1) The document describes an ontology-based knowledge base that is used for advanced driver assistance systems.
2) Ontologies were designed for concepts like maps, vehicle control, and cars to represent the driving environment.
3) An intelligent speed adaptation system and intelligent decision making system were developed that use the knowledge base to detect overspeed situations and make decisions at intersections.
4) An experiment evaluated the systems using real sensor data and found that inferences were made within milliseconds.
1. ONTOLOGIES FOR ADVANCED
DRIVER ASSISTANCE
SYSTEMS
Presentation by Lihua ZhaoSWO2015
Lihua Zhao, Toyota Technological Institute
Ryutaro Ichise, National Institute of
Informatics
Seiichi Mita, Toyota Technological Institute
Yutaka Sasaki, Toyota Technological Institute
SIG-SWO-035-03
2. Outline
Motivation
Related Work
Ontology-Based Knowledge Base
Advanced Driver Assistance ADAS Systems
(ADAS)
Experiment
Conclusion & Future Work
2
3. Advanced Driver Assistance Systems (ADAS)
Perceive driving environment by processing sensor data.
Make driving decisions in different traffic situations.
Machine Understandable Ontology-based Knowledge Base
Advanced Digital Map
Road information, speed limits, etc.
Traffic Regulations
Right-of-Way Rules
Motivation
3
4. Automation level ontology and situation assessment ontology are
designed for co-driving. [Pollard, 2013]
Use ontology and 14 SWRL rules to enable the vehicle to understand the
context information when it approaches road intersections. [Armand, 2014]
A complex intersection ontology (car, crossing, road connection, and sign
at crossing) is introduced for fast reasoning. [Hulsen, 2011]
An ontology-based traffic model that can represent typical traffic
scenarios such as intersections, multi-lane roads, opposing traffic, and bi-
directional lanes is introduced. [Regele,2008]
Related Work
4
6. Ontology: Machine-understandable knowledge representation
Classes: called as Concepts, defined by owl:Class.
Properties: owl:ObjectProperty and owl:DatatypeProperty.
Instances: individuals of a domain, defined by owl:Thing.
Rules: describe logical inferences, with if-then sentence.
Ontology Editor
Protégé ontology editor
Ontologies
6
8. Describe the path of autonomous cars. (34 Classes)
ObjectProperty (15)
control:nextPathSegment
(intersection or lane)
control:giveWay
control:collisionWarningWith
control:approachTo
DataProperty (2)
control:pathSegmentID
control:nodePos
Control Ontology
8
9. Concepts of vehicles and devices such as sensors.
(33 Classes)
ObjectProperty (3)
car:usedSensor
car:isRunningOn
car:currentPath
DataProperty (15)
car:car_length
car:car_ID
car:velocity
Car Ontology
9
10. Instances are also known as individuals that model
abstract or concrete objects based on the ontologies.
Tempaku Map Instance
Path Instance
Car Instance
Instances
10
12. Constructed based on the Tempaku map and control ontology.
Path: E -> A -> G
Path Instance
12
13. Describe a car and devices installed on
the car.
Car Instance
13
14. Semantic Web Rule Language (SWRL) is used to express rules.
Pellet reasoner is used for ontology reasoning.
SWRL Rules
14
At an intersection, the
car turning right should
give way to the other
car which is going
straight.
Identify driving direction.
15. Retrieve the next path segment based on current path
segment. (pathSegmentID: 0, 1, 2, …, n)
SPARQL Query I
15
16. Retrieve the speed limit of current path segment.
SPARQL Query II
16
17. If a car’s average velocity in the past 500ms exceeds its
own speed limit. (i.e. maxSpeed:120km/h)
RANGE: duration to receive sensor stream data
STEP: frequency of a sensor receiver.
C-SPARQL Query
17
18. Intelligent Speed Adaptation (ISA) System
Detect overspeed situations.
Intelligent Decision Making System
Make driving decisions at uncontrolled
intersections.
ADAS Systems
18
19. Input
Sensor Data
GPS-IMU sensor
Knowledge Base
Ontology-based data
Output
Overspeed warning
Intelligent Speed Adaptation
System
19
20. Intelligent Decision Making
System
20
1. Send sensor data to SPARQL Query
Engine & SWRL Rule Reasoner.
2. Retrieve current lane, next lane, and
driving direction, etc.
3. SWRL rule reasoner adds some
additional information such as
collision warning and the other vehicle's
position, velocity, and driving direction .
21. Intelligent Decision Making
System
21
4. Ontology reasoning on the updated
Knowledge Base.
5. The SPARQL query engine retrieves
the commands and the vehicles that
our vehicle should give way to.
6. The decision signals are sent to the
path planning system to update driving
path or driving behavior.
7. Newly added inferred knowledge is
removed from the ontology-based
Knowledge Base.
22. Data Format
Evaluation of ISA System
Evaluation of Decision Making System
Experiment
22
26. Ontology-Based Knowledge Base
Advanced Driver Assistance Systems (ADAS)
Intelligent Speed Adaptation System
Intelligent Decision Making System
Experiment with real sensor data.
Conclusion
26
27. Speed up execution time
Use part of Knowledge Base for reasoning.
Add more rules to cover other situations
Driving on a corner or on private roads.
Future Work
27