Method and apparatus for providing annotated traffic area data

Abstract

To provide a method for limiting complicatedness in adding annotations, by adding annotations only to areas in which also vehicles actually travel.SOLUTION: A method 300 includes: a step 310 of reading traffic area data representing a respective section of a traffic area utilized by road users, and reading automatically detected positional data of the road users in the traffic area; and a step 320 of associating the traffic area data where the road users corresponding to the detected positional data are located at present with at least one respective annotation data set. Accordingly, annotated traffic area data is acquired that uses signals to inform about other road users' use and / or moving possibilities in the traffic area represented by the traffic area data. The annotation data set is generated by using a classifier based on a mechanical learning method and / or a mechanical learning algorithm.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The invention is based on an apparatus or a method as defined in the preamble of the independent claims. The subject of the invention is also a computer program. [Background technology]

[0002] To enable autonomous driving, huge amounts of expensive annotation data are currently being generated (e.g., as is done in third-generation video signal processing), which suffer from additional errors due to "label noise" because human annotators must solve annotation problems that require estimation (e.g., correct estimation of road user classes) without direct quality control. These estimations require post-hoc quality control of the annotation data, where annotation errors may be overlooked.

[0003] An approach to solving such problems could consist in annotating only those areas that are actually driven over by the vehicle, so that the complexity of making these annotations would be limited to at least the relevant areas. Summary of the Invention [Means for solving the problem]

[0004] Against this background, the approach presented here presents methods according to the main claims, further devices for using these methods and finally corresponding computer programs. Advantageous developments and improvements of the devices according to the independent claims are possible by the measures recited in the dependent claims.

[0005] According to the approach presented herein, a method for providing annotated traffic area data is presented, the method comprising the following steps. - reading traffic area data representative of each section of a traffic area used by a road user and reading automatically detected location data of the road user within the traffic area, and - obtaining annotated traffic area data signalling the use and / or mobility of the traffic area represented by the traffic area data by other road users by associating at least one annotation dataset, respectively, to the traffic area data in which the road user corresponding to the detected position data is currently located, wherein in particular these annotation datasets have been generated using a machine learning method and / or a classifier based on a machine learning algorithm.

[0006] Traffic area data can be understood as data representing the environment in the area of ​​a road user, for example as a map. A section of a traffic area can be understood as a subdivision representing a part of a traffic area, which section can be for example two-dimensional and can represent a part of a road surface or a sidewalk. A traffic area can be understood as an area that is open to the public and is used for the movement of road users. A road user can be understood here as a person, such as a pedestrian, or an object, such as a vehicle. An annotation dataset can be understood here as metadata that is associated with the respective traffic area data describing a section of a traffic area and that contains further information on properties or parameters that are or may be relevant for another road user who is going to move in the section of the traffic area represented by the traffic area data. An annotation for traffic area data can be understood as an annotation dataset that is associated with the traffic area data for the section in which the road user is located, for example. Location data can be understood as data representing the current location of a road user, provided for example as coordinates of a global satellite navigation system. A road user class of road users may be understood as a group of road users with as similar characteristics or types as possible, such as, for example, the road user classes of cars, motorbikes, trucks, buses, bicycles or pedestrians as road users. A classifier based on a machine learning method or a machine learning algorithm may be understood as an algorithm which, by means of a previously carried out training phase, has reached a state in which, when supplied with data, it is able to make a decision or assessment as to whether a corresponding section of a traffic area currently travelled or commonly used by a road user can also be used by another road user.

[0007] The approach presented here is based on the realization that if the traffic area data, which often result from the automatic mapping of the road user's environment, are linked with the automatically detected location data of the road user, annotated traffic area data can be provided very quickly, extensively and reliably, and thus it is recognizable from the annotated traffic area data via the annotation data set associated with the traffic area data that the road user has actually used or traveled through the relevant section of the traffic area. In this way, it can be automatically detected which area or section of the traffic area a road user (particularly road users of this particular road user class) has used, so that it can be inferred that the relevant section of the traffic area can also be used in the future by road users of this particular road user class. It is therefore advantageous that there is no need to again correct or edit the traffic area data representing the sections of the traffic area by manual or human intervention, or to annotate all of these data with annotation data.

[0008] According to a particularly preferred embodiment of the approach proposed here, in the matching step for obtaining the annotated traffic area data, the at least one adjacent annotation data set can also be further matched to adjacent traffic area data representing respective sections of the traffic area adjacent to the traffic area data, in particular these adjacent annotation data sets being generated using a classifier based on machine learning methods and / or machine learning algorithms. Here, the adjacent annotation data set can be understood as the annotation data set associated with the adjacent traffic area data. This adjacent annotation data set can then again also contain metadata, as already detailed in connection with the annotation data set. Here, the adjacent traffic area data can be understood as the traffic area data representing sections directly connected to the section used by the road user in the traffic area. An advantage of such an embodiment of the approach proposed here is that in this way, when a road user uses an adjacent section of the traffic area, information can be determined or calculated as to whether, for example, a side area of ​​the roadway or a pedestrian walkway can also be used by this road user or another road user, in particular a specific road user class. In addition, using machine learning methods or classifiers based on machine learning algorithms, e.g. neural networks, it is possible to calculate the relevance of whether and possibly to what extent sections of a traffic area that are (e.g. directly) connected to sections in the traffic area that are actually being used by a road user can also be used by that road user or another road user.

[0009] An embodiment of the approach proposed here is further preferred, in which in the reading step, adjacent data sets for adjacent traffic area data representing traffic areas that are not currently used or cannot be used by the road user are read in. In this way, not only is a determination about the availability of the section of the traffic area that is actually used by the road user achieved, but also the availability of further areas adjacent to the section currently used by the particular road user can be stated.

[0010] According to a further embodiment, in the reading step, road user classes representative of types of road users can be read, and in the matching step, the annotation data set associated with the traffic area data is calculated using the road user classes. The road user classes can be, for example, classes classifying road users into pedestrians, cyclists, cars, trucks, buses or motorbike drivers. Such an embodiment of the presented approach has the advantage that information from the annotated traffic area data can be very easily used when determining the availability of the traffic area by other road users, in particular since it is possible to determine the significance of the annotated traffic area data for other road users. For example, this allows to determine that a traffic area used by trucks as road users providing the annotated traffic area data is also likely to be available for passenger cars as road users.

[0011] An embodiment of the proposed approach can further be envisaged, in which in the reading step a height profile of the road user, which represents the contour of the road use above the traffic area used by the road user, and / or a silhouette, which represents the contour of the road use above the traffic area used by the road user, are read and in the mapping step an annotation data set, which is respectively mapped to the traffic area data, is calculated using the height profile and / or the silhouette. Such a height profile or silhouette or contour can be understood as a spatial or three-dimensional area required or occupied by a road user when using a section of the traffic area. In this way it is advantageously possible to ensure that a particular road user leaves a specific spatial requirement for the respectively used section of the traffic area, so that it is possible to indicate, for example, to road users following these respectively used sections that these sections have already been used once by the corresponding road user. For example, in this way it can be very easily stored in the traffic area data that a road is very narrow due to lateral green cover, so that for example passenger cars can use this road, but trucks with a different height profile or different silhouette or contour can no longer use this road.

[0012] An embodiment of the approach presented here is particularly easily implementable, in which in the reading step traffic area data are read that are representative of a section of the road user traffic area used by the road user and that are calculated using images recorded by a camera and / or in which position data are read that are calculated using an odometry unit and / or a satellite navigation system. Since such cameras or odometry units are now often equipped as standard in vehicles, by using these images recorded by the camera as traffic area data or as position data provided to the odometry unit, traffic area data can be generated very quickly and efficiently and the corresponding sections subdivided.

[0013] Particularly relevant for the use of traffic area data with at least one annotation is an embodiment of the approach presented here, in which in the reading step traffic area data is read that represents a section of a road user traffic area used by a road user that is designed and / or identifiable as a roadway for at least one vehicle or as a footpath. Thus, in particular for the field of classification or determination of the availability of roadways or footpaths for motor-driven vehicles or also bicycles, the method proposed here can be used very preferably.

[0014] According to a further embodiment of the proposed approach, at least the reading and / or matching steps can be performed in the road user's mobile unit and / or in a central fixed computing unit, and / or a step of transmitting the annotated traffic area data to a central fixed computing unit after the matching step is provided. The central fixed computing unit can be understood as a cloud server, for example. Such an embodiment has the advantage that good availability of the provided traffic area data with at least one annotation can be ensured, although numerically or circuit-wise more complex algorithms can be implemented at the respective preferred location depending on the available computing power or data transmission capacity.

[0015] Furthermore, according to a further embodiment of the proposed approach, a method for evaluating the utilization of a section of a traffic area available to a road user may be provided, said method comprising the following steps: - reading the annotated traffic area data produced according to a variant of the method presented herein and location data representative of current or future locations of road users, and - checking the availability of at least one section of the traffic area by road users using the annotation dataset and the location data.

[0016] Such an embodiment of the approach proposed herein has the advantage that the provided annotated traffic area data can be advantageously used to, for example, determine in advance whether a particular section of the traffic area can be used by road users (e.g. of a particular road user class) or whether this is not possible or only possible with significant problems based on the experience / information of road users who have previously used this section of the traffic area.

[0017] According to a further embodiment of the presently proposed approach, a method for training a classifier based on a machine learning algorithm may be provided, comprising the following steps: - reading first traffic area data representative of a respective section of a traffic area used or available by a first road user and first automatically detected position data of the first road user, reading second traffic area data representative of a respective section of a traffic area used or available by a second road user and second automatically detected position data of the second road user, and - training elements of a classifier using the first and second traffic area data and the first and second location data.

[0018] Such an embodiment of the approach proposed here has the advantage that by using a classifier based on a machine learning algorithm, e.g. a neural network, this classifier can be trained for traffic area data for a particular section or sections of a traffic area provided by multiple road users, such that when matching the annotation dataset to the traffic area data, it is also possible to state whether sections of the traffic area not currently used by the current road user but adjacent to the section of the traffic area used by this road user can be used by similar road users (e.g. of the same road user class).

[0019] The variants of these methods presented here can be implemented, for example, in software or hardware or in a mixed form of software and hardware, for example in a control device or apparatus.

[0020] Moreover, the approach presented herein produces an apparatus configured to execute, drive or perform the steps of a variant of the method presented herein in a corresponding device, and the apparatus form of this variant of implementation of the invention allows for a rapid and efficient solution of the problem underlying the invention.

[0021] For this purpose, the device may have at least one arithmetic unit for processing signals or data, at least one memory unit for storing signals or data, at least one interface to sensors or actuators for reading sensor signals from sensors or outputting data or control signals to actuators, and / or at least one communication interface for reading or outputting data embedded in a communication protocol. The arithmetic unit may be, for example, a signal processor, or a microcontroller, etc., and the memory unit may be a flash memory, an EEPROM, or a magnetic memory unit. The communication interface may be configured to read or output data wirelessly and / or wired, and a communication interface capable of reading or outputting wired data may read or output these data from or to a corresponding data transmission line, for example electrically or optically.

[0022] Here, a device can be understood as an electrical device that processes sensor signals and outputs control and / or data signals accordingly. The device may have an interface that can be configured by hardware and / or software. In the case of a hardware configuration, the interface can be, for example, part of a so-called system ASIC that includes the various functions of the device. However, it is also possible that the interface is a separate integrated circuit or is at least partially composed of discrete components. In the case of a software configuration, the interface can be, for example, a software module that resides on a microcontroller next to other software modules.

[0023] A computer program product or computer program having a program code which can be stored on a machine-readable carrier or storage medium, such as a semiconductor memory, hard disk memory or optical memory, and which is used to execute, perform and / or drive control the steps of the method according to any of the above-described embodiments, is also advantageous, particularly when the program product or program is implemented on a computer or device.

[0024] An embodiment of the approach presented herein is illustrated in the drawings and detailed in the following description. [Brief description of the drawings]

[0025] [Figure 1] FIG. 2 illustrates a scene in which annotated traffic area data is provided by a road user, e.g. a vehicle. [Diagram 2] 13 is a schematic operation diagram of a matching unit 160 having a classifier 178 according to a further embodiment; [Diagram 3] 1 is a flowchart of an example implementation of a method for providing traffic area data having at least one annotation. [Figure 4] 2 is a flow chart of an embodiment of a method for evaluating the utilization of a section of a traffic area available to road users; [Diagram 5] 1 is a flow chart of an embodiment of a method for training a classifier based on a machine learning algorithm. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0026] In the following description of preferred embodiments of the invention, elements that appear in multiple figures and have similar effect will be designated with the same or similar reference numerals, and repeated description of these elements will be omitted.

[0027] Fig. 1 is an illustration of a scenario in which annotated traffic area data 105 are provided by a road user 100, for example a vehicle. Here, firstly traffic area data 110 are read in via an interface 115, which traffic area data 110 can then be emitted, for example, as image data of an optical camera 120 recording a surrounding 125 around the road user 100. In the example shown in Fig. 1, as the road user 100 travels along a road as a traffic area 130, the camera 120 images a section 135 that is used by the vehicle as road user 100, i.e. that is traveled or will be used shortly thereafter.

[0028] Alternatively or additionally, the traffic area data 110 may be retrieved from a stored digital map 140 in which sections 135 of roads available for vehicular use by road users 100 are stored as traffic areas 130 .

[0029] Furthermore, adjacent traffic area data 145 is also provided, for example by the camera 120 and / or from the digital map 140, which represents adjacent sections 150 adjacent to the section 135 used, here traveled or immediately used, by the vehicle as road user 100.

[0030] Additionally, automatically detected position data 155, which represent the current position of the road user 100, here a vehicle, are further provided via the interface 115. The position data 155 can be obtained, for example, from an odometry unit 157 by evaluating the movement trajectory of the road user 100, possibly with the aid of data from a digital map 140. Alternatively or additionally, the position data 155 can also be obtained using a receiver of a satellite navigation system 158, for example a GPS system.

[0031] Then, in the association unit 160, for example, the traffic area data 110 is associated with the respective annotation dataset 162, which is generated in an annotation dataset generation unit 165 using the current position of the road user 100. For this purpose, further information can be used, for example a height profile 167 or a silhouette, which indicates the spatial extent of the road user 100. It is also conceivable that further information of the road user class 168 of the road user 100, in this case for example the road user class of a passenger car as road user, is inserted into the annotation dataset 162. Such a procedure allows to equip the annotated traffic area data 105 with information on the availability of the respective section 135 by the road user 100, by associating metadata, also called annotations, as annotation datasets 162 to the individual sections 135 actually used by the road user 100 and identifiable by the current position according to the position signal 155. Here, for example, if the annotated traffic area data 105 is transmitted to a central processing unit 170, for example a cloud server, and from there to another road user 172, the other road user 172 can evaluate the annotated traffic area data 105 using its current position and recognize whether the section 135 represents a route that can also be used by itself as another road user 172. It can also take into account whether the other road user 172 has the same road user class 168 as the road user 100. In this way, for example, if the road user 100 is a truck and the other road user 120 is a car, it can be recognized that the section 135 can most likely also be used or traveled by the other road user 172 due to the smaller external dimensions of the other road user 172. In contrast, if the road user 100 is a car and the other road user 172 is a truck, it cannot be recognized unambiguously that the section 135 can also be used by the other road user 172, or at least it cannot be warned.

[0032] Here, in order to be able to evaluate the availability of the selected section 135, which is usually very close to the spatial dimensions of the current road user 100, by other road users 172, but also to be able to take into account the availability of the adjacent sections 150 adjacent to the section 135 together when automatically detecting and evaluating the availability of the traffic area 130, the adjacent annotation dataset 175 can be associated by the association unit 160 with the adjacent traffic area data 145 representing the respective adjacent section 150, for example the adjacent annotation dataset 175 being arranged in the annotation dataset generation unit 165 and provided by a classifier 178 based on a machine learning algorithm. In this case, these adjacent traffic area data 145 with the adjacent annotation dataset 175 are marked with reference number 180 in Fig. 1 and can be associated with the annotated traffic area data 105 as well.

[0033] The classifier 178 can use, for example, a trained neural network or another algorithm based on machine learning, to generate this adjacent annotation data set 175, which is associated with the adjacent traffic area data 145, taking into account the current position of the road user 100, which can for example be derived from the position data 155, and / or the section 135 currently used by the road user 100. In this way, it can also be stated with a high degree of certainty about the availability by other road users 172 of the adjacent sections 150 adjacent to the section 135, and this information can be forwarded to the other road users 172 as well, for example again via the central processing unit 170.

[0034] Alternatively or additionally, the classifier 178 can also generate the annotation data set 162, which is associated with the traffic area data 110, taking into account the current position of the road user 100, which can for example be derived from the position data 155, and / or the section 135 currently used by the road user 100. In this way, it is possible to state with a high degree of certainty about the availability of the section 135 by other road users 172, and this information can be forwarded to the other road users 172 as well, for example again via the central processing unit 170.

[0035] This classifier 178 should now be trained with the aforementioned training method, so that the annotation dataset 162 and / or the adjacent annotation dataset 175 provided by the classifier 178 already allows a fairly good prediction of the availability of the section 135 and / or the adjacent section 150 by other road users 172. This training or drilling can be done using multiple sets of annotated traffic area data 105, i.e. with at least a first annotated traffic area data 105 originating from a first road user, such as road user 100, and a second annotated traffic area data 105' originating from a second road user 100 that utilizes the traffic area 130, for example at a later time, here on a slightly different section 134. Here, for example, if the height profile 167, silhouette and / or road user class 168 of each road user 100 or 100' are also taken into account, the classifier 178 can be trained very accurately using currently available algorithms, so that when evaluating a section 135 of a traffic area 130 during the travel of the road user 100, the availability of the respective section 135 and / or adjacent sections 150 by other road users 172 can be predicted very well.

[0036] It is also conceivable that the matching unit 160 is not accommodated in a unit of the road user 100, i.e. here for example in a unit of the vehicle, but is instead arranged for example in the central processing unit 170. Alternatively or additionally, the interface 115 can also be arranged in the central processing unit 170. In this case, the traffic area data 110 and / or the position data 155 and possibly also the adjacent traffic area data 145 with the adjacent annotation data set 175, together with further information such as for example the height profile 167, the silhouette and / or the road user class 168 of the road user 100, would therefore have to be transmitted to the central processing unit 170, unless these data are already available in the central processing unit 170 for the road user 100, in particular at its current position.

[0037] Fig. 2 shows a schematic method of operation of the matching unit 160 with a classifier 178 according to a further embodiment, where the classifier 178 is configured as a module with the functionality of artificial intelligence, for example. This matching unit 160 with classifier 178 is then supplied with a parameter set, possibly in the form of a height profile 167 or a silhouette of the road user 100, for example with information on the vehicle shape of the road user 100 from Fig. 1. Additionally, further data, such as for example the traffic area data 110 or the position data 155, can also be supplied here as sensor data to the matching unit 160 with classifier 178 according to the diagram from Fig. 1. The classifier 178 or the matching unit 160 can then output annotated traffic area data 105 which corresponds to or represents a three-dimensional diagram of the space actually occupied by the road user 100. Additionally, the matching unit 160, in particular here the classifier 178, can also output adjacent traffic area data 145 having an adjacent annotation dataset 175, in this case referenced 180 in FIG. 1, which corresponds to the amount of free space of the traffic area 130 in the three-dimensional view that can be utilized by the road user 100 and another road user 172.

[0038] In summary, it can be seen that an important goal of the approach presented here can be seen in approximating the traversable / drivable / passable free space (as traffic area available for road users) in time, using classifiers such as neural networks, for example, without the need for manually annotated data for that purpose. An important aspect of the approach presented here is to calculate the area where dynamic objects can move as road users (e.g. vehicles or pedestrians) without manual annotation data. For that purpose, automatically recorded data such as traffic area data, position data such as odometry data, and vehicle parameters such as vehicle shape are used. A further important aspect can be seen in particular in how to train such a network or such a classifier. Thus, the approach presented here allows for an optimal automatic annotation of the traversable free space in time for dynamic objects, using deep neural networks as artificial intelligence, for example, according to one embodiment.

[0039] The advantages of using one embodiment of the approach presented here can be summarized as follows: - "Label noise" due to human annotation errors is no longer expected. - Maximum annotation data quality can be achieved. It is possible to limit the annotation cost to the recording of a sequence (any sensor). - There is no need to generate annotations that have no usefulness, e.g. empty spaces that no road users, e.g. vehicles, can enter. - There is no need to generate annotations that create dangerous situations, for example driving too close to a still-moving pedestrian. The vehicle driver will assume a corresponding safety distance during the sequence recording. - The free space calculation already takes into account the motion of other dynamic objects, eliminating the need to evaluate the optical flow.

[0040] Without limiting generality, an embodiment of the approach presented here can be elaborated using the case of vehicles as road users, without considering the description of 3D annotations that can be based on various known strategies (e.g. voxels). The procedure can be implemented with any dynamic object.

[0041] 1. First, the recording of sequence and vehicle / odometry data takes place, here for example traffic area data and position data. Furthermore, as the amount of free space depends on it, for each vehicle used to input annotation data, the road user 100 can generate a vehicle model (here a height profile 167 or a silhouette, possibly in conjunction with a road user class 168), as in the diagram from FIG. 1.

[0042] 2. For each image (here, for example, from the traffic area data 110 from the camera 125), calculate the volume that the vehicle will occupy in the further course as a road user 100, i.e. which section 135 of the traffic area 130 the vehicle will use (this corresponds to the trajectory of the volume of movement of the road user 100). This is possible by using odometry data. Alternatively, the retrieved trajectory can also be calculated using image processing means, i.e. by evaluating the image data provided by the camera 120 (keyword: "visual odometry"). This allows to calculate, for each sequence and for the corresponding vehicle, for each image, the area / section 135 into which the vehicle has moved and thus can be safely traveled. However, what is to be calculated is the range within which the vehicle can safely travel as a road user, which can be approximated using these data. This corresponds in FIG. 1 to the traffic area data 105 with annotations together with the adjacent traffic area data 145 with the adjacent annotation data set 175.

[0043] 3. In one very simple variant, only the current path is displayed as a drivable area, excluding areas that may have been driven by another driver. This problem can be solved by backpropagating only the error of the amount of trajectories at training to each annotation, and not backpropagating the error to the rest of the 3D space. In this case, it is only necessary to take care that the weights of the network or classifier 178 or neural network do not remove relevant knowledge. This can happen because the weights are only changed with respect to the amount of trajectories. To solve this problem, for example, the learning rate of the (neural) network as classifier 178 can be significantly reduced, which would be the simplest approach, or methods to prevent "catastrophic forgetting" are used.

[0044] The trained network now provides the drivable space for each image or each section 135 of the traffic area 130 .

[0045] Fig. 3 shows a flow chart of an embodiment of a method 300 for providing traffic area data with at least one annotation. The method 300 comprises a step 310 of reading traffic area data representative of respective sections of the traffic area used by a road user and reading automatically detected position data of the road user in the traffic area. Furthermore, the method 300 comprises a step 320 of associating at least one annotation dataset to the traffic area data in which the road user is currently located corresponding to the obtained position data, in particular these annotation datasets having been generated using a machine learning method and / or a classifier based on a machine learning algorithm, in order to obtain annotated traffic area data signalling the use and / or mobility possibility by another road user of the traffic area represented by the traffic area data.

[0046] Figure 4 shows a flow chart of an embodiment of a method 400 for assessing the availability of a section of a traffic area for use by a road user. The method 400 comprises a step 410 of reading at least annotated traffic area data produced by a method according to any of the preceding claims and location data representative of a current or future location of a road user. Furthermore, the method 400 comprises a step 420 of checking the availability of at least one section of the traffic area by a road user using the annotation data set and the location data.

[0047] Fig. 5 shows a flow chart of an embodiment of a method 500 for training a classifier based on a machine learning algorithm. The method 500 comprises a step 510 of reading first traffic area data representing respective sections of a traffic area used or available by a first road user and first automatically detected position data of the first road user, and reading second traffic area data representing respective sections of a traffic area used or available by a second road user and second automatically detected position data of the second road user. Furthermore, the method 500 comprises a step 520 of training elements of the classifier using the first and second traffic area data and the first and second position data. The elements of the classifier can be understood here, for example, as in particular nodes of a neural network.

[0048] The advantages described herein may also be realized quickly and efficiently by such embodiments of the approaches presented herein.

[0049] When an example includes an "and / or" connection between a first feature and a second feature, this can be read as the example having both the first feature and the second feature in some embodiments, and only the first feature or only the second feature in some further embodiments.

Claims

1. A method (300) for providing annotated traffic area data (105), said method (300) comprising the steps of: - reading (310) traffic area data (110) representative of each section (135) of a traffic area (130) used by a road user (100) and reading (310) automatically detected position data (155) of said road user (100) within said traffic area (130); - associating (320) at least one annotation data set (162) with the traffic area data (110) in which the road user (100) corresponding to the detected position data (155) is currently located, thereby obtaining the annotated traffic area data (105) signalling the use and / or mobility potential of the traffic area (130) represented by the traffic area data (110) by other road users (100), the annotation data set (162) comprising having been generated using a classifier (178) based on machine learning methods and / or machine learning algorithms; having 13. A method (300) according to claim 12, characterized in that in the reading step (310) a height profile (167) representative of a contour of a road use on the traffic area (130) used by the road user (100) and / or a silhouette representative of the contour of the road user on the traffic area (130) used by the road user (100) is read, and in the association step (320) the annotation dataset (162) respectively associated to the traffic area data (110) is calculated using the height profile (167) and / or the silhouette.

2. 2. The method (300) of claim 1, characterized in that in the matching step (320) to obtain the annotated traffic area data (105), at least one adjacent annotation dataset (175) is matched to adjacent traffic area data (150) representing respective sections (135) of the traffic area (130) adjacent to the traffic area data (110), the adjacent annotation dataset (175) including those generated using machine learning methods and / or classifiers (178) based on machine learning algorithms.

3. 3. The method (300) according to claim 2, characterized in that in the reading step (310), the adjacent annotation dataset (175) for adjacent traffic area data (150) representing traffic areas (130) not utilized or unavailable by the road user (100) is read.

4. The method (300) according to any one of claims 1 to 3, characterized in that in the reading (310) a road user class (168) representative of a type of the road user (100) is read and in the associating (320) the annotation dataset (162) associated with the traffic area data (110) is calculated using the road user class (168).

5. Method (300) according to any one of claims 1 to 4, characterized in that in the reading (310) traffic area data (110) representative of a section (135) of the traffic area (130) of the road user (100) used by or available to the road user (100) are read, calculated using images recorded by a camera (120) and / or in that in the reading (310) position data (155) calculated using an odometry unit (157) and / or a satellite navigation system (158) are read.

6. The method (300) according to any one of claims 1 to 5, characterized in that in the reading (310) traffic area data (110) are read which are representative of a section (135) of the traffic area (130) of the road user (100) used by the road user (100), which is designed and / or identifiable around a periphery of the road user (100) as a carriageway (130) for at least one vehicle or as a footpath.

7. Method (300) according to any one of claims 1 to 6, characterized in that at least the reading (310) and / or the matching (320) steps are performed in mobile units of the road users (100) and / or in a central, fixed computing unit (170) and / or that after the matching (320) a step of transmitting the annotated traffic area data (110) to a central, fixed computing unit (170) is provided.

8. A method (400) for assessing the availability of a section (135) of said traffic area (130) available to a road user (100), said method (400) comprising the following steps: - reading (410) annotated traffic area data (105) produced by the method (300) according to any one of claims 1 to 7 and location data (155) representative of the current or future location of said road users (100), and - checking (420) the availability of at least one section (135) of said traffic area (130) by said road users (100) using said annotation data set (162) and said location data (155) of said annotated traffic area data (105), The method (400).

9. A method (500) for training a classifier (178) based on a machine learning algorithm, said method (500) comprising the steps of: - reading (350) first traffic area data (110) representative of a respective section (135) of a traffic area (130) used or available by a first road user (100) and first automatically detected position data (155) of said first road user (100) and reading (510) second traffic area data (110) representative of a respective section (135) of a traffic area (130) used or available by a second road user (100') and second automatically detected position data (155) of said second road user (100'), - training (520) elements of said classifier (178) using said first and second traffic area data (100) and said first and second location data (155); The method (500).

10. An apparatus (115, 160) configured to perform and / or drive and control the steps (310, 320, 410, 510, 520) of the method (300, 400, 500) according to any one of claims 1 to 9 in a corresponding unit.

11. A computer program arranged to carry out and / or drive the steps of the method (300, 400, 500) according to any one of claims 1 to 9.

12. A machine-readable storage medium having stored thereon the computer program of claim 11.

Images