Homotopy-Aware RRT* : Toward Human-Robot Topological Path-Planning - - PowerPoint PPT Presentation

homotopy aware rrt toward human robot topological path
SMART_READER_LITE
LIVE PREVIEW

Homotopy-Aware RRT* : Toward Human-Robot Topological Path-Planning - - PowerPoint PPT Presentation

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion Homotopy-Aware RRT* : Toward Human-Robot Topological Path-Planning Daqing Yi Michael A. Goodrich Kevin D. Seppi Computer Science


slide-1
SLIDE 1

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Homotopy-Aware RRT* : Toward Human-Robot Topological Path-Planning

Daqing Yi Michael A. Goodrich Kevin D. Seppi

Computer Science Department Brigham Young University

slide-2
SLIDE 2

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Outline

1

Introduction Human-Robot Path-Planning Homotopy

2

Related Work

3

Homotopy Identification Map Decomposition Homotopic Equivalence

4

Homotopy-Aware RRT*

5

Experiments and Application Experiments Single Homotopy Class Multiple Homotopy Classes Application

6

Conclusion

slide-3
SLIDE 3

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Outline

1

Introduction Human-Robot Path-Planning Homotopy

2

Related Work

3

Homotopy Identification Map Decomposition Homotopic Equivalence

4

Homotopy-Aware RRT*

5

Experiments and Application Experiments Single Homotopy Class Multiple Homotopy Classes Application

6

Conclusion

slide-4
SLIDE 4

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Human-Robot Collaboration

Instruct a robot executing a task

slide-5
SLIDE 5

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Collaborative Path-Planning

T

  • pology space

Metric space

slide-6
SLIDE 6

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Homotopy

Definition A path can can be continuously deformed into another without encroaching any obstacle σ1 ≃ σ2 σ1 ≃ σ2 Homotopy Class A set of paths that are homotopic

slide-7
SLIDE 7

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Homotopy-based Path Planning

Find the optimal paths of several homotopy classes σ(0) = xinit σ(1) = xgoal H ⊆ ❍(xinit, xgoal) ∀hi ∈ H, σ∗

hi =

arg maxσ∈Xfree∧h(σ)=hi Cost(σ)

slide-8
SLIDE 8

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Framework

Map

Topology Constraint Paths

Homotopic DFA

Strings

TrimREP

Homotopy Classes

HARRT*

Optimal Paths Homotopy Identification

Map

slide-9
SLIDE 9

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Outline

1

Introduction Human-Robot Path-Planning Homotopy

2

Related Work

3

Homotopy Identification Map Decomposition Homotopic Equivalence

4

Homotopy-Aware RRT*

5

Experiments and Application Experiments Single Homotopy Class Multiple Homotopy Classes Application

6

Conclusion

slide-10
SLIDE 10

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Spatial Semantic Hierarchy

Kuipers, “The spatial semantic hierarchy,” Tech. Rep. AI99-281, 29, 1999.

slide-11
SLIDE 11

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Approaches of Homotopic Path-Planning

(a) Space-cut approaches 1, 2 (b) Homology approaches 3, 4 (c) Reference-frame approaches 5, 6

(a) (b) (c) 1Hershberger et al., “Computing minimum length paths of a given homotopy class,” Computational Geometry, 1994. 2Grigoriev et al., “Polytime algorithm for the shortest path in a homotopy class amidst semi-algebraic

  • bstacles in the plane,” ISSAC 98.

3Bhattacharya, “Search-based path planning with homotopy class constraints,” AAAI 2010. 4Florian et al., “Multiscale topological trajectory classification with persistent homology,” RSS 2014. 5Hernandez et al., “A topologically guided path planner for an AUV using homotopy classes,” ICRA 2011. 6Kim et al., “Path planning for a tethered mobile robot,” ICRA 2014.

slide-12
SLIDE 12

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Outline

1

Introduction Human-Robot Path-Planning Homotopy

2

Related Work

3

Homotopy Identification Map Decomposition Homotopic Equivalence

4

Homotopy-Aware RRT*

5

Experiments and Application Experiments Single Homotopy Class Multiple Homotopy Classes Application

6

Conclusion

slide-13
SLIDE 13

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Map Decomposition

S1-1 S1-0 S2-0 S3-0 S3-1 S4-0 Sampling representative point in each obstacle Sampling the center point Generating a radial structure Reference frames separate regions An ID for a reference frame A string for a sequence of reference frames

slide-14
SLIDE 14

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Topology of DFA

S1-1 S1-0 S2-0 S3-0 S3-1 S4-0

S1-1 S2-0 S3-1 S3-0 S1-0 S4-0

Region ⇒ Node Reference frame ⇒ Edge Deterministic Finite Automata Path ⇒ String String block A set of paths that generates the same string by DFA

slide-15
SLIDE 15

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Homotopic DFA

S1-1 S2-0 S3-1 S3-0 S1-0 S4-0

S1-1 S2-0 S3-1 S3-0 S1-0 S4-0

Homotopic equivalence in crossing the center point Merge regions that connect with the center point

slide-16
SLIDE 16

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Hierarchy of Path Partition

slide-17
SLIDE 17

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

REP String

Definition A string made up of recursive embedded palindromic substrings Example

Before deform After deform

A path crosses several reference frames and backtrack Example:

ε ww R uvv Rww RuR

REPTrim()

trim REP string

slide-18
SLIDE 18

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Homotopic Equivalence Identification

Theorem Removing all the REP substrings of ▼h(σ) yields the v∗

i for which

Γ▼h(σ) ≃ Γv∗

i .

Corollary REPTrim(▼h(σi)) = REPTrim(▼h(σj)) iff σi ≃ σj.

slide-19
SLIDE 19

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Outline

1

Introduction Human-Robot Path-Planning Homotopy

2

Related Work

3

Homotopy Identification Map Decomposition Homotopic Equivalence

4

Homotopy-Aware RRT*

5

Experiments and Application Experiments Single Homotopy Class Multiple Homotopy Classes Application

6

Conclusion

slide-20
SLIDE 20

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Homotopy-Aware RRT*

Homotopic identification of each branch Bidirectional RRT*

goal

Reference Frame

Homotopic identification start goal Bidirectional RRT*

slide-21
SLIDE 21

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Flow

Sample Add Vertex Add Vertex Rewire Vertices Rewire Vertices Iteration Sample Paths of Homotopy Classes Find a Path Find a Path Start Tree Goal Tree

slide-22
SLIDE 22

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Outline

1

Introduction Human-Robot Path-Planning Homotopy

2

Related Work

3

Homotopy Identification Map Decomposition Homotopic Equivalence

4

Homotopy-Aware RRT*

5

Experiments and Application Experiments Single Homotopy Class Multiple Homotopy Classes Application

6

Conclusion

slide-23
SLIDE 23

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Human-Robot Interactive Path-Planning

Quickly go from point A to point B through a sequence of specific regions. 1 2 3 4 Quickly go from point A to point B making sure to visit some regions and avoid other regions. In quickly going from point A to point B, I prefer some types of paths over

  • thers, but I recognize that tradeoffs may be needed.
slide-24
SLIDE 24

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Single Homotopy Class

slide-25
SLIDE 25

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Multiple Homotopy Classes

Unidirectional Tree Bidirectional Tree

slide-26
SLIDE 26

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Multiple Homotopy Classes

slide-27
SLIDE 27

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Interactive Graphical Interface

Sketch reference Planned path

slide-28
SLIDE 28

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Nature Language Interface

“Go between box 8 and box 6, and avoid the left of box 1”.

slide-29
SLIDE 29

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Outline

1

Introduction Human-Robot Path-Planning Homotopy

2

Related Work

3

Homotopy Identification Map Decomposition Homotopic Equivalence

4

Homotopy-Aware RRT*

5

Experiments and Application Experiments Single Homotopy Class Multiple Homotopy Classes Application

6

Conclusion

slide-30
SLIDE 30

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Conclusion and Futurework

Conclusion Encode path into string Homotopic identification Find optimal paths of homotopy classes Futurework Integrate with multiple performance objectives Integrate with moving objects

slide-31
SLIDE 31

Introduction Related Work Homotopy Identification Homotopy-Aware RRT* Experiments and Application Conclusion

Thank you