Creating an entity structure describing the environment of an autonomous vehicle

Abstract

A method for generating an entity structure (20) describing the environment of an autonomous vehicle comprises: classifying entities (34, 34a) based on sensor data in the environment of the vehicle from sensor data detected by the vehicle, wherein an entity (34) based on sensor data comprises at least one position and one type of entity; and assigning the entities (34, 34a) based on sensor data into entity classes (24) based on the type of entities (34), wherein the entity classes (24) are sorted into a class hierarchy (36) that subordinates certain entity classes (24) to other entity classes (24); wherein the entity classes (24) are assigned safety rules (28) for an autonomous driving system (22), which are inherited by the subordinate entity classes (24).

Description

[0001] The invention relates to a method, a computer program, and a computer-readable medium for generating an entity structure describing the environment of an autonomous vehicle. The invention further relates to a method and system for controlling an autonomous vehicle.

[0002] Autonomous or automated vehicles are required to respond appropriately to emerging traffic scenarios. The key challenge is to define how a given scenario will develop and how to respond accordingly. Another challenge is to clarify the extent of the vehicle's or the given automated system's responsibility and where the responsibility of other road users or human participants (e.g., drivers) begins.

[0003] To date, known automated systems predominantly fall into the SAE AD Levels 1 and 2 categories according to SEA J3016. Responsibility for these systems essentially lies with the humans behind the wheel. Some systems, such as automatic emergency braking systems, cannot formally be classified into these levels, but they follow the same concept of responsibility. Essentially, humans are responsible for ensuring that traffic rules are followed. The systems largely limit themselves to not doing anything that a human cannot control. Existing Level 3 systems are rare and limited to very narrow ODDs (Operational Design Domains), so that the scenarios that arise are manageable and generalizable principles are less important.

[0004] However, the development of advanced automation, especially for AD levels L4 or even L5, requires rule-based decision criteria for the corresponding systems to make reproducibly appropriate decisions. The scientific article "On a Formal Model of Safe and Scalable Self-driving Cars" by Shai Shalev-Shwartz, Shaked Shammah, and Amnon Shashua, Mobileye, 2017, arXiv:1708.06374, explores what such rules might look like.

[0005] The object of the invention is to provide a rule-based system for an autonomous vehicle that can be flexibly adapted and that leads to a high level of safety.

[0006] This object is achieved by the subject matter of the independent claims. Further embodiments of the invention emerge from the dependent claims and the following description.

[0007] One aspect of the invention relates to a method for generating an entity structure describing the environment of an autonomous vehicle. The entity structure comprises entities assigned to entity classes. Entities are data objects or instances of the entity classes. The entity structure can be created and / or stored as a data structure of an object-oriented programming language. The entity classes comprise variables, properties, and / or functions assigned to the respective entities. All entities of an entity class have the same variables, properties, and / or functions.

[0008] An autonomous vehicle is a vehicle that can target and / or navigate to a destination independently without driver intervention. It can react to other road users and adhere to traffic regulations. This particularly applies to vehicles with AD levels L4 or L5 or higher.

[0009] An autonomous vehicle has a control system that performs autonomous or self-driving. The method described herein can be performed automatically by this control system. The entity structure can be stored in the control system, and in particular, in its memory.

[0010] According to one embodiment, the method comprises classifying entities in the environment of the vehicle based on sensor data from sensor data detected by the vehicle, wherein a sensor data-based entity comprises at least a position, a geometry and / or a motion state and a type of entity. Real entities in the environment of the vehicle are detected and classified by sensors of the vehicle. It should be understood that the term entity is used both for the (real) entities and for the associated data structures, i.e., data entities. The environment of the vehicle can be the area around the vehicle that is maximally detectable by the sensors. The classification can be carried out in a known manner, for example using image recognition and / or machine learning algorithms. The sensor data can comprise image data from a camera, radar data from a radar, and / or lidar data from a lidar of the autonomous vehicle.

[0011] At the end of the classification, each entity has at least a position and a type. The entity may have further spatial properties, such as orientation, fill volume, velocity, acceleration, etc. The type of the entity is another property of the entity and / or usually comes from a classifier that performs the classification. The type of the entity can be a generic term for the entity and / or a type of the entity. Types of entities can be, for example, "other vehicle," "pedestrian," "drivable surface," "road marking," etc.

[0012] According to one embodiment, the method further comprises: assigning the sensor data-based entities to entity classes based on the entity type, wherein the entity classes are sorted into a class hierarchy that subordinates certain entity classes to other entity classes. The entity classes are all assigned to one or more entity types. Thus, each entity can be created as an instance of one or more entity classes.

[0013] The class hierarchy is structured according to entity types, meaning that an entity that actually represents a generalization of a more specific entity is also superior to the specific entity in the class hierarchy. Examples include vehicle - truck, vehicle - parked vehicle, and non-driveable obstacle - traffic island. In this example, the entity of a parked truck can be assigned to the entity classes truck and parked vehicle.

[0014] According to one embodiment, safety rules for an autonomous driving system are assigned to the entity classes, which are inherited by the subordinate entity classes. Thus, each type of entity is assigned specific safety rules. A safety rule can be a property of the entity class from which the autonomous driving system can derive safety conditions for the entity. A safety rule can, for example, be a rule for a minimum distance or a maximum relative speed. The condition is determined from the rule and is then a value, for example, the minimum distance or the maximum relative speed.

[0015] The entity structure can also be used by the autonomous vehicle to control the vehicle, for example, to find a route to reach a destination. The safety rules are used by the autonomous vehicle to achieve driving behavior that leads to safer driving behavior and driving behavior that complies with traffic regulations. By "safer driving behavior" we mean that safety rules can be encoded that a human could also implement and that are also understandable to a human. The behavior of a trained machine learning algorithm in all possible situations is generally difficult to predict. It is not easy to understand how the autonomous vehicle will behave in certain situations. The safety rules ensure that the vehicle can behave algorithmically and comprehensibly with regard to its driving behavior.For example, rules such as “Drive so as not to rear-end a vehicle in front of you” or “Brake appropriately when a vehicle in front of you changes lane” can be coded and then implemented by the vehicle.

[0016] According to one embodiment, a safety rule comprises a function, a mathematical formula, and / or an algorithm, which, or the properties assigned to the entity, generate safety conditions for the autonomous vehicle. The safety rule is assigned to the entity class and generates a safety condition from variables or values ​​and / or properties of the entity and optionally other values ​​(e.g., of the autonomous vehicle). The safety condition can comprise one or more values, such as a minimum distance from the entity, a maximum relative speed to the entity, a maximum acceleration of the autonomous vehicle, etc.

[0017] According to one embodiment, the entity classes comprise at least classes for traffic-relevant and non-traffic-relevant entities and / or traversable and non-traversable entities.

[0018] For example, the entity classes for drivable objects, such as beverage cans, small animals, etc., and for drivable surfaces, such as roads, lanes, etc., are subordinate to the entity class of drivable entities.

[0019] For example, the entity classes for road users, obstacles, potentially dangerous entities, and / or the boundaries of drivable areas are subordinate to the entity class for non-drivable entities. Other class hierarchies are also conceivable.

[0020] For example, the safety rules for drivable areas include at least staying on the drivable area and / or the safety rules for non-drivable entities include at least maintaining a minimum distance from the non-drivable entities.

[0021] For example, a stop line can be assigned to the entity classes "traffic sign" and "road marking," which in turn can be subordinate to the traffic-relevant entities. Safety rules can also be situational, meaning they depend on previous events, since the stop line may be crossed after stopping. For example, a lane marking may only be crossed when changing lanes.

[0022] A traffic sign can be assigned to the corresponding entity class Traffic Sign and additionally to the entity class Drivable Entity if it is painted on the road, or to the entity class Non-Drivable Entity if it is a mounted sign.

[0023] According to one embodiment, the method further comprises: generating anticipated (e.g., non-visible) entities from the entities based on sensor data; and assigning the anticipated entities to the entity classes based on the type of the entities. Once all entities that can be classified from the sensor data have been generated, anticipated entities can be generated, in particular algorithmically, for example, using appropriate anticipation rules. For example, a ball rolling out from behind a stationary vehicle can lead to a child behind the vehicle with the corresponding safety rules. Or a bus can lead to people behind the bus. Furthermore, a cyclist who disappears behind a truck and was present in the entity structure at an earlier point in time can be retained there.

[0024] According to one embodiment, the anticipated entities describe objects that cannot be detected by the vehicle's sensors. Some entities can be hidden behind other entities, at least temporarily, meaning they cannot be detected by the vehicle's sensors. These entities can be inserted into the entity structure, for example, using rules. Thus, the security rules of the corresponding entity classes are also taken into account.

[0025] According to one embodiment, the method further comprises: performing a security assessment of the entities, wherein an entity is classified as suspicious if its properties and / or criteria for classification change and / or its position changes in a way that is not possible for the class of the entity. The assessment can be performed using assessment rules, in particular algorithmically. Each assessment rule can output "suspicious" or "non-suspicious" as a result. For example, one rule may state that a pedestrian can only move at a certain maximum speed, or another rule may state that a bicycle cannot suddenly change direction by 90°.

[0026] The classification criteria can be determined from sensor data for classification and / or can be values ​​and / or quantities that contribute to the classification. For example, a vehicle cannot change its size. All properties of entities, such as position, geometry (e.g., size, shape, other features), acceleration, speed, etc., can be used.

[0027] According to one embodiment, the method further comprises: if an entity is classified as suspicious, moving the entity to a higher-level entity class; or moving the entity from its entity class to another entity class with different security rules. The higher-level or other entity class will typically have stricter security rules, such as a larger minimum distance.

[0028] Another aspect of the invention relates to a method for controlling an autonomous vehicle, the method comprising: generating the entity structure; inputting the safety rules into an autonomous driving system; generating a trajectory of the autonomous vehicle by the autonomous driving system, wherein the trajectory includes a location, speed, and / or acceleration of the autonomous vehicle in all mechanical degrees of freedom such that a driving goal is implemented and the safety rules are observed; and controlling the autonomous vehicle along the trajectory.

[0029] In addition to the safety rules, there can also be a driving goal, such as reaching a specific location, which can be divided into further sub-goals, such as the route points of a trajectory that can be provided by a navigation system. The autonomous driving system attempts to implement the driving goal and, for example, reach the destination. It can control the vehicle's actuators so that the autonomous vehicle follows the trajectory. The safety rules can be used to limit the permissible parameter space for the autonomous driving system. For example, no trajectory is permitted that violates a safety rule (e.g., a minimum relative distance to an obstacle).

[0030] A further aspect of the invention relates to a computer program which, when executed on at least one processor, performs the method according to one of the preceding claims. The processor can be a processor of a control system. A further aspect of the invention relates to a computer-readable medium on which such a computer program is stored, such as the memory of the control system. A computer-readable medium can be a hard disk, a USB storage device, a RAM, a ROM, an EPROM, or a FLASH memory. A computer-readable medium can also be a data communications network, such as the Internet, which enables the download of program code.

[0031] Another aspect of the invention relates to a vehicle system for an autonomous vehicle, comprising: at least one sensor for acquiring sensor data; actuators for controlling the vehicle; and a control system configured to receive the sensor data, perform the method as described herein, and control the actuators for controlling the vehicle. The control system may comprise a classification system and an autonomous driving system. It should be understood that features of the method, as described above and below, may also be features of the vehicle system, and vice versa.

[0032] In the following, embodiments of the invention are described in detail with reference to the accompanying figures. Fig. 1 schematically shows an autonomous vehicle with a control system according to an embodiment of the invention. Fig. Figure 2 shows a schematic diagram of an entity class. Fig. 3 shows an entity class hierarchy. Fig. 4 shows a flowchart for a method for controlling an autonomous vehicle according to an embodiment of the invention.

[0033] The reference symbols used in the figures and their meaning are summarized in the list of reference symbols. Identical or similar parts are generally provided with the same reference symbols.

[0034] Fig. 1 shows a vehicle system 10 for an autonomous vehicle, comprising sensors 12a, 12b for acquiring sensor data, actuators 16a, 16b, 16c for controlling the autonomous vehicle, and a control system 14 for controlling the vehicle. The sensors 12a, 12b comprise, for example, a camera, a radar, and / or a lidar. The sensor data can comprise image data from the camera, radar data from the radar, and / or lidar data from the lidar of the autonomous vehicle. The actuators 16a, 16b, 16c can be a drive, a brake, a steering system, etc., of the autonomous vehicle.

[0035] The control system 14 can comprise multiple software modules in multiple computing units, for example, directly within the sensors 12a, 12b. Shown are a classification system 18 that generates an entity structure 20 as described herein, and an autonomous driving system 22 that, among other things, controls the vehicle system 10 and thus the autonomous vehicle based on the entity structure 20.

[0036] Fig. Figure 2 schematically shows an entity class 24. Each entity class 24 has variables, functions and / or general properties 26 and one or more security rules 28. Furthermore, each entity class 24 can be subordinate to one or more other entity classes 24, as indicated by the arrow. Further elements of an entity class 24 can be one or more anticipation rules 30 and / or one or more evaluation rules 32. The meaning and function of the various rules is explained in relation to the Fig. 4 and the procedure is explained in more detail.

[0037] The entity structure 20 from the Fig. 1 comprises entities 34, in particular a hierarchy of entities 34 assigned to entity classes 24. Entities 34 are data objects or instances of the entity classes 24. The entity structure 20 can be created and / or stored as a data structure of an object-oriented programming language. All entities 34 of an entity class 24 have the same variables, functions, properties 26, and rules 28, 30, 32. The hierarchy of entities 34 is generated according to the hierarchy of entity class 24.

[0038] The Fig. 3 shows an example of a possible class hierarchy 36 for the entity structure 20, which is constructed from entity classes 24. All entity classes 24 are subordinate to a base class 24a, i.e., a fundamental safety-relevant entity class. At least classes for traffic-relevant entities 24b and non-traffic-relevant entities 24c are subordinate to the base class 24a. A non-traffic-relevant entity 24c can, for example, be a house next to the road that cannot be directly reached by the autonomous vehicle. Non-traffic-relevant entities 24c can be further subdivided into non-traffic-relevant entities with a possible influence on sensor technology and classification 24l or non-traffic-relevant entities without influence on sensor technology and classification 24m.For example, the sun may blind a camera even though a collision with the sun is excluded, so that the sun can be classified as a non-traffic-relevant entity with a possible influence on sensors and classification 24l, while a stone wall at a large distance from the road can be classified as a non-traffic-relevant entity with no influence on sensors and classification 24m.

[0039] Class 24 for traffic-relevant entities (24b) is subordinate to classes for traversable entities (24d), non-traversable entities (24e), and traffic signs (24f). Entity classes 24 for traversable objects (24g) and for drivable surfaces (24h) are subordinate to entity class 24 for traversable entities (24d). Traversable objects (24g) can be a mouse or a beverage can. Drivable surfaces (24h) can be a part of the road and a lane.

[0040] The entity class 24 of non-driveable entities 24e is assigned to the entity classes 24 for road users 24i (such as car, truck, pedestrian), driving area limitation 24j (such as road marking, solid line, stop line, transition from asphalt to grass) and potentially dangerous entity 24k (such as boulder, fallen tree, wild boar).

[0041] Each of the entity classes 24 has corresponding safety rules 28. For example, a safety rule 28 for a drivable area 24h may specify remaining on the drivable area 24h. A safety rule 28 for non-drivable entities 24e may specify maintaining a minimum distance, a maximum relative speed, and / or a maximum relative acceleration relative to the non-drivable entities 24e.

[0042] Fig. Figure 4 shows a flowchart for a method for generating an entity structure 20 describing the environment of an autonomous vehicle and for controlling the autonomous vehicle. The method is executed automatically by the control system 14. In particular, steps S10 to S14 are executed by the classification system 18, and step S16 is executed by the driving system 22.

[0043] In step S10, the sensors 12a, 12b generate sensor data, which is forwarded to the classification system 18. The classification system 18 classifies entities 34, 34a in the vehicle's surroundings based on the sensor data. During classification, the position, geometry, state of motion, and type of the entity can also be determined and set as properties 26 of the entity. In step S10, a list of classified entities 34a and their properties is created. It is also possible for the class hierarchy 36 to have an "unclear class" entity class; for example, the base class 24a can be used for this purpose.

[0044] In step S10, real entities in the vehicle's surroundings are detected by sensors 12a, 12b of the vehicle and subsequently classified. The classification can be performed in a known manner, for example, using image recognition and / or machine learning algorithms. The sensor data can include image data from a camera, radar data from a radar, and / or lidar data from a lidar of the autonomous vehicle.

[0045] At the end of the classification, each entity 34a has at least a position, a geometry, a motion state, and a type as properties 26. An entity 34 may have further spatial properties. The type of the entity 34 is a further property of the entity 34 and / or typically originates from a classifier that performs the classification. The type of the entity 34 may be a generic term for the entity and / or a type of the entity. Types of entities may be, for example, "other vehicle," "pedestrian," "drivable surface," "road marking," etc.

[0046] The classification system 18 further assigns the entities 34, 34a based on sensor data to the entity classes 24, in particular based on the type of the entities 34. The entity classes 24 are all assigned to one or more entity types. Thus, each entity 34 can be created as an instance of one or more entity classes 24.

[0047] As already mentioned in relation to the Fig. 2, the entity classes 24 are assigned safety rules 28 for an autonomous driving system 22, which are inherited by the subordinate entity classes 24. Thus, each type of entity 34 is assigned specific safety rules 28. A safety rule 28 can be a property of the entity class 24 from which the autonomous driving system 22 can derive safety conditions for the entity 34. A safety rule 28 can, for example, be a rule for a minimum and / or maximum distance, a maximum / minimum relative speed, or a maximum and / or minimum relative acceleration. The safety condition is determined from the safety rule 28 and is a value, for example the minimum and / or maximum distance, the maximum and / or minimum relative speed, or a maximum / minimum relative acceleration.

[0048] A safety rule 28 is a function, a mathematical formula, and / or an algorithm that generates safety conditions for the autonomous driving system 22 from the properties 26 assigned to the entity 34. The safety rule 28 is assigned to the entity class 24 and generates a safety condition from variables or values ​​and / or properties 26 of the entity 34 and optionally other values ​​(for example, of the autonomous vehicle). The safety condition can include one or more values, such as a minimum distance from the entity, a maximum relative speed to the entity, a maximum relative acceleration or absolute acceleration of the autonomous vehicle, etc.

[0049] In step S12, the classification system 18 generates anticipated entities 34b from the entities 34a based on sensor data. The classification system 18 then assigns the anticipated entities 34b to the entity classes 24 based on the type of the anticipated entities 34. Once all entities 34a that can be classified from the sensor data have been generated, anticipated entities 34b can be generated, in particular, algorithmically, for example, using corresponding anticipation rules 30. For example, a detected and classified ball rolling out from behind a stationary vehicle can lead to an anticipated child behind the vehicle with the corresponding safety rules. Or a detected and classified bus can lead to anticipated people behind the bus.Furthermore, a detected and classified cyclist who disappears behind a truck and was present in the entity structure 20 at an earlier point in time can be retained therein and its reasonably foreseeable behavior can be anticipated based on its properties 26.

[0050] The anticipated entities 34b describe objects that cannot be detected by the vehicle's sensors 12a, 12b. Some entities 34 may be at least temporarily hidden behind other entities 34, i.e., not detectable by the vehicle's sensors 12a, 12b. These entities 34 can be inserted into the entity structure 20, for example, via rules 30. Thus, the security rules 28 of the corresponding entity classes 24 are also taken into account.

[0051] In step S14, the classification system 18 performs a security assessment of the entities 34, 34a, 34b, wherein an entity 34 is classified as suspicious if its properties 26 and / or criteria for classification change in a way that is not possible for the class 24 of the entity 34. The assessment can be performed using assessment rules 32, in particular algorithmically. Each assessment rule 32 can output "suspicious" or "non-suspicious" as a result. For example, one rule 32 can require that a pedestrian can only move at a certain maximum speed, or another rule 32 can require that a bicycle cannot suddenly change direction by 90°.

[0052] Furthermore, the classification system 18 moves the entity 34 in the entity structure 20 if an entity 34 is classified as suspicious. The entity 34 can be moved to a higher-level entity class 24 or, more generally, to another entity class 24 with different security rules 28. The higher-level or other entity class 24 will generally have stricter or more stringent security rules 28, such as a larger minimum distance.

[0053] In step S16, the entity structure 20 and, in particular, the safety rules 28 are transferred to the autonomous driving system 22. The autonomous driving system 22 anticipates or calculates a further development of the scenario based, among other things, on the entity structure 20. This creates potential future locations for the entities 34, 34a, 34b over time. Furthermore, the autonomous driving system 22 anticipates or calculates the performance of the actuators 16a, 16b, 16c under the conditions of the scenario in order to take into account its own available scope of action.

[0054] In step S16, the autonomous driving system 22 further generates a trajectory 38 of the autonomous vehicle, wherein the trajectory 38 includes a location, speed, and / or acceleration of the autonomous vehicle, so that a driving destination is implemented and the safety rules 28 are observed. The entity structure 20 is used by the autonomous driving system 22 to find a route with which the driving destination can be reached.

[0055] In addition to the safety rules 28, there may also be a driving goal, such as reaching a specific location, which can be divided into further sub-goals, such as the route points of a trajectory 38, which can be provided by a navigation system. The autonomous driving system 22 attempts to implement the driving goal and, for example, reach the destination. The safety rules 28 and the safety conditions derived therefrom can be used to limit the permissible parameter space for the autonomous driving system. For example, no trajectory 38 is permissible that violates a safety rule (e.g., a minimum relative distance to an obstacle).

[0056] The safety rules 28 are used to achieve driving behavior that leads to safer driving behavior and traffic-compliant driving behavior, taking into account the performance of the actuators 16a, 16b, 16c. Safety conditions are derived for all entities 34 from the safety rules 28, the determined properties of the entities 34, the anticipated scenario development, and the anticipated performance of the entities. These safety conditions must be considered when selecting a possible trajectory 38.

[0057] At the end of step S16, the autonomous driving system 22 steers the autonomous vehicle along the trajectory 38. The autonomous driving system 22 can control the actuators 16a, 16b, 16c of the vehicle such that the autonomous vehicle follows the trajectory 38.

[0058] Additionally, it should be noted that "comprising" does not exclude other elements or steps, and "a" or "an" does not exclude a plurality. Furthermore, it should be noted that features or steps described with reference to one of the above embodiments may also be used in combination with other features or steps of other embodiments described above. Reference signs in the claims are not to be considered as limitations. Reference symbol 10 Vehicle system 12a Sensor 12b Sensor 14 Tax system 16a Actuator 16b Actuator 16c Actuator 18 Classification system 20 Entity structure 22 Driving system 24 Entity class 24a Basic Class 24b transport-relevant entity 24c non-traffic-relevant entity 24d traversable entity 24e non-driveable entity 24f traffic signs 24g drivable objects 24-hour accessible area 24i road users 24j driving range limitation 24k potentially dangerous entity 24l non-traffic-relevant entity with possible influence on sensor technology and classification 24m non-traffic-relevant entity without influence on sensors and classification 26 Property 28 Safety Rule 30 Anticipation rule 32 Valuation rule 34 Entity 34a Entity based on sensor data 34b anticipated entity 36 Class hierarchy 38 Trajectory QUOTES CONTAINED IN THE DESCRIPTION

[0000] This list of documents submitted by the applicant was generated automatically and is included solely for the convenience of the reader. This list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions. Cited non-patent literature

[0000] On a Formal Model of Safe and Scalable Self-driving Cars" by Shai Shalev-Shwartz, Shaked Shammah, Amnon Shashua, Mobileye, 2017, arXiv:1708.06374

[0004]

Claims

[1] A method for generating an entity structure (20) describing the environment of an autonomous vehicle, the method comprising: Classifying sensor data-based entities (34, 34a) in the environment of the vehicle from sensor data acquired by the vehicle, wherein a sensor data-based entity (34) comprises at least a position and a type of entity; Assigning the sensor data-based entities (34, 34a) into entity classes (24) based on the type of entities (34), wherein the entity classes (24) are sorted into a class hierarchy (36) that subordinates certain entity classes (24) to other entity classes (24); wherein the entity classes (24) are assigned safety rules (28) for an autonomous driving system (22), which are inherited by the subordinate entity classes (24). [2] Method according to claim 1, wherein a safety rule (28) comprises a function, a mathematical formula and / or an algorithm that generates safety conditions for the autonomous driving system (22) from the properties (26) assigned to the entity (34). [3] Method according to claim 1 or 2, wherein the entity classes (24) comprise at least classes for traffic-relevant entities (24b) and non-traffic-relevant entities (24c) and / or traversable entities (24d) and non-traversable entities (24e); wherein the entity classes for traversable objects (24g) and for traversable surfaces (24h) are subordinate to the entity class of traversable entities (24d); wherein the entity classes for road users (24i) are subordinate to the entity class of non-driveable entities (24e). [4] Method according to claim 1 or 2, wherein the safety rules (28) for drivable areas (24h) at least include remaining on the drivable area (24h); and / or wherein the safety rules (28) for non-overridable entities (24e) at least comprise maintaining a minimum distance and / or a maximum and / or minimum relative speed and / or a maximum and / or minimum relative acceleration to the non-overridable entities (24e). [5] Method according to one of the preceding claims, further comprising: Generating anticipated entities (34b) from the sensor data-based entities (34a) and / or the property (26) associated with the entity; Assigning the anticipated entities (34b) to the entity classes (24) based on the type of the anticipated entities (34). [6] Method according to claim 5, wherein the anticipated entities (34b) describe undetectable and / or detected objects by sensors of the vehicle. [7] Method according to one of the preceding claims, further comprising: Carrying out a security assessment of the entities (34), wherein an entity (34) is classified as suspicious if its properties (26) and / or criteria for classification change in a way that is not possible for the class (24) of the entity (34); if an entity (34) is classified as suspicious: Moving the entity (34) to a parent entity class (24); or Moving the entity (34) from its entity class (24) to another entity class (24) with different security rules (28). [8] Method according to one of the preceding claims, wherein the sensor data comprise image data from a camera, radar data from a radar and / or lidar data from a lidar of the autonomous vehicle. [9] A method for controlling an autonomous vehicle, the method comprising: Creating the entity structure (20) using the method according to one of the preceding claims; Entering the safety rules (28) into an autonomous driving system (22); generating a trajectory (38) of the autonomous vehicle by the autonomous driving system (22), wherein the trajectory (38) comprises a location, speed and / or acceleration of the autonomous vehicle such that a driving destination is implemented and the safety rules (28) are observed; Controlling the autonomous vehicle along the trajectory (38). [10] A computer program which, when executed on at least one processor, carries out the method according to any one of the preceding claims. [11] A computer-readable medium having stored thereon a computer program according to claim 8. [12] Vehicle system (10) for an autonomous vehicle, comprising: at least one sensor (12a, 12b) for detecting sensor data; Actuators (16a, 16b, 16c) for controlling the vehicle; a control system (14) designed to receive the sensor data, to carry out the method according to one of claims 1 to 9 and to control the actuators (16a, 16b, 16c) for controlling the vehicle.

Images