![Page 1: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/1.jpg)
Planning Under Topological Constraints Using Beam-Graphs
Venkatraman Narayanan, Paul Vernaza, Maxim Likhachev and Steven M. LaValle
Robotics Institute, Carnegie Mellon UniversityDepartment of Computer Science, University of Illinois, Urbana
Topological Constraints
Related Work
Beam-Graph Construction
Formalizing Topological Constraints
Experiments
![Page 2: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/2.jpg)
Topological Constraints
“Find a circuit path for an UAV such that it obtains 360° views of certain regions of interest, in a particular order”
![Page 3: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/3.jpg)
Topological Constraints
“Find the optimal path from start to goal such that it goes above region A and below region B”
Start
Goal
A
B
![Page 4: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/4.jpg)
Topological ConstraintsFind the optimal loop that encloses the regions of interest in some specified order
Find the optimal loop that encloses the regions of interest in any order
![Page 5: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/5.jpg)
Topological ConstraintsFind the optimal loop that obtains 360° views of the regions of interest in some specified order
Find the optimal loop that obtains 360° views of the regions of interest in any order
![Page 6: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/6.jpg)
Related Work
• S. Bhattacharya, M. Likhachev, and V. Kumar. Topological constraints in search-based robot path planning. Autonomous Robots, 2012• P. Vernaza, V. Narayanan, and M. Likhachev. Efficiently finding optimal winding-constrained loops in the plane. In Proceedings of Robotics: Science and Systems, 2012• H. Gong, J. Sim, M. Likhachev, and J. Shi. Multihypothesis motion planning for visual object tracking. In Thirteenth International Conference on Computer Vision, 2011
Methods inspired by complex analysis, electromagnetic theory and flow theory:
![Page 7: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/7.jpg)
Related Work
• S. Bhattacharya, M. Likhachev, and V. Kumar. Topological constraints in search-based robot path planning. Autonomous Robots, 2012• P. Vernaza, V. Narayanan, and M. Likhachev. Efficiently finding optimal winding-constrained loops in the plane. In Proceedings of Robotics: Science and Systems, 2012• H. Gong, J. Sim, M. Likhachev, and J. Shi. Multihypothesis motion planning for visual object tracking. In Thirteenth International Conference on Computer Vision, 2011
Designed for “homology” constraints. Cannot handle general topology constraints
Mathematically and numerically intensive
Methods inspired by complex analysis, electromagnetic theory and flow theory:
![Page 8: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/8.jpg)
Related Work
• S. Cabello, Y. Liu, A. Mantler, and J. Snoeyink. Testing homotopy for paths in the plane. In Proceedings of the eighteenth annual symposium on Computational Geometry, 2002• A. Efrat, S. Kobourov, and A. Lubiw. Computing homotopic shortest paths efficiently. Computational Geometry
Computational geometry approaches:
![Page 9: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/9.jpg)
Related Work
• S. Cabello, Y. Liu, A. Mantler, and J. Snoeyink. Testing homotopy for paths in the plane. In Proceedings of the eighteenth annual symposium on Computational Geometry, 2002• A. Efrat, S. Kobourov, and A. Lubiw. Computing homotopic shortest paths efficiently. Computational Geometry
Not easily applicable to robotics--difficult to incorporate kinodynamic constraints
Cannot handle arbitrary cost functions
Computational geometry approaches:
![Page 10: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/10.jpg)
Related Work
B. Tovar, F. Cohen, and S. M. LaValle. Sensor beams, obstacles, and possible paths. In G. Chirikjian, H. Choset, M. Morales, and T. Murphey, editors, Algorithmic Foundations of Robotics, VIII
Idea of ‘virtual beams’:
“Virtual beams” construction - captures topological information, intuitive
Our work shows how virtual beams can be integrated into existing search-based motion planning frameworks for robotics
![Page 11: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/11.jpg)
Beam-Graph Construction
• Sequences of beam-crossings capture ‘topological’ structure
• Hallucinate ‘virtual’ beams to differentiate topologically different paths
Path 1: ab
Path 2: aa’ab
Path 3: cc
![Page 12: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/12.jpg)
Beam-Graph Construction
• For every region of interest, construct a beam starting at the lowest point on the ROI, and terminating at another ROI, or the boundary
• Beams are directed
Beam crossing: aBeam crossing: c’
![Page 13: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/13.jpg)
Beam-Graph Construction
A state in the graph: s - [x(s) b(s)]
Spatial configuration of the robot
(e.g: x(s) = (x y θ))
History of beam crossings
(e.g: b(s) = aabc’de’)
![Page 14: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/14.jpg)
Beam-Graph Construction
A state in the graph: s - [x(s) b(s)]
0 1 2 3 4 50
1
23
4
5
(2, 4, 𝛆)b
a
![Page 15: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/15.jpg)
Beam-Graph Construction
A state in the graph: s - [x(s) b(s)]
0 1 2 3 4 50
1
23
4
5
(2, 4, a)b
a
![Page 16: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/16.jpg)
Beam-Graph Construction
A state in the graph: s - [x(s) b(s)]
0 1 2 3 4 50
1
23
4
5
(2, 4, ab)b
a
![Page 17: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/17.jpg)
Beam-Graph Construction
A state in the graph: s - [x(s) b(s)]
0 1 2 3 4 50
1
23
4
5
(2, 4, aa)b
a
![Page 18: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/18.jpg)
Beam-Graph Construction
A state s’ is a successor of s iff. :
• There exists a kinodynamically feasible motion primitive between s and s’
• b(s’) is a concatenation of b(s) and the string of beam crossings formed by the motion primitive, i.e, b(s’) = b(s) + w(s,m)
Pivtoraiko, M., Kelly, A., “Generating Near Minimal Spanning Control Sets for Constrained Motion Planning in Discrete State Spaces”,
Proceedings IROS 2005, Edmonton Alberta, Canada, August 2005
![Page 19: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/19.jpg)
Beam-Graph Construction
A state s’ is a successor of s iff. :
• There exists a kinodynamically feasible motion primitive between s and s’
• b(s’) is a concatenation of b(s) and the string of beam crossings formed by the motion primitive, i.e, b(s’) = b(s) + w(s,m)
0 1 2 3 4 50
1
23
4
5b
a
(3,1, 𝛆)
(4,1,b)(3,1,a)
(4,1,ab)
(3,1,b)
(4,1,bb)
![Page 20: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/20.jpg)
Formalizing Topological Constraints
For a topological space X, the set of all loops based at some x0 ∈ X, is the first homotopy group or the fundamental group
![Page 21: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/21.jpg)
Formalizing Topological Constraints
The multi-punctured plane
![Page 22: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/22.jpg)
Formalizing Topological Constraints
The multi-punctured plane
For the multi-punctured plane, the fundamental group is the free group on n letters, Fn
x0
𝛔1𝛔2
𝛔3
![Page 23: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/23.jpg)
Formalizing Topological Constraints
The multi-punctured plane
For the multi-punctured plane, the fundamental group is the free group on n letters, Fn
“Free group” because there are no relations, other than the group axioms
x0
𝛔1𝛔2
𝛔3
F3 = {All words formed from letters in }
![Page 24: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/24.jpg)
Formalizing Topological Constraints
Same as the homotopy group, but also satisfies commutative property
Eg: 𝛔1𝛔2 and 𝛔2𝛔1 belong to the same homology group, whereas they belong to different homotopy groups
Homology
x0
Homology constraints are winding constraints
x0
![Page 25: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/25.jpg)
Formalizing Topological Constraints
HomologyHomotopy
Inverse
Commutativity
Yes Yes
YesNo
Inverse
𝛔i𝛔i-1= 𝛔i-1𝛔i = 𝛆
Identity Yes Yes
Commutativity
𝛔i𝛔j= 𝛔j𝛔i
Identity
𝛔i𝛆= 𝛆𝛔i = 𝛔i
![Page 26: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/26.jpg)
Relating Beam Crossings and Topology
How do beam crossings relate to homotopy and homology?
Recollect construction of virtual beams
![Page 27: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/27.jpg)
Relating Beam Crossings and Topology
a a’
bb’x
x’
Beam Crossings Letters of Fn
Spanning tree ensures “minimally sufficient collection of sensor beams”
Mapping from beam crossings to elements of Fn
![Page 28: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/28.jpg)
Relating Beam Crossings and TopologyApplying equivalence relations on the string of beam crossings yields corresponding topological constraints
Inverse
aa’ ~ a’a ~ 𝛆Commutativity
ab ~ ba
Example reduction: b(s) = abcc’b’ca’
Inverse (Homotopy): abcc’b’ca’ ~ abb’ca’ ~ aca’
Inverse and Commutativity (Homology): abcc’b’ca’ ~ abb’ca’ ~ aca’ ~ aa’c ~ c
![Page 29: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/29.jpg)
Relating Beam Crossings and Topology
“Find a circuit path for an UAV such that it obtains 360° views of n regions of interest, in a particular order”
“Find a loop in the n-punctured plane, that belongs to a particular element of the fundamental group Fn”
“Find a loop that has a particular sequence of beam crossings”
![Page 30: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/30.jpg)
Homotopy vs Homology
Homotopy Homology
Order matters Order does not matter
Goal string: abccdd
![Page 31: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/31.jpg)
Experiments
UAV surveillance in dynamic environments
• Gather information about ROI in the environment
• Avoid hostile patrollers
• 5D (x, y, θ, t, b) planning problem
• A* search
• 2D (x,y) Dijkstra search to precompute admissible heuristics
• Cost function: Length of the path + Risk associated with getting close to
a radar installation
![Page 32: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/32.jpg)
Experiments
UAV surveillance in dynamic environments
![Page 33: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/33.jpg)
Experiments
Loop planning for object modeling
• Obtain 3D models of objects in partially structured environments
• Mobile robot equipped with RGB-D sensor circumnavigates areas of
interest
• 4D (x, y, θ, b) planning problem
![Page 34: Planning Under Topological Constraints Using Beam-Graphsvenkatrn/papers/icra13_slides.pdf · theory and flow theory: Related Work ... Algorithmic Foundations of Robotics, VIII Idea](https://reader033.vdocuments.us/reader033/viewer/2022050607/5fae3fc4115c4a6adf2aafeb/html5/thumbnails/34.jpg)
Experiments
Scaling Experiments
1 2 3 4 5 6 7 8 9 100
10
20
30
40
50
60
70
80
90