WO2018115321A1

WO2018115321A1 - System and method for constructing a virtual horizon

Info

Publication number: WO2018115321A1
Application number: PCT/EP2017/084159
Authority: WO
Inventors: Thomas LIENNARD; Brouk WEDAJO
Original assignee: Valeo Schalter Und Sensoren Gmbh
Priority date: 2016-12-21
Filing date: 2017-12-21
Publication date: 2018-06-28
Also published as: FR3060737A1

Abstract

A system for constructing a virtual horizon, comprises: - a processing unit (12) designed to identify, by analysis of data captured (CAPT) by at least one sensor (2; 4) fitted to a vehicle, events (E_i) occurring in an environment of the vehicle; - a storage unit (14) for storing said events each in association with an information item tagging the event concerned along a path followed by the vehicle; and - an assembly (16, 20) for determining at least one road environment characteristic (C1; C2) to be allocated to the path on the basis of at least two stored events (E_i). An associated method is also proposed.

Description

System and method for building a virtual horizon

TECHNICAL FIELD TO WHICH THE INVENTION RELATES The present invention relates to driving assistance systems and the recognition, in this context, of a context of evolution of a vehicle.

It relates more particularly to a system and a method of constructing a virtual horizon.

BACKGROUND

In order to know a context of evolution of a vehicle, one frequently uses nowadays a cartographic database and a device of geolocalisation of the vehicle.

In fact, the device for geolocation of the vehicle makes it possible to know its current position and consequently to extract from the cartographic database the characteristics of the immediate environment of the vehicle.

It is thus possible to build on the basis of these cartographic data an electronic horizon in order to anticipate the situations likely to be encountered by the vehicle.

However, this solution can not work correctly when the geolocation device fails (for example because of a reception problem, typically in a tunnel) or when the environment is changed without this change being reflected in the base cartographic data, for example in case of modification of the traffic plan or during road works.

OBJECT OF THE INVENTION

In this context, the present invention proposes a system for constructing a virtual horizon, comprising a processing unit designed to identify, by analyzing data captured by at least one sensor fitted to a vehicle, events occurring in a vehicle environment; a unit for storing said events each associated with information identifying the event concerned along a path taken by the vehicle; and a set of determining at least one road environment characteristic to be assigned to the path based on at least two stored events.

This builds a virtual horizon that describes the road environment and can be used as the electronic horizon mentioned above, but which is built in real time and therefore does not depend on a map database or vehicle location using an external system.

According to other characteristics that can be envisaged as optional (and therefore not limiting):

the system comprises a unit of determination, based on the stored events, of information associated with a characteristic of the environment for a plurality of points of the path;

the determination unit is designed to produce a probability associated with said information;

the system comprises an automaton comprising internal registers and arranged to successively traverse the points of the path, to update said internal registers according to the information associated with said characteristic for the current point and to produce a context detection information according to said internal registers;

the controller is designed to produce, for each current point, a confidence index associated with the context detection information;

the information identifying the event is a curvilinear abscissa along the path;

the storage unit is designed to determine said curvilinear abscissa as a function of a current curvilinear abscissa of the vehicle delivered by an odometer;

the information identifying the event is temporal information;

the system includes a driver assistance system adapted to activate or deactivate a feature according to the determined road environment characteristic;

- the feature is a stand-alone driving feature. The invention also proposes a method for constructing a virtual horizon, comprising the following steps:

- identification by means of a processing unit, by analyzing data captured by at least one sensor fitted to a vehicle, events occurring in a vehicle environment;

storage of said events each associated with information identifying the event concerned along a path taken by the vehicle; determining at least one road environment characteristic to be assigned to the path on the basis of at least two memorized events.

DETAILED DESCRIPTION OF AN EXEMPLARY EMBODIMENT The following description with reference to the accompanying drawings, given by way of non-limiting examples, will make it clear what the invention consists of and how it can be implemented.

In the accompanying drawings:

FIG. 1 represents an example of an onboard system according to the teachings of the invention;

FIG. 2 diagrammatically represents a first type of processing performed within the system of FIG. 1;

- Figure 3 schematically shows a second type of treatment performed within the system of Figure 1; and

FIG. 4 schematically represents a vehicle equipped with an onboard system such as that of FIG. 1.

FIG. 1 represents an example of an on-board system, here in a motor vehicle (such as that of FIG. 4), in accordance with the teachings of the invention.

This embedded system comprises various sensors, here an image sensor 2 (such as a video camera) and a flight time sensor 4 (such as a laser range finder).

These sensors produce captured CAPT data representative of the vehicle environment. In the case of the image sensor 2, the captured data CAPT are representative of the light received by the image sensor 2 at each of its pixels (associated respectively with different directions of the field of view of the image sensor 2 ). In the case of the flight time sensor 4, the CAPT captured data are representative of the distance of the first object encountered by the transmitted beam, for a plurality of directions of the environment.

The onboard system further comprises an odometer 5 which provides a measure of the distance traveled by the vehicle along its path, that is to say the curvilinear abscissa s along the path of the vehicle.

The on-board system also includes a locating device 6, for example a geolocation device (or GPS for "Global Positioning Such a locating device 6 produces a POS positioning information of the vehicle in a given reference frame, here the terrestrial reference system.

The onboard system includes a cartographic database 7 containing descriptive elements of the environment that the vehicle is likely to borrow. These descriptive elements include in particular descriptive data of the roads present in this environment, in particular for example an identifier of the road, the geographical location of the road, and, for different portions of this road, the number of traffic lanes and the speed authorized limit for the portion concerned.

The aforementioned descriptive elements may furthermore include descriptive data of other elements of the environment.

It should be noted that the descriptive elements present in the cartographic database 7 describe the environment as it existed during the construction of the database; this environment may, however, have been modified at certain points during the actual passage of the vehicle and the map database 7 therefore does not always perfectly represent the environment traversed by the vehicle.

The map database 7 is here stored in the vehicle, for example on a hard disk or an optical disk present on board the vehicle. Alternatively, the map database 7 could be stored on a remote and accessible server of the vehicle by means of a telecommunication system equipping the vehicle.

The onboard system comprises a navigation module 8 which extracts, from the map database 7, ADAS data relating to the environment close to the vehicle, according to the current position of the vehicle given by the positioning information POS produced by the locating device 6.

The location device 6, the map database 7 and the navigation module 8 form a navigation unit 9.

The onboard system of FIG. 1 finally comprises a recognition system 10 making it possible to identify at least one context of evolution of the vehicle (here several contexts C1, C2 of evolution of the vehicle) or, more generally, to determine a characteristic of the vehicle. road environment attribute to the route taken by the vehicle, as described below.

The recognition system 10 is for example built on the basis of a microprocessor architecture: in this case, a memory associated with the microprocessor stores computer program instructions whose execution by the microprocessor allows the implementation of different functional units described now.

The recognition system 10 thus comprises a processing unit 12 which analyzes the captured CAPT data (for example by means of shape recognition algorithms), recognizes (by means of this analysis) objects of the environment (for example road signs, traffic lights, fixed or mobile objects, traffic lanes adjacent to the route taken, floor markings, etc.) and deduces certain events E, to be listed as explained below. (Such an event E is, for example, the presence of a speed limitation panel authorized at a given speed.)

The recognition system 10 comprises a storage unit 14 which receives on the one hand the events E, produced by the processing unit 12 and on the other hand the curvilinear abscissa s measured by the odometer 5, and can thus store each of the events E, in association with information identifying the event concerned along the path taken by the vehicle (this information here being the curvilinear abscissa along the path).

Note that the curvilinear abscissa recorded in association with a given event E is not necessarily equal to the current curvilinear abscissa s provided by the odometer 5, but is determined according to the current curvilinear abscissa provided by the odometer and the position of the event E, relative to the vehicle (an event E, such as a traffic sign which can for example be detected by the image sensor 2 at the front of the vehicle, eg 100 m in front of the vehicle). The position of an event E with respect to the vehicle is for example in this case provided by the processing unit 12 as data attached to the event E ,.

According to another conceivable possibility, the information identifying the event concerned along the path taken could be temporal.

Anyway, the storage unit 14 thus builds a virtual horizon vHor which lists the events E, past or future (within the scope of sensors 2, 4) along the vehicle path.

In order to limit the memory consumption due to the memorization of past events, the storage unit 14 can keep in memory only the events E, relating to curvilinear abscissa whose distance to the current curvilinear abscissa is less than one predetermined threshold (this threshold being for example equal to 1000 m in practice).

The recognition system 10 comprises a pattern matching unit 1 6 (in English "pattern matching") which analyzes the events E, stored in the virtual horizon vHor and deduces, for a plurality of characteristics PROF1, PROF2 of the vehicle environment, information 11, 12, which is possibly associated a probability P1, P2 (which represents a level of confidence associated with the information 11, 12 concerned).

For a characteristic PROF1 corresponding to the number of lanes of the road taken, information 11 is for example an indication of passage from 2 to 3 lanes, or an indication of passage of 3 to 2 lanes of traffic. Each piece of information 11, 12 delivered by the pattern recognition unit 1 6 may be associated with information identifying this information 11, 12 along the path taken by the vehicle (here the curvilinear abscissa s).

FIG. 2 schematically shows the operation of the pattern recognition unit 16.

The pattern recognition unit 1 6 implements a plurality of algorithms ALGO1, ALGO2, each algorithm ALGO1, ALGO2 producing (at each point of the path taken by the vehicle) information 11, 12 of realization of a given characteristic PROF1, PROF2 (associated with this algorithm) based on the events E, stored in the virtual horizon vHor. At least some of these algorithms can also produce a probability P1, P2 (or confidence index) associated with the information 11, 12 produced by the algorithm concerned.

In the following, we will denominate "profi each characteristic PROF1, PROF2 thus evaluated because this feature PROF1, PROF2 generally allows to identify a particular road profile.This characteristic PROF1, PROF2, or profile, can be for example the number of traffic lanes on the road, a given speed limit, or a particular road context (city, highway, junction, etc.). The algorithms ALG01, ALG02 can be, for example, probabilistic algorithms based on the Dempster-Shafer theory.

The pattern recognition unit 1 6 thus makes it possible to determine (and store), for each of the monitored PROF1, PROF2 characteristics, an embodiment information 11, 12 and, optionally, a probability value P1, P2 associated with this information. embodiment 11, 12, and this for each point of the path taken by the vehicle (the points of the path thus considered being for example separated by a fixed pitch).

The recognition system 10 furthermore comprises a formatting unit 18 designed to construct information 13, 14 relating to characteristics PROF3, PROF4 of the environment, this time on the basis of the ADAS data extracted by the navigation module 8. such features are usually referred to as "electronic horizon" or "e Horizon".

The characteristics PROF3, PROF4 followed by the formatting unit 18 may include characteristics (such as the characteristics PROF1, PROF2) also monitored by the pattern recognition unit 1 6 and / or characteristics distinct from those monitored by the unit. pattern recognition unit 1 6.

Like the pattern recognition unit 1 6, the formatting unit 18 makes it possible to determine (and thus to store) for each of the monitored PROF3, PROF4 characteristics, information 13, 14 and, optionally, a probability value P3. P4 associated with each of these information 13, 14, and for each point of the path taken by the vehicle. These probability values may, for example, be determined as a function of the update date of the map database 7 and / or the quality of the signal processed by the location device 6.

The recognition system 10 finally comprises a synthesis unit 20 which uses the information 11, 12, 13, 14 (as well as possibly the probability values P1, P2, P3, P4) determined respectively for the different characteristics PROF1, PROF2, PROF3, PROF4 of the environment (either by the pattern recognition unit 1 6, or by the formatting unit 18) to produce different road environment characteristics to be attributed to the path taken, that is to say here different context detection information C1, C2. As schematically shown in FIG. 1, the synthesis unit 20 can optionally receive directly {Le. without processing by the pattern recognition unit 1 6) the events E, and use them in the same way as the information 11, 12, 13, 14 (as described below).

The synthesis unit 20 here comprises a controller (for example a state machine or, more generally, a controller designed to implement a particular processing algorithm) specific for each road environment characteristic (in practice for each context). C1, C2 to watch. Figure 3 shows the operation of one of these automata.

Such an automaton can be called "visiteui" because, as schematically represented in FIG. 3, this automaton traverses all the points t of the path and takes into account, at each point t of the path, information 11, 12, 13, 14 (as well as optionnally probability values P1, P2, P3, P4) respectively associated with the various characteristics PROF1, PROF2, PROF3, PROF4 as explained above.

Precisely, at each point t of the path, the automaton or visitor updates (ie evolves) its internal registers R according to information 11, 12, 13, 14 (and, if appropriate, probability values P1, P2, P3, P4) respectively associated with the different characteristics PROF1, PROF2, PROF3, PROF4 for the point t concerned, and determines a detection information of a given context Ci (associated with the visitor) according to the state of its registers Internal R.

The context detection information C1, C2 is for example a binary value which indicates whether or not the context concerned is identified.

In addition, at least some visitors can produce a confidence index associated with the context detection information C1, C2 delivered.

Such a context Ci (or road environment characteristic) is, for example, the traffic of the vehicle on a multi-lane road, the presence of works, the end of a lane on a multi-lane road, the approach of a road. toll or authorized speed limit.

The characteristics (such as PROF1, PROF2, PROF3, PROF4) traversed (and scanned) by a visitor depend on the context that the visitor seeks to identify. For example, a visitor designed to detect a multi-lane road may notably use a lateral distance feature of the traffic signs; a visitor designed to detect the presence of works may use a feature relating to the presence of certain specific signs and a feature relating to the color of the markings on the ground; a visitor designed to detect an end of taxiway may use in particular a characteristic relating to the presence of arrows marked on the ground; a visitor designed to detect the approach of a toll may use a feature relating to the presence of more than 3 traffic lanes and a feature relating to the presence of other vehicles in the vicinity.

Note also that in practice, the detection information delivered (at each moment) by a visitor can be used at the input of another visitor (as a representative value of a characteristic of the environment).

The context detection information C1, C2 delivered by a visitor may also possibly be inscribed in the virtual horizon vHor (by the storage unit 14), this visitor playing somehow the role of a virtual sensor.

The use of PROF1, PROF2 characteristics (or profiles) based on the vHor virtual horizon (and thus on the detections made in real time by the sensors 2, 4), in addition to the PROF3, PROF4 characteristics from the electronic horizon , makes it possible to obtain (by merging information 11, 12, 13, 14 associated with these different characteristics PROF1, PROF2, PROF3, PROF4 by means of the visitor concerned) particularly reliable context detection information C1, C2, which take into account in particular any modifications made in the environment since the establishment of the cartographic database 7. The production of such detection information may then cause the generation of an alert or a deactivation of an autonomous driving mode, as explained below.

For example, a visitor designed to detect a toll approach may use features based on the virtual horizon (a feature relating to the presence of more than three lanes of traffic and a feature relating to the presence of other nearby vehicles as indicated above) and one or more characteristics based on the electronic horizon (for example a characteristic relating to the presence of the toll).

The context detection information (or road environment characteristics) C1, C2 provided by the recognition system 10 (precisely here by the synthesis unit 20) can indeed be used by various systems of the vehicle.

In the example described here, two distinct context detection information C1, C2 are used by an autonomous driving management system 30 to allow or prohibit (according to the values of the information C1, C2) the implementation of a feature. autonomous driving of the vehicle.

The context detection information C1 is for example information indicative of a "highway" type situation and the context detection information C2 is for example information indicative of a "branching" situation.

The autonomous driving management system 30 can then authorize a fast autonomous driving mode of the vehicle if the context detection information C1 indicates a "highway" type situation and if the context detection information C2 does not indicate a situation of type "branching.

On the other hand, since the context detection information C1 no longer indicates a "highway" type situation or the context detection information C2 indicates a "branching" situation, the autonomous driving management system 30 disables the fast autonomous driving mode (and can then switch to a low-speed autonomous driving mode, or to a manual driving mode after warning the driver and receiving a confirmation thereof, for example).

In another conceivable embodiment, the context detection information (delivered by a visitor) could be information indicative of a given speed limit (or alternatively information indicative of the approach of a toll); such information can then for example be used by a system for controlling the effective speed of the vehicle (for example in the case of a self-driving vehicle, but also in the case of a vehicle with a speed control function) in order to to command a reduction in the speed of the vehicle.

Claims

1. System for constructing a virtual horizon (vHor), comprising:

a processing unit (12) designed to identify, by analysis of captured data (CAPT) by at least one sensor (2; 4) fitted to a vehicle, events (E,) occurring in an environment of the vehicle;

- a storage unit (14) of said events each associated with information identifying the event concerned along a path taken by the vehicle; and

an assembly (1 6, 20) for determining at least one characteristic

(C1; C2; Ci) of road environment to be allocated to the path on the basis of at least two memorized events (E,).

2. System according to claim 1, comprising a determination unit (1 6), based on the memorized events (E,), of a piece of information

(11; 12) associated with a feature (PROF1, PROF2) of the environment for a plurality of points of the path.

The system of claim 2, wherein the determining unit is arranged to produce a probability associated with said information (11; 12).

4. System according to claim 2 or 3, comprising an automaton comprising internal registers (R) and arranged to successively traverse the points of the path, to update said internal registers (R) according to the information (11; ) associated with said characteristic for the current point (t) and producing a context detection information (Ci) as a function of said internal registers (R).

The system of claim 4, wherein the controller is adapted to generate, for each current point (t), a confidence index associated with the context detection information (Ci).

6. System according to one of claims 1 to 5, wherein the information identifying the event is a curvilinear abscissa along the path.

7. System according to claim 6, wherein the storage unit is designed to determine said curvilinear abscissa according to a current curvilinear abscissa (s) of the vehicle delivered by an odometer (5).

8. System according to one of claims 1 to 5, wherein the information identifying the event is a time information.

9. System according to one of claims 1 to 8, comprising a driving assistance system (30) designed to activate or deactivate a feature according to the determined road environment characteristic (C1; C2).

The system of claim 9, wherein the feature is a stand-alone driving feature.

1 1. A method of constructing a virtual horizon (vHor), comprising the steps of:

- identification by means of a processing unit (12), by analysis of captured data (CAPT) by at least one sensor (2; 4) fitted to a vehicle, events (E,) occurring in an environment of the vehicle;

- Storing said events (E,) in association each with information identifying the event concerned along a path taken by the vehicle;

determination of at least one characteristic (C1; C2; Ci) of road environment to be assigned to the path on the basis of at least two memorized events (E,).