CN115695473A - Ontology construction method for intelligent networking automobile simulation test scene - Google Patents
Ontology construction method for intelligent networking automobile simulation test scene Download PDFInfo
- Publication number
- CN115695473A CN115695473A CN202211241729.9A CN202211241729A CN115695473A CN 115695473 A CN115695473 A CN 115695473A CN 202211241729 A CN202211241729 A CN 202211241729A CN 115695473 A CN115695473 A CN 115695473A
- Authority
- CN
- China
- Prior art keywords
- scene
- ontology
- traffic
- semantic
- test
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Traffic Control Systems (AREA)
Abstract
The invention discloses a body construction method of an intelligent network-connected automobile simulation test scene, which comprises the steps of designing a basic test scene, determining a function to be tested, adjusting model parameters and expanding the basic test scene; analyzing, sorting and classifying the scene information, and conceptualizing the scene information by using the ontology; classifying information contained in the scene; constructing a traffic scene ontology model by taking the concept of the scene as a root node; using OWL2 to complete the construction of a scene concept ontology and establish an ontology knowledge base; introducing a semantic network rule language SWRL to carry out semantic description to obtain constraint; and checking and evaluating the constructed traffic scene body model, and correcting and optimizing according to the result. The invention provides an ontology-based automatic driving scene semantic modeling method, which is used for constructing a scene library for virtual simulation test of an intelligent networked automobile and ensuring a formalized and standardized test scene.
Description
Technical Field
The invention relates to the technical field of intelligent networking automobile simulation test, in particular to a body construction method of an intelligent networking automobile simulation test scene.
Background
Ontology is a branch of philosophy, and mainly studies the nature of existence. In the computer and related fields, ontology refers to a basic method of applying ontology, and the method analyzes concepts and models the real world
The entities in (1) are abstracted into a set of concepts and theories of relationships between the concepts. Since the 90 s of the 20 th century, ontologies have become one of the important research directions in the computer field, and are now widely applied to a plurality of fields such as knowledge engineering, multi-agent systems, system modeling, semantic Web, heterogeneous information integration and the like.
Ontology (Ontology) is used for modeling a traffic scene, and introduces two types of languages OWL2 and Semantic network Rule Language (SWRL) for constructing Ontology Semantic descriptions, which are respectively used for representing knowledge bases and rules of domain ontologies.
The existing automatic driving simulation test scene library is constructed by using a representation method which only considers entity space positions, such as a grid map, a Bayesian network and the like. A road network is constructed in space, various static entities such as buildings, mark lines, signal lamps and the like are added on the road network, then traffic participants are added, and the relationship among objects is additionally added.
The performance representation of the automatic driving automobile has uncertainty, and is a result of mutual coupling of people, the automobile and the external environment, and the traditional test method is not suitable for use. According to the application characteristics and the feasibility of operation of the automatic driving automobile, the scene test in the virtual environment becomes an important means for functional verification of the automatic driving system, and the key of the scene test is how to construct a test scene which follows rationalization, formalization and standardization and how to generate a test case which is high in efficiency, wide in coverage and reusable according to the scene.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a body construction method of an intelligent networked automobile simulation test scene, which is used for carrying out structured analysis on a traffic scene in an automatic driving virtual test environment, establishing a basic hierarchical framework and defining related terms of scene elements; and designing a traffic scene ontology model based on the concept, finally, expanding semantic analysis, constructing an ontology model of a basic scene, establishing a corresponding scene knowledge base through ontology semantic description, representing elements in the scene as concepts, and formally analyzing the attributes and relationships of the concepts.
The purpose of the invention is realized by the following technical scheme:
a body construction method for an intelligent networking automobile simulation test scene comprises the following steps:
step 1: firstly, designing a typical and representative basic test scene which can be applied to an intelligent networking automobile simulation test, determining a function to be tested, and then expanding the basic test scene through the increase of scene elements and the adjustment of input parameters;
step 2: analyzing, sorting and classifying the scene information, and conceptualizing the scene information by using an ontology; classifying information contained in the scene;
and step 3: taking the concept of a scene as a root node, and constructing a traffic scene ontology model from top to bottom based on the structure and the hierarchical description of the scene;
and 4, step 4: using OWL2 to complete the construction of a scene concept ontology, completing semantic expression, and using OWL2 to establish an ontology knowledge base;
and 5: a semantic network rule language SWRL is introduced to enhance the rule description capability, and the SWRL can directly refer to concepts, attributes and examples in a local knowledge base to carry out semantic description on traffic rules to obtain constraints;
step 6: and checking and evaluating the constructed traffic scene body model, and correcting and optimizing according to the result.
Specifically, step 1 specifically includes: firstly, designing a basic test scene according to an expected verified automatic driving function, wherein the scene has typicality and representativeness; starting from functional requirements, designing a typical and representative basic test scene; then expanding the basic test scene through the increase of scene elements and the adjustment of input parameters, and combining a plurality of test cases; and modeling and analyzing the accident scene, selecting key scene elements as the input number of the test scene, combining different scene elements, and deriving a plurality of test cases from one accident scene.
Specifically, step 2 specifically includes the following substeps:
s201, analyzing a scene structure, and obtaining a total composition structure of a traffic scene by analyzing specific definitions of the traffic scene;
s202, according to the definition and structure of scene terms, the traffic scene is decomposed into four layers, wherein the first layer is a road network, the second layer is a static object, the third layer is a dynamic object, and the fourth layer is an environment.
Specifically, step 3 specifically includes the following substeps:
s301, analyzing the structure of the scene, uniformly defining terms of the scene, standardizing the meanings of the terms, distinguishing scene concepts with different meanings, and providing a consistent expression for the construction of a later scene model;
s302, constructing a traffic scene ontology model from top to bottom by using a scene as a root concept based on the structure and the hierarchical description of the scene and adopting an ontology method;
and S303, performing scene semantic analysis, creating an example according to the traffic scene ontology model, extracting partial concepts in the macroscopic traffic scene ontology model, and defining attributes and relations required by the scene.
Specifically, step 4 specifically includes the following substeps:
s401, analyzing the relation of 'human-vehicle-environment' by the traffic scene ontology model through OWL2, and performing formal description on semantic relation and attributes in a scene entity;
s402, establishing an ontology knowledge base by using OWL2 according to the semantic relation and the attribute in the formally described scene entity.
Specifically, step 5 specifically includes: based on concepts, attributes and examples of an ontology knowledge base established by OWL2, a semantic network rule language SWRL is introduced to describe the semantics of traffic rules, and the traffic rules and the implicit driving order are established to obtain the constraints of a traffic scene ontology model.
Specifically, step 6 specifically includes: carrying out consistency and integrity inspection on the traffic scene ontology model; integrity check checks the integrity of the examples; consistency checking is to verify the consistency of a relationship, checking whether the relationship definition is logically incorrect.
The invention has the beneficial effects that:
1. the ontology is applied to the construction of a scene library of the intelligent networking automobile virtual simulation test, and a formalized and standardized test scene is guaranteed.
2. An automatic driving test case generation method based on a scene is designed, and the semantic expression of a traffic scene is completed by using OWL2 and SWRL languages.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a schematic view of a scene information analysis, sorting and classification flow;
FIG. 3 is a schematic diagram of a scene;
FIG. 4 is a flow chart of traffic scene ontology model construction;
FIG. 5 is a schematic view of a traffic scene;
FIG. 6 is a schematic diagram of the construction of a concept ontology of a scenario;
FIG. 7 is a semantic model diagram of a scene.
Detailed Description
The following detailed description is given to select technical features, objects and advantages of the present invention in order to make the technical features, objects and advantages of the present invention more clearly understood. It should be understood that the embodiments described are illustrative of some, but not all embodiments of the invention, and should not be taken to limit the scope of the invention. All other embodiments that can be obtained by a person skilled in the art based on the embodiments of the present invention without any inventive step are within the scope of the present invention.
The first embodiment is as follows:
in this embodiment, as shown in fig. 1, a method for constructing an intelligent networked automobile simulation test scenario includes the following steps:
step 1: firstly, a typical and representative basic test scene which can be applied to an intelligent networking automobile simulation test is designed, a function to be tested is determined, and then the basic test scene is expanded through the increase of scene elements and the adjustment of input parameters.
Step 2: analyzing, sorting and classifying the scene information, and conceptualizing the scene information by using an ontology; information contained in a scene is classified into: objects, tasks, attributes, constraints. The process is referred to traffic regulations and driving experience.
And 3, step 3: the concept of the scene is used as a root node, and a traffic scene ontology model is constructed from top to bottom based on the structure and the hierarchical description of the scene. And standardizing and unifying the definition and the naming of the attributes in the body construction, determining a semantic relation frame of the scene, determining the semantic relation of various information in the scene, and determining the data attributes and the object attributes in the body.
And 4, step 4: and (5) using OWL2 to complete the construction of the scene concept ontology and complete semantic expression. Ontology knowledge base was built using OWL 2.
And 5: the semantic network rule language SWRL is introduced to enhance the rule description capability, and the SWRL can directly refer to concepts, attributes and examples in the OWL knowledge base so as to realize the semantic description of the traffic rules.
And 6: and checking and evaluating the constructed traffic scene body model, and correcting and optimizing according to the result.
2. In step 1, basic scene generation needs to be completed, and the method specifically comprises the following steps:
firstly, a basic test scene is designed according to the expected verified automatic driving function, and the scene is typical and representative. Starting from functional requirements, a typical and representative basic test scenario (reference may be made to the scenarios in ADAS and test specifications) is designed.
And then expanding the basic test scene through the increase of scene elements and the adjustment of input parameters, and combining a plurality of test cases. The accident scene can be modeled and analyzed, key scene elements are selected as input numbers of the test scene, different scene elements are combined, and a plurality of test cases can be derived from one accident scene.
3. In step 2, the scene information is analyzed, sorted and classified, and is processed conceptualized by using the ontology, and the method specifically comprises the following steps:
3.1, analyzing the scene structure, and analyzing the specific definition of the traffic scene to obtain the overall composition structure of the traffic scene, as shown in fig. 2, which consists of five parts.
The scene comprises the components of objects, tasks, attributes and constraints, and the scene itself is also composed of one or a plurality of scenes (working conditions) with time sequence. When a scene is constructed by a language, the scene needs to be defined and declared according to the structure.
3.2 in order to describe the real environment in a targeted manner, according to the definition and structure of the scene terms, the traffic scene is divided into four layers, the element composition in the scene is described more clearly and intuitively, and the inclusion range of each layer of elements is classified uniformly. Depending on the functional requirements of the desired verification and the characteristics of the actual scenario, one or more of four layers may be extracted to describe the instance, flexibly combining scenarios:
layer 1: road network
The road network layer represents the geometry and topology of roads using basic elements in road design, such as straight lines, curves and clothoids, where straight lines are described by length, curves are described by constant radius, curvature and length, and lane markings are specified by width, color and style (e.g., solid or dashed lines). In addition, the state of the road surface must be specified, such as dry, slippery, snowy, road wear, and the like. The design parameters of expressways, urban roads and rural roads are different, and the design parameters refer to the technical standards of highway engineering. An intersection can be connected with a plurality of basic roads, so that the number of the connected roads and the number of lanes of the connected roads need to be defined, and lanes for left-turning, right-turning or straight-going should be marked.
Layer 2: static objects
The static object layer represents a static environment around a road or a static element on the road, and is similar to the definition of a "landscape" term, and mainly comprises traffic infrastructure of a roadside, such as a traffic sign, a traffic light, a road separation barrier and the like. According to specific requirements, elements such as houses, fences, separation belts, street lamps and the like can be added around the basic road to generate a near-real urban environment.
Layer 3: dynamic objects
The dynamic object layer defines the number and behavior of dynamic elements at various levels of abstraction. At a higher level of abstraction, random traffic flows may be generated on the specified roads to simulate congested or clear traffic conditions; at a lower level of abstraction, dynamic objects may be represented as concrete traffic participant instances, such as following vehicles, parallel vehicles, lane changing vehicles, pedestrians, etc. These traffic participants will interact or interfere with the autonomous automobile in the scene for verification of the responsiveness and decision-making capabilities of the autonomous system.
Layer 4: environment(s) of
The environmental layer mainly represents a climate environment and an information environment. The climate environment comprises illumination conditions, and possible states of the climate environment comprise dawn, day, dusk, night and the like, so that the sensing function of the automatic driving system under different illumination conditions can be tested, and the climate environment also comprises weather conditions, wherein the possible states comprise sunny days, cloudy days, rainy and snowy days, haze and the like, and the weather conditions can influence the illumination conditions, the road surface state and the road visibility. The information environment is mainly the strength of 5G signals, and 5G is an indispensable part of a high-grade automatic driving automobile and is the key for ensuring the communication quality and the information transmission speed of the automatic driving automobile.
The structure and the layering of the scene mainly aim at ensuring the consistency of the traffic scene body model and constructing a reasonable, formalized and standardized test scene.
4. In step 3, the concept of the scene is taken as a root node, and a traffic scene ontology model is constructed from top to bottom based on the structure and the hierarchical description of the scene, and the specific steps are as follows:
4.1. structural analysis of a scene
Firstly, the terms of the scene are uniformly defined, and the meaning of the terms is specified. As shown in FIG. 3, the scene concepts with different meanings are necessarily distinguished, and consistent expression is provided for the following scene model construction.
Scene (Scene) is a snapshot describing the environment, i.e. representing the environment at a certain moment, including the external scenery, the dynamic elements, the self-characterization of all traffic participants and the relations between the various scenes. The dynamic elements contain the state and attributes of the dynamic object, such as the coordinate positions, relative distances, velocities, etc. of the pedestrian and vehicle. The self-characterization of traffic participants refers to driving behavior and traffic events, such as acceleration, deceleration, car following, passing, etc.
There are many possible relationships between various scene elements, such as positional relationships and combinatorial relationships.
Landscape (Scenery) represents all Scenery that is geospatially static, including road networks, static elements, and weather conditions. The road network comprises various road sections and intersection intersections, and the emphasis represents the topological structure of the road. The static elements include obstacles, curbs, traffic signs, traffic lights, etc., wherein the traffic lights and partially variable traffic signs are static elements because they are stationary in their geographical locations and do not have subjective changes in state, although there is a change in phase. The weather conditions include cloudy, sunny, rainy and snowy states of weather and illumination conditions, and influence the friction coefficient of the road surface and the perception function of the automatic driving automobile.
Scenario (setup) is a state reflection of traffic Situation at a certain moment, the intentions and behaviors of all subjects (autonomous cars and other dynamic traffic participants) in the scenario belong to the traffic Situation, the scenario contains all relevant conditions, choices and determinants of the intentions and behaviors, and is spread through an event until the event is finished. The scenes comprise task events such as overtaking, lane changing, avoiding, crossing passing and the like, and in addition, the simulated traffic accidents also belong to the scene category, so the scenes have certain time span and subjectivity. A certain operating condition can be understood.
A scene (Scenario) is a key concept concerned in modeling of the present embodiment, and the test method based on the scene is first applied to the field of software testing. In the field of automatic driving, a scene is a comprehensive reflection of traffic scenes and driving scenes in a certain time and space range, information such as road networks, infrastructure traffic facilities, weather conditions, behaviors and states of traffic participants and automatic driving automobiles and the like jointly form the concept of the whole scene, the specific composition of the elements is determined by the expected verified automatic driving function, and the elements contained in different actual scenes are different. The scenes have the characteristics of certain time span and dynamic change, each scene starts from an initial scene, the driving behaviors and traffic events of the automatic driving automobiles and other traffic participants can be appointed, and therefore various scenes are introduced, and different scene test cases are derived.
4.2 scene ontology construction
The complexity of autonomous vehicles and the variety of conditions that may be encountered during driving results in a significant increase in the amount of testing effort required. By using the scene semantic description based on the body, clear traffic scene semantic expression can describe scene information through human and machine readable grammars, generalization and reuse can be carried out in a large number of traffic scenes, scene knowledge in the automatic driving field is standardized and organized, theoretical support and reference are provided for subsequent use case design, and a bridge can be built for mapping between a formalized semantic scene and a specific scene of a virtual driving platform.
Traffic scenes may be modeled by various methods such as a representation method occupying a grid map, a bayesian network, etc. considering only the spatial positions of entities, and although these methods can finely design a road network, when a large number of entities are involved, the model becomes very complicated and it is difficult to express semantic relationships between the entities and traffic rules.
The ontology is used as a specification for defining, sharing and conceptualizing, provides easily-understood shared knowledge for software systems and developers, can define a hierarchy structure among concepts, clearly represents context entities and domain knowledge in a knowledge-intensive system, and is a formal expression method which can interoperate and is easy to understand.
In the embodiment, an ontology method is selected to construct a traffic scene model, ontology knowledge is expressed by using OWL2, and a theoretical basis is provided for the following test scene design and test case generation.
As shown in fig. 4, in the present embodiment, based on the structure and the hierarchical description of the scene, the traffic scene ontology model is constructed from top to bottom by taking the scene as a root concept.
The scene is defined by terms of the scene, and the scene comprises static scenery, dynamic traffic participants and driving tasks in corresponding scenes, so that the scenery, the traffic participants and the tasks are used as a primary concept of an ontology and form a combined relationship with the scene; it can be defined by the term of scenery, scenery represents static elements, mainly including weather environment, road network and static elements, so the above 3 kinds of concepts form a combined relationship with scenery; next, the known concepts are further refined layer by layer, so that a traffic scene ontology model with a macroscopic view angle can be obtained, which is intended to cover most possible scenes in the urban traffic environment, wherein the concept including the inheritance relationship may have other subclasses in a specific scene, which are not expanded one by one, as shown in fig. 4.
1) RoadNetwork (road network): this concept is the most important part of representing the road geometric topology,
2) Staticientity (static entity): the entity represents an element with static geographic position, describes the layer 2 of a traffic scene, and mainly comprises sub concepts of TrafficLight, trafficSign, obstacle and the like. Common traffic lights are motor car lights, direction indicator lights, flashing warning lights, and the like; the traffic signs are mainly classified into warning signs, prohibition signs, indication signs, direction signs, and the like. The obstacles can be water horses or cone marks for dividing the road surface, other vehicles for reasons staying on the road, trees or telegraph poles falling down, and the like, and the road surface pit, the inspection well and the like also belong to obstacles and need vehicles to go around.
3) Participant (traffic Participant): the entity represents a dynamic element, describes the layer 3 of a traffic scene, and mainly comprises the sub-concepts of EgoVehicle (automatic driving vehicle), otherVehicle (other motor vehicles or non-motor vehicles), pedestrian (Pedestrian) and the like. When the automatic driving system identifies other traffic participants, some special vehicles need to be identified and collaborative avoidance is adopted, such as a road operation vehicle, a fire engine, an ambulance and the like; if in rural highways, it may also be desirable to avoid passing animals.
4) Weather (Weather environment): the entity describes the 4 th layer of a traffic scene, including cloudy, sunny, rainy and snowy, which can influence the intensity of illumination and interfere the perception capability of an automatic driving system; in addition, rainfall and snowfall both affect the road conditions and indirectly affect the braking capacity of an autonomous vehicle. Weather factors are also a key contributor in some autodrive test scenarios.
5) Task: task entities are key parts describing a scene, and tasks can be decomposed into specific Act (driving behavior) to be executed by traffic participants. According to the statement of the test regulations (trial) of the automatic driving function of the intelligent networked automobile, the main test tasks of the automatic driven automobile comprise 14 items, namely identification and response of traffic signs and marking lines, identification and response of traffic lights, identification and response of the driving state of a vehicle in front, identification and response of obstacles, identification and avoidance of pedestrians and non-motor vehicles, following driving, stopping by roadside, overtaking, merging, crossing passing, annular crossing passing, automatic emergency braking, manual taking-over operation and networking communication. Taking the lane change task as an example, the method can be decomposed into a plurality of steps: (1) turning on a steering lamp; (2) cutting into an adjacent lane; and (3) aligning the steering lamp. In each scenario, a traffic participant (mainly an autonomous vehicle) needs to perform a task or a series of task sets.
4.3 scene semantic analysis
Each concept in an ontology may be described by attributes, which are divided into two classes, object attributes and data attributes.
As shown in table 1 below, the object attributes are used to represent concepts and relationships between the concepts, and a list of object attributes based on traffic scene elements is established according to attribute names, definition domains and value domains in the table. The object properties may represent affiliation, e.g. the road network is part of a landscape, denoted isPartOf; each road section at least has one lane which is represented by hasLane; the junction is usually provided with sidewalks and traffic lights, denoted as hasTrafficLight and hasSidewalk; the object attributes may represent a positional relationship, such as a link meeting a meeting point, represented by connectitto; the proximity relation among a plurality of lanes is represented by hasLeft/RightLane; the position and direction (front, back, left and right relationship) of the automatic driving automobile and the road obstacle are represented by hasFront/backObstationary and hasLeft/rightObstationary; the position of the road where the scene element is located is indicated by isOn.
TABLE 1 object Attribute Table
As shown in Table 2, data attributes are used to represent individual attributes of a concept instance, a domain is defined as a concept or class, and a value domain is a data type. Data for an autonomous vehicle includes speed (hasSpeed), distance to other road elements (distTo), direction of travel (currentDirection), intent or task (hasintetion), and the like. The data attribute of the traffic light is color (hasColor), and the value range can comprise red, red flashing, green and green flashing. The data attribute of the lane includes lane width (hasLaneWidth), speed limit information (hasMax/MinSpeed), and the like. The data attribute of the traffic marking has meaning (hasProperty) and wear degree (webDegreee), wherein the wear degree is expressed in percentage. The data attribute of the road segment includes road surface condition (facecondition), and if the road segment is a curve, radius of curvature (curvatureRadius).
TABLE 2 data Attribute Table
Different attributes can be obtained from different angles, so that all attribute information cannot be listed blindly in the research process, and the attributes with certain influence factors are selected for analysis according to scene characteristics and function requirements.
Specific scenes in a real environment can create examples according to the traffic scene body model, part of concepts in a macroscopic traffic scene body model are extracted, attributes and relations required by the scenes are defined, the detailed degree of the body depends on the complexity degree of the actual scene and the functional requirements of an automatic driving system, and for example, for verifying the automatic emergency braking function, the attributes such as the density degree of traffic flow, the road surface condition (friction coefficient), the vehicle speed and the like need to be focused; for verifying the function of parking alongside, the road topology is of great concern, whereas the weather factor may be an irrelevant item.
5. And in the step 4, constructing a scene concept ontology by using OWL2 to finish semantic expression, wherein the specific example is as follows.
The semantic description aims to collect and construct relevant test parameters, so as to lay a foundation for generating test cases, and the semantic description of a specific scene is described in detail through an example.
As shown in fig. 5, it represents a multi-lane t-junction scene with traffic lights. The scene comprises 1T-shaped intersection, 3 road sections, 6 lanes, 3 road dotted lines, 3 sidewalks, 1 vehicle, 1 motor vehicle signal lamp and 1 pedestrian. The scene elements in the graph are named respectively and labeled with name serial numbers. In this scenario, the task of autodrive the car is to cross the intersection.
As can be seen from the scene diagram, the scene has 2 traffic participants, namely pedestrian1 (pedestrian) and egoVehicle (autonomous vehicle); trafficLight1 (traffic light) belongs to a unique static view and is positioned in the middle of an intersection; the intersection is connected with 3 road sections in the east and west and south directions, each road section is provided with 2 lanes, and the lanes are separated by dotted lines; egoVehicle is located in the driveway westLane2, pedestrian1 is located in the sidewalk3. A traffic scene ontology model based on the scene can thus be built, as shown in fig. 6.
The next step is to analyze the object attributes and the data attributes of the concepts according to the concepts in the traffic scene ontology model. The attributes of the intersection-related entities can be clearly and intuitively displayed by means of a semantic model diagram, as shown in fig. 7, red oval boxes represent dynamic elements (ego vehicle, pedestrian 1) such as traffic participants, black oval boxes represent landscape elements (traffic light, junction1, side walk3, westLane 2) such as road networks and traffic facilities, blue arrows represent object attributes (isOn, has, connect, currentLane) between entities, and black dashed boxes represent data attributes (hasSpeed, disttostepline, hasintuition) of the entities.
Here we assume that the speed of the egoVehicle is 20km/h, and the egoVehicle is 30 meters from the stop line of the crossing and intends to turn right.
Combining the above analysis, the traffic scene ontology model of the T-junction can be described in OWL2 language, which is exemplified as follows:
# statement partial concept (class)
Declaration(Class(:EgoVehicle));Declaration(Class(:Lane));Declaration(Class(:Junction));
# declaration section object properties
Declaration(ObjectProperty(:currentLane));
Declaration(ObjectProperty(:connectTo));
# declaration partial data attribute
Declaration(DataProperty(:hasSpeed));
Declaration(DataProperty(:distToStopLine));
Declaration(DataProperty(:hasIntention));
# statement section example Individual
Declaration(NamedIndividual(:junction1));
Declaration(NamedIndividual(:egoVehicle));
Declaration(NamedIndividual(:westLane2));
Declaration(NamedIndividual(:eastLane1));
Defining field and value field of # defining object attribute
ObjectPropertyDomain(:currentLane:EgoVehicle);
ObjectPropertyRange(:currentLane:Lane);
ObjectPropertyDomain(:connectTo:RoadSegment);
ObjectPropertyRange(:connectTo:Junction);
# Definitions Domain and value Domain of data Properties
DataPropertyDomain(:hasSpeed:EgoVehicle);
DataPropertyRange(:hasSpeed xsd:double);
DataPropertyDomain(:distToStopLine:EgoVehicle);
DataPropertyRange(:distToStopLine xsd:double);
DataPropertyDomain(:hasIntention:EgoVehicle);
DataPropertyRange(:hasIntention xsd:string);
# denotes an assertion set, exemplified by an autonomous vehicle only
ObjectPropertyAssertion(:currentLane:egoVehicle:westLane2);
ObjectPropertyAssertion(:connectTo:westLane2:junction1);
DataPropertyAssertion(:hasSpeed:egoVehicle“20.0”^^xsd:double);
DataPropertyAssertion(:distToStopLine:egoVehicle“30.0”^^xsd:double);
DataPropertyAssertion(:hasIntention:egoVehicle“turnRight”^^xsd:string)。
The traffic scene ontology analyzes the relation of 'human-vehicle-environment' through OWL2, formally describes the semantic relation and the attribute in the scene entity, and realizes effective organization and standard integration of scene information.
6. In step 5, a semantic network rule language SWRL is introduced to enhance rule description capability, and the specific steps are as follows:
the ontology modeling and semantic analysis of the traffic scene specifically expound and integrate 4 parts of objects, tasks, attributes and scenes in the overall structure, and formalized description of constraints is needed. The constraint is mainly expressed as a traffic rule and an implicit driving order, and is a knowledge basis for making a scene decision by an automatic driving automobile, namely, the situation of the automatic driving automobile is specified. The constraint can be realized by constructing a rule base, on one hand, the traffic laws can be analyzed in a syntactic manner and information can be extracted through a natural language processing technology, and on the other hand, rules can be established by using a rule description language.
An ontology knowledge base can be established by using OWL2, and in view of the defects of OWL2 in an inference mechanism, a semantic network rule language SWRL can be introduced to enhance the description capability of the rule.
The integrity and reasoning capability of the automatic driving test scene model can be improved through the regular language formalized representation constraint. The decision reasonability of the automatic driving system facing different scenes or executing different driving tasks can be verified through the construction rule, and meanwhile, expected test performance can be deduced according to scene description, so that a solution is provided for automatic generation of future test scenes.
7. In step 6, the constructed traffic scene body model needs to be checked and evaluated, and the method specifically comprises the following steps:
and (3) carrying out consistency and integrity check on the body: integrity check mainly checks the integrity of examples, for example, the specific examples under the child concepts are not defined under any parent concept, which indicates that the parent concept is incomplete or the child concept is added with errors; the consistency check mainly verifies the consistency of the relation, such as the check of the 'hasPart' and the 'ispartOf', and since the 'hasPart' relation has transmissibility and is a reciprocal relation, whether the relation definition is logically wrong is checked.
The evaluation of the scene ontology is developed on the basis of the above. The completeness, accuracy and level of the whole scene are clear, the defined relation is complete and correct, and whether the model can realize the expected function is determined.
In this embodiment, to complete the construction of a scene, a structured analysis is first performed on a traffic scene in an autopilot virtual test environment, a basic hierarchical framework is established, and relevant terms of scene elements are defined (these are the basis of the whole application, and accuracy, universality and complete coverage are to be ensured); designing a traffic scene body model based on the method, finally, developing semantic analysis, constructing the traffic scene body model of a basic scene, establishing a corresponding scene knowledge base through body semantic description, representing elements in the scene as concepts, and formally analyzing the attributes and the relations of the concepts; use OWL2 for modeling of traffic scenarios and SWRL for description of traffic laws; acquiring influence factors related to the automatic driving function, and taking attributes with the influence factors as input parameters; and preprocessing the value range of the input parameter by using methods such as equivalence class division, boundary value analysis and the like.
The foregoing shows and describes the general principles, essential features, and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are described in the specification and illustrated only to illustrate the principle of the present invention, but that various changes and modifications may be made therein without departing from the spirit and scope of the present invention, which fall within the scope of the invention as claimed. The scope of the invention is defined by the appended claims and equivalents thereof.
Claims (7)
1. A body construction method for an intelligent networking automobile simulation test scene is characterized by comprising the following steps:
step 1: firstly, designing a typical and representative basic test scene which can be applied to an intelligent networking automobile simulation test, determining a function to be tested, and then expanding the basic test scene through the increase of scene elements and the adjustment of input parameters;
and 2, step: analyzing, sorting and classifying the scene information, and conceptualizing the scene information by using an ontology; classifying information contained in the scene;
and 3, step 3: taking the concept of a scene as a root node, and constructing a traffic scene ontology model from top to bottom based on the structure and the hierarchical description of the scene;
and 4, step 4: using OWL2 to complete the construction of a scene concept ontology, completing semantic expression, and using OWL2 to establish an ontology knowledge base;
and 5: a semantic network rule language SWRL is introduced to enhance the rule description capability, and the SWRL can directly refer to concepts, attributes and examples in a local knowledge base to carry out semantic description on traffic rules to obtain constraints;
step 6: and checking and evaluating the constructed traffic scene body model, and correcting and optimizing according to the result.
2. The ontology construction method for the intelligent networked automobile simulation test scenario according to claim 1, wherein the step 1 specifically comprises: firstly, designing a basic test scene according to an expected verified automatic driving function, wherein the scene has typicality and representativeness; starting from functional requirements, designing a typical and representative basic test scene; then expanding the basic test scene through the increase of scene elements and the adjustment of input parameters to combine a plurality of test cases; and modeling and analyzing the accident scene, selecting key scene elements as the input number of the test scene, combining different scene elements, and deriving a plurality of test cases from one accident scene.
3. The ontology construction method for the intelligent networked automobile simulation test scenario according to claim 1, wherein the step 2 specifically comprises the following substeps:
s201, analyzing a scene structure, and obtaining a total composition structure of a traffic scene by analyzing specific definitions of the traffic scene;
s202, according to the definition and structure of scene terms, the traffic scene is decomposed into four layers, wherein the first layer is a road network, the second layer is a static object, the third layer is a dynamic object, and the fourth layer is an environment.
4. The ontology construction method for the intelligent networked automobile simulation test scenario according to claim 1, wherein the step 3 specifically comprises the following substeps:
s301, analyzing the structure of the scene, uniformly defining terms of the scene, standardizing the meanings of the terms, distinguishing scene concepts with different meanings, and providing a consistent expression for the construction of a later scene model;
s302, constructing a traffic scene ontology model from top to bottom by using a scene as a root concept based on the structure and the hierarchical description of the scene and adopting an ontology method;
and S303, performing scene semantic analysis, creating an example according to the traffic scene ontology model, extracting partial concepts in the macroscopic traffic scene ontology model, and defining attributes and relations required by the scene.
5. The ontology construction method for the intelligent networked automobile simulation test scenario according to claim 1, wherein the step 4 specifically comprises the following substeps:
s401, analyzing the relation of 'human-vehicle-environment' by the traffic scene ontology model through OWL2, and performing formal description on semantic relation and attributes in a scene entity;
s402, establishing an ontology knowledge base by using OWL2 according to the semantic relation and the attributes in the formally described scene entities.
6. The ontology construction method for the intelligent networked automobile simulation test scenario according to claim 1, wherein the step 5 specifically comprises: based on concepts, attributes and examples of an ontology knowledge base established by OWL2, a semantic network rule language SWRL is introduced to describe the semantics of traffic rules, and the traffic rules and implicit driving orders are established to obtain the constraints of a traffic scene ontology model.
7. The ontology construction method for the intelligent networked automobile simulation test scenario according to claim 1, wherein the step 6 specifically comprises: carrying out consistency and integrity inspection on the traffic scene ontology model; integrity check checks the integrity of the examples; the consistency check is to verify the consistency of the relation and check whether the relation definition is logically wrong; and correcting and optimizing the traffic scene body model according to the inspection result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211241729.9A CN115695473A (en) | 2022-10-11 | 2022-10-11 | Ontology construction method for intelligent networking automobile simulation test scene |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211241729.9A CN115695473A (en) | 2022-10-11 | 2022-10-11 | Ontology construction method for intelligent networking automobile simulation test scene |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115695473A true CN115695473A (en) | 2023-02-03 |
Family
ID=85064274
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211241729.9A Pending CN115695473A (en) | 2022-10-11 | 2022-10-11 | Ontology construction method for intelligent networking automobile simulation test scene |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115695473A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116597690A (en) * | 2023-07-18 | 2023-08-15 | 山东高速信息集团有限公司 | Highway test scene generation method, equipment and medium for intelligent network-connected automobile |
-
2022
- 2022-10-11 CN CN202211241729.9A patent/CN115695473A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116597690A (en) * | 2023-07-18 | 2023-08-15 | 山东高速信息集团有限公司 | Highway test scene generation method, equipment and medium for intelligent network-connected automobile |
CN116597690B (en) * | 2023-07-18 | 2023-10-20 | 山东高速信息集团有限公司 | Highway test scene generation method, equipment and medium for intelligent network-connected automobile |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110597711B (en) | Automatic driving test case generation method based on scene and task | |
Scholtes et al. | 6-layer model for a structured description and categorization of urban traffic and environment | |
Algers et al. | Review of micro-simulation models | |
Buechel et al. | Ontology-based traffic scene modeling, traffic regulations dependent situational awareness and decision-making for automated vehicles | |
CN101359434A (en) | Process for constructing road net model for traffic navigation system and traffic emulation system | |
CN112560277B (en) | Automobile automatic driving scene modeling method based on domain specific modeling language ADSML | |
CN114354219A (en) | Test method and device for automatic driving vehicle | |
Labetski et al. | A proposal for an improved transportation model in CityGML | |
CN113076638A (en) | BIM and VISSIM microscopic simulation-based bidirectional traffic diversion simulation method | |
CN115695473A (en) | Ontology construction method for intelligent networking automobile simulation test scene | |
Park et al. | Development and evaluation of a calibration and validation procedure for microscopic simulation models. | |
Gangel et al. | Modelling the zalazone proving ground: a benchmark of state-of-the-art automotive simulators prescan, ipg carmaker, and vtd vires | |
Chen et al. | Generating autonomous driving test scenarios based on openscenario | |
CN116597690B (en) | Highway test scene generation method, equipment and medium for intelligent network-connected automobile | |
Hou et al. | Twin scenarios establishment for autonomous vehicle digital twin empowered sotif assessment | |
EP3920128A1 (en) | Operational design domains in autonomous driving | |
Medrano-Berumen et al. | Development of a validation regime for an autonomous campus shuttle | |
Favreau et al. | What are intersections for pedestrian users? | |
Li | A scenario-based development framework for autonomous driving | |
Zhang et al. | An ODD-Based Scalable Assurance Framework for Automated Driving Systems | |
Che et al. | An open vehicle-in-the-loop test method for autonomous vehicle | |
Kanakagiri | Development of a virtual simulation environment for autonomous driving using digital twins | |
Desta et al. | Impacts of autonomous vehicle driving logics on heterogenous traffic and evaluating transport interventions with microsimulation experiments | |
Liu et al. | Baseline microscopic and macroscopic models: Deliverable D4. 1 of the CoEXist project | |
Chen | Formal Modeling and Automatic Generation of Test Cases for the Autonomous Vehicle |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |