EP4251485A1

EP4251485A1 - Method and device for determining a virtual boundary line between two traffic lanes

Info

Publication number: EP4251485A1
Application number: EP21810655.7A
Authority: EP
Inventors: Badreddine ABOULISSANE; Ismail Abouessire
Original assignee: PSA Automobiles SA
Current assignee: Stellantis Auto SAS
Priority date: 2020-11-25
Filing date: 2021-10-20
Publication date: 2023-10-04
Also published as: FR3116501B1; WO2022112672A1; FR3116501A1

Abstract

The invention relates to a method and a device for determining a virtual boundary line between a traffic lane referred to as the ego-lane, and a traffic lane which is adjacent to the ego-lane and referred to as the adjacent lane, in order to adjust the position of an autonomous vehicle travelling in the ego-lane, the method comprising the steps of: - receiving (201) information from at least a first detected object travelling in the ego-lane; - receiving (202) information from at least a second detected object travelling in the adjacent lane; - for each of the two objects, if the existence probability is above a predetermined existence threshold and if at least a predetermined number of geolocation points is received (203, 204), then - constructing (205, 206) a line for each object on the basis of the gelocation points; and - determining (207) the virtual line.

Description

DESCRIPTION

Title: Method and device for determining a virtual boundary line between two traffic lanes.

The present invention claims the priority of French application 2012108 filed on 25.11.2020, the content of which (text, drawings and claims) is incorporated herein by reference.

The invention is in the field of autonomous vehicle driving assistance systems. In particular, the invention relates to a method and a device for determining a virtual boundary line between a traffic lane, called ego-lane, and a traffic lane adjacent to the ego-lane, called adjacent lane, for regulating a position of an autonomous vehicle traveling on the ego-lane.

“Vehicle” means any type of vehicle such as a motor vehicle, moped, motorcycle, warehouse storage robot, etc. “Autonomous driving” of an “autonomous vehicle” means any process capable of assisting the driving of the vehicle. The method can thus consist in partially or totally directing the vehicle or providing any type of assistance to a natural person driving the vehicle. The process thus covers all autonomous driving, from level 0 to level 5 in the OICA scale, for Organization International des Constructeurs Automobiles.

The processes capable of assisting the driving of the vehicle are also called ADAS (from the English acronym “Advanced Driver Assistance Systems”), ADAS systems or driver assistance systems. The systems capable of regulating a position of an autonomous vehicle traveling on a traffic lane are known ADAS systems, in particular those called LPA (from the acronym “Lane Positioning Assist”). Conventionally, these systems include on-board cameras capturing images of the vehicle's environment as well as sensors of the RADAR, LIDAR, ultrasound type, etc., capturing data from the vehicle's environment.

Processing of the captured images and data detects, first of all, objects (such as, for example, a vehicle, a truck, a cyclist, a pedestrian, an obstacle, etc.) present in the vehicle's environment (in a vicinity of the vehicle). Next, these detected objects are identified. Each object is tracked over time in order to assign it a probability of existence, to classify it (for example, assign a class of vehicle such as car, truck, bus, motorcycle, ... or a class of non-vehicle such as pedestrian, animal , panel, ...), assign a probability of belonging to a given class, determine a state of motion (in particular, determine if the object is stationary/static, rolling in the same direction as the vehicle, driving in the opposite direction of the vehicle). A determination of the position (in particular relative to the vehicle) of each object is also determined. In particular, this determination identifies on which traffic lane the detected objects are circulating. Generally, this determination is a function of the lateral deviation of the detected object with respect to the vehicle and also of another processing of the images and of the data captured. In case of doubt, the probability of existence of the object is reduced.

This further processing of captured images and data identifies and recognizes lane markings and other lane delineations, and thereby identifies traffic lanes. In order to regulate the position of an autonomous vehicle circulating on a traffic lane, ground markings are essential. Indeed, the width of traffic lanes generally depends on the country, the roads and their maximum speeds. Regulation of the position of the vehicle then takes place with respect to the center line between the line representing a marking to the left of the traffic lane and a line representing a marking to the right of said traffic lane.

However, road markings are sometimes poorly recognized during image processing for reasons, for example, of luminosity (low sun, wet or snowy road, etc.) or, quite simply, of worn, erased or concealed.

In the absence of marking, ADAS systems regulating the position of an autonomous vehicle are more capable of automatically positioning the vehicle and must abruptly return control of the vehicle's operation to a driver. This situation is partially stressful, the driver being provided only at the last moment. Furthermore, the availability of the ADAS system is reduced, which gives a bad brand image.

An object of the present invention is to remedy the aforementioned problem, in particular to virtually reconstruct a dividing line between two tracks. So the operation of ADAS systems capable of regulating a position of an autonomous vehicle is more robust and regular.

To this end, a first aspect of the invention relates to a method for determining a virtual boundary line between a traffic lane, called ego-lane, and a traffic lane adjacent to the ego-lane, called adjacent lane , to regulate a position of an autonomous vehicle traveling on the ego-lane, said method comprising the steps of:

• Reception of information from at least a first detected object circulating on the ego-lane, the information characterizing a probability of existence of the first object and a geolocation point of the first object in the circulated lane;

• Reception of information from at least a second detected object traveling on the adjacent lane, the information characterizing a probability of existence of the second object and a geolocation point of the second object in the traffic lane;

• For each of the two objects, if the probability of existence is greater than a predetermined existence threshold and if at least a predetermined number of geolocation points is received, then o Construction of a line for each object from the points geolocation, the line representing the path traveled by the object; o Determination of a virtual line delimiting the ego-path of the adjacent path from the line constructed for the first object and from the line constructed for the second object.

Thus, in the event of wear, concealment or erasure of a marking on the ground separating the ego-lane and the adjacent lane, the latter is reconstructed and the regulation of the position of the autonomous vehicle in the ego-lane continuous path. Control of the vehicle is not returned to a driver of said vehicle. The driver assistance system continues to fulfill its role of assisting reliably.

The determination of the virtual line is based on the path traveled by the first objects circulating on the ego-lane and on the path traversed by the second objects circulating on the adjacent lane, thus forming two lines. This virtual line can be a median line between the two lines then naturally representing a separation between the ego-path and the adjacent path. The determination of the virtual line is robust and reliable with respect to measurement uncertainties. This determination takes into account a plurality of geolocation points, about twenty for example, per detected object, and each object is detected reliably thanks to a probability of existence greater than an existence threshold (for example 70% ).

Advantageously, the information received for the first and for the second object further comprises a class of membership of the object, a probability of membership of the class, and/or a state of movement, and in which the step of construction of the line for each object is carried out only if the class of membership is a member of a list of predetermined classes, if the probability of membership of the class is greater than a predetermined membership threshold, and/or if the motion state indicates that the object is moving.

Thus, the determination of the virtual line is even more robust and more reliable with respect to measurement uncertainties. In one operating mode, only objects being vehicles in motion, in the direction of the vehicle or coming in the opposite direction, are taken into account.

Advantageously, the construction of the line for each object is based on a parametric adjustment of a polynomial, the polynomial modeling the path traveled by the object.

Thus, the reconstruction of a line representing the path traveled by an object is smoother. Indeed, the measurements are marred by uncertainties and the objects have slight transverse deviations.

Advantageously, when an additional object circulating on the ego-path is detected, the information received is combined with the information of the first object to complete the information of said first object.

Advantageously, when an additional object traveling on the adjacent lane is detected, the information received is combined with the information of the second object to complete the information of said second object.

Thus, the determination of the line is even more robust and more reliable with respect to measurement uncertainties. Indeed, vehicles traveling on the same lane travel the same path. Advantageously, in which each piece of information received is timestamped, and only the data received over a predetermined time interval are used in the construction of a line for each object from the geolocation points. Thus, a global, continuous and periodic processing is carried out determining a virtual line over a given time horizon, for example over 3 seconds.

A second aspect of the invention relates to a device comprising a memory associated with at least one processor configured to implement the method according to the first aspect of the invention.

The invention also relates to a vehicle comprising the device.

The invention also relates to a computer program comprising instructions adapted for the execution of the steps of the method, according to the first aspect of the invention, when said program is executed by at least one processor.

Other characteristics and advantages of the invention will emerge from the description of the non-limiting embodiments of the invention below, with reference to the appended figures, in which:

[Fig. 1] schematically illustrates a device, according to a particular embodiment of the present invention.

[Fig. 2] schematically illustrates a method for determining a virtual line, according to a particular embodiment of the present invention.

The invention is described below in its non-limiting application to the case of an autonomous motor vehicle traveling on a road or on a traffic lane. Other applications such as a robot in a storage warehouse or a motorcycle on a country road are also possible.

FIG. 1 represents an example of a device 101 included in the vehicle, in a network (“cloud”) or in a server. This device 101 can be used as a centralized device in charge of at least certain steps of the method described below with reference to FIG. 2. In one embodiment, it corresponds to an autonomous driving computer.

In the present invention, the device 101 is included in the vehicle. This device 101 can take the form of a box comprising printed circuits, of any type of computer or else of a mobile telephone (“smartphone”).

The device 101 comprises a random access memory 102 for storing instructions for the implementation by a processor 103 of at least one step of the method as described above. The device also comprises a mass memory 104 for storing data intended to be kept after the implementation of the method.

The device 101 may also include a digital signal processor (DSP) 105. This DSP 105 receives data to shape, demodulate and amplify, in a manner known per se, this data.

Device 101 also includes an input interface 106 for receiving data implemented by the method according to the invention and an output interface 107 for transmitting data implemented by the method according to the invention.

FIG. 2 schematically illustrates a method for determining a virtual boundary line between a traffic lane, called ego-lane, and a traffic lane adjacent to the ego-lane, called adjacent lane, to regulate a position of an autonomous vehicle traveling on the ego-lane, according to a particular embodiment of the present invention.

Step 201, "Recep 1", is a step where the device 101, for example, receives information from at least one first detected object circulating on the ego-path, the information characterizing a probability of existence of the first object and a geolocation point of the first object in the route travelled.

In one embodiment, processing of images acquired by a camera and processing of data from sensors of the RADAR, LIDAR, ultrasound type, etc., have detected an object present in the ego-path. For example, by merging the processed data, a geolocation point of the first object in the route traveled is determined and transmitted to the device 101. The geolocation point of a detected object represents for example the projection on the route of the center of gravity image of a detected vehicle. In another example, the geolocation point represents the lateral and longitudinal distance from the autonomous vehicle. A probability of existence of the first object is associated with the geolocation point. It represents the plausibility of the first detected object. For example, this probability has a value between 0 and 1 as its value, and is provided by a conventional perception module.

In another embodiment, the device 101 also receives a class of membership of the first object, a probability of class membership. For example, the detected object is first classified according to a mobile class or not, this class identifying the objects which move. Then, if necessary, the detected object is classified according to a subclass representing the number of wheels (0 wheel,

1 wheel, 2 wheels 4 wheels or more). Then, if necessary, the detected object is classified according to a sub-class representing a personal vehicle, a truck, a bus, etc. This classification is carried out by a module external to the invention conventionally present in a perception module. A class probability is associated with this classification.

In another embodiment, device 101 receives a motion status that identifies whether the object is currently moving in the direction of autonomous vehicle travel, if the object is currently moving in the opposite direction of vehicle travel standalone, or if the object is currently static.

In another embodiment, when an additional object circulating on the ego-path is detected, the information received is combined with the information of the first object to complete the information of said first object. In some use cases, several objects circulating in the ego-path are detected. The information concerning these other objects completes the data of the first object thus multiplying the number of information received.

In another embodiment, each item of information received is timestamped. This makes it possible to use data over a predetermined time interval such as 3 to 5 seconds, other values being possible. For example, the data is stored in memory 102.

Step 202, "Recep 2", is a step where the device 101 for example receives information from at least one second detected object traveling on the adjacent lane, the information characterizing a probability of existence of the second object and a geolocation point of the second object in the traffic lane. This step is similar to step 201, but only concerns objects traveling on the adjacent lane.

Step 203, "Test 1?" is a step that tests whether the information received during step 201 meets certain criteria. For example, to take into account the data relating to the geolocation points in a following step 205 described below, the probability of existence of an object must be greater than a predetermined existence threshold. The existence threshold has a value of 0.7 (i.e. 70%), but can take any other value. In an operating mode, the existence threshold varies according to the use case, such as the type of road on which the autonomous vehicle travels.

For example, to take into account the data relating to the geolocation points in the next step 205, a minimum number of geolocation points must be received. In an operating mode, this number is 20, but it can take any other value and varies according to the use case. For example, if only one first object is detected, the minimum number is 15. If another object circulating in the ego-path is detected, the overall minimum number (first object and other object) is 20.

For example, to take into account the data relating to the geolocation points in the next step 205, the detected object must belong to a class of vehicle with at least 4 wheels and the probability of belonging to this class must be greater than 70% (or another value that may vary depending on the use case).

Thus, “2-wheel” type vehicles are not taken into account. Indeed, the two wheels do not necessarily travel close to the middle of the traffic lane as would a wider vehicle.

For example, to take into account the data relating to the geolocation points in the following step 205, the motion state of the detected object must indicate that the object is in motion. Indeed, in the next step 205, a line representing the path traveled by the object is constructed. If the object is stationary, the line cannot be constructed. In an operating mode, the state of movement also includes a direction of circulation. For example, objects traveling on the ego-lane must travel in the same direction as the autonomous vehicle.

The different examples constitute different criteria for accepting geolocation points. If any combination of these criteria is not met, then there is no not enough data validates to be able to move on to the next step 205. We return to step 201 in order to receive new information.

Step 204, "Test2?" is a step that tests whether the information received during step 202 meets certain criteria.

For example, to take into account the data relating to the geolocation points in a following step 206 described below, the probability of existence of an object must be greater than a predetermined existence threshold. The existence threshold has a value of 0.7 (i.e. 70%), but can take any other value. In an operating mode, the existence threshold varies according to the use case, such as the type of road on which the autonomous vehicle travels.

For example, to take into account the data relating to the geolocation points in the next step 206, a minimum number of geolocation points must be received. In an operating mode, this number is 20, but it can take any other value and varies according to the use case. For example, if only one second object is detected, the minimum number is 10. This number is not necessarily the same as that of step 203 because, in certain cases of use, the vehicles circulating on the lane adjacent travel in the opposite direction with respect to the direction of travel of the autonomous vehicle. If another object traveling in the adjacent lane is detected, the global minimum number (second object and other object traveling in the adjacent lane) is 15. The numbers indicated are examples and can take other values depending on the case. use.

For example, to take into account the data relating to the geolocation points in the next step 206, the detected object must belong to a class of vehicle with at least 4 wheels and the probability of belonging to this class must be greater than 70% (or another value that may vary depending on the use case).

For example, to take into account the data relating to the geolocation points in the following step 206, the motion state of the detected object must indicate that the object is in motion. Indeed, in the next step 206, a line representing the path traveled by the object is constructed. If the object is stationary, the line cannot be constructed. In an operating mode, the state of movement also includes a sense of traffic. For example, objects traveling on the adjacent lane must all travel in the same direction (direction of the autonomous vehicle, or opposite direction) in order to be taken into consideration. In the latter case, the information of the objects carrying out a prolonged overtaking are not taken into account.

The different examples constitute different criteria for accepting geolocation points. If a combination of these criteria is not respected, then there is not enough valid data to be able to move on to the next step 206. We return to step 202 in order to receive new information.

Step 205, “Li 1”, is a step for constructing a line for the first object from the geolocation points which have been retained after step 203, the line representing the path traveled by the object.

In one embodiment, the geolocation points also include the geolocation points of another object traveling on the ego-lane. The line thus constructed represents the path traveled by a fictitious vehicle comprising all the geolocation points.

Thanks to step 203, the geolocation points are reliable and are representative of a center line of the ego-path.

In one embodiment, the construction of the line for each object is based on a parametric fitting of a polynomial, the polynomial modeling the path traveled by the object. For example, in a straight line by a conventional method of linear regression, the best straight line passing through all the geolocation points which have been retained after step 203 is determined. In a preferred operating mode, the order of the polynomial is greater than 1, for example 4, in order to take into account the curvature of the road. The order of the polynomial is to be chosen to meet the compromise between precision and speed of calculation.

Step 206, “Li 2”, is a step for constructing a line for the second object from the geolocation points which have been retained after step 204, the line representing the path traveled by the object.

In one embodiment, the geolocation points also include the geolocation points of another object traveling on the adjacent lane. Line thus constructed, represents the path traveled by a fictitious vehicle comprising all the geolocation points.

Thanks to step 204, the geolocation points are reliable and are representative of a center line of the ego-path.

Step 207, "VirtLi" is the step of determining a virtual line delimiting the ego-path of the adjacent path from the line constructed for the first object, line resulting from step 205, and from from the line constructed for the second object, line resulting from step 206.

The lines resulting from step 205 and from step 206 respectively represent the path traveled by a fictitious vehicle on the ego-lane and on the adjacent lane. In one embodiment, the virtual line is a median line between the line constructed from the first object and the line constructed from the second object, thus clearly indicating the separation between the ego-path and the adjacent path.

In another mode of operation, the method receives information of a width of the ego-path and the construction of the virtual line is based on a half width of the ego-path, and on the line constructed for the first object or on the line constructed for the second object.

In one operating mode, the constructed virtual line replaces a line representing the detection of the marking on the ground delimiting the ego-lane of the adjacent lane in order to regulate the position of the autonomous vehicle by a driving assistance system.

The present invention is not limited to the embodiments described above by way of examples; it extends to other variants. Thus, an exemplary embodiment has been described above in which the determination of the virtual line is carried out after a separate processing of the information received concerning the ego-channel and of the information received concerning the adjacent channel. In an operating mode, the virtual line is built from all the information received and which respects a combination of the stated criteria (existence, number of points, belonging to a class, state of movement, number of points, etc. ). For this, the virtual line can use techniques of the SVM type (acronyms for “Support Vector Machine”) being capable of solving discrimination problems at the cost of making calculations more complex.

Thus, an embodiment has been described above in which the information received comes from a processing of images and data from sensors. In one operating mode, all or part of this information comes from a transmission ensured by a radiofrequency link, such as a communication from vehicle to vehicle (V2V) or communication from vehicle to any other communicating object (V2X).

Claims

1. Method for determining a virtual boundary line between a traffic lane, called ego-lane, and a traffic lane adjacent to the ego-lane, called adjacent lane, to regulate a position of an autonomous vehicle circulating on the self-track, said method comprising the steps of:

- Reception (201) of information from at least a first detected object circulating on the ego-lane, the information characterizing a probability of existence of the first object and a geolocation point of the first object in the circulated lane;

- Reception (202) of information from at least a second detected object circulating on the adjacent lane, the information characterizing a probability of existence of the second object and a geolocation point of the second object in the circulated lane;

- For each of the two objects, if the probability of existence is greater than a predetermined existence threshold and if at least a predetermined number of geolocation points is received, (203, 204), then o Construction (205, 206) a line for each object from the geolocation points, the line representing the path traveled by the object; o Determination (207) of the virtual line delimiting the ego-path of the adjacent path from the line constructed for the first object and from the line constructed for the second object.

2. Method according to claim 1, in which the information received for the first and for the second object further comprises a class of membership of the object, a probability of belonging to the class, and/or a state of motion , and in which the step of constructing (205, 206) the line for each object is carried out only if the class to which it belongs is a member of a list of predetermined classes, if the probability of belonging to the class is greater than a predetermined membership threshold, and/or if the motion state indicates that the object is in motion.

3. Method according to one of the preceding claims, in which the construction (205, 206) of the line for each object is based on a parametric adjustment of a polynomial, the polynomial modeling the path traveled by the object.

4. Method according to one of the preceding claims, in which, when an additional object circulating on the ego-path is detected, the information received is combined with the information of the first object to complete the information of said first object.

5. Method according to one of the preceding claims, in which, when an additional object traveling on the adjacent lane is detected, the information received is combined with the information of the second object to complete the information of said second object.

6. Method according to one of the preceding claims, in which each item of information received is timestamped, and only the data received over a predetermined time interval are used in the construction of a line for each object from the geolocation points.

7. Device (101) comprising a memory (102) associated with at least one processor (103) configured to implement the method according to one of the preceding claims.

8. Vehicle comprising the device according to the preceding claim.

9. Computer program comprising instructions adapted for the execution of the steps of the method according to one of claims 1 to 6 when said program is executed by at least one processor (103).