robot navigation based on the mapping of coarse qualitative route descriptions to route graphs
DESCRIPTION
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs. Motivation. Outline. Wheelchair Rolland as mobile robotic platform. Need for appropriate HRI is essential, especially for a wheelchair-bound person. - PowerPoint PPT PresentationTRANSCRIPT
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Motivation
Outline
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
• Wheelchair Rolland as mobile robotic platform.
• Need for appropriate HRI is essential, especially for a wheelchair-bound person.
• Application scenario in an office-like environment allows user to command his vehicle by natural language input.
Conclusion
1
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Motivation
Outline
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Route Graph
Local path planner
Formalized Coarse Route Descriptions
MMC global localizer
Conclusion
Natural LanguageCoarse Route Descriptions
localizes within m
appe
d on
to
translated to
target sequence
navigates within
2
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Motivation
Outline
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
• Preprocessed bitmap from CAD-blueprint.
• Voronoi-Graph based RouteGraph-layer represents navigable space.
• Semantic RouteGraph-layer stores nodes for rooms and regions,
• as well as for annotated landmarks.
Conclusion
3
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Motivation
Outline
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Conclusion
• Geometric path planner models obstacle-free paths by means of cubic Bezier curves.
• Benefits versus DWA and behavioural approaches:
- explicit modelling of necessary haul-off movements
- small search space due to 2 free parameters
- consideration of heading in goal-pose
4
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Motivation
Outline
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Turn around, drive through the interaction lab, turn left, go to the kitchen and stop at the first junction to the right.
Introductory example:
Conclusion
5
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Motivation
Outline
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Introductory example:
Turn around, drive through the interaction lab, turn left, go to the kitchen and stop at the first junction to the right.
Conclusion
5
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Motivation
Outline
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Introductory example:
Turn around, drive through the interaction lab, turn left, go to the kitchen and stop at the first junction to the right.
Conclusion
5
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Motivation
Outline
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Introductory example:
Turn around, drive through the interaction lab, turn left, go to the kitchen and stop at the first junction to the right.
5
Conclusion
<CoarseRouteDesc.> ::= { <CoarseRouteSegmentDesc.> }<CoarseRouteSegmentDesc.> ::= ( <ControllerOnRoutesegment>,
<RouterOnRouteSegment>, <ActionAtEndOfRouteSegment> )
<ControllerOnRouteSegment> ::= { ( [Preposition] <Attributed Place> ) | <DistancePredicate> }<RouterOnRouteSegment> ::= { [Preposition] <AttributedPlace> }<ActionAtEndOfRouteSegment> ::= <TurnAction> | „Stop“<Preposition> ::= „Through“, „OutOf“, „Along“,
„NaturalLanguageDirection“, „After“, „Past“, „Between“, „Towards“, ...
<AttributedPlace> ::= ( <Place>, <AttributeType>, <AttributeValue> )
<DistancePredicate> ::= ( <NumericalValue>, <LengthUnit> | <TimeUnit> )<TurnAction> ::= „NaturalLanguageDirection“
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Motivation
Outline
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Conclusion
Turn around, drive through the interaction lab, turn left, go to the kitchen and stop at the first junction to the right.
6
Turn around, drive through the interaction lab, turn left,
go to the kitchen and stop at the first junction to the right.
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Motivation
Outline
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Conclusion
<CoarseRouteDesc.> ::= { <CoarseRouteSegmentDesc.> }<CoarseRouteSegmentDesc.> ::= ( <ControllerOnRoutesegment>,
<RouterOnRouteSegment>, <ActionAtEndOfRouteSegment> )
<ControllerOnRouteSegment> ::= { ( [Preposition] <Attributed Place> ) | <DistancePredicate> }<RouterOnRouteSegment> ::= { [Preposition] <AttributedPlace> }<ActionAtEndOfRouteSegment> ::= <TurnAction> | „Stop“<Preposition> ::= „Through“, „OutOf“, „Along“,
„NaturalLanguageDirection“, „After“, „Past“, „Between“, „Towards“, ...
<AttributedPlace> ::= ( <Place>, <AttributeType>, <AttributeValue> )
<DistancePredicate> ::= ( <NumericalValue>, <LengthUnit> | <TimeUnit> )<TurnAction> ::= „NaturalLanguageDirection“
6
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Conclusion
Motivation
Outline
<CoarseRouteDesc.> ::= { <CoarseRouteSegmentDesc.> }<CoarseRouteSegmentDesc.> ::= ( <ControllerOnRoutesegment>,
<RouterOnRouteSegment>, <ActionAtEndOfRouteSegment> )
<ControllerOnRouteSegment> ::= { ( [Preposition] <Attributed Place> ) | <DistancePredicate> }<RouterOnRouteSegment> ::= { [Preposition] <AttributedPlace> }<ActionAtEndOfRouteSegment> ::= <TurnAction> | „Stop“<Preposition> ::= „Through“, „OutOf“, „Along“,
„NaturalLanguageDirection“, „After“, „Past“, „Between“, „Towards“, ...
<AttributedPlace> ::= ( <Place>, <AttributeType>, <AttributeValue> )
<DistancePredicate> ::= ( <NumericalValue>, <LengthUnit> | <TimeUnit> )<TurnAction> ::= „NaturalLanguageDirection“
Turn around, drive through the interaction lab, turn left,
go to the kitchen and stop at the first junction to the right.6
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Conclusion
Motivation
Outline
Turn around, drive through the interaction lab, turn left,
go to the kitchen and stop at the first junction to the right.
<CoarseRouteDesc.> ::= { <CoarseRouteSegmentDesc.> }<CoarseRouteSegmentDesc.> ::= ( <ControllerOnRoutesegment>,
<RouterOnRouteSegment>, <ActionAtEndOfRouteSegment> )
<ControllerOnRouteSegment> ::= { ( [Preposition] <Attributed Place> ) | <DistancePredicate> }<RouterOnRouteSegment> ::= { [Preposition] <AttributedPlace> }<ActionAtEndOfRouteSegment> ::= <TurnAction> | „Stop“<Preposition> ::= „Through“, „OutOf“, „Along“,
„NaturalLanguageDirection“, „After“, „Past“, „Between“, „Towards“, ...
<AttributedPlace> ::= ( <Place>, <AttributeType>, <AttributeValue> )
<DistancePredicate> ::= ( <NumericalValue>, <LengthUnit> | <TimeUnit> )<TurnAction> ::= „NaturalLanguageDirection“
6
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
<CoarseRouteDesc.> ::= { <CoarseRouteSegmentDesc.> }<CoarseRouteSegmentDesc.> ::= ( <ControllerOnRoutesegment>,
<RouterOnRouteSegment>, <ActionAtEndOfRouteSegment> )
<ControllerOnRouteSegment> ::= { ( [Preposition] <Attributed Place> ) | <DistancePredicate> }<RouterOnRouteSegment> ::= { [Preposition] <AttributedPlace> }<ActionAtEndOfRouteSegment> ::= <TurnAction> | „Stop“<Preposition> ::= „Through“, „OutOf“, „Along“,
„NaturalLanguageDirection“, „After“, „Past“, „Between“, „Towards“, ...
<AttributedPlace> ::= ( <Place>, <AttributeType>, <AttributeValue> )
<DistancePredicate> ::= ( <NumericalValue>, <LengthUnit> | <TimeUnit> )<TurnAction> ::= „NaturalLanguageDirection“
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Conclusion
Motivation
Outline
Turn around, drive through the interaction lab, turn left,
go to the kitchen and stop at the first junction to the right.6
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Conclusion
Motivation
Outline
<CoarseRouteDesc.> ::= { <CoarseRouteSegmentDesc.> }<CoarseRouteSegmentDesc.> ::= ( <ControllerOnRoutesegment>,
<RouterOnRouteSegment>, <ActionAtEndOfRouteSegment> )
<ControllerOnRouteSegment> ::= { ( [Preposition] <Attributed Place> ) | <DistancePredicate> }<RouterOnRouteSegment> ::= { [Preposition] <AttributedPlace> }<ActionAtEndOfRouteSegment> ::= <TurnAction> | „Stop“<Preposition> ::= „Through“, „OutOf“, „Along“,
„NaturalLanguageDirection“, „After“, „Past“, „Between“, „Towards“, ...
<AttributedPlace> ::= ( <Place>, <AttributeType>, <AttributeValue> )
<DistancePredicate> ::= ( <NumericalValue>, <LengthUnit> | <TimeUnit> )<TurnAction> ::= „NaturalLanguageDirection“
Turn around, drive through the interaction lab, turn left,
go to the kitchen and stop at the first junction to the right.6
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
<CoarseRouteDesc.> ::= { <CoarseRouteSegmentDesc.> }<CoarseRouteSegmentDesc.> ::= ( <ControllerOnRoutesegment>,
<RouterOnRouteSegment>, <ActionAtEndOfRouteSegment> )
<ControllerOnRouteSegment> ::= { ( [Preposition] <Attributed Place> ) | <DistancePredicate> }<RouterOnRouteSegment> ::= { [Preposition] <AttributedPlace> }<ActionAtEndOfRouteSegment> ::= <TurnAction> | „Stop“<Preposition> ::= „Through“, „OutOf“, „Along“,
„NaturalLanguageDirection“, „After“, „Past“, „Between“, „Towards“, ...
<AttributedPlace> ::= ( <Place>, <AttributeType>, <AttributeValue> )
<DistancePredicate> ::= ( <NumericalValue>, <LengthUnit> | <TimeUnit> )<TurnAction> ::= „NaturalLanguageDirection“
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Conclusion
Motivation
Outline
Turn around, drive through the interaction lab, turn left,
go to the kitchen and stop at the first junction to the right.6
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Conclusion
Motivation
Outline
<CoarseRouteDesc.> ::= { <CoarseRouteSegmentDesc.> }<CoarseRouteSegmentDesc.> ::= ( <ControllerOnRoutesegment>,
<RouterOnRouteSegment>, <ActionAtEndOfRouteSegment> )
<ControllerOnRouteSegment> ::= { ( [Preposition] <Attributed Place> ) | <DistancePredicate> }<RouterOnRouteSegment> ::= { [Preposition] <AttributedPlace> }<ActionAtEndOfRouteSegment> ::= <TurnAction> | „Stop“<Preposition> ::= „Through“, „OutOf“, „Along“,
„NaturalLanguageDirection“, „After“, „Past“, „Between“, „Towards“, ...
<AttributedPlace> ::= ( <Place>, <AttributeType>, <AttributeValue> )
<DistancePredicate> ::= ( <NumericalValue>, <LengthUnit> | <TimeUnit> )<TurnAction> ::= „NaturalLanguageDirection“
Turn around, drive through the interaction lab, turn left,
go to the kitchen and stop at the first junction to the right.6
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Motivation
Outline
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Conclusion
Common spatial relations: example 1
Along 2-valued
7
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Motivation
Outline
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Conclusion
Common spatial relations: example 3
Front-Left 2-valued
8
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Motivation
Outline
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Conclusion
Common spatial relations: further directions
…9
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Motivation
Outline
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Turn around, drive through the interaction lab, turn left, go to the kitchen and stop at the first junction to the right.
Conclusion
10
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Motivation
Outline
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Turn around, drive through the interaction lab, turn left, go to the kitchen and stop at the first junction to the right.
Conclusion
10
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Motivation
Outline
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Turn around, drive through the interaction lab, turn left, go to the kitchen and stop at the first junction to the right.
Conclusion
10
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Motivation
Outline
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Turn around, drive through the interaction lab, turn left, go to the kitchen and stop at the first junction to the right.
Conclusion
10
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Motivation
Outline
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Turn around, drive through the interaction lab, turn left, go to the kitchen and stop at the first junction to the right.
Conclusion
10
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Motivation
Outline
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Turn around, drive through the interaction lab, turn left, go to the kitchen and stop at the first junction to the right.
Conclusion
10
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Motivation
Outline
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Turn around, drive through the interaction lab, turn left, go to the kitchen and stop at the first junction to the right.
5,...,0
55.0pp
Conclusion
Ø 36.0
10
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Motivation
Outline
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Conclusion
11
• For an illustrative video of experimental results visit:
http://www.informatik.uni-bremen.de/rolland/videos/Interpretation_of_Coarse_Route_Description_08_06_06.mpg
Robot Navigation based on the Mapping of Coarse Qualitative Route Descriptions to Route Graphs
Motivation
Outline
Global World Knowledge
Local Path Planning
Coarse Route Descriptions
• Formalization
• Interpretation
• Mapping to Route Graphs
Experimental Results
Conclusion
• Interpretation of coarse route descriptions facilitated by their mapping onto multi-layered Route Graphs.
• Key-concept is the fuzzy interpretation of common spatial relations.
To do:
• Extension of the set of supported spatial relations.
• Benchmarking of the system against available corpora.
12