FR3120694A1 - Method and device for determining the reliability of a base definition cartography. - Google Patents

Abstract

The invention relates to a method and a device for determining the reliability of a base definition map to make the activation of at least one driving assistance system of an autonomous vehicle more reliable, said autonomous vehicle traveling on a road and comprising a navigation system and a perception system, the navigation system comprising said mapping and providing a list of road signs mapped around said vehicle, called a mapped list, the perception system providing a list of measured road signs, called a list measured, each list comprising, for each road sign in the list, a type, a location and contextual information, said method being updated periodically when said vehicle is circulating. The process includes the steps of: Determination (201) of a number indicator; Determination (202) of a contextual indicator; Calculation (203) of a correlation value; Determination (204) of a reliability indicator, reliable or unreliable. Figure for abstract: Figure 2

Description

Method and device for determining the reliability of a base definition map.

The invention is in the field of autonomous vehicle driving assistance systems. In particular, the invention relates to determining the reliability of a base-definition map to make the activation of at least one driver assistance system of an autonomous vehicle more reliable.

“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.

Among the driving assistance systems are known, for example, lane keeping devices, lane change devices, adaptive speed regulation devices, etc.

A vehicle, comprising one of these devices, comprises numerous sensors such as a camera, a RADAR, a LIDAR, ultrasounds, accelerometers, an inertial unit, position or localization sensors, speed sensors, acceleration, etc. Processing of information carried out by at least one on-board computer in the vehicle which makes it possible to perceive the environment is known. By environment, we mean the exterior and the interior of the vehicle. This perception of the environment makes one able to identify and then recognize road signs, also called signs, characteristics of a road and/or a lane, etc.

Furthermore, this vehicle also includes a navigation system which includes a means for locating and a map. There are different types of mapping. A high definition cartography, called HD cartography, characterizes the lanes and attributes related to the road such as a number of lanes, curvatures of the road, a slope of the road, road signs, etc. Moreover, with HD cartography, the navigation system, which is connected with servers outside the vehicle via telecommunications links, regularly updates the map. Thus, the attributes related to the route are thus very up to date. Moreover, in an HD cartography, the attributes are well positioned (resolution of 1 meter).

In an SD cartography, simple definition, the characterization of the ways is less precise and the attributes are not always well digitized (lack or less precise positioning of the order of several meters). Mapping is also less regularly updated. If this update is carried out manually and annually from a database of a mapping supplier, the data is then on average dated 6 months.

Through regular updating and accuracy of attributes, HD mapping is intrinsically reliable. Thus, it is known that ADAS systems of a vehicle comprising HD mapping, use the information from the mapping to improve the functionalities of these ADAS systems. For example, these functionalities are able to act on the longitudinal and transverse dynamics of the vehicle by activating a driving assistance system (for example modifying a set speed, modifying the positioning of the vehicle in the lane, etc.).

Unfortunately, the cost of onboard HD mapping is very high compared to the manufacturing cost of a vehicle. If the vehicle is fitted with an SD cartographer, it is dangerous to condition activations or modifications to the instructions of the driving assistance systems which act directly on the direction and/or on the speed of the vehicle. An SD map is not reliable enough.

In order to use map information by a driving assistance system, devices are known that calculate an instant map reliability indicator. In particular, the location precision given by a device of the GPS type (from the English “Global Positioning System”) is used. However, these devices calculating an instantaneous map reliability indicator, on the one hand, assume that the information to calculate the indicator is continuously available and, on the other hand, remain quite conceptual in their implementation.

Unfortunately, the perception of the environment requires a large volume of information to be processed, and therefore a high computational load. Thus in certain cases of use (many vehicles around the autonomous vehicle, strong accelerations or decelerations, etc.) or weather conditions, certain perception information is regularly missing or arrives late. Moreover, the perception information resulting from image processing is also regularly missing due to the available visibility (for example, a truck can hide road signs on the side of the road). Thus, the calculations of an instantaneous map reliability indicator periodically when the vehicle is traveling on a road vary suddenly and randomly. The indicator is then unstable, unreliable and unusable, for safety reasons, in driving aid systems.

An object of the present invention is to remedy the aforementioned problem, in particular to make the information, characteristics of the track and attributes, resulting from an SD map reliable.

To this end, a first aspect of the invention relates to a method for determining the reliability of a base definition map to make the activation of at least one driver assistance system of an autonomous vehicle more reliable, said autonomous vehicle traveling on a road and comprising a navigation system and a perception system, the navigation system comprising said cartography and supplying a list of road signs mapped around said vehicle, called the mapped list, the perception system supplying a list of signs measured road signs, called measured list, each list comprising, for each road sign in the list, a type, a location and contextual information, said method being updated periodically when said vehicle is traveling and comprising the steps of:

• Determination, from the mapped list and the measured list, of a number indicator indicating a number of road signs matching in type and location between the two lists;
• Determination, from the mapped list and the measured list, of a contextual indicator, indicating a concordance on the context between the two lists;
• Calculation of a correlation value from the number indicator, the context indicator and a previously calculated correlation value, the correlation value increases or decreases by a gradient depending on the number indicator and the contextual indicator;
• Determination of a reliability indicator, reliable or unreliable, the indicator being reliable if the correlation value is greater than a predetermined threshold.

Thus, if there is good reliability, this means that for a certain time, a certain number of periods, the road signs detected by the camera are substantially the same as those resulting from the cartography. Thus, the navigation system positions the vehicle well on the map, and also the map is locally reliable (from the detection distance of the camera). There is no need to use location precision information given by a GPS-type device, which remains too basic. A driver assistance system that controls the longitudinal and/or lateral dynamics of the vehicle can rely on data from the map.

Using both indicators makes a correlation value update predictive and responsive. It is predictive thanks to the number indicator which is based on a distant vision of a camera for example (up to 300 meters for example). It is reactive thanks to the contextual indicator which is based on a vision very close to the vehicle, the recognition of the speed is only known in the immediate environment of the vehicle (between 0 and 10 meters for example).

Moreover, this determination of reliability is also robust to short random errors of detection and localization due to a computational overload. This robustness is due to the recursion of the calculation of the correlation value and to a determination by level.

Advantageously, the determination of the reliability indicator is on three levels, if the correlation value is greater than a predefined high threshold, then the reliability is currently good, if the correlation value is greater than a predefined intermediate threshold lower than the high threshold then the reliability is previously good, otherwise the reliability is insufficient.

Thus, if good reliability has previously been determined, then if the correlation value decreases in the absence of detection of new road signs, we do not pass directly to an insufficiency of reliability which risks stopping a driving aid suddenly.

Advantageously, the gradient is a value between -1 and 1, and the correlation value is initialized at 0 then is bounded between 0 and 1.

Advantageously, the larger the number indicator, the higher the gradient.

Advantageously, if the contextual indicator indicates a match then the gradient is positive.

Advantageously, if the contextual indicator indicates a mismatch then the gradient is negative, unless the number indicator is greater than a predefined threshold where in this case the gradient is positive.

Thus, when several road signs agree, this indicates that said vehicle is correctly positioned on the map even if there is no speed agreement. The cartography is locally reliable, it can be used, for example by a trajectory control system. The gradient can be positive, which increases the correlation value even if there is doubt about the speed. The doubt may be due to poor recognition at a given time (road sign too far away) or to a recent or temporary change.

Advantageously, if the number indicator is zero while the list of road signs measured includes at least one road sign that is not zero, then the gradient is negative.

Thus, this indicates either incomplete, imprecise or outdated mapping, or a local and temporary modification, for example construction site, of the road on which said vehicle is traveling.

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:

schematically illustrates a device, according to a particular embodiment of the present invention.

schematically illustrates a method for determining the reliability of a base definition map, 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.

There 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 the . 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 even 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.

There schematically illustrates a method for determining the reliability of a base definition map, according to a particular embodiment of the present invention.

The method is implemented by a device 101 in an autonomous vehicle which comprises at least one driver assistance system. Said autonomous vehicle travels on a road and includes a navigation system and a perception system. Said navigation system includes said cartography and provides a list of road signs mapped around said vehicle, called mapped list. Said perception system provides a list of measured road signs, called the measured list.

Each list, mapped list and measured list, includes, for each road sign in the list, a type, a location and contextual information.

By type of signaling, we mean, without being limiting, signaling by panels, signaling by lights, signaling by marking of the roadways, signaling by beaconing, signaling by demarcation, and/or signaling by closing devices. In one embodiment, a type of signage is more specific and includes, as for example for signs, signs indicating a danger, signs comprising an absolute prescription, signs comprising a simple indication or indicating a direction, relative signs at intersections and priority schemes, and/or placards. Also, a type is capable of comprising a shape, such as for example a panel, a triangle, an inverted triangle, a circle, a square, a rectangle, a hexagonal shape, or any other geometric shape. For example, one type is a speed limit sign. Using a camera on board the vehicle and associated image processing, it is known to recognize a type of signaling at a distance of 300 meters for example or even more.

By contextual information, we mean a meaning, a valuation, an interpretation and/or giving a more precise meaning to a signaling recognized by a camera and its associated image processing. For example, without being limiting, contextual information is a number of traffic lights, a speed limit value of a speed limit sign, a mandatory right direction at the next intersection, an identification of the image (car, truck , pedestrian cyclist, …) … Using a camera embedded in the vehicle and an associated image processing, it is known to recognize the contextual information of a signaling at a distance of 10 meters for example, and often, the recognition information is only available when the vehicle is at the signaling level, therefore without anticipation.

By location, we mean a relative distance, longitudinally, laterally and/or vertically, of the location of the road signs in relation to the current position of the vehicle. In a preferred operating mode, a list comprises the locations around the vehicle in a horizon covering, for example, from 3000 meters in front of the vehicle to −500 meters (therefore behind the vehicle). A measured list and a mapped list include the same horizon. Thus the measured list comprises a memory of nearby past detections. For example, if, at a time T, a traffic sign with a speed limit value has been identified and located 0 meters from the vehicle, at time T+3 seconds, for a vehicle traveling at 120 km/h, the road sign with the speed limit value is located at -100 meters (therefore 100 meters behind the vehicle).

When the vehicle is moving, said method is updated periodically, for example every 100 ms, but another value is possible. In one embodiment, the update is also made on a detected event or on a request from one of the driving assistance systems.

Step 201, DetInbr, is a step for determining, from the mapped list and the measured list, a number indicator indicating a number of road signs matching in type and location between the two lists. The concordance is a simple comparison between the two lists. For example, for each type of signaling in the mapped list, it is checked whether the same type of signaling with the same location is present in the measured list. If so, there is a match and the count indicator increases by a value of 1, the count indicator being initialized to 0 at the start of the procedure.

Since the data of a cartography are imprecise (common location error of the order of 2 meters) and since the measurements are always marred by errors (location of the order of 1 to 30 meters depending on the relative distance from the signaling recognized with respect to the vehicle), the location of a mapped signaling and of a measured signaling is the same if the difference in location is less than a predefined threshold. For example, this threshold is 5 meters and/or is variable as a function of the relative distance with respect to the vehicle (for example, the further a location is from the vehicle, the greater the threshold).

In this step, the number indicator is higher than the number of matches found. A type match is equally robust against errors on an SD map. For example, if locally a speed limit has been changed, from 90 km/h to 80 km/h on national roads, while the cartography is not up to date, the concordance in type and location is valid for this same signaling if the speed limit value is not equal between the measured signage and the mapped signage.

Step 202, DetIctx, is a step for determining, from the mapped list and the measured list, a contextual indicator, indicating a concordance on the context between the two lists. The determination of the contextual indicator can be done by several methods. For example, one method is to create a boolean flag, context match or no match. There is a cotext match if a signal from the mapped list matches in type, location and context with a signal from the measured list. A match in type or context for a signal means the same value in both lists. A match in location means a proximity, a location deviation below a predetermined threshold or variable according to the relative distance with respect to the vehicle as described above.

In a preferred mode of operation, the context indicator has three values. The context indicator takes a first value, called current context match, if there is a match in type, location and context in a close vicinity of the vehicle. The close neighborhood being, for example, a distance of 50 meters around the vehicle, in particular in front of and behind the vehicle. Location matching is determined to within a threshold. The first value corresponds to a situation where the road sign is close to the vehicle and therefore easily recognized by the perception device or which has been recognized a few moments previously. In a pocket neighborhood of the vehicle, context detection is very reliable.

In the absence of a determination of a present context match, the context indicator takes a second value, called a past context match, if there is a match in type, location and in context outside the close vicinity of the vehicle. Beyond the close vicinity, if the signaling is in front of the vehicle, if a context is recognized (rare case), this recognition is less reliable. If the sign is behind the vehicle, it has been previously recognized for some time and/or distance. This recognition is also less reliable. The second value corresponds to a less reliable context match. In the absence of determination of a past context match, the match indicator takes a third value corresponding to an absence of context match.

Step 203, ValCor, is a step of calculating a correlation value from the number indicator, the context indicator and a previously calculated correlation value, the correlation value increases or decreases d 'a gradient function of the number indicator and the context indicator. The gradient is a value between -1 and 1, and the correlation value is initialized to 0 and then bounded between 0 and 1.

Different methods can be used to calculate the correlation value. In a first example, a numerical value is assigned to the contextual indicator then, depending on the product between the contextual indicator and the number indicator, a gradient value is assigned. For example, the higher the product, the higher the gradient.

In a preferred mode of operation, the gradient is determined by a predetermined table depending on the number indicator and the context indicator as, for example in the table below. Gradient value Context indicator Lack of agreement Past match Current match Number indicator 0 -0.001 0 0.001 1 -0.001 0.001 0.002 2 0.001 0.002 0.003 >3 0.002 0.004 0.006

Thus the gradient is even higher when there is agreement in type, location and context in the two lists. Conversely, the gradient is all the more small and/or negative in absence or low agreement. Advantageously, the larger the number indicator, the higher the gradient. Advantageously, if the contextual indicator indicates a match then the gradient is positive. Advantageously, if the contextual indicator indicates a mismatch then the gradient is negative, unless the number indicator is greater than a predefined threshold where in this case the gradient is positive.

Other rules are also possible to determine at each period a gradient value according to the concordances. For example, if the perception device fails to recognize the type of signaling, the gradient is chosen negative.

Step 204, DetIfia, is a step for determining a reliability indicator, reliable or unreliable, the indicator being reliable if the correlation value is greater than a predetermined threshold.

As the correlation value increases or decreases by the gradient determined at each period, a correlation value close to 1 indicates numerous matches over a fairly long period (several tens of seconds), the correlation between the map and the measurements is very good.

Conversely, in the absence of agreement or little agreement the gradient is negative, the correlation value tends towards 0 indicating a poor correlation between the map and the measurements.

For example, if the correlation value is greater than a threshold of 0.66, the map is reliable. The reliability indicator indicates a reliable card. Conversely, the reliability indicator indicates an untrusted card.

In a preferred operating mode, the determination of the reliability indicator is on three levels. If the correlation value is greater than a predefined high threshold, for example 0.66, then the reliability is currently good. Otherwise, if the value of correlation is higher than a predefined intermediate threshold lower than the high threshold, for example 0.33, then the reliability is previously good. Otherwise the reliability is insufficient. Thus, the history is taken into account and random and unstable measurement errors are filtered on the one hand by a small variation of the gradient and on the other hand by thresholding.

The present invention is not limited to the embodiments described above by way of examples; it extends to other variants. For example, the number indicator takes a value indicating non-identification of a signal. The gradient is then determined as a function also of this new value.

In another example, numerical values have been given in particular for the gradients. Other values are possible and depend, for example, on the update period of the procedure.

Claims (10)

Method for determining the reliability of a base definition map to make the activation of at least one driver assistance system of an autonomous vehicle more reliable, said autonomous vehicle traveling on a road and comprising a navigation system and a perception system, the navigation system comprising said cartography and supplying a list of road signs mapped around said vehicle, called the mapped list, the perception system supplying a list of measured road signs, called the measured list, each list comprising, for each road signs in the list, a type, a location and contextual information, said method being updated periodically when said vehicle is traveling and comprising the steps of:

• Determination (201), from the mapped list and the measured list, of a number indicator indicating a number of road signs matching in type and location between the two lists;
• Determination (202), from the mapped list and the measured list, of a contextual indicator, indicating a concordance on the context between the two lists;
• Calculation (203) of a correlation value from the number indicator, the contextual indicator and a previously calculated correlation value, the correlation value increases or decreases by a gradient depending on the indicator number and contextual indicator;
• Determination (204) of a reliability indicator, reliable or unreliable, the indicator being reliable if the correlation value is greater than a predetermined threshold.

Method according to claim 1, in which the determination (204) of the reliability indicator is on three levels, if the correlation value is greater than a predefined high threshold, then the reliability is currently good, if the correlation value is higher than a predefined intermediate threshold lower than the high threshold then the reliability is previously good, otherwise the reliability is insufficient. Method according to one of the preceding claims, in which the gradient is a value between -1 and 1, and the correlation value is initialized at 0 and then is bounded between 0 and 1. Method according to one of the preceding claims, in which the larger the number indicator, the higher the gradient. Method according to one of the preceding claims, in which if the contextual indicator indicates a match then the gradient is positive. Method according to one of the preceding claims, in which if the contextual indicator indicates a mismatch then the gradient is negative, unless the number indicator is greater than a predefined threshold in which case the gradient is positive. Method according to one of the preceding claims, in which if the number indicator is zero while the list of road signs measured includes at least one road sign is non-zero, then the gradient is negative. 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. Vehicle comprising the device according to the preceding claim. Computer program comprising instructions adapted for the execution of the steps of the method according to one of claims 1 to 7 when said program is executed by at least one processor (103).