path planning in expansive c-spaces d. hsuj.-c. latomber. motwani cs dept., stanford university,...

24
Path Planning in Expansive C- Path Planning in Expansive C- Spaces Spaces D. Hsu J.-C. Latombe R. Motwani CS Dept., Stanford University, 1997

Post on 22-Dec-2015

216 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

Path Planning in Expansive C-Path Planning in Expansive C-SpacesSpaces

D. Hsu J.-C. Latombe R. Motwani

CS Dept., Stanford University, 1997

Page 2: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

What we Want: Good Connectivity

For each connected component of the free space, there should beonly one connected component of the roadmap.

Page 3: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

What we Want: Good Coverage

Given a pre-computed roadmap, it should be easy to connect newstart and goal configurations.

Page 4: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

Main Result

If the C-space is expansive, then we can build a roadmapthat has both good connectivity and good coverage.

Page 5: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

Definition - -goodness

A free configuration q is -good if it sees an -fraction of thevolume of the free space F.

F is -good if every free configuration is -good.

qA is 1-good

qB is ½-good

F is ½-good

Page 6: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

Definition - Lookout of a Subset

The -lookout of a subset S of F is the subset of points of S that seea -fraction of the volume of F\S.

Page 7: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

Definition - Expansiveness

A free space F is expansive if every subset S of F has a large lookout.

More formally:The free space F is (, , )-expansive if:1. F is -good2. For every subset S of F, the volume of a -lookout of S is an

-fraction of the volume of S.

Page 8: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

Main Result

If the C-space is expansive, then we can build a roadmapthat has both good connectivity and good coverage.

Page 9: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

Definition: Linking Sequence

Pt+1 is chosen from the lookout of the subset of points seen byp0, p1, …, pt.

Page 10: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

Linking Sequences in Expansive Spaces

Any milestone of a roadmap is likely to have a linking sequence ofarbitrary length t, provided the roadmap is big enough.

For large t, the linking sequence of any milestone spans a large fraction of the volume of F.

Hence, the intersection of two linking sequences is likely to containa milestone of the roadmap.

Page 11: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

Theorem 1: Roadmap Connectivity

The probability that a roadmap fails to achieve good connectivityin an expansive space decreases exponentially with the number of milestones.

Page 12: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

Theorem 2: Roadmap Coverage

The probability that a roadmap fails to achieve good coveragein an expansive space decreases exponentially with the number ofmilestones.

Page 13: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

In Practice…

How to build linking sequences ?

Problem: lookouts cannot be easily computed.

However, we know that lookouts occupy a large fraction of the free space.

Hence, linking sequences can be found by sampling uniformly at random, and by keeping only those points that see a large portion of the free space.

Page 14: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

The New Planner

We grow two trees from qinit and qgoal, respectively.

New nodes are selected by sampling uniformly at random around thealready existing nodes.

We incorporate the nodes that are most likely to see a large portionof the free space.

A path is found when the two trees can be connected.

qinitqgoal

Page 15: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

The Weight Function

We incorporate the nodes that are most likely to see a large portionof the free space.

For each node x of the tree T, w(x) is equal to the number of samplednodes of T that lie in a fixed neighborhood of x.

Selecting nodes with probability 1/w(x) ensures the tree spreads uniformly in the free space.

xw(x)=3

T

Page 16: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

The Expansion Algorithm

Pick a node x from T with probability 1/w(x)Sample K points from a fixed neighborhood of xFor each sampled configuration y, retain y with probability 1/w(y)If:

1. y is retained2. y has no collision3. x and y see each other

Then, place an edge between x and y

x

T

Page 17: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

The Expansion Algorithm

Pick a node x from T with probability 1/w(x)Sample K points from a fixed neighborhood of xFor each sampled configuration y, retain y with probability 1/w(y)If:

1. y is retained2. y has no collision3. x and y see each other

Then, place an edge between x and y

x

T

Page 18: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

The Expansion Algorithm

Pick a node x from T with probability 1/w(x)Sample K points from a fixed neighborhood of xFor each sampled configuration y, retain y with probability 1/w(y)If:

1. y is retained2. y has no collision3. x and y see each other

Then, place an edge between x and y

x

Ty

Page 19: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

The Expansion Algorithm

Pick a node x from T with probability 1/w(x)Sample K points from a fixed neighborhood of xFor each sampled configuration y, retain y with probability 1/w(y)If:

1. y is retained2. y has no collision3. x and y see each other

Then, place an edge between x and y

x

Ty

Page 20: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

The Connection Algorithm

For every x in Tinit and y in Tgoal such that dist(x,y)<L do:If x and y see each other, then connect x and y

qinitqgoal

Tinit Tgoal

xy

Page 21: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

The Connection Algorithm

For every x in Tinit and y in Tgoal such that dist(x,y)<L do:If x and y see each other, then connect x and y

qinitqgoal

Tinit Tgoal

xy

Page 22: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

The Connection Algorithm

For every x in Tinit and y in Tgoal such that dist(x,y)<L do:If x and y see each other, then connect x and y

qinitqgoal

Tinit Tgoal

xy

Page 23: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

Example

Page 24: Path Planning in Expansive C-Spaces D. HsuJ.-C. LatombeR. Motwani CS Dept., Stanford University, 1997

Conclusion

If the C-space is expansive, then we can efficiently build a roadmapthat has both good connectivity and good coverage.

Suggested improvements:Find a parametrization of the C-space that maximizes expansivenessApply geometric transforms that increase expansivenessDecompose the free space into expansive components