aspect of a guidance instruction formulation. For example, this could be formulations such as "Stay in your lane", "First, we will go towards", "Then, turn left". This concept also refers to language elements with a safety function (for example, "Pay attention to the turn").

[0062] The concept of "geographical reference" will designate the position of a visual clue in relation to a target, and can be expressed as follows: "at the end of", "just after", "between", "completely to the right", "slightly to the right", "opposite", "straight ahead", "in the direction of"... Step S2: Video acquisition

[0063] Once the vehicle has started, the camera 10 acquires images of the vehicle's surroundings and transmits them to the processing module 25. It is possible to envisage that all the images acquired by the camera 10 are processed by the processing module 25, or only part of them, at a frequency higher than one Hertz.

[0064] The processing module 25 is programmed to process these images so as to develop voice guidance instructions which enable the driver to know, in particular, which maneuver he must perform next. It is programmed to develop at least part of these instructions taking into account the environment visible to the driver.

[0065] When the vehicle follows the route defined by the driver, the processing module 25 follows the progress of this vehicle using the navigation and geolocation system 22. This allows it to predict, upstream of each intersection, whether the vehicle must perform a maneuver and, if so, which maneuver it must perform.

[0066] Each maneuver that the vehicle must perform is then broken down by the processing module 25 into five successive phases during each of which a voice guidance instruction can, if necessary, be developed and then issued.

[0067] The first phase is an anticipation phase that tells the driver how to position his vehicle on the road in order to approach the next event (intersection, roundabout, etc.) in the best possible way.

[0068] The second phase is a description phase allowing the driver to be informed of the maneuver that he will have to carry out, so that he can understand it before having to execute it.

[0069] The third phase is a support phase dedicated to assisting the driver during the maneuver in order to build their confidence.

[0070] The fourth phase is a reassurance phase allowing the driver to confirm that he has carried out the correct maneuver.

[0071] The fifth phase, which occurs when two maneuvers are separated from each other by a distance or duration greater than a predefined threshold, is a reassurance phase which allows the driver to be shown that the processing module 25 is still active (after a prolonged absence of guidance instruction).

[0072] A guidance instruction can then be developed during each of these five phases.

[0073] The generation of these guidance instructions is carried out sequentially, in several successive steps which generally consist of detecting static and moving objects visible on the images, identifying the roads and the route on each acquired image, positioning the detected objects on a map from the cartographic database 20, associating the detected static objects with the known static objects from the cartographic database 20, and applying statistical and / or continuity reasoning on the objects to develop the voice guidance instruction.

[0074] In this presentation, we will first describe in detail how the guidance instructions are developed. In a second part, we will specify how these guidance instructions can be reinforced by the use of environmental objects detected on the acquired images. Finally, in a third part, we will indicate how these instructions are precisely generated according to the phase in which the vehicle is (anticipation, description, accompaniment, etc.). Step S3: Object detection

[0075] The step of detecting moving and static objects on the image acquired by the camera 10 and the attributes characterizing each of these objects is carried out by the processing module 25, using existing artificial intelligence technologies by learning.

[0076] It should be noted that an object will be defined here as an element of the environment which is potentially or actually visible to the driver of the vehicle.

[0077] The list of attributes depends on the type of object detected.

[0078] Thus, as an illustrative example, if the object is of the type “4-wheeled motor vehicle”, the attributes are as follows: category (car, van, van, truck, bus, etc.), main body color, direction of travel, side seen by the driver (front, rear, left side, right side), state (parked, stopped, moving), orientation of the object relative to the vehicle, percentage of visibility.

[0079] The detected object and each attribute have a confidence index expressed as a percentage. The closer the value of this confidence index is to 100%, the more the object and attribute can be used. Below a predetermined threshold, the attribute or object is no longer taken into account. The more attributes there are for the same object, the more this object can be recognized at each scene analysis step (i.e. on each processed image). Step S4: Positioning the objects

[0080] To identify the object or objects to be used in formulating the voice guidance instructions, the processing module 25 matches the objects detected in the processed image with the data stored in the map database 20, thereby making it possible to understand the relationship of each object to the environment and the driver's perception.

[0081] To do this, the processing module 25 uses the data stored in the cartographic database 20, and in particular the objects referenced therein (point of interest, road sign, traffic light, bus stop, etc.) and their descriptions (name, type, color, etc.).

[0082] Then, the processing module 25 identifies the roads and the route on the processed image.

[0083] Then, the processing module 25 positions each detected object on the map by matching the referenced objects with the detected objects. It also positions the detected and non-referenced objects (vehicle, pedestrian, etc.) on the map. Step S6: classification

[0084] At this stage, the processing module 25 classifies all detected objects.

[0085] A first classification technique consists of using existing artificial intelligence technologies through learning.

[0086] This technique could involve training the artificial intelligence to identify at least ten detected objects on its own, focusing on the right area of ​​the image if the vehicle is to turn right, the central area if it is to continue straight ahead, and the left area if it is to turn left, and then ranking these ten objects in order of importance.

[0087] A second classification technique consists of a statistical calculation to assign a weight to each detected object.

[0088] This weight can be calculated based on criteria from the following list: the persistence of the object in the scene, image after image (the longer the object is visible, the more its weight increases, if the presence of the object alternates between visible and non-visible, the weight decreases), the surface occupied by the object in the scene and its percentage of visibility (a truck is more visible than a car), the salience of the object (the more its color stands out from the rest of the environment, the higher its weight; for example, a red vehicle is more visible than a gray vehicle), the uniqueness of the object (the use of a single blue vehicle is less ambiguous than the use of a red vehicle among other red or orange vehicles), the position of the object in relation to the maneuver to be performed (the closer it is, the higher its weight), the phase in which the vehicle is located (as will be detailed later in this presentation, depending on the phase, classes of objects are more suitable for guidance than others),

[0089] A third technique relates to a strategy of continuity of visual clues along a path.

[0090] An object is classified as a priority for use during the preparation of the instruction of step S7 described below, if an attribute of the object has been identified by the processing module 25 as being an attribute common to several objects or an index already used during the preparation of a previous guidance instruction preceding the current route and / or potentially usable for the preparation of a subsequent guidance instruction on this same route.

[0091] Advantageously, the common attribute has been identified by the processing module 25 when determining the path S1.

[0092] For example, when determining the route S1, the processing module 25 identifies several road signs on which the name “Versailles” is written. The common attribute identified is the name “Versailles” written thereon. During the video acquisition step S2, the camera 10 acquires an image comprising a sign. During the object detection step S3, the processing module 25 identifies the sign on the image and detects that it is characterized by the common attribute previously identified. The detection of the attribute can be carried out by deciphering in the image the name written on the sign or by correspondence via geolocation, with a sign listed in the cartographic database 20, said sign being associated with attributes also listed in the cartographic database 20.During classification step S5, the processing module 25 classifies said panel as a priority so that it can be used for the development of the next guidance in step S7.

[0093] Thus, if on the route, there is a sign at several consecutive intersections whose inscription includes the same name, the guidance instructions relating to these intersections will favor the use of the same visual clue, ensuring continuity of visual clues guaranteeing consistency and facilitating user guidance while reducing their cognitive load.

[0094] For example, on a route including a first roundabout followed by an intersection and a road on which you must stay on the main lane, you have the following succession of guidance instructions: "At the roundabout, turn right towards Versailles", "Continue straight ahead towards the Versailles sign", "Stay in your lane always towards Versailles". Step S7: preparation of the instruction

[0095] The guidance instruction is then developed, possibly using an index that is adapted to the context and the maneuver to be performed and which depends on the attributes of the selected object.

[0096] One goal is to describe when and in what context an object should be used to develop the voice guidance instruction, so as to best help the driver to orient himself.

[0097] The context depends in particular on the complexity of the maneuver to be carried out. This complexity is estimated here based on at least one of the following criteria, and preferably based on each of them: the position of the vehicle in relation to the maneuver (if a lane change is necessary, the situation is considered more complex), the configuration of the road infrastructure (the more lanes an intersection has, the more complex it is considered), the driving context (driver tension, i.e. stress; heavy traffic, etc.), the presence of a large number of objects in the driver's field of vision, the driver's knowledge of the road infrastructure, the phase in which the vehicle is located, and the statistical weight of the different visual cues detected.

[0098] Here, preferably, a visual cue is only used in the voice guidance instruction if this proves necessary.

[0099] If the situation is simple, that is, if the level of complexity is below a first complexity threshold, we will prefer not to use them so as not to increase the driver's cognitive load. Thus, we will prefer to say "Turn left at this intersection" rather than "Turn left behind the bus stop", because the latter instruction would require the driver to locate the bus stop and the intersection rather than the intersection alone.

[0100] On the other hand, if the level of complexity is higher than this first threshold of complexity, the instructions are developed using temporal or visual cues (or "references"), which improves the understanding of the maneuver to be carried out. This is called "reinforcement of the voice guidance instruction."

[0101] This reinforcement can be carried out in various ways, depending on different parameters.

[0102] One way to reinforce a voice guidance instruction is to include a notion of time in the instruction to facilitate understanding of the action to be performed. This can be illustrated with several examples using contextual clues.

[0103] In a first example, if the traffic is heavy or if an object prevents the maneuver or hides the purpose of the maneuver, the processing module 25 is programmed to integrate a contextual clue of the type “as soon as possible” in the middle of the guidance instruction, so that the driver feels free to reposition the vehicle at the best time.

[0104] In a second example, if the maneuver can be carried out and must be immediate, the processing module 25 is programmed to integrate another contextual clue of the type “now” or “immediately”, so that the driver repositions his vehicle without waiting.

[0105] In a third example, if the maneuver can be performed without needing to be immediate, the processing module 25 is programmed to integrate a contextual clue of the type “as soon as possible”, so that the driver feels free to reposition his vehicle at the best time.

[0106] Each guidance instruction is then structured by combining at least one action, one contextual clue and one attribute.

[0107] In the case of a guidance instruction concatenated with another guidance instruction, the processing module 25 is programmed to replace the contextual clue "as soon as possible" with "then" or "after".

[0108] In this case, the complete guidance instruction is structured by concatenating an action and an egocentric attribute, concatenated with an action, a contextual clue, and another attribute.

[0109] For example, this complete guidance instruction could be expressed as "Turn right then, then position yourself to the left", or "Turn right then position yourself to the left", or "Turn right then follow the direction of Versailles".

[0110] A second way to reinforce a voice guidance instruction is to provide a moving landmark to facilitate understanding of the action to be performed. This method is used when the object detected in the acquired image that has the greatest weight is a moving landmark.

[0111] We can consider that the processing module 25 has managed to: detecting the different types of moving landmarks (cars, trucks, buses, vans, bicycles, motorcycles), positioning the moving landmarks on the lanes of the map from the cartographic database 20, detecting the attributes of these moving landmarks to differentiate them from each other, identifying the single moving landmark that will be seen by the driver without ambiguity by characterizing it at least by its type, its color, its position, and identifying that the single moving landmark is stable (i.e. that it appears on the different images successively processed.

[0112] In this case, the processing module 25 is programmed to use this unique moving marker to help the driver correctly position his vehicle.

[0113] The complete guidance instruction is then structured by combining an action, a geographic marker, a visual clue and an attribute.

[0114] For example, this guidance instruction might be expressed as "Move into the same lane as the red car."

[0115] A third way to reinforce a voice guidance instruction is to provide a static landmark such as a road infrastructure to facilitate understanding of the action to be performed. This method is used when the object detected in the acquired image that has the greatest weight is a static landmark.

[0116] For this, we can consider that the processing module 25 has managed to: detecting the different types of static markers (bridges, traffic lights, etc.), associating the detected static markers with the objects stored in the cartographic database 20, detecting the attributes of these static markers to differentiate them from each other, identifying the unique static marker that will be seen by the driver without ambiguity by characterizing it at least by its type, its color, its position, and identifying that the marker is stable.

[0117] In this case, the processing module 25 is programmed to use this unique static marker to help the driver correctly position his vehicle.

[0118] The complete guidance instruction is then structured by combining an action, a target, a geographic landmark and a visual clue.

[0119] For example, this complete guidance instruction might be expressed as "Take the exit just after the bridge."

[0120] A fourth way to reinforce a voice guidance instruction is to use a directional sign to facilitate understanding of the action to be performed. This method is used when the object detected in the acquired image that has the greatest weight is a sign.

[0121] For this, we can consider that the processing module 25 has managed to: detect the different types of signs with their attributes (traffic lights, stop, etc.), associate the detected signs with the signs listed in the cartographic database 20 in particular with the aim of determining the attributes of each sign (color, text, direction indicated, etc.), thus determining the attributes of each sign to differentiate them, identify the unique sign that will be seen by the driver without ambiguity (thanks to its color, the text written on it, etc.), identify that the sign is stable (and in particular that the text or shape displayed cannot be modified in the short term).

[0122] In this case, the processing module 25 is programmed to use this unique panel with its attributes to help the driver correctly position his vehicle.

[0123] The complete guidance instruction is then structured by combining an action, a geographical marker, an egocentric attribute and a purpose.

[0124] For example, this complete guidance instruction might be expressed as "Move into the left lane toward the green sign."

[0125] A fifth way to reinforce a voice guidance instruction is to use a significant location (restaurant, pharmacy, etc.) to facilitate understanding of the action to be performed. This method is used when the object detected in the acquired image that has the greatest weight is a significant location.

[0126] For this, we can consider that the processing module 25 has managed to: detecting the different types of static landmarks forming significant locations, associating the detected static landmarks with the landmarks listed in the cartographic database 20, detecting the attributes of these landmarks to differentiate them from each other, identifying the unique landmark that will be best seen by the driver without ambiguity given its attributes (landmark type, color and position), and identifying that the landmark is stable.

[0127] In this case, the processing module 25 is programmed to use this significant location with its attributes to help the driver correctly position his vehicle.

[0128] The complete guidance instruction is then structured by combining an action, an egocentric attribute, a geographic landmark, and a visual cue.

[0129] For example, this guidance instruction might be expressed as "Turn right just after the restaurant."

[0130] A sixth way to reinforce a voice guidance instruction is to use a notable place (a known significant building whose function will not change in the coming years, such as a monument, a cathedral, etc.) to facilitate understanding of the action to be performed. This method is used when the object detected in the acquired image that has the greatest weight is a notable place.

[0131] In this case, the processing module 25 is programmed to use the name of this remarkable place, without adding any attribute, to help the driver position his vehicle correctly.

[0132] The complete guidance instruction is then structured by combining an action, an egocentric attribute, a geographic landmark, and a visual cue.

[0133] For example, this complete guidance instruction might be expressed as "Turn right just after the Eiffel Tower."

[0134] A seventh way to reinforce a voice guidance instruction is to use two significant and / or notable locations to facilitate understanding of the action to be performed. This method is used when the detected object in the acquired image that has the greatest weight is a notable or significant location, and one of the other detected objects is also a notable or significant location.

[0135] For this, we can consider that the processing module 25 has managed to: detect the different types of static markers forming significant or remarkable places, associate the detected static markers with the markers listed in the cartographic database 20, detect the attributes of these markers to differentiate them from each other, identify that the target of the maneuver is located between 2 detected static markers which will be seen by the driver without ambiguity, identify that these markers are stable.

[0136] In this case, the processing module 25 is programmed to use these remarkable or significant places with their attributes to help the driver correctly position his vehicle.

[0137] The complete guidance instruction is then structured by combining an action, an egocentric attribute, a geographic marker, a first visual cue and a second visual cue.

[0138] For example, this complete guidance instruction might be expressed as "Turn right between the church and the pharmacy."

[0139] An eighth way to reinforce a voice guidance instruction is to use road features to facilitate understanding of the action to be performed.

[0140] In the cartographic database 20, information is in fact stored on the characteristics of the road axes which can be used within the framework of the present invention.

[0141] We can then consider that the processing module 25 has managed to detect an attribute differentiating the road axis to be taken from the other road axes visible to the driver.

[0142] In this case, the processing module 25 is programmed to use this attribute to help the driver correctly position his vehicle.

[0143] The complete guidance instruction is then structured by combining an action, an egocentric attribute and a target, or even an attribute of the target.

[0144] For example, this complete guidance instruction might be expressed as "Turn right into the alley" or "Turn right into the small street" or "Turn right into the cobblestone street."

[0145] A ninth way to reinforce a vocal guidance instruction is used when reinforcement of the guidance instruction was advocated in the anticipation or description phases but no visual cue was then available, whereas, during the accompaniment phase, a visual cue becomes available.

[0146] In this case, the processing module 25 is programmed to use this visual cue to generate a reinforced voice guidance instruction as soon as possible.

[0147] This visual cue can be used while the driver does not yet see it (in the description or even anticipation phase), or as soon as he sees it. In particular, it is possible to envisage that the processing module 25 orders a voice guidance instruction to be repeated by completing it using this visual cue.

[0148] A tenth way to reinforce a voice guidance instruction is used depending on the complexity of the maneuver to be performed and the level of hesitation detected.

[0149] Thus, when the complexity of the maneuver is judged to be very high (greater than a second predetermined complexity threshold) or when the driver's level of hesitation exceeds a predetermined threshold, it is planned to combine several types of reinforcement.

[0150] The level of driver hesitation can be assessed in various ways known elsewhere, for example based on the trajectory taken (which deviates from the planned trajectory), steering wheel oscillations, the way the steering wheel is held, the driver's position, etc.

[0151] A first example of association consists of using in the voice guidance instruction a reinforcement using a notion of time with any other type of reinforcement.

[0152] A second example of association is to use reinforcement using a significant location or a notable place with reinforcement using the characteristics of the route to be taken.

[0153] In this example, the complete guidance instruction is then structured by combining an action, an egocentric attribute, a geographical marker, a visual cue, a target and a target attribute. This complete guidance instruction can then be expressed in the form "Turn right just after the Eiffel Tower into the small alley".

[0154] A third example of association is to use reinforcement that uses directional signs with reinforcement that uses a significant location or notable place.

[0155] A fourth example of association is to use two reinforcers that use a significant location or a notable location.

[0156] In this example, the complete guidance instruction is then structured by combining an action, an egocentric attribute, a visual cue, a geographical marker, and another visual cue. This complete guidance instruction can be expressed in the form "Turn right after the Eiffel Tower between the church and the pharmacy."

[0157] A fourth example of association is to employ reinforcement that uses a moving landmark with reinforcement that uses a fixed landmark or road infrastructure.

[0158] In this example, the complete guidance instruction is then structured by combining an action, an egocentric attribute, a first visual cue, another visual cue, and an attribute of the cue. This complete guidance instruction can be expressed as "Turn left after the light like the red car."

[0159] It has been taught above that in a low complexity situation, the voice guidance instruction is not reinforced.

[0160] It was also taught that the level of complexity of a maneuver depended on several variables.

[0161] Preferably, it will be possible to predict that the level of complexity is also a function of another parameter, which is the cost that would be caused, in terms of time or distance, by a driver error (if he took a wrong exit).

[0162] Thus, when the processing module 25 predicts that if an error were made by the driver during the maneuver, this error would add a lot of time to the remaining route to be covered, it is programmed to increase the value of the complexity level, which results in the reinforcement of the voice guidance instruction.

[0163] We therefore understand that in this scenario, it is also planned to resort more frequently, even systematically, to the combination of several reinforcements, as detailed above.

[0164] The processing module 25 must in fact guarantee the driver the success of the maneuver if the failure to carry out this maneuver would have a significant impact on the estimated arrival time at the destination. This situation arises in particular on motorways, or on roads where there are few entrances and exits, or on one-way roads.

[0165] In practice, the processing module 25 is programmed to calculate the additional time (or distance) required to reach the destination if the driver were to take the wrong lane during the maneuver to be performed. Beyond a predetermined threshold (expressed in terms of time or distance or percentage of time remaining before arrival at the destination), the processing module 25 will then generate a more reinforced voice guidance instruction than if the calculated time were below this threshold.

[0166] Preferably, the processing module 25 may further use a firmer intonation than usual to speak the voice guidance instruction so that the driver understands the importance of performing the correct maneuver.

[0167] The cost of not carrying out the maneuver (and therefore the choice to generate a more reinforced voice guidance instruction) can be calculated also based on other parameters (type of journey, type of road, etc.) and can be increased in clearly identified cases, for example in cases where: the vehicle is not correctly positioned to carry out the maneuver, the vehicle is correctly positioned in relation to the maneuver to be carried out but the indicators are not activated, the driver has not applied the instruction before the end of the anticipation phase or before the end of the description phase.

[0168] In these cases, the processing module 25 may be programmed to repeat the voice guidance instruction by adding an insistent formulation such as “Now position yourself on the lane” or by adding another reinforcement to emphasize the maneuver to be carried out such as “Now position yourself on the lane of ... in the direction of ...”.

[0169] As the vehicle moves forward, the context may change, making a guidance instruction at best ambiguous, at worst obsolete.

[0170] We can thus consider two cases.

[0171] The first case is when an element comes between the vehicle and the object that was selected to develop a guidance instruction, preventing the driver from viewing this object.

[0172] The second case is when another object, a counterpart to the one selected to generate a guidance instruction, appears in the driver's field of vision. This situation can occur, for example, when a bus leaves its location and discovers a second bus stop.

[0173] Both situations involve increased stress and cognitive load for the driver, as well as a risk of error or late maneuvering.

[0174] In both cases, the processing module 25 will then be programmed to generate a new instruction adapted to the new context.

[0175] As introduced above, the voice guidance instruction to be given will not be the same depending on the phase the vehicle is in. We can now describe how this guidance instruction will be formulated in each of these phases. Anticipation phase

[0176] As a reminder, the anticipation phase occurs before the maneuver to be performed. It begins and ends at a distance from the area where the maneuver will be performed. It allows the driver to position the vehicle as best as possible and / or prepare to perform this maneuver (especially when the target of the maneuver will be visible late).

[0177] In such an anticipation phase, the reinforcement of voice guidance instructions is recommended, but the use of remarkable or significant places is avoided.

[0178] The processing module 25 is programmed to issue a so-called anticipation voice guidance instruction when the section of road taken contains several lanes.

[0179] How to reinforce this anticipatory guidance instruction may vary depending on the situation encountered.

[0180] Basically (i.e. in situations other than those considered below), reinforcement is achieved in the following way.

[0181] If the vehicle is located on a multi-lane road axis and is positioned in the lane most compatible with carrying out the next maneuver, the processing module 25 is programmed to generate an anticipatory voice guidance instruction reassuring the driver.

[0182] The voice guidance instruction is then structured by combining an action, a familiarity criterion, a geographical marker and an egocentric attribute.

[0183] For example, this voice guidance instruction might be expressed as "Stay in your lane."

[0184] If the vehicle is located on a multi-lane road and is positioned in a lane that is compatible with performing the next maneuver (typically, if no solid white line is located between the lane it is on and the location of the next maneuver) but which requires a lane change, the processing module 25 is programmed to generate a voice-anticipatory guidance instruction instructing the driver to change lanes quickly.

[0185] The voice guidance instruction is then structured by combining an action, a geographical marker and a contextual clue.

[0186] For example, this voice guidance instruction might be expressed as "Move to lane x now (or as soon as possible).

[0187] If the vehicle is to merge into a multi-lane road and the vehicle is in a limited length merge lane, the processing module 25 is programmed to generate a look-ahead voice guidance instruction instructing the driver to change lanes.

[0188] The voice guidance instruction is then structured by combining at least one action and a geographical marker.

[0189] For example, this voice guidance instruction might be expressed as "Join the lane."

[0190] If the vehicle has to enter a multi-lane road and the vehicle is on an insertion lane of unlimited length (which then becomes a lane of the main road) and this lane is the most compatible for carrying out the next maneuver, the processing module 25 is programmed to generate an anticipatory voice guidance instruction instructing the driver to stay on this lane.

[0191] The voice guidance instruction is then structured by combining at least one action, a geographical marker and an egocentric attribute.

[0192] For example, this guidance instruction might be expressed as "Stay in your lane."

[0193] If the vehicle is to enter a multi-lane road and the vehicle is on an insertion lane of unlimited length and this lane is not the most compatible with the next maneuver to be performed, the processing module 25 is programmed to count the number of lane changes necessary to reach the most compatible lane, then to generate a voice guidance instruction instructing the driver to change lanes.

[0194] The voice guidance instruction is then structured by combining at least one action and attribute.

[0195] For example, this voice guidance instruction might be expressed as "Move to the nth lane."

[0196] In this situation, the way to formulate which lane the vehicle should be placed on can be determined as follows.

[0197] In the absence of any constraint on the lane to be taken among the several lanes of the road axis, the processing module 25 is programmed to generate a voice guidance instruction instructing the driver to merge.

[0198] The complete guidance instruction is then structured by combining at least one action and one target.

[0199] For example, this guidance instruction might be expressed as "Merge onto...". The road name or type can then be specified.

[0200] If the road axis has only two traffic lanes, the processing module 25 is programmed to generate a voice guidance instruction indicating whether the vehicle should move into the “right” or “left” lane.

[0201] The guidance instruction is then structured by combining at least one action, a geographical marker and an egocentric attribute.

[0202] If the road axis has exactly three traffic lanes, the processing module 25 is programmed to generate a voice guidance instruction indicating whether the vehicle should move into the “right” or “left” or “middle” lane.

[0203] The guidance instruction is then structured by combining at least one action, a geographical marker and an egocentric attribute.

[0204] If the road axis has exactly four traffic lanes and only one of them is compatible with the maneuver to be carried out, the processing module 25 is programmed to generate a voice guidance instruction indicating whether the vehicle should move into the “rightmost” or “leftmost” lane or “the second lane” or “the third lane”.

[0205] The guidance instruction is then structured by combining at least one action, a geographic marker and an exo-centric attribute.

[0206] If the road axis has exactly four traffic lanes and several of them are compatible with the maneuver to be carried out, the processing module 25 is programmed to generate a voice guidance instruction indicating whether the vehicle should move to the “rightmost” or “leftmost” or “middle” lanes.

[0207] The guidance instruction is then structured by combining at least one action, a geographic marker and an exo-centric attribute.

[0208] If the road axis has an odd number of traffic lanes greater than four and only the middle one is compatible with the maneuver to be carried out, the processing module 25 is programmed to generate a voice guidance instruction indicating that the vehicle must move into the “middle” lane.

[0209] The guidance instruction is then structured by combining at least one action, a geographic marker and an exo-centric attribute.

[0210] If the road axis is made up of several lanes and at least one of them is prohibited to vehicle traffic, the processing module must include it in the lane count.

[0211] When the vehicle is outside a built-up area, the reinforcement of the voice guidance instruction in the anticipation phase can preferentially indicate to the driver the reason for the positioning (the purpose) by describing the next maneuver to be carried out.

[0212] The destination is thus indicated according to the distance between the vehicle and the maneuver, according to the type of maneuver to be carried out (motorway exit, slip road, roundabout, etc.) and the road infrastructure visible to the driver (direction signs, etc.).

[0213] On the roads, the processing module 25 must use a visual marker which is preferably static (of the direction sign type), otherwise (in the absence of such a sign) a mobile marker, otherwise (in the absence of a sign and a mobile marker) a road infrastructure element.

[0214] We can then clarify this rule with the following examples.

[0215] If the processing module 25 detects a sign and has the ability to read the inscriptions marked thereon, it is programmed to generate an anticipatory voice guidance instruction instructing the driver to follow the direction indicated thereon.

[0216] The complete guidance instruction is then structured by combining at least one action, one geographical marker, one egocentric attribute, one purpose, one geographical marker, one visual clue and one attribute.

[0217] For example, this voice guidance instruction might be expressed as "Move into the left lane to take the exit toward...".

[0218] If the processing module 25 detects a sign, that it does not have the capacity to read the inscriptions marked there but that the cartographic database 20 has this information, it operates in the same way as above.

[0219] If the processing module 25 detects a sign whose color is unique among the other signs, it is programmed to generate a voice guidance instruction directing the driver to follow the direction of the sign of the identified color.

[0220] The voice guidance instruction is then structured by combining at least one action, one geographical marker, one egocentric attribute, one purpose and one attribute.

[0221] For example, this guidance instruction may be expressed as "Move to the left lane to follow the blue sign." If the writing on the sign is legible or known, this instruction may be supplemented by indicating the direction indicated there.

[0222] When the vehicle is in a built-up area, the signs are more frequently hidden or covered by vegetation. Therefore, the processing module 25 must identify the most relevant visual cue among other cues, including: mobile landmarks, elements of road infrastructure, static landmarks other than signs, and significant or notable places.

[0223] This situation can be illustrated by the following four examples.

[0224] In a first example, the processing module 25 develops the following guidance instruction: “Position yourself in the same lane as the red car”. In this example, the guidance instruction is structured by combining at least one action, a geographical marker, a visual cue and an attribute.

[0225] In a second example, the processing module 25 develops the following guidance instruction: “Position yourself in the left lane towards the green sign”. In this example, the guidance instruction is structured by combining at least one action, a geographic marker, an egocentric attribute, a geographic marker, a visual cue and an attribute.

[0226] In a third example, the processing module 25 develops the following guidance instruction: “Position yourself in the left lane then at the roundabout take the 3rd exit”. In this example, the guidance instruction is structured by combining at least one action, a geographical marker, a visual cue, a concatenation element, another visual cue, an action and an exo-centric attribute.

[0227] In a fourth example, the processing module 25 develops the following guidance instruction: “Position yourself in the right lane to turn right at the intersection”. In this example, the guidance instruction is structured by combining at least one action, a geographic marker, a visual cue, a purpose and another visual cue. Maneuver description phase

[0228] The second phase is the description phase. During this second phase, the voice guidance instruction called description characterizes the action that will have to be carried out to reach the target.

[0229] In this instruction, the processing module 25 delivers one or more pieces of information necessary for the driver to locate himself and to locate the target in relation to the environment (and in particular in relation to the detected visual landmarks).

[0230] When the vehicle arrives at an intersection, the processing module 25 calculates the angle of each road leaving the intersection, relative to the road taken by the vehicle. This angle is expressed for example in degrees and is between -180° and 180°.

[0231] The guidance instruction can then be reinforced by a geographic attribute adapted to the intersection which can be “hard right” or “ahead” (or “straight ahead”).

[0232] The way to reinforce the guidance instruction may vary depending on the type of intersection.

[0233] In a T-intersection, in the case where the vehicle arrives at the foot of the T on a road forming the crosspiece of the T, the processing module 25 may be programmed to use the word "intersection".

[0234] If the angle between the base and the crosspiece of the T is substantially perpendicular (for example, within 10, 20 or 30 degrees), the processing module 25 will use the action verbs "Turn to" or "Take to".

[0235] If the angle between the base and the crosspiece of the T is not approximately perpendicular, two cases must be considered.

[0236] The first case is when the vehicle, in the intersection, must deviate by an angle significantly less than 90 degrees from its initial trajectory. In this case, the processing module 25 will supplement the verbs used with the expression "slightly to". It will also favor the use of the action verb "Take to" over the verb "Turn to".

[0237] The second case is when the vehicle, in the intersection, must deviate at an angle significantly greater than 90 degrees from its initial trajectory. In this case, the processing module 25 will supplement the verbs used with the expression "completely to". It will also favor the use of the action verb "Turn to" over the verb "Take to".

[0238] A fork intersection is an intersection consisting of one entrance lane and exactly two exit lanes located in the extension of the entrance lane.

[0239] The way to formulate a voice guidance instruction in such an intersection will then depend on the configuration of the premises.

[0240] In the case where both exit lanes are to the right of the entry lane, if the vehicle is to take the rightmost lane, the instruction will be formulated by an expression such as "Take the rightmost lane". On the other hand, if the vehicle is to take the other exit lane, it will be necessary to reinforce the guidance instruction to indicate to the driver which one to take.

[0241] Conversely, if both exit lanes are to the left of the entry lane, if the vehicle is to take the leftmost lane, the instruction will be formulated by an expression such as "Take the leftmost lane". On the other hand, if the vehicle is to take the other exit lane, it will be necessary to reinforce the guidance instruction to indicate to the driver which one to take.

[0242] In the case where the two exit lanes are on either side of the entry lane, the instruction will be formulated by an expression such as "Take the lane on the left" or "Take the lane on the right".

[0243] In a roundabout-type intersection, the voice guidance instruction will be defined based in particular on the number of exits from the roundabout.

[0244] A roundabout consists of a ring road and roads leading into this central road, hereinafter referred to as "adjacent roads". It has entrances and exits, and one entrance and one exit may be located on the same adjacent road.

[0245] It should be noted that the complexity of a roundabout will depend on the following criteria: angle between the vehicle's arrival axis and the exit axis, angle between the exit to be taken and those nearby, number of adjacent roads, difference between the number of entrances and the number of exits, number of lanes in the annular road, outer diameter of the roundabout.

[0246] A roundabout will be considered simple if it has as many entrances as exits distributed in pairs on adjacent roads, if the number of adjacent roads is less than or equal to four, if these adjacent wheels are oriented in clearly distinct directions, and if priority in the roundabout is given to vehicles traveling on the annular road. In other cases, the roundabout will be considered complex.

[0247] A first example of a simple roundabout will be a roundabout with four adjacent roads, angularly spaced approximately 90 degrees apart. This is called an “orthogonal roundabout”. In this example, the processing module 25 may be programmed to give a voice guidance instruction such as “At the roundabout, exit to the right (or left or straight ahead)”.

[0248] A second example of a simple roundabout will be a roundabout with three or four adjacent roads, spaced angularly two by two at angles of which at least one is substantially different from 90 degrees. In this example, the processing module 25 may be programmed to pronounce a voice guidance instruction reinforced by the number of the exit lane to be taken. This instruction may be expressed in the form "At the roundabout, take the first exit on the right".

[0249] A third example of a simple roundabout will be a small diameter roundabout. In this example, the processing module 25 will interpret this roundabout as a simple intersection (see below).

[0250] A voice guidance instruction for maneuvering through a complex roundabout will always be reinforced.

[0251] Various examples can be given in which the instructions will be reinforced.

[0252] In a first example, the roundabout does not have four orthogonal exits two by two.

[0253] So, a first case to consider is that in which the exit is orthogonal (within 10 degrees) to the entry lane by which the vehicle arrives and corresponds to the first or second exit. In this eventuality, the processing module 25 is programmed to preferentially pronounce a voice guidance instruction reinforced by the number of the exit to be taken. This instruction could be expressed in the form "At the roundabout, take the first exit on the right".

[0254] A second case to consider is that in which the exit is not orthogonal to the entry lane by which the vehicle arrives and corresponds to the first or second exit. In this event, the processing module 25 is programmed to necessarily pronounce a voice guidance instruction reinforced by the number of the exit to be taken. This instruction could be expressed in the form "At the roundabout, take the first exit on the right".

[0255] A third case to consider is one in which the exit corresponds to the third or a later exit. In this eventuality, the processing module 25 is programmed to pronounce a voice guidance instruction reinforced by an exocentric attribute. This instruction could be expressed in the form "Take the exit on the left just before the restaurant".

[0256] A second example of a complex roundabout is one in which the numbers of entrances and exits differ.

[0257] In particular, we can consider the case in which there is one entrance more than exits, this entrance being located on an adjacent one-way road separate from the road by which the vehicle arrives.

[0258] In this example, the processing module 25 applies the rules defined in the previous example, sometimes refraining from indicating the exit to be taken by its number.

[0259] So, if the exit to be taken is after the one-way entrance, the processing module 25 is programmed to pronounce a voice guidance instruction reinforced by an exocentric attribute, without indicating an exit number. This instruction can be expressed in the form "Take the exit on the right at...".

[0260] On the other hand, if the exit to be taken is before the one-way entrance, the processing module 25 is programmed to speak a voice guidance instruction in the same way as in the first complex roundabout example given above.

[0261] A third example of a complex roundabout is one in which the exits are very close in distance or angle to each other. In particular, a roundabout is considered complex when the angle between two exits is less than or equal to 30°.

[0262] In this example, the processing module 25 applies the same rules as those set out in the first complex roundabout example given above, by necessarily reinforcing the voice guidance instructions.

[0263] A fourth example of a complex roundabout is one in which vehicles arriving at the roundabout have priority over those already circulating around the roundabout (this is called a "French-style roundabout", the one at Place de l'Etoile being the best known).

[0264] In this example, the processing module 25 applies the same rules as those set out in the first complex roundabout example given above, by necessarily reinforcing the voice guidance instructions with a warning such as: “Beware of right-of-way in the roundabout”.

[0265] In an intersection of the crossroads type, with at least four lanes leading from this intersection, the processing module 25 applies the same rules as those set out in the various roundabout examples given above, replacing the term “roundabout” with “crossroads” or “intersection”.

[0266] The processing module 25 can in particular carry out a count of the exits from the lane taken by the vehicle, this time counting all the lanes, even those in one direction or no entry.

[0267] If the exit lane to be taken is offset from the other exit lanes, the processing module 25 is programmed to add to the instruction a command of the type "Proceed through the intersection to take the 2nd exit on the left towards...". Support phase in the maneuver

[0268] The accompaniment phase is essentially intended to indicate to the driver the exact moment at which he can begin the actual maneuver, i.e. the change of direction. During this phase, the processing module 25 only issues a so-called accompaniment voice guidance instruction if crossing the intersection is deemed complex.

[0269] In this situation, the choice of whether or not to reinforce the voice guidance instruction and how to reinforce it depends on the type of intersection.

[0270] Thus, when several T-intersections follow one another on the road and no visual marker (infrastructure, sign, notable place, etc.) is present on the site to differentiate them, the processing module 25 will be programmed to pronounce an accompanying voice guidance instruction. This instruction will include a contextual clue and may be formulated as follows: “Now turn right…”.

[0271] In a fork-type intersection, no accompanying voice guidance instructions will be issued.

[0272] In a roundabout type intersection, if the diameter of the roundabout is less than a predetermined threshold and the exits are not orthogonal two by two, the processing module 25 will be programmed to generate an accompanying voice guidance instruction during the maneuver either by repeating an instruction already issued during the maneuver description phase, or by using a new distinctive or remarkable location or road infrastructure that was not visible until then. The distinctive or remarkable location or road infrastructure used must of course be visible, unique and immediately identifiable by the driver.

[0273] This instruction could be formulated as follows: “Now exit the roundabout, just before Hotel X”.

[0274] If the processing module 25 detects that the driver has completed a full circle of the roundabout, it may be programmed to generate an accompanying voice guidance instruction by repeating the voice guidance instruction issued during the description phase.

[0275] If the voice guidance instruction issued during the description phase did not refer to any visual reference because no reference was then clearly visible and the vehicle slows down significantly (either due to hesitation on the part of the driver or due to a traffic jam), the processing module 25 is programmed to generate an accompanying voice guidance instruction using a distinctive or remarkable location or road infrastructure that has become visible.

[0276] The slowdown here must be greater than a predetermined threshold. Reinsurance Phase

[0277] Once the maneuver has been carried out, the processing module 25 assesses, based on the complexity of the maneuver and any hesitations on the part of the driver, the need to reassure the driver about the quality of the maneuver he has just carried out.

[0278] For this purpose, if the processing module 25 detects hesitation during the maneuver or when exiting the maneuver, it is programmed to generate a reinforcement voice guidance instruction. The decision to generate this instruction may also depend on other factors such as the complexity of the maneuver being performed, the immediate presence of another maneuver to be performed, etc.

[0279] This instruction is preferably reinforced with a visual cue located near the end of the maneuver or visible only when the maneuver is completed, this cue preferably being distinct from the visual cues used until then. Continuity Phase

[0280] In the absence of a maneuver to be carried out over a distance or duration greater than a threshold (for example more than one minute or more than one kilometer), it is preferable for the processing module 25 to express to the driver that it is still active.

[0281] The voice guidance instruction can then be formulated differently depending on the configuration of the road axis taken and the environment visible to the driver.

[0282] It is thus possible to envisage the case where at least one intersection is detected between the position of the vehicle and the next maneuver to be carried out, where no change of road is planned, and where no change of lane is possible. In this situation, the processing module 25 is programmed to generate a voice guidance instruction which depends on the curvature of the road.

[0283] Thus, if the radius of curvature of the road is greater than a predetermined threshold, the voice guidance instruction is structured by adding at least one action, one geographical marker, one egocentric attribute, one purpose and one geographical marker. It can be expressed in the form "Continue straight ahead through the intersections...".

[0284] On the contrary, if the radius of curvature of the road is less than this threshold, the voice guidance instruction is structured by adding at least one action, one geographical marker, one egocentric attribute, one purpose, one geographical marker and one other egocentric attribute. It can be expressed in the form "Follow the main road through the next intersections...".

[0285] It is also possible to envisage the case where at least one intersection is detected between the position of the vehicle and the next maneuver to be carried out, where no change of road is planned, and where no change of lane is desirable given the next maneuver to be carried out. In this situation, the processing module 25 is again programmed to generate a voice guidance instruction which depends on the curvature of the road.

[0286] Thus, if the radius of curvature of the road is greater than a predetermined threshold, the voice guidance instruction is structured by adding at least one action, one geographical marker, one egocentric attribute, one purpose, one geographical marker, concatenated with another action, one familiarity criterion, another geographical marker, and another egocentric attribute. It can be expressed in the form "Continue straight through intersections and stay in your lane...".

[0287] On the contrary, if the radius of curvature of the road is less than this threshold, the voice guidance instruction is structured by adding at least one action, one geographical marker, one attribute, one purpose, one geographical marker, concatenated with another action, one familiarity criterion, another geographical marker, and another egocentric attribute. It can be expressed in the form "Follow the main road through the next intersections and stay in your lane".

[0288] Finally, we can consider the case where only an intersection is detected between the position of the vehicle and the next maneuver to be carried out, where no change of road is planned, but where a change of lane is necessary given the next maneuver to be carried out. In this situation, the processing module 25 is again programmed to generate a voice guidance instruction which depends on the curvature of the road.

[0289] Thus, if the radius of curvature of the road is greater than a predetermined threshold, the voice guidance instruction is structured by adding at least one action, a geographical marker, an attribute, a purpose, a geographical marker, concatenated with another action, a familiarity criterion, a target. It can be expressed in the form "Continue straight ahead at the intersection and then position yourself on lane X".

[0290] On the contrary, if the radius of curvature of the road is less than this threshold, the voice guidance instruction is structured by adding at least one action, one geographical marker, one attribute, one purpose, one geographical marker, concatenated with another action, one familiarity criterion, and one target. It can be expressed in the form "Follow the main road to the next intersection and then position yourself on lane X".

[0291] When the distance between the vehicle position and the next maneuver to be performed (or the time required to reach the maneuver area) is greater than a predetermined threshold (for example 2 km), the processing module 25 is programmed to integrate a notion of distance into the voice guidance instruction. This can be expressed in the form “Continue straight ahead for x km…”.

[0292] During the so-called continuity phase, the voice guidance instructions may also mention a road infrastructure (roundabout, bridge, tunnel, intersection) or a distinctive or remarkable location (train station, gas station, illuminated sign) which will be visible (and unique) when approaching the maneuver to be performed. Here again, two cases can be considered.

[0293] If the road curvature radius is greater than a predetermined threshold, the voice guidance instruction may be expressed as "Continue straight ahead to the gas station."

[0294] Conversely, if the road's curvature radius is less than this threshold, the voice guidance instruction may be expressed as "Follow the main road to the next bridge."

[0295] If the same maneuver can be performed at several locations, the voice guidance instruction can be expressed as: "Continue straight until the next instruction" or "Continue straight to the end of the street" or "Continue straight to the large tower...".

Claims

1. Method for generating voice guidance instructions intended for an individual, comprising steps of: - determining, by means of a computer (25), a route to be taken, - acquiring, by means of an image-acquiring means (10), at least one image of the environment of the individual, - processing said image with a view to detecting therein at least one object and to characterizing said object, and - establishing and transmitting a voice guidance instruction informing the individual how to perform a manoeuvre in order to follow said route, characterized in that provision is made for a step of determining a level of complexity of said manoeuvre depending on the position of the vehicle with respect to the zone in which said manoeuvre must be performed, in order to take into account whether a change of lane is necessary or not, in that, in the establishing step, said voice guidance instruction is formulated using a cue deduced from the characterization of said object only if the level of complexity of said manoeuvre is higher than a first threshold, and in that if the level of complexity of said manoeuvre is higher than a second threshold strictly higher than said first threshold, said voice guidance instruction is generated using two cues.

2. Generating method according to the preceding claim, wherein the level of complexity is computed depending on at least one of the following criteria: - the type of road infrastructure present in the zone in which said manoeuvre must be performed, - the tension felt by the individual, - the density of road traffic, - the number of objects detected in at least one portion of the acquired image, - the individual's knowledge of the road infrastructure, - the phase in which the vehicle is currently, and, if the road infrastructure is a roundabout, also depending on at least one of the following criteria: - the angle between an axis of arrival of the individual at the roundabout and an axis of exit of the individual from the roundabout, - the angle between at least two roads leading to the roundabout, - the number of roads leading to the roundabout, - the difference between the number of entrances to the roundabout and the number of exits from the roundabout, - the number of lanes present in the roundabout, and - the outside diameter of the roundabout.

3. Generating method according to one of the preceding claims, wherein, said cue being a visual cue visible to the individual, if, between the moment at which the voice guidance instruction was transmitted vocally and the moment when the manoeuvre must be performed, the visual cue disappears from the environment visible to the individual, provision is made for a step of generating a new voice guidance instruction not using said visual cue.

4. Generating method according to one of the preceding claims, wherein, said cue being a visual cue visible to the individual, if, between the moment at which the voice guidance instruction was transmitted vocally and the moment when the manoeuvre must be performed, another visual cue that is indistinguishable from the visual cue used in the voice guidance instruction appears in the environment visible to the individual, provision is made for a step of generating a new voice guidance instruction different from the voice guidance instruction that was transmitted vocally.

5. Generating method according to one of the preceding claims, wherein provision is made to determine a cost relating to an additional time or distance that an error in following said route would cause, and wherein the level of complexity is determined depending on said cost.

6. Generating method according to the preceding claim, wherein provision is made to distinguish between at least two distinct and successive phases prior to the performance of said manoeuvre, namely a manoeuvre-anticipation phase, in which a voice guidance instruction may be transmitted to indicate to the individual how to position himself to approach the manoeuvre, and a description phase, in which a voice guidance instruction is transmitted to indicate to the individual how to perform the manoeuvre, and wherein the voice guidance instructions transmitted in these phases use distinct cues, provision further being made for at least one other phase which follows the description phase and in which a voice guidance instruction may be generated, said other phase being: - an accompaniment phase simultaneous to said manoeuvre, or - a reassurance phase subsequent to said manoeuvre and allowing it to be indicated to the individual whether he has performed the manoeuvre correctly, or - a continuity phase allowing the individual to be informed of the active character of said generating method.

7. Generating method according to the preceding claim, wherein, if the voice guidance instruction transmitted during the description phase did not refer to any visual indicator and the vehicle slows down, provision is made in the accompaniment phase to transmit a voice guidance instruction using a cue.

8. Generating method according to either one of the two preceding claims, wherein, in the reassurance phase, provision is made to transmit a voice guidance instruction using a visual indicator distinct from the one or more visual cues used in the one or more voice guidance instructions transmitted during the preceding phases.

9. Generating method according to one of the three preceding claims, wherein, in the continuity phase, provision is made to generate a different voice guidance instruction depending on whether the radius of curvature of the road taken is lower or higher than a threshold.

10. Generating method according to any one of the preceding claims, wherein, the step of determining, by means of a computer (25), a route to be taken comprising retrieving, from a map database (20), static visual indicators located on said route and attributes associated with said static visual indicators, provision is made to determine an attribute common to a plurality of static visual indicators, the voice guidance instruction being formulated in the establishing step using a cue comprising the common attribute when the object is characterized as comprising the common attribute in the image-processing step.

11. Generating method according to the preceding claim, wherein, the static visual indicators being road signs, the common attribute comprises a name that is written on a road sign or a background colour of a road sign, the object being a road sign.

12. Generating method according to either one of the two preceding claims, wherein, the step of processing said image comprising detecting a plurality of objects, the generating method further comprises a step of classifying the detected objects, an object characterized as having the common attribute being prioritized, the object selected for the establishing step being the prioritized object.

13. Device for generating route guidance instructions that are tailored to the environment, comprising: - a means (10) for acquiring images of the environment, - a map database (20), - a navigation and geolocation system (22), - a processing module (25) suitable for receiving and combining elements originating from the acquiring means (10), from the map database (20) and from the navigation and geolocation system (22), and - a voice transmitter (11) suitable for transmitting an indication to the individual, characterized in that the processing module (25) is programmed to implement a method according to any one of the preceding claims.