28
1 Efficient planning of informative paths for multiple robots Amarjeet Singh * , Andreas Krause + , Carlos Guestrin + , William J. Kaiser * , Maxim Batalin * * Center for Embedded Networked Sensing, University of California, Los Angeles + Machine Learning Department, Carnegie Mellon University

1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

  • View
    213

  • Download
    0

Embed Size (px)

Citation preview

Page 1: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

1

Efficient planning of informative paths for

multiple robots

Amarjeet Singh*, Andreas Krause+, Carlos Guestrin+, William J. Kaiser*, Maxim Batalin*

* Center for Embedded Networked Sensing, University of California, Los Angeles+ Machine Learning Department, Carnegie Mellon University

Page 2: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

2

Predicting spatial phenomena in large environments

Constraint: Limited fuel for making observations

Fundamental Problem: Where should we observe to maximize the collected information?

Biomass in lakes Salt concentration in rivers

Page 3: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

3

How to quantify collected information?

Mutual Information (MI): reduction in uncertainty (entropy) at unobserved locations

[Caselton & Zidek, 1984]

MI = 4Path length = 10

MI = 10Path length = 40

Page 4: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

4

Selecting the sensing locations

Lake Boundary

G1

G2

G3

G4

Greedy selection of sampling

locations is (1-1/e) ~ 63% optimal [Guestrin et. al, ICML’05]

Result due to Submodularity of MI: Diminishing returns

Greedy may lead to longer paths!

Greedily select the locations that provide the most amount of information

Page 5: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

5

Greedy - reward/cost maximization

Available Budget = B

s

Reward = B

Cost = B

rewardcost

= 2

rewardcost

= 1

Page 6: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

6

Greedy - reward/cost maximization

Available Budget = B-

s

B

B

BToo far!

Greedy Reward = 2

Page 7: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

7

Greedy - reward/cost maximization

Available Budget = 0

s

B

B

Greedy Reward = 2

Optimal Reward = B

Greedy can be arbitrarily poor!

Page 8: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

8

Informative path planning problem

maxp MI(P) MI – submodular function

Lake Boundary

Start- sFinish- t

P

C(P) · B Informative path planning – special

case of Submodular Orienteering Best known approximation algorithm –

Recursive path planning algorithm [Chekuri et. Al, FOCS’05]

Page 9: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

9

Recursive path planning algorithm

[Chekuri et.al, FOCS’05]

Start (s)Finish (t)

vm

Recursively search middle node vm

P1

P2

Solve for smaller subproblems P1 and P2

Page 10: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

10

vm2

Recursive path planning algorithm

[Chekuri et.al, FOCS’05]

Start (s)Finish (t)

P1vm1

vm3

Maximum reward

Recursively search vm C(P1) · B1

Lake boundary

vm

Page 11: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

11

Recursive path planning algorithm

[Chekuri et.al, FOCS’05]

Start (s)Finish (t)

P1

vm

Recursively search vm C(P1) · B1

Commit to the nodes visited in P1

Recursively optimize P2 C(P2) · B-B1

P2

Maximum reward

Committing to nodes in P1 before optimizing P2 makes the algorithm greedy!

Page 12: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

12

Quasi-polynomial running time O(B*M)log(B*M)

B: Budget

RewardChekuri ¸RewardOptimal

log(M) M: Total number of nodes in the graph

60 80 100 120 140 160

Cost of output path (meters)

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

Exe

cutio

n T

ime

(S

eco

nd

s)OOPS!

Small problem with 23 sensinglocations

Recursive path planning algorithm[Chekuri et.al, FOCS’05]

Page 13: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

13

60 80 100 120 140 16010

0

105

102

103

104

101

Exe

cutio

n T

ime

(se

con

ds)

Cost of output path (meters)

Almost a day!!

Recursive path planning algorithm[Chekuri et.al, FOCS’05]

Quasi-polynomial running time O(B*M)log(B* M)

B: Budget

RewardChekuri ¸RewardOptimal

log(M) M: Total number of nodes in the graph

Small problem with 23 sensinglocations

Page 14: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

14

Our contributions

Algorithm with significantly improved running time exploiting recursive path planning

Spatial decomposition of sensing region Branch and bound - Calculating bounds using

submodularity and other heuristics to prune search space

Extended single robot path planning to multiple robots with strong approximation guarantee

Extensive empirical evaluation on several real world sensing datasets

Including data collected using robotic boat at Lake Fulmor, California

Page 15: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

15

Lake Boundary

Spatial decomposition into cells

Ending node t

Starting node s

Ending Cell Ct

Starting Cell Cs

Search for middle Cell Cm

Perform recursive path planning on cells

P1P2

Page 16: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

16

Ending Cell Ct

Starting Cell Cs

Middle Cell Cm

P1P2

Greedily select locations without path cost constraint: 1-1/e optimal

Node selection inside the cell

Incoming path Exiting path

G1

G2 G4

G3

Tradeoff: Larger cell size ) Faster Execution, Increased additional traveling

cost Smaller cell size ) Slower Execution, Reduced additional traveling

cost

Small cells: Traveling cost

inside cell can be ignored

Additional cost for traveling to the

sensing locations

Page 17: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

17

Recursive Path Planning

Approximation guarantees

Running time:

O((B*N)log B*N)

Required budget:O(B)

Collected Reward ¸(1-1/e) RewardOptimal

log(N) N: Total number of cells in the graph

80 100 120 140 160Cost of output path (meters)

6010

0

105

102

103

104

101

Exe

cutio

n T

ime

(se

con

ds)

Efficient Path Planning

Approx. a day

Approx. 2 min.

Small problem – 23 sensing locations

Too slow for larger problems!

Page 18: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

18

Further improvement in running time

Search space represented as SUM-MAX tree (similar to AND-OR tree)

102

103

104

105

Exe

cutio

n T

ime

(se

con

ds)

200 250 300 350 400 450Cost of output path (meters)

Upto 400 meters calculated within approx. 15 min.

Pruned search space using branch and bound Upper bound exploiting

submodularity Lower bound exploiting

known heuristic [Chao et. al’

96] Several other tricks – see

paper

Larger problem – 109 sensing locations

Page 19: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

19

Multi robot informative path planning problem

maxP1,P2,P3 MI(P1 U P2 U P3)

MI – submodular function

s t

C(P1) · B; C(P2) · B; C(P3) · B

P2

P3

P1

Page 20: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

20

Multi robot path planning – Simple sequential allocation approach

s t

P2

P3

P1

Use algorithm for single robot instance to find path P1 for the first robot

Optimize for second robot (P2) committing to nodes in P1

Optimize for third robot (P3) committing to nodes in P1 and P2

Page 21: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

21

Performance evaluation

Works for any single robot path planning algorithm Independent of number of robots used

RewardMR¸

RewardOptimal

1 +

Greedy selection of

nodes with no path cost constraintArbitrarily Poor

Recursive Greedy path planning

RewardRG ¸RewardOpt

(=log(M))

Sequential allocation for multiple robots – Greedy over paths

??

Page 22: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

22

Efficient multi-robot information path planning

Spatial Decomposition Obtain cell path exploiting submodularity, branch and bound

A

B C

D

Greedy node selection within visited cells to get node path

Sequential Allocation for multi-robot path planning

Page 23: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

23

Robotic boat measuring surface temperature and chlorophyll at

Lake Fulmor, California

Empirical evaluation

SERVER

LAB

KITCHEN

COPYELEC

PHONEQUIET

STORAGE

CONFERENCE

OFFICEOFFICE50

51

52 53

54

46

48

49

47

43

45

44

42 41

3739

38 36

33

3

6

10

11

12

13 14

1516

17

19

2021

22

242526283032

31

2729

23

18

9

5

8

7

4

34

1

2

3540

52 sensor motes used to monitor temperature at Intel Research

laboratory, Berkeley

Precipitation data collected from 167 regions in Pacific NW, during

the years 1949-1994

Page 24: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

24

102

103

104

105

Exe

cutio

n T

ime

(sec

onds

)

10 15 20 25 30 35Cost of output path (meters)

Empirical evaluation – varying the cell size

Precipitation Dataset

No. of cells = 36 No. of cells = 25 No. of cells = 16

Low

er is be

tter

2

3

4

5

6

7

8

Hig

her

info

rma

tion

qu

ality

10 15 20 25 30 35Cost of output path (meters)

Page 25: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

25

4

5

6

7

8

9

10

Hig

her

info

rma

tion

qu

ality

60 80 100 120 140 160Cost of output path (meters)

ChekuriAlgorithm

Proposed EfficientAlgorithm

Intel Laboratory Temperature Dataset

Empirical evaluation – reward comparison

• Reduced execution time by several factors• Similar collected reward

Page 26: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

26

Empirical evaluation – heuristic comparison

Lake Temperature Dataset

4

6

8

10

12

14

Hig

her

info

rma

tion

qu

ality

200 250 300 350 400 450Cost of output path (meters)

Efficient informative path planning algorithm

Known heuristic [Chao et. al’ 96]

Page 27: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

27

8

9

10

11

12

13

14

15

16

To

tal R

MS

Err

or

200 250 300 350 400 450

Cost of output path per robot (meters)

Empirical evaluation – multi robot

Robot-1

Robot-3

Robot-2

1 Robot

2 Robots

3 Robots

Low

er is be

tter

Page 28: 1 Efficient planning of informative paths for multiple robots Amarjeet Singh *, Andreas Krause +, Carlos Guestrin +, William J. Kaiser *, Maxim Batalin

28

Conclusions

First efficient multi robot informative path planning algorithm with performance guarantee Exploits spatial decomposition Exploits submodularity and other heuristics for branch

and bound Near optimal extension of single robot path

planning algorithm to multiple robots Extensive empirical evaluation on several

real world sensor network datasets Including data collected using robotic boat in real lake Planning on a deployment at Lake Merced, California

with robotic boat in February