26
Localization for Mobile Sensor Networks ACM MobiCom 2004 ACM MobiCom 2004 Philadelphia, PA Philadelphia, PA 28 September 2004 28 September 2004 University of Virginia Computer Science Lingxuan Hu and David Evans You are here

Localization for Mobile Sensor Networks

  • Upload
    verdi

  • View
    43

  • Download
    0

Embed Size (px)

DESCRIPTION

Lingxuan Hu and David Evans. Localization for Mobile Sensor Networks. You are here. ACM MobiCom 2004 Philadelphia, PA 28 September 2004. University of Virginia Computer Science. Location Matters. Sensor Net Applications Mapping Environment monitoring Event tracking - PowerPoint PPT Presentation

Citation preview

Page 1: Localization for Mobile Sensor Networks

Localization for Mobile Sensor Networks

ACM MobiCom 2004ACM MobiCom 2004Philadelphia, PAPhiladelphia, PA28 September 200428 September 2004

University of VirginiaComputer Science

Lingxuan Hu and David Evans

You are here

Page 2: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 2

Location Matters

• Sensor Net Applications– Mapping– Environment monitoring– Event tracking

• Geographic routing protocols

Page 3: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 3

Determining Location• Direct approaches

– GPS• Expensive (cost, size, energy)• Only works outdoors, on Earth

– Configured manually• Expensive• Not possible for ad hoc, mobile networks

• Indirect approaches– Small number of seed nodes

• Seeds are configured or have GPS

– Other nodes determine location based on messages received

Page 4: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 4

Hop-Count TechniquesDV-HOP [Niculescu & Nath, 2003]Amorphous [Nagpal et. al, 2003]

Works well with a few, well-located seeds and regular, static node distribution. Works poorly if nodes move or are unevenly distributed.

r

1

1

2

23

3

33

4

4

4

44

5

5

6

7

8

Page 5: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 5

Local TechniquesCentroid [Bulusu, Heidemann, Estrin, 2000]:Calculate center of all heard seed locations

APIT [He, et. al, Mobicom 2003]:Use triangular regionsDepend on a high density of

seeds (with long transmission ranges)

Page 6: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 6

Our Goal

• (Reasonably) Accurate Localization in Mobile Networks

• Low Density, Arbitrarily Placed Seeds

• Range-free: no special hardware • Low communication (limited

addition to normal neighbor discovery)

Page 7: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 7

Scenarios

NASA Mars TumbleweedImage by Jeff Antol

Nodes moving, seeds stationary

Nodes and seeds moving

Nodes stationary, seeds moving

Page 8: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 8

Our Approach: Monte Carlo Localization

• Adapts an approach from robotics localization

• Take advantage of mobility:– Moving makes things harder…but

provides more information– Properties of time and space limit

possible locations; cooperation from neighbors

Frank Dellaert, Dieter Fox, Wolfram Burgard and Sebastian Thrun. Monte Carlo Localization for Mobile Robots. ICRA 1999.

Page 9: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 9

MCL: Initialization

Initialization: Node has no knowledge of its location.

L0 = { set of N random locations in the deployment area }

Node’s actual position

Page 10: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 10

MCL Step: Predict

Node’s actual position

Predict: Node guesses new possible locations based on previous possible locations and maximum velocity, vmax

Page 11: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 11

Prediction

p(lt | lt-1) = c if d(lt, lt-1) < vmax

0 if d(lt, lt-1) ≥ vmax

Assumes node is equally likely to move in any direction with any speed between 0 and vmax.

Can adjust probability distribution if more is known.

Page 12: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 12

MCL Step: Predict

Node’s actual position

Predict: Node guesses new possible locations based on previous possible locations and maximum velocity, vmax

Filter

Filter: Remove samples that are inconsistent with observations

Seed node: knowsand transmits location

r

Page 13: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 13

Filtering

Indirect SeedIf node doesn’t hear a seed, but one of your neighbors hears it, node must be within distance (r, 2r] of that seed’s location.

Direct SeedIf node hears a seed,the node must (likely) bewith distance r ofthe seed’s location

S S

Page 14: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 14

Resampling

Use prediction distribution to create enough sample points that are consistent with the observations.

Page 15: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 15

Recap: AlgorithmInitialization: Node has no knowledge of its location. L0 = { set of N random locations in the deployment area }

Iteration Step: Compute new possible location set Lt based on Lt-1, thepossible location set from the previous time step, and the new observations. Lt = { } while (size (Lt) < N) do R = { l | l is selected from the prediction distribution } Rfiltered = { l | l where l R and filtering condition is met } Lt = choose (Lt Rfiltered, N)

Page 16: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 16

Results Summary• Effect of network parameters:

– Speed of nodes and seeds– Density of nodes and seeds

• Cost Tradeoffs:– Memory v. Accuracy: Number of samples– Communication v. Accuracy: Indirect seeds

• Radio Irregularity: fairly resilient• Movement: control helps; group motion

hurts

Page 17: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 17

Convergence

Node density nd = 10, seed density sd = 1

The localization error converges in first 10-20 steps

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

0 5 10 15 20 25 30 35 40 45 50

Est

imate

Err

or

(r)

Time (steps)

vmax=.2 r, smax=0

vmax=r, smax=0

vmax=r, smax=r

Page 18: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 18

Speed Helps and Hurts

Increasing speed increases location uncertainty ̶R but provides more observations.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0.10.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

Est

imat

e E

rror

(r)

vmax (r distances per time unit)

sd=1, smin=0, smax=vmax

sd=1, smax=smin=r

sd=2, smax=vmax

sd=2, smax=smin=r

Node density nd = 10

Page 19: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 19

00.20.40.60.81

1.21.41.61.82

2.22.42.62.83

0.1 0.5 1 1.5 2 2.5 3 3.5 4

Est

imate

Err

or

(r)

Seed Density

MCL

Centroid

Amorphous

Seed Density

nd = 10, vmax = smax=.2r

Better accuracy than other localization algorithms over range of seed densities

Centroid: Bulusu, Heidemann and Estrin. IEEE Personal Communications Magazine. Oct 2000.

Amorphous: Nagpal, Shrobe and Bachrach. IPSN 2003.

Page 20: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 20

Cost Tradeoff: Samples Maintained

00.10.20.30.40.50.60.70.80.91.0

1.2

1 2 5 10 20 50 100 200 5001000

Est

imate

Err

or

(r)

Sample Size (N)

sd=1, vmax=smax=.2r

sd=1, vmax=smax=r

sd=2, vmax=smax=.2rsd=2, vmax=smax=r

1.1 nd = 10

Good accuracy is achieved with only 20 samples (~100 bytes)

Page 21: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 21

Cost Tradeoff: Impact of Indirect Seeds

0

0.2

0.40.6

0.8

11.2

1.41.6

1.8

22.2

2.4

2.62.8

3

0.1 0.5 1 1.5 2 2.5 3 3.5 4

Est

imate

Err

or

(r)

Seed Density

Direct seeds onlyDirect and Indirect seeds

Indirect seeds help, and cost is low if neighbor discovery is required.

nd = 10, vmax = smax=.2r

Page 22: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 22

Radio Irregularity

nd = 10, sd = 1, vmax = smax=.2r

Insensitive to irregular radio pattern

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

0 0.1 0.2 0.3 0.4 0.5

Est

imate

Err

or

(r)

Degree of Irregularity (r varies ±dr)

MCL

Centroid

Amorphous

Page 23: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 23

Motion

nd=10, vmax=smax=r

Adversely affected by consistent group motion

00.51

1.52

2.53

3.54

4.55

5.56

0 0.5 1 2 4 60

0.51

1.52

2.53

3.54

4.55

5.56

0 0.5 1 2 4 6

Est

imate

Err

or

(r)

Maximum Group Motion Speed (r units per time step)

sd =.3

sd =1

sd =2

0

1

2

3

4

0 20 40 60 80 100 120 140 160 180 200Est

imate

Err

or

(r)

Time

Random, vmax=smax=.2r

Area Scan

Random, vmax=0, smax=.2r

Scan

Stream and Currents Random Waypoint vs. Area Scan

Controlled motion of seeds improves accuracy

Page 24: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 24

Future Work: Security• Attacks on localization:

– Bogus seed announcements• Require authentication between seeds and

nodes

– Bogus indirect announcements• Retransmit tokens received from seeds

– Replay, wormhole attacks• Filtering has advantages as long as you get

one legitimate announcement

• Proving node location to others

Page 25: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 25

Summary• Mobility can improve localization:

– Increases uncertainty, but more observations

• Monte Carlo Localization– Maintain set of samples representing

possible locations– Filter out impossible locations based on

observations from direct and indirect seeds– Achieves accurate localization cheaply with

low seed density

Page 26: Localization for Mobile Sensor Networks

www.cs.virginia.edu/mcl 26

Thanks!

http://www.cs.virginia.edu/mcl

People: Tarek Abdelzaher, Tian He, Anita Jones, Brad Karp, Kenneth Lodding, Nathaneal Paul, Yinlin Yang, Joel Winstead, Chalermpong Worawannotai

Funding: NSF ITR, NSF CAREER, DARPA SRS