CN116324907A

CN116324907A - Method for determining the probability of the presence of a possible element in a motor vehicle environment, driver assistance system and motor vehicle

Info

Publication number: CN116324907A
Application number: CN202180070273.1A
Authority: CN
Inventors: J·弗里克; C·普拉赫特卡; B·雷赫
Original assignee: Volkswagen AG
Current assignee: Volkswagen AG
Priority date: 2020-11-03
Filing date: 2021-10-12
Publication date: 2023-06-23
Also published as: DE102020213831B4; DE102020213831A1; WO2022096236A1

Abstract

The invention relates to a method for determining the presence probability (W) of a possible element (14) in the environment (16) of a motor vehicle (10). The method comprises providing map data (18 a) and sensor data (D) of an environment (16) of the motor vehicle (10), comparing the map data (18 a) and the sensor data (D) with respect to the presence of possible elements (14), and determining a presence probability (W) from the result of the comparison, wherein the possible elements (14) are classified as present at least on the basis of the map data (18 a) and/or the sensor data (D). A rule database (36) is provided, in which information about the traffic space is stored, wherein the presence probability (W) is determined on the basis of at least one first information provided by the rule database (36), at least in the case that a divergence in the presence of at least one possible element (14) is ascertained when comparing the map data (18 a) with the sensor data (D).

Description

Method for determining the probability of the presence of a possible element in a motor vehicle environment, driver assistance system and motor vehicle

The invention relates to a method for determining the probability of the presence of at least one possible element in a motor vehicle environment, wherein map data of a map of the motor vehicle environment are provided, said map data relating to a first environment region comprising the at least one possible element; providing sensor data by means of at least one environment detection device of the motor vehicle, the sensor data relating to a second environment region comprising at least one possible element, and wherein the map data and the sensor data are compared with respect to the presence of the at least one possible element. Furthermore, a presence probability is determined and output as a function of the result of the comparison, wherein at least one possible element is classified when present based at least on map data and/or sensor data. The invention further relates to a driver assistance system and a motor vehicle.

Currently, the automatic driving function of a motor vehicle generally relies on driving-related information stored in a high-definition map, such as a lane topology or road right rules, without checking it. However, some of the information stored in such high definition maps may be erroneous, for example if an error occurs in the map generation or not updated when the environment changes (e.g., in the case of a construction site). In order to detect possible errors in such a map, environmental detection may be performed using vehicle sensors. The sensor detection and the divergence between the map data represent a possible error in terms of either environmental detection or map data, respectively. However, in this case, it is generally not possible or not particularly reliable to determine whether the error is present in the map data or in the environment detection. However, for precautions, the relevant elements in the map are judged to be problematic, and in order to avoid dangerous situations arising therefrom, the motor vehicle that is performing the autopilot function is shifted to a safe state in connection with the stopping of the vehicle. However, stopping of the vehicle may also be a safety hazard, as for example a rear-end collision may be initiated. Furthermore, it is very limited in comfort, as it does not achieve the intended destination. Accordingly, it is desirable to be able to evaluate the existence probability of elements in such a map more reliably.

Document WO 2018/127461A1 describes a method of providing a high definition map, in particular a first high definition map resulting from a data fusion of sensor data and map data of a second high definition map. The generated data can thus be supplemented by additional information provided by the sensor. But also map data can be replaced by sensor data. It follows that the sensor data is more accurate and reliable. The reliability of the data is limited here by the reliability of the vehicle sensor system. However, situations in which the sensor data may be erroneous or may not be sufficiently accurate are not considered here.

Furthermore, document WO 2018/126215 A1 describes a method for updating a high-definition map based on detected sensor data of a plurality of motor vehicles travelling through a geographical area. If there is a divergence or difference in the collected sensor data, an online system can be evaluated, to which the data is transferred.

In principle, fleet vehicles may be used to detect errors that may occur in map data. However, there is a general problem in that the map updating according to this cannot be carried out in real time, but only at significantly longer time intervals, for example weeks or months.

The object of the present invention is therefore to provide a method for determining the probability of the presence of at least one possible element in the environment of a motor vehicle, a driver assistance system and a motor vehicle, which enable a possible element in the environment of a motor vehicle to be evaluated as being present or absent as quickly and reliably as possible, preferably in real time during driving.

The object is achieved by a method, a driver assistance system and a motor vehicle having the features according to the independent claims. Advantageous embodiments of the invention are the subject matter of the dependent claims, the description and the figures.

In a method according to the invention for determining the probability of presence of at least one possible element in the environment of a motor vehicle, map data of an environment map of the motor vehicle are provided, which map data relate to a first environment area comprising the at least one possible element. Furthermore, the sensor data is provided by means of at least one environment detection device of the motor vehicle, which relates to a second environment region comprising the at least one possible element. Further, the map data and the sensor data are compared in terms of the presence of at least one possible element, wherein the at least one possible element is classified as present based on at least the map data and/or the sensor data, and the presence probability is determined and output according to the result of the above comparison. Furthermore, a rule database is provided, in which information relating to the traffic space is stored, wherein the presence probability is determined from at least one first information provided by the rule database, at least in the case that a divergence regarding the presence of at least one possible element is ascertained when comparing the map data and the sensor data.

The invention is based on the insight that, in order to resolve the divergence between the sensor data and the map data, knowledge about the traffic space, in particular knowledge about driving and safety-related aspects of the traffic space, can be used, which knowledge can be stored in a rule database in the form of rules. Thus, knowledge explicitly expressed in the rules database can be advantageously used as additional information in order to resolve possible contradictions between the sensor data and the map data. The knowledge can be confirmed in particular in the form of rules which can include traffic rules on the one hand, but which also allow conclusions from facts detected exclusively or recorded by the map on the other hand. For example, the rules database allows conclusions to be drawn: crossroads without traffic signs mean "right before left". This allows modeling of road weights for each branch of the intersection. Furthermore, a speed limit of 50km/h can be deduced from the entry signs. Reliable or unreliable associations between elements may also be validated by the rules database. It is thereby possible to confirm which landmarks can coexist in the same place and which landmarks cannot. For example, prohibited overtaking may be combined with speed limiting. However, the combination of the stop flag and the forward flag is contradictory. Such association rules may be stored in a rules database and used in case of a divergence in terms of possible elements between the sensor data and the map data in order to be able to more accurately determine the existence probability of the element in order to then correctly judge, for example, based on the existence probability, with a higher probability, whether the sensor data or the map data is correct in terms of the existence probability of the element. In summary, this knowledge of the rules database can be used in order to ultimately determine the probability of the presence of a particular element, here for example a landmark. If, for example, the element is given by a map and is not detected based on sensor data, this may likewise have a number of reasons. For example, it may be that the element is exactly obscured by a traffic participant traveling in front. Conversely, if such a forward traveling traffic participant is detected by means of sensor data at the location where the element should be detected, it is considered more likely that: the element is still present but is simply occluded by another traffic participant, rather than the element being absent. Such conclusions may be specified by the rules database. That is to say, in particular, the presence of a possible element can be checked for compliance with at least one rule of the rule database. The result of this check can then be used for the determination of the probability of existence. If the existence of the possible element does not meet the rule, the existence probability of the possible element is degraded, and vice versa, i.e., if the existence of the possible element meets the rule, the existence probability of the possible element is promoted. Thus, by providing a rule database, an additional device is provided from which the probability of the presence of an element can be determined in addition to the sensor data and the map data. By comparing the environmental information provided by the sensor data and the map data with such a rule frame, a generally more reliable conclusion can thus be provided about the presence of such elements in the environment of the motor vehicle. In other words, the probability of the presence of a possible element in the automotive environment can be determined more accurately by additional consideration of the information available in the rules database. This ultimately results in the fact that the presence probability thus determined maximally achieves a more definite conclusion. The presence assumption can thus be checked and corrected if necessary, in particular in real-time and thus during the driving or normal operation of the vehicle. This makes it possible to resolve possible divergences between the sensor data and the map data without having to interrupt the automatic driving for safety reasons, and thus to achieve safe automatic continued driving of the vehicle even in the event of such a more frequent occurrence, i.e., in the event of divergences, which in turn increases road safety and improves driving comfort.

The probability of existence itself may be determined from a predetermined metric. Such metrics are known in a sufficient manner and are therefore not specified in more detail within the scope of the present invention. For example, the presence probability may be provided as a weighted average of a plurality of different parameters that have an influence on the presence probability. In addition to the knowledge provided by the rules database as such parameters, a variety of other parameters may be used in the averaging of the presence probabilities, such as the reliability of the respective sensor data and map data, the degree of matching between the sensor data and map data, the type of environment detection device providing the sensor data, the number of environment detection means used for environment detection and providing the same result of the presence of the element, for example, etc. All these parameters can be used accordingly for weighted averaging, for example such a predetermined measure, in order to finally provide the probability of existence of the possible element concerned.

Possible elements relate to elements classified as present at least according to map data and/or sensor data, as defined above. This means that either the map data indicates the presence of the element, wherein the element is not necessarily also detected by the environment detection means, for example a sensor of a motor vehicle, or vice versa, i.e. the element has been detected by the at least one environment detection means, irrespective of whether the presence of the element is also proven in the map data. In particular, it may be provided that the presence probability is determined from the information provided by the rule database only if a divergence regarding the presence of at least one possible element is confirmed. This means that if the elements are classified as stored or present, for example, based on map data and sensor data, the rule database need not be used. When the sensor data matches the map data, it can be assumed that the probability of the element being present is very high, so no further verification is required for this. However, it is preferred that, even so, the (logical) consistency of the map elements is checked in the general context, i.e. with consideration of all existing information or elements, by means of a rule database.

Further, the map data provided may be map data of a high definition map to which a description is input. Such high definition maps may contain a large amount of high resolution information about the environment of the motor vehicle, in particular information about the route of roads and lanes and driving rules, in particular traffic rules, applicable to these roads and lanes.

In order to provide sensor data, the motor vehicle may have not only one environment sensor or an environment detection device, but also a plurality of environment sensors. The at least one environment detection device may furthermore be an ultrasonic sensor and/or a laser radar (light detection and ranging), such as a laser scanner and/or a radar and/or a camera. The motor vehicle may in particular comprise a plurality of such sensors in any combination and number. In particular, a 360 ° region around the motor vehicle can be detected by the at least one environment detection device.

The results of the method in determining the probability of existence may be sent to a back-end, i.e. an internet server and/or a service provider, e.g. a service provider for maps, which itself manages the map data and is able to update the map data according to the communicated probability of existence. For example, existing maps, for example maps stored in a vehicle, can be verified or updated by the service provider according to the existence probabilities thus determined. For example, if it is confirmed according to the described method that a possible element classified as present according to the map still has a very low probability of existence, the element may be removed from the map. Vice versa, an element is not present in the map, however, since the method classifies the probability of the presence of the element as high, the element can be supplemented into the map. In this case, the detection of individual vehicles can be summarized at the rear end and checked for temporary (e.g. construction site) or permanent (e.g. new sign) vehicles. This is especially possible when multiple vehicles transmit such information to such a back-end independently of each other. The high definition map cannot be (permanently) modified by detection of only a single vehicle. However, it is also conceivable to temporarily change the map data in the vehicle in accordance with the determined existence probability.

Based on the determined existence probabilities, a currently valid static environment model may be automatically generated in the vehicle. This means that the high definition map itself, even a single vehicle, does not have to be modified. A new model of the environment is composed of only said information, which model is valid only for the current time. The actual modification may then be made at the back end, which follows a map update for all vehicles.

However, it is particularly advantageous if the finally determined probability of existence can be provided directly to other systems of the motor vehicle. The probability of existence determined in this way can also be used in particular directly for other driver assistance systems of the motor vehicle, for example for scene interpretation or for driving planning in the automatic driving category. A great further advantage of the method according to the invention is also shown here, since by determining the presence probability, it is now possible to provide a map, in particular a high-definition map, with the respective elements, which have been validated according to the method if necessary, for a subsystem, for example for the scene interpretation and/or driving planning, but also to assign the individual elements their associated presence probabilities. The background is that not every element or its presence is equally important for every subsystem in a motor vehicle. In this case, each subsystem also has an inherent requirement for the reliability of the information used in the high-definition map, so that the probability of the generated hypothesis, i.e. the evaluation of the presence of possible objects, can be determined on the basis of the current vehicle environment system-specific and optionally element-specific basis. For example, information about a sidewalk parallel to a lane is more important to the scene interpretation subsystem than to the path planning subsystem because it supports classifying dynamic objects as pedestrians because areas outside the driving lane are not considered in the trajectory planning. Which requirements are currently set by the respective subsystems for the respective existence probabilities of the respective elements (e.g. which thresholds of the existence probabilities must be exceeded or undershot in order to be able to make reliable assumptions) are then decided by the subsystems themselves or can be defined individually for the respective subsystems.

Overall, the reliability of the evaluation of the presence probability of elements in the environment of the motor vehicle is significantly improved and the function of the individual driver assistance systems and subsystems is also significantly improved.

The traffic space covered by the information provided by the rules database is preferably not a geographical space here, but rather an abstract space with elements, rules and other knowledge related to traffic.

The information stored in the rules database may preferably be divided as follows: contextual knowledge, action knowledge, misbehavior, and traffic rules. For example, the information about traffic rules relates to rules specified by the road traffic law. The context knowledge then allows defining, for example, a set of contexts. For a particular set of contexts, such as crosswalks, highway entrances, loop intersections, etc., a minimum set of elements defining the set of contexts may be specified by context knowledge. For example, a minimum set of elements belonging to the group of situations crosswalk and/or crosswalk signs and/or pedestrian traffic lights and/or adjacent pedestrian roads. For example, if it is assumed that a crosswalk exists at a certain position based on map data, but the crosswalk is not detected based on sensor data, it can still be concluded that: if, for example, crosswalk signs and crosswalk traffic lights, i.e. a plurality of elements assigned to a crosswalk context group, are detected on the basis of sensor data, then a crosswalk is highly probable to be present even if the crosswalk or zebra crossing itself is not recognized. Furthermore, the action knowledge can ascertain how to interpret the movement profile of the other traffic participants, which can be detected by the at least one detection device of the motor vehicle. Such a movement profile (also referred to as a trajectory) can be used, for example, to deduce the course of a lane, the presence of a stop sign, etc. Rules for such conclusions may also be specified in the action knowledge. For example, it may be provided that the parking of other traffic participants means that there is a right of way, a zebra crossing, a stop sign, etc. at the location. Knowledge about the misbehaviour may also be stored in the rules database. This means that the incorrect behaviour of other traffic participants can also be taken into account when determining the probability of existence. For example, it follows that many traffic participants exceed the speed limit. For example, even if many traffic participants travel at a speed of 90km/h, a speed limit of 80km/h may still be effective at present, and should not be rejected as impossible simply because the sensor data detects many other traffic participants at speeds slightly faster than 80 km/h. In other words, certain errors may be accepted in the motivational-based verification that take into account the erratic behavior of other traffic participants.

Furthermore, the knowledge stored in the rules database may be interrelated to enable further conclusions to be drawn. For example, if a plurality of traffic participants park at a particular location where, for example, there should be a zebra crossing based on map data, but not directly detected based on sensor data, then a conclusion may be drawn that other traffic participants park at that location based on sensor data: the probability of zebra crossings being present is still extremely high. Thus, knowledge of actions may also be combined with knowledge of context as defined above. For example, it is possible that there are many traffic participants parked at a particular location, and if other elements of the crosswalk context group are detected at that location as well, then there is a zebra crossing at that particular location, rather than a stop sign.

These are just some examples of how knowledge or information provided by a rules database may be utilized to improve the determination of the probability of existence of possible elements in the environment of a motor vehicle, particularly in the case of differences between sensor data and map data. Furthermore, the rule database may be stored in a memory of the motor vehicle and/or provided outside the vehicle, in which case the information may be recalled from the motor vehicle at any time. It is theoretically possible that the above-described method steps can be carried out outside the motor vehicle, except for the provision of sensor data. The result of the method, i.e. the probability of existence, can then be transmitted back to the motor vehicle accordingly. However, it is preferred that the entire method is carried out in a motor vehicle, since the sensor data are provided in the motor vehicle, so that a large amount of data need not be transmitted to the motor vehicle external device. This also advantageously makes it possible to make driving-related decisions in real time during driving based on the probability of existence, since the calculation time is sufficiently short and there is no additional data transfer time. Furthermore, the result of this method can also be used directly in the motor vehicle, in particular in real time, which significantly increases the safety of the autopilot and significantly reduces the possibility that the motor vehicle has to be converted into a "safe state".

In a further advantageous embodiment of the invention, a divergence regarding the presence of at least one possible element is confirmed when comparing the map data and the sensor data if the at least one possible element is classified as a newly detected element. This means that the possible elements are detected based on the sensor data and are not present in the map data. If at least one possible element is classified as an unevaluated element, which is present in the map data but not detected from the sensor data, the presence of a divergence is also confirmed. This means that the element is not necessarily not detected at all from the sensor data, but cannot be evaluated, for example, based on the sensor data, for example, because the sensor system does not see the element, for example, because it is blocked, or because of errors in the algorithm, or because of poor sensor data quality, for example, due to bad weather, etc. In this case, precisely, a divergence defined in this way is ascertained between the sensor data and the map data, in which case it is particularly advantageous to have further evaluation criteria, namely a rule database, from which it can be determined whether a possible element is present.

Furthermore, the at least one possible element may be a physical element related to an object that may be present in the environment. Examples of such solid elements include traffic lights, road signs, road markings, lane markings, sidewalks, crosswalks, and the like. For example, in the case of a construction site, these physical elements may be added or disappeared. Particularly in this case, the lanes are frequently modified, new road signs are placed or road right rules are changed. Therefore, it is very important to correctly identify the physical elements, especially in the case of autopilot.

However, not only the existence probability of the entity element but also the existence probability of the semantic element can be determined. Which accordingly represents another advantageous design of the invention, if at least one possible element is a semantic element, it represents a logical group of entity elements or at least one traffic rule. Additional information in the form of semantic elements is also included in, for example, maps (such as the maps described above), in particular high definition maps. Such semantic elements may represent traffic rules, such as road right rules for intersections. Furthermore, individual elements, in particular individual entity elements, may be logically grouped and assigned to higher-level groups. Which may also be referred to as a combination. Such supergroups may also represent semantic elements. Relationships between elements may also be used and derived. For example, the dashed line and the solid line may be combined into a lane. In addition, the traffic islands and lanes may be combined into a ring intersection. Accordingly, lanes and annular intersections are semantic elements. Furthermore, a plurality of semantic elements are combined, i.e. a combination between elements, for example a combination between a traffic light and a lane assigned to the red-green light. Another example is to assign a park line to a park mark, for example. In addition, as previously described, the right-of-way rule is also a semantic element. For example, a road right rule is also stored as a semantic element for each intersection in the map. However, such semantic elements may not only be stored in the map, but may also be detected by at least one detection device of the motor vehicle or derived from the detected sensor data. Individual physical elements that can be assigned to a semantic supergroup can be detected by motor vehicle sensors. For example, if the traffic islands and lanes are detected by at least one environment detection device, the ring intersection as a semantic element may be considered to be detected by at least one environment detection device, or rather by an algorithm that processes sensor data from the environment detection devices and classifies the ring intersection as detected accordingly. Thus, not only the presence of entity elements can be evaluated, but also many semantic elements related to traffic and driving can be evaluated.

Furthermore, according to a further advantageous embodiment of the invention, it is provided that if the possible element is classified as a newly detected semantic element on the basis of the comparison, the probability of existence of the possible element is determined from the existence of the newly detected entity element. In this case, the probability of the presence of a newly detected entity element can be evaluated in advance, i.e. only the newly detected entity element that may be present is considered. Therefore, if a new element not included in the map data is detected from the sensor data, the existence probability of the newly detected entity element may be first evaluated, for which purpose the rule database is used as described above. However, the newly detected semantic elements which have likewise been detected by the vehicle sensors but are not contained in the map data can be evaluated in terms of their probability of existence on the basis of the first information provided by the rule database. Furthermore, newly detected entity elements, at least those that may exist, may be considered in determining their probability of existence. The probability of a pedestrian crossing, a newly detected semantic element, may also be considered high, for example, if the pedestrian crossing and pedestrian traffic light are redetected as solid elements and evaluated as possible according to the method. The evaluation and determination of the relevant existence probabilities can thereby be further improved.

The described method with respect to at least one possible element, in particular for all possible elements contained in a map or map part, can similarly be implemented in an environment area in which sensor data provided by at least one environment sensor or at least one environment detection device can also be used.

The first and second ambient regions described above may be identical but may also be different from each other, however, having at least one overlapping region, as both relate to an ambient region comprising at least one possible element. Typically, the second environmental area is smaller than the first environmental area, and the second environmental area may be entirely encompassed by the first environmental area. This is because the map can cover a very large geographical area, whereas sensor data can typically only be provided in a small environment around the vehicle, at least at the current time. Then, the comparison of the map data and the sensor data may be correspondingly limited to the overlapping area where the map data and the sensor data of the environment coexist.

According to a further advantageous embodiment of the invention, in the comparison of the map data and the sensor data, a comparison is carried out in the overlapping region of the first and second environment regions with respect to all elements present in the overlapping region on the basis of the map data and/or the sensor data, wherein all elements present in the overlapping region on the basis of the map data and/or the sensor data are classified into at least one of the following groups by classification as a result of the comparison. When an element is detected not only based on sensor data, but also exists in map data, it is classified as a first set of "consistent elements"; when an element is detected based on the sensor data, but is not present in the map data, classifying to a second set of "deviations, in particular newly detected elements"; when an element does not exist in the map data and is detected as not existing based on the sensor data, the third group of "non-evaluated elements" is classified. Such unevaluated elements are in particular only map elements that cannot be detected on the basis of sensor data due to occlusion or due to algorithm errors. It is thus uncertain whether the element is present or not. Conversely, map elements that are not detected by the sensor in the "clear view" and error-free algorithms are also most likely not present (no longer present) and are therefore considered a subset of another offset element, rather than the third set of unevaluated elements defined herein. This does not mean that the element is classified as absent based on the sensor data, but rather that the presence of the element cannot be reliably determined based on the sensor data. For example, the element may be obscured by the object, and thus the area to be detected in which the element is or should be located may be detected directly by the detection means, the area to be detected being detectable in the area to be detected.

In this way, the elements currently existing in the map data and all elements detected based on the sensor data are classified into the group. Here, for example, the newly detected element group may be a subgroup of the deviation elements. In addition, a group may be assigned to other elements that cannot be categorized in any of the above groups. However, all possible cases are generally covered by the above groups. The elements of the second and third groups represent possible elements accordingly, since the elements assigned to these groups are not detected on the basis of both map data and sensor data, so that there is a certain divergence between the map data and the sensor data in terms of these elements, which may question the presence of these elements. Instead, the elements of the first group may be considered likely to be present, as their presence is verified on the basis of map data and sensor data.

The elements contained in the map data and detected based on the sensor data may be filtered according to the above-described division of groups. These element groups with associated elements contained therein may be passed to other modules for subsequent determination of the respective existence probabilities, as will be further explained below. Basically, the elements in the related group can also be divided into entity elements and semantic elements, as defined above. Thus, according to a further advantageous embodiment of the invention, each element assigned to the second or third group represents a possible element whose probability of existence is determined. The determination of the existence probabilities of the possible entity elements is performed by a first evaluation module to which all entity elements of the second and third groups are passed, wherein the first evaluation module determines the existence probabilities of all newly detected entity elements and all non-evaluated entity elements from the passed elements. The determination of the probability of existence is performed in particular as described above, in particular using at least one first information item from a rule database. Furthermore, the determination of the existence probabilities of possible semantic elements is performed by a second evaluation module, which delivers all entity elements of the second group and all semantic elements of the third group, wherein the second evaluation module determines the existence probabilities of all unevaluated semantic elements and all newly detected semantic elements, i.e. possible new semantic connections, from the delivered elements. The probability of existence is determined as described above, in particular using at least one first information from a rule database. It is furthermore particularly advantageous here if a predetermined newly detected entity element is also taken into account when determining the probability of the existence of a possible semantic element, in particular a newly detected semantic element. Because it is through this newly detected entity element that a new semantic connection is formed, which in turn leads to a new detected semantic element or increases its probability of existence. Thus, in evaluating or determining the existence probability of a particular possible element, not only map data and sensor data about that element and the rule database and information contained therein are acted upon, but also maps and sensor data about other elements. In other words, not only are each element considered individually, but also elements in possible combinations between elements. Additional information about the element can thus be obtained, which allows conclusions to be drawn about its probability of existence.

Finally, a consistency check can also be performed, which checks whether there is a possibility of contradiction in the results thus obtained. It thus represents a further advantageous embodiment of the invention if, after the determination of the probability of existence, a verification of the consistency is carried out by the verification module, wherein all elements of the first group and all elements evaluated as possible by the first and second evaluation modules are passed to the verification module, whose existence and/or coexistence is carried out by the verification module in accordance with at least one second information provided by the rule database, without a spearhead verification, wherein, if a contradiction with respect to at least one element is detected, the probability of existence of the relevant element is reevaluated in accordance with at least one third information provided by the rule database.

By means of the latter step, inconsistencies and thus elements that may be erroneously evaluated can advantageously be found. The extent of erroneous evaluations can thereby be avoided or at least greatly reduced. This furthermore achieves a significant increase in the reliability of the result.

Optionally, the input data may also be filtered, especially for data aspects that may be erroneous. For example, map data provided by the map may be compared in advance with information and rules contained in the rules database in terms of consistency, for example, to clear logically invalid connections in the high definition map. This is because in principle high definition map elements are provided whose (logical) connection to each other is not verified itself. For example, the light signal is associated with a lane. The cause of errors in existing connection assumptions may be errors in manual processing, in particular due to carelessness and unknowing and map creation algorithms. The connection between map elements appears as a system-specific security-related aspect. For example, knowledge of the semantic groupings of the loop intersections affects interpretation of the scene and trajectory planning, as different traffic rules apply. The rule database can thus be used to check for inconsistencies in the initially existing connection assumptions between the elements provided by the map data. This also allows to further improve the reliability of the method. It is preferred, however, that no pre-inspection of the map data is performed, as it can be assumed that the high-definition map is initially, i.e. from the map manufacturer, without logical inconsistencies.

The invention further relates to a driver assistance system for a motor vehicle, wherein the driver assistance system is designed for carrying out the method of the invention or one of its embodiments. The invention further relates to a motor vehicle having such a driver assistance system. The driver assistance system may be, for example, an assistance system for the automatic driving of a motor vehicle.

The invention also relates to a development of the driver assistance system according to the invention, which has the technical features described above in connection with the development of the method according to the invention. The corresponding improvements of the method according to the invention are therefore not described in detail here.

The invention also includes combinations of features of the described embodiments.

Embodiments of the present invention are described below. In the drawings:

fig. 1 shows a schematic illustration of a motor vehicle with a driver assistance system for determining the probability of the presence of possible elements in the vehicle environment according to an embodiment of the invention; and is also provided with

Fig. 2 shows a schematic diagram of the functional components of an auxiliary system according to an embodiment of the invention.

The embodiments explained below are preferred embodiments of the present invention. In this embodiment, the components described as components are each individual features of the invention which can be seen independently of one another and which also form an extension of the invention independently of one another and can therefore also be regarded as components of the invention individually or in different combinations than those shown. Furthermore, the described embodiments may be supplemented by other features of the invention already described.

In the drawings, functionally identical elements are provided with the same reference numerals, respectively.

Fig. 1 shows a schematic illustration of a motor vehicle 10 with a driver assistance system 12 according to an embodiment of the invention for driving the probability W of the presence of a possible element 14 in an environment 16 of the motor vehicle 12. The driver assistance system 12 may in particular be an assistance system for automatically driving the motor vehicle 10. To implement the autopilot function, but also for other purposes, where applicable, a map 18 is used, in particular a high definition map 18 comprising driver related information. For this purpose, roads, lanes, in particular lane topologies, driving directions matched to the respective lane, road weight rules, etc. can be included. The map 18 is inserted, i.e. stored, in a memory 20 of the motor vehicle 10, in particular in a memory 20 associated with a control device 22 of the motor vehicle 10. Furthermore, the control device 22 is designed to carry out the method described in more detail below, in particular to determine the probability W of the presence of a possible element 14 in the environment 16 of the motor vehicle 10. In addition, the motor vehicle 10 may have a system 24 for self-positioning within the high definition map 18. The system 24 may have a GPS receiver. The key is that the system 24 preferably uses sensor data D provided by at least one environment detection device 26 of the motor vehicle 10. A plurality of such environment detection devices 26, in particular sensors 26 of motor vehicle 10, are exemplary. However, the vehicle 10 may include more or fewer sensors 26. Such a sensor 26 may be designed, for example, as a camera and/or a laser scanner and/or a radar and/or ultrasonic sensor. Advantageously, a plurality of, in particular different types of, sensors 26 are used for detecting the environment. With the GPS sensor, the vehicle 10 can generally determine its own position within the map 18. Due to the environmental detection provided by the sensors 26, the vehicle 10, and in particular the system 24, may be self-locating within the high definition map 18 with great accuracy. This includes matching the current position of the motor vehicle 10 and the current orientation 10 within the map 18.

However, the sensor data D provided by the sensor 26 can be used not only for self-locating, but also in particular for checking the information provided by fig. 18 about the environment 16 of the motor vehicle 10. As described above, the map 18 includes relevant information about the environment 16 of the motor vehicle 10, which is particularly relevant to traffic or driving. Such information is hereinafter referred to as an element. As an example of such elements, a road marking 14a and a lane marking 14b, in particular for speed limitation, are shown in fig. 1. However, there are many other such elements 14 which are relevant for driving and which can be detected by the sensor 26 on the one hand and which can also be stored in the map 18 in the form of map data on the other hand. However, not only the entity elements as shown, but also semantic elements belong to such elements 14. In addition to context-dependent relationships between entity elements, such semantic elements may also relate to, for example, traffic rules.

Optionally, the motor vehicle 10 may also have a sensor system for detecting and recording the trajectories of other traffic participants, i.e. for recording the movement profile of other traffic participants. The above-described environment sensor 26 can also be used in particular for detecting other traffic participants in the environment 16 of the motor vehicle 10 and for detecting movement trajectories. The provision or determination of the movement profile of the other traffic participants on the basis of these detected sensor data D can be performed beforehand by the respective sensor 26 and transmitted as a result to the control device 22 or supplemented by the control device 22 as a function of the data D provided by the sensor.

In general, both the sensor data D and the map data provided by the map 18 may be incorrect or no longer correspond to the current state. This may occur, for example, when the course of the road changes, for example, from building site to building site. It is correspondingly advantageous to verify the correctness of such map assumptions, which assume the presence of a specific element 14 in the environment 16 of the motor vehicle 10. For this purpose, map deviations are first detected. This is implemented by the system 28 for detecting map deviations and for filtering. The system may also be, for example, only a functional component of the control device 22. For this purpose, i.e. for detecting map deviations, the sensor data D are compared with map data of the same overlap region provided by the map 18 with respect to the environment 16. Here, the results of the comparison may be filtered and divided into different groups. If an element in the environment 16 is detected, either based on the sensor data D or based on the map 18, then the element 14 constitutes a consistent element and may be assigned to the first group. If elements 14 are detected on the basis of sensor data D, but are not present in map 18, they can be assigned to a second group, i.e. a biased, in particular newly detected group of elements. If the element 14 is present in the map data of the map 18, however, it is detected as absent on the basis of the sensor data, for this purpose the implementation element is assigned to a third group, i.e. an unevaluated element group. The element 14 thus filtered can then be passed on to other modules in order to determine the corresponding probability of existence W. A first evaluation module 30 is provided here, which can likewise be a functional unit of the control device 22. The first evaluation module 30 here assumes the task of determining the probability of existence W of all newly detected entity elements and all unevaluated entity elements. For this purpose, all entity elements of the second and third groups are passed to the first evaluation module.

Furthermore, a second evaluation module 32 is provided, which can likewise be a functional unit of the control device 22. All entity elements of the second group and all semantic elements of the third group are passed to the second evaluation module 22, wherein the second evaluation module 22 determines the existence probabilities W of all unevaluated semantic elements and all newly detected semantic elements. A check of the consistency can also be carried out by the checking module 34 after the determination of the existence probability W. The checking module checks whether all elements that are ultimately considered to be possible, i.e. all elements of the first group and all elements 14 that are evaluated as possible by the first and

second evaluation modules

30, 32, are present and/or co-exist that would result in a contradiction. When a possible contradiction is found, a re-purposeful inspection of this single element 14 can be performed and the resulting result and the contradiction are finally cleared. Subsequently, the existence probabilities W determined for the respective elements 14, for example, may be output as a result. This may also be output as a result if it is confirmed during the method that some elements 14 contained in the map 18 have been evaluated as impossible, and vice versa, i.e. when some elements 14 are determined to be possible, but not contained in the map data of the map 18.

The great advantage of the invention is that the determination of the presence probability W and the checking of the map assumption are not carried out solely on the basis of the comparison of the sensor data D with the map 18, but rather that, precisely when the presence probability W is determined, if a divergence occurs, i.e. a map deviation is detected, a rule database 36 can also be used, which can likewise be inserted into the memory 20, for example. Such a rule database 36 may for example also be referred to as a knowledge database or a presence database (Ontologie). In addition, such a rule database 36 may also be created at the front end, i.e. online. Here, aspects related to driving unsafe, such as element groups in intersections, element groups in ring intersections, permitted combinations of elements 14, traffic rules, etc., are summarized in such knowledge databases by formal descriptive logic. The element 14 on which it is based and the possible associations between the elements 14 are defined in the scope of the subject matter, i.e. in the complete traffic space. In addition, rules are determined how elements 14 and associations are combined in order to be able to extend knowledge database 36. Thus, the presences are tools that enable logical deductions, i.e. deducing conclusions about the new facts of the existing element 14 by comparison with previous knowledge, and plausibility checks, i.e. checking the consistency of the modulation hypotheses. The information placed in the rules database 36 may be, for example, divided into four groups of context knowledge, action knowledge, indications about misbehavior, and traffic rules. The contextual knowledge specifies a minimum set of elements 14 defining a particular contextual group, such as crosswalks, highway entrances, loop intersections, crossroads, and the like. Action knowledge describes rules by which conclusions can be drawn from the detected motion profiles of other traffic participants. That is, if the lane cannot pass the sensor 26 or is only poorly detected, then the trajectories of other traffic participants may be used. The course of a lane, the presence of specific traffic signs, for example stop signs, traffic lights, etc., can be deduced from the driving behavior of other traffic participants. Rules for deriving conclusions may be specified in the action knowledge accordingly.

The knowledge about misbehavior is to be taken into account here that other traffic participants do not always have to follow the defined traffic regulations exactly, as is often the case in speed limits. This can be taken into account by predefined tolerances in the conclusion of the drawing. The rules database 36 also includes all available traffic rules, signs and the meaning that the signs apply in the environment 16 where the motor vehicle 19 happens to stay.

At this point, if a divergence in the presence of a particular element 14 occurs between the sensor data D and the map data of the map 18, the rules database 36 may advantageously be invoked to make a determination: whether this relevant element 14 is currently possible. By means of the knowledge stored in the rule database 36, a more reliable and more accurate determination of the probability W of the presence of the relevant element 14 in the environment 16 of the motor vehicle 10 can be provided explicitly. In particular, the first and

second evaluation modules

30, 32 can take into account the knowledge stored in the rule database 36 when determining the relevant probability of existence W. And the consistency check finally performed by the checking module 34 may also be performed by means of the rules database 36. This allows a more reliable interpretation of all available information, i.e. the sensor data D and the map data of the map 18, in terms of its compliance and probability of conforming to the current reality. Furthermore, the method allows for providing corrected high-definition map hypotheses in the form of corrected map data at run-time in a particularly advantageous manner, so that the driver assistance system 12 can directly utilize the corrected map data to continue to operate the motor vehicle 10 correctly and autonomously, without stopping such autonomous driving due to excessive unreliability in terms of environmental interpretation and detection.

Fig. 2 again shows a schematic diagram of the functions implemented by the driver assistance system 12 according to an embodiment of the invention. In particular, the individual functional units of the method are shown here.

In particular, map data 18a and sensor data D are provided again here first. Map deviations are then detected in a function block 28 a. This is done by comparing the map data 18a with the detected sensor data D. The result of this comparison may then be filtered in a filtering module 28 b. The filtering module 28b may perform the assignment of individual elements 14 to the groups described above. In other words, elements 14 relating to the following modules, in particular the map elements to be evaluated and the necessary connections in terms of driving and safety, can be filtered out in this way. The data cost and the calculation amount can be reduced through the filtering. In addition, the filtered elements 14 may then be imported into a later module for reprocessing. In this case, the evaluation of the physical element 14 is first carried out by the first evaluation module 30. Here, the element group "unevaluated entity element 14" and "newly detected element 14" are passed. The following two subtasks are then carried out in the first evaluation module 30 using the knowledge database 36: on the one hand, the existence probability W of the unevaluated entity element 14 is estimated, and on the other hand, the existence of the newly detected entity element 14 is estimated. The corresponding probability of existence is then output as a result according to defined metrics 38, which can be defined and upon which the module 40 models the environment. Subsequently, an evaluation of the semantic elements 14 is taken in a second evaluation module 32. To this end, individual elements 14, and in particular entity elements 14, may be logically grouped and matched to supergroups, and associations between elements 14 may be used and established. The different elements 14 may also be combined with each other, for example a traffic light and a corresponding lane. And the right of way rule may cause portions of the semantic element 14. The road right rule may also be stored as a semantic element 14, for example, for each intersection in the map 18. In this case, a new connection between map elements can also be established, in particular by reconstructing partially occluded map elements. What can be achieved by means of the presence is that, given the known rules and prior knowledge, spatial interpolation can be provided, for example, reconstructing a road marking, which is represented by a partially occluded implementation or a dashed line, as a lane marking. But also by combining map elements a connection, in particular a new connection, between map elements can be established. The presence theory allows not only "high-level" groupings of entity elements 14, such as lane markers, to be categorized into lane marker lines, but also "low-level" combinations of semantic elements 14 themselves, such as traffic islands and lane lines, to make up a circular intersection. And these map elements may be combined with each other. And thereby a connection of the messages between the map elements can be established. The combination of map elements can also be achieved by the presence theory. The logical relationships and associations specifically described in the presences, such as the arrangement and prescribed distance between elements 14, enable conclusions to be drawn regarding the logical association of elements 14, such as the logical association of road signs with lane lines or light signals with lanes. And traffic rules may be deduced. Here, the presence theory allows to logically infer new information by explicitly described combinations of associations, e.g. an intersection without a road sign means right before left, whereas an entrance sign means a speed limit of 50 km/h.

The element group passed to the second evaluation unit 32 is here a subgroup of the unevaluated semantic elements 14 and the newly detected entity elements 14. The following two subtasks are performed by the second evaluation module 33 using the knowledge database 36: the new detected entity elements are used to estimate the existence probability B of the unevaluated semantic elements 14 and to estimate the existence probability W of the new semantic traffic rules. The corresponding probability of existence W determined from the defined metric 38 is then output. A consistency check is also performed in the final module 34. In this case, all correctly mapped elements, i.e. elements which are detected in agreement on the basis of the sensor data D and on the basis of the map data 18a, and elements which are evaluated by the two

other modules

30, 32, are passed to the checking module 34. Inconsistencies should be found and cleared in the inspection module 34 using the knowledge database 36. For example, it may occur at a construction site that two traffic lights are found, one being normal and one being construction site traffic lights, and both are classified as highly likely to be present. However, in the present case normal traffic lights are not active. However, the sensor device 26 of the motor vehicle does not recognize a warning which is correspondingly arranged at the traffic light. Thus, these two different traffic lights are classified as present and valid at the intersection. Due to the consistency check by means of the knowledge database 36, contradictions can be deduced therefrom, since two traffic lights cannot be valid for one lane at the same time. The rules may be described and stored, for example, by knowledge database 36. In addition, other knowledge in knowledge database 36 may be used, such as describing yellow building site traffic lights as having priority over conventional traffic lights. This information provided by knowledge database 36 may be advantageously utilized in the present case to "clear" the normal traffic lights. In other words, it can therefore be considered that this normal traffic light may not be present, at least in terms of its effectiveness.

After the consistency check, the finally provided presence probability W and the generally determined assumptions are provided to the other systems, on the one hand inside the motor vehicle and on the other hand outside the motor vehicle. The presence probability thus determined can be used inside the motor vehicle by a corresponding subsystem 42, which is only schematically shown in fig. 2. Such a system may be, for example, path planning and/or scene interpretation of the driver assistance system 12 for autopilot. But may also provide map verification and map updating in the vehicle 10. The result of the method can however be transmitted to the rear end 44, which likewise uses the transmitted data for updating the map data, which can then be provided to the motor vehicle 10 again as a new map. In addition, the back end 44 may also collect data for a plurality of vehicles and then update the map data if the probability of the plurality of vehicles is sufficient. It is, however, particularly advantageous if the map verification is implemented in the motor vehicle 10 by means of this method, which includes error measures for satisfying the safety objectives of the automated driving as a whole. In this way, errors or changes in the map data can be determined reliably and quickly, in particular in real time, so that safe and autonomous further travel is possible. The present invention generally enables solutions that are not only online but also in real-time. Furthermore, the knowledge database 36 is also scene independent, meaning that it models traffic regulations as a whole, rather than just a portion of road traffic, such as only intersections. Furthermore, a system-specific metric can be provided and used to set the reliability of the generated hypothesis, i.e. the probability of existence W. Map elements may advantageously be partitioned and assembled into semantic groups, e.g., intersections, ring intersections, and modeled in detail. An explicit correspondence to the map format is also achieved, which allows modeling of temporary changes in high definition maps. In addition, map elements critical to driving and safety may be prioritized and emphasis may be placed on safely performing driving functions and reducing data and computation. By combining high definition maps, sensing means and knowledge explicitly described by a knowledge database, inferences beyond the working range of the sensor can also be made. Furthermore, this method can also be used particularly advantageously for temporary and short-term changes, for example in the case of construction sites.

The invention thus advantageously allows an automated process of providing corrected map assumptions during travel. For this purpose, knowledge is used to build a knowledge-based model, i.e. a rule database, more precisely, knowledge explicitly described in all driving and safety critical aspects, for interpretation and consistency checking of all existing information. Thereby, a maintenance driving function is also achieved at map deviations (both in terms of entities and semantic elements). This is preferably to provide current and validated map assumptions (including all information required for driving functions, such as current lane conditions and road right rules).

List of reference numerals

10. Motor vehicle

12. Driver assistance system

14. Element(s)

14a traffic sign

14b lane marking

16. Environment (environment)

18. Map(s)

18a map data

20. Memory device

22. Control device

24. System for self-positioning

26. Environment sensor

28. System for detecting map bias and filtering

28a module for detecting map deviations

28b module for filtration

30. First evaluation module

32. Second evaluation module

34. Inspection module

36. Rule database

38. Metrics (MEM)

40. Module for environmental modeling

42. Subsystem

44. Rear end

D sensor data

Probability of W existence

Claims

1. A method for determining a probability (W) of presence of at least one possible element (14) in an environment (16) of a motor vehicle (10), the method comprising the steps of:

-providing map data (18 a) of a map (18) of an environment (16) of the motor vehicle (10), the map data relating to a first environment area comprising at least one possible element (14);

providing sensor data (D) by means of at least one environment detection device (26) of the motor vehicle (10),

the sensor data relates to a second environmental area comprising at least one possible element (14);

-comparing the environmental data (18 a) and sensor data (D) in the presence of the at least one possible element (14);

-determining a presence probability (W) from the result of the comparison; and is also provided with

-outputting a probability of presence (W);

-wherein the at least one possible element (14) is classified as present based at least on map data (18 a) and/or sensor data (D);

it is characterized in that the method comprises the steps of,

A rule database (36) is provided, in which information about the traffic space is stored, wherein the probability of existence (W) is determined at least for the case that a divergence in the existence of at least one possible element (14) is ascertained when comparing the map data (18 a) and the sensor data (D), on the basis of at least one first information provided by the rule database (36).

2. The method according to claim 1,

it is characterized in that the method comprises the steps of,

a divergence is confirmed when at least one possible element (14) is classified as a newly detected element (14) that has been detected based on the sensor data (D) but is not present in the map data (18 a), or when at least one possible element (14) is classified as an unevaluated element (14) that is present in the map data (18 a) but is detected by the sensor data (D) as not present.

3. The method according to any of the preceding claims,

it is characterized in that the method comprises the steps of,

at least one possible element (14) is a physical element (14) that relates to an objective object that may be present in the environment.

4. The method according to any of the preceding claims,

it is characterized in that the method comprises the steps of,

at least one possible element (14) represents a semantic element (14) representing a logical group of entity elements (14) or at least one traffic rule.

5. The method according to any of the preceding claims,

it is characterized in that the method comprises the steps of,

when a possible element (14) has been classified as a newly detected semantic element (14) based on the comparison result, the existence probability (W) of the possible element (14) is determined from the existence of the newly detected entity element.

6. The method according to any of the preceding claims,

it is characterized in that the method comprises the steps of,

in comparing the map data (18 a) and the sensor data (D), a comparison is carried out in the overlapping region of the first and second environment regions with respect to all elements (14) which are present in the overlapping region on the basis of the map data (18 a) and/or the sensor data (D), wherein all elements (14) which are present in the overlapping region on the basis of the map data (18 a) and/or the sensor data (D) are classified into at least one of the following groups by classification according to the result of the comparison:

-classifying into a first set of "consistent elements" when an element (14) is not only detected based on the sensor data (D) but also present in the map data (18 a);

-classifying into a second set of "deviations, in particular newly detected elements", when an element (14) is detected based on the sensor data (D), but not present in the map data (18 a);

-classifying into a third set of "unevaluated elements" when the element (14) is not present in the map data (18 a) and is detected as not present based on the sensor data (D).

7. The method according to claim 6, wherein the method comprises,

it is characterized in that the method comprises the steps of,

each element (14) classified as a second group or a third group is a possible element (14) for which the existence probability (W) has been confirmed, wherein,

-performing, by the first evaluation module (30), a determination of the presence probability (W) of the possible entity elements (14), all possible entity elements of the second and third groups being passed to the first evaluation module, wherein the first evaluation module (30) determines the presence probability for all newly detected entity elements (14) and all unevaluated entity elements (14) from the passed elements (14); and is combined with

And is also provided with

-wherein the determination of the existence probabilities of possible semantic elements (14) is performed by the second evaluation module (32), all entity elements (14) of the second group and all semantic elements (14) of the third group being passed to the second evaluation module, wherein the second evaluation module (32) determines the existence probabilities for all unevaluated semantic elements (14) and all newly detected semantic elements (14) from the passed elements (14).

8. The method according to claim 6 or 7,

it is characterized in that the method comprises the steps of,

after determining the existence probability (W), a check of the consistency is carried out by a checking module (34), wherein all elements (14) of the first group and all elements (14) evaluated as possible existence by the first and second evaluation modules (30, 32) are passed to the checking module, wherein the existence and/or coexistence of the elements is checked in terms of consistency by the checking module (34) on the basis of the second information provided by the at least one rule database (36), wherein, when a contradiction about the at least one element (14) is detected, the existence probability (W) of the relevant element (14) is reevaluated on the basis of the third information provided by the at least one rule database (36).

9. Driver assistance system (12) for a motor vehicle (10), wherein the driver assistance system (12) is provided for carrying out the method according to any one of the preceding claims.

10. A motor vehicle (10) having a driver assistance system (12) according to claim 9.