Fine-Grained Localization in Sensor and Ad-Hoc Networks

You might also like

Download as ppt, pdf, or txt
Download as ppt, pdf, or txt
You are on page 1of 56

Fine-Grained Localization in

Sensor and Ad-Hoc Networks


Ph.D. Dissertation Defense
David Goldenberg
Dissertation Advisor: Y. Richard Yang

Committee Members: Jim Aspnes, A. Stephen Morse,


Avi Silberschatz, Nitin Vaidya (UIUC)
2

Overview
 This dissertation provides a theoretical basis for
the localization problem, demonstrating
conditions for its solvability and defining its
computational complexity.
 We apply our fundamental results on localization
to identify conditions under which the problem is
efficiently solvable and to develop localization
algorithms for a broader class of networks than
previous approaches could localize.
3

Collaborators (2003-2006)
 Brian D.O. Anderson (Australia National University and NICTA)
 James Aspnes
 P.N. Belhumeur (Columbia University)
 Pascal Bihler
 Ming Cao
 Tolga Eren
 Jia Fang
 Arvind Krishnamurthy
 Jie (Archer) Lin
 Wesley Maness
 A. Stephen Morse
 Brad Rosen
 Andreas Savvides
 Walter Whiteley (York University)
 Y. Richard Yang
 Anthony Young
4

Outline
 Introduction to Localization

 Conditions for Unique Localization

 Computational Complexity of Localization

 Localization in Sparse Networks


5

Why are Locations Important?


 Wireless ad-hoc networks are an important emerging technology
 Small, low-cost, low-power, multi-functional sensors will soon be a reality.
 Accurate locations of individual sensors are useful for many applications
 “Sensing data without knowing the sensor location is meaningless.” [IEEE
Computer, Vol. 33, 2000]
 New applications enabled by availability of sensor locations.

 Location-aware computing
 Resource selection (server, printer, etc.).
 Location aware information services (web-search, advertisement, etc.).

 Sensor network applications


 Inventory management, intruder detection, traffic monitoring,
emergency crew coordination, air/water quality monitoring,
military/intelligence apps.
6

Example: Great Duck Island Sensor Network

 Monitoring breeding of Leach’s


Storm Petrels without human
presence.
 15 minute human visit leads to 20%
offspring mortality.
 Sensors need to be small to avoid
disrupting bird behavior.
7

Great Duck Island Deployment Goals


 Occupancy pattern of nests? Single hop weather
 Environmental changes around the Single hop burrow
nests over time? Multi hop weather
 Environmental variation across nests? Multi hop burrow
 Correlation with breeding success?

 Light, temperature, infrared, and


humidity sensors installed.
 Infrared sensors detect presence of
birds in nests.

 Sensor locations critical to interpreting data.


10m
 Locations determined by manual
configuration, but this will not be possible in
the general case.
8

Example: ZebraNet Sensor Network


 Biologists want to track animals to study:
 Interactions between individuals.
 Interactions between species.
 Impact of human development.

 Current tracking technology: VHF collar transmitters


 Wishlist:
 24/7 position, data, and interaction logs.
 Wireless connectivity for mobility.
 Data storage to tolerate an intermittent base station.

 ZebraNet:
 Mobile sensor net with intermittent base station.
 Records position using GPS every 3 minutes.
 Records Sun/shade info.
 Detailed movement information (speed, movement signature)
3 minutes each hour.
 Future: head up/head down, body temperature, heart rate,
camera.
 Goal, full ecosystem monitoring (zebras, hyenas, lions…).
9

Military Applications
 Intelligence gathering (troop movements, events of interest).
 Detection and localization of chemical, biological, radiological,
nuclear, and explosive materials.
 Sniper localization.
 Signal jamming over a specific area.

 Visions for sensor network deployment:


 Dropped in large numbers from UAV.
 Mortar-Launched.

!
10

Why is Localization a Non-Trivial Problem?

 Manual configuration
 Unscalable and sometimes impossible.

 Why not use GPS to localize?


 Hardware requirements vs. small sensors.
 Obstructions to GPS satellites common.
 GPS satellites not necessarily overhead.
 Doesn’t work indoors or underground.
 GPS jammed by sophisticated adversaries.
 GPS accuracy (10-20 feet) poor for short
range sensors.
11

Fine-Grained Localization (Savvides, 2001)


 Physically:
 Network of n nodes, m of which have known location, existing in space at
locations: {x1…xm,xm+1,…,xn}.
 Set of some pair-wise inter-node distance measurements.
 Usually between proximal nodes (iff d < r in unit disk networks).

 Abstraction
 Given: Graph GN, {x1,...,xm}, and δ, the edge weight function.
 Find: Realization of the graph.

3 1 4 {x1,x2,x3}
2
5 {d14, d24, d25, d35, d45}
3 5
4 2

1
{x4, x5}

Beacons: nodes with known position


Regular nodes: nodes with unknown position
12

Ranging Systems
 TDoA – Time Difference of Arrival
 Uses ultrasound and radio signals
to determine distance.

 Range of meters, cm accuracy.

 Possible to increase sensing range MIT cricket mote

by increasing transmission power.

UCLA medusa mote 2 (2002)

Yale ENALAB XYZ Motes UCLA medusa mote (2001)


13

Our Contributions
 Graph-theoretic conditions for the unique solvability of the localization
problem in the plane.

 Proof that the problem is NP-complete even for the idealized case of
unit-disk networks.

 Constructive characterization of classes of uniquely localizable and


easily localizable networks for the plane and 3D.

 A localization algorithm that localizes a wider class of networks than


was possible with existing approaches.

 In-depth study of the localizability properties of random networks:


 New adaptive localizability-optimizing deployment strategies.
 Impact of non-uniquely localizable nodes on network performance.
14

Outline
 Introduction to Localization

 Conditions for Unique Localization

 Computational Complexity of Localization

 Localization in Sparse Networks


15

Unique Localizability
 Network is uniquely localizable if there is exactly one set
of points {xm+1,…,xn} consistent with GN, {x1,…,xm} and
δ:E to R.

 Can we determine localizability by graph properties alone?


(as opposed to the properties of δ).

 In the plane, yes (more or less). Properties of the graph


determine solvability in the generic case.
 Probability 1 for randomly generated node locations.
16

Degenerate Cases Fool Abstraction


2
{x1, x2, x3}
2
4 {d14, d24, d34}
4 1
1 3
3
In general, this network is
probability 1 case: uniquely localizable.

1 4 first case: {x4}


second case: ???
2 ?
1
3
2

? 3

probability 0 case: In degenerate case, it is not:


The constraints are redundant.
17

Continuous Non-Uniqueness

 Continuous non-uniqueness:
 Can move points from one configuration to another while
respecting constraints.
 Excess degrees of freedom present in configuration.
 A formation is RIGID if it cannot be continuously deformed.
18

Condition for Rigidity


 Purely combinatorial characterization of
generic rigidity in the plane.
 2n-3 edges necessary for rigidity, and:

Laman’s condition:
A graph G with 2n-3 edges is rigid in two dimensions
if and only if no subgraph G’ has more than 2n’-3 edges*.
* where n’ is the number of vertices in G’

Laman’s condition is a statement that any rigid graph with n vertices must have a set
of 2n-3 well-distributed edges.

Not enough edges Enough edges but not well distributed Just right
19

Discontinuous Non-Uniqueness in Rigid Graphs


2 2 3
3 3
1 1
1
2
Flip Ambiguities: 0 4
0 4 0 4

6 5

6 5 6 5
214 config
configs

1 1

5 4 1 5 4
1
Discontinuous Flex 1
2 2
Ambiguities:
2
2
3 2
3
20

Unique Graph Realization


Solution:

G must be rigid.

G must be 3-connected.

b c b
G must be redundantly rigid:
f d It must remain rigid upon
a
e e removal of any single edge.
a c
d f
A graph has a unique realization in the plane iff it is redundantly
rigid and 3-connected (globally rigid). Hendrickson, ‘94
21

Is The Network Uniquely Localizable?


 Problem: By looking only at the physical connectivity structure, we would neglect
our a priori knowledge of beacon positions.
 Solution: The distances between beacons are implicitly known!
 By adding all edges between beacons to GN, we get the Grounded
Graph of the network, whose properties determine network
localizability.
 Theorem: A network is generically uniquely localizable iff its grounded
graph is globally rigid and it contains at least three beacons.

 By augmenting graph structure in this way, we fully express all available


constraint information in a graph.

5 Is this localizable?
1 4
4 2
2

1 3 5
22

Examples of GR graphs - Constructions


 Every globally rigid graph has a spanning subgraph that is minimally globally rigid.
 Every minimally globally rigid graph can be constructed inductively starting from K4 by a
series of extensions (Berg-Jordan ‘01):
 New node w and edges uw and vw replace edge uv.
 Edge wx added for some node x distinct from u, v.
 Minimal globally rigid graphs have 2n-2 edges.

5 5

1 2 1 2
1 2
6

3 4 3 4 3 4

5
5
1 2
1 2 7
7 6
6 8

3 4
3 4

Light edges are those subdivided by the extension operation.


23

Examples of Global Rigidity


Globally rigid components in green.

Random network – avg node degree 6. Regularized random network – avg node degree 4.5.
24

Construction Using Trilateration


 A position is uniquely determined by three distances to three non-collinear references.
 Minimal trilateration graphs formed by trilateration extension:
 New node w and edges uw, vw, xw added, for u, v, x distinct.
 Minimal trilateration graphs are globally rigid.
 Minimal trilateration graphs have 3n-6 edges.

5 5

1 2 1 2
1 2
6

3 4 3 4 3 4

5 5

1 2 1 2
7 6 7 6
8 …
3 4 3 4

Light edges are those removed in extension for minimally GR graph but not in trilateration.
25

Trilateration Graphs
 A trilateration graph G is one with an trilaterative ordering: an ordering
of the vertices 1,...,n such that the complete graph on the initial 3 vertices
is in G and from every vertex j > 3, there are at least 3 edges to vertices
earlier in the sequence.
 Trilateration graphs are globally rigid.

Hand-made trilateration – avg degree 6. Trilateration graph from mobile network – avg degree 9.
26

“Tripled” Connected Graphs are


Trilateration Graphs
 Theorem:
Let G = (V,E) be a connected graph.
Let G3 = (V,E E2  E3) be the graph formed from G by
adding an edge between any two vertices connected by
paths of 2 or 3 edges in E.
Then G3 is a trilateration graph.

Example where
G is a path.
27

“Doubled” 2-connected Graphs are


Globally Rigid in 2D
 Theorem:
Let G be a 2-connected graph.
One gets G2 by doubling sensing radius or
Then G2 is globally rigid. measuring angles between adjacent edges.

Example where
G is a cycle.

Minimally GR graph
by extension: Doubled cycles always
have two edges more than
a minimally GR graph, so
they are globally rigid.
Doubled cycle:
28

“Tripled” Biconnected Graphs are


Globally Rigid in 3D

 There is no known generic characterization


of global rigidity in 3D, but our result on
doubled graphs extends to 3D.

 Theorem:
Let G be a 2-connected graph.
Then G3 is globally rigid in 3D.
29

Summary of Constructive Characterization


of Globally Rigid Graphs
 2D
 3-connectivity necessary for GR.
 G2 GR if G 2-connected.
 G3 GR if G connected.

 3D
 G3 GR in 3D if G 2-connected.
 G4 GR in 3D if G connected.

 Unique localizability by increasing sensing range, given


initial connectivity.
 Conditions under which additional information can help.
30

Outline
 Introduction to Localization

 Conditions for Unique Localization

 Computational Complexity of Localization

 Localization in Sparse Networks


31

Localization
3

5
Decision problem Search problem
4 2

1 1 4 {x1,x2,x3}
1 4 2
{d14, d24, d25, d35, d45}
2 3 5

3 5
This graph has a
Does this have a unique realization.
unique realization? What is it?
This problem is in ???
Rigidity
theory general NP-hard.

{x4,x5}
Yes/No
32

Computational Complexity
 Intuitively, reflection possibilities are linked with
computational complexity
Suppose all edge
distances known …and reflection
for small triangles. possibilities are only
sorted out when one
Localization goes
gets to another beacon.
working out from
any beacon.
Triangle reflection
possibilities grow
exponentially….
33

Complexity of GR Graph Realization


 If a network is localizable, how does one go about localizing it?
 It is NP-hard to localize a network in R2 even when it is known to
be uniquely localizable.

 We will use two tools in our argument:


 The NP-hard set-partition problem.
 The globally rigid wheel graph Wn.

The set partition problem:


Input: A set of numbers S.
Output: Can S be partitioned into two subsets A
and S-A such that the sum of numbers
in each subset is equal?
W6
34

NP-hardness of Realization
Theorem:
Realization of globally rigid weighted graphs that are realizable is NP-hard

Proof sketch:
Assume we have algorithm X that takes as input a realizable globally rigid
weighted graph and outputs its unique realization.

We will find the set-partition of the partitionable set S scaled w.l.o.g so that the
sum of elements in S is less than π/2 by using calls to X.
Suppose we have S={s1,s2,s3,s4} with a set-partition. Construct a graph G along
with its edge weights for X:
Even without Set Partition,
1 we have the edge weights of G:
3 2 di,i+1=2sin(si/2)
s1+s4=s2+s3 that uniquely determine the
s4 realization .
s2
rights = lefts When G is realized, we obtain
4 s1 the picture on the left, from
s3 This is a realizationwhich
of W5we! obtain set partition!
0
35

Localization Complexity for Sparse Networks

 Problem with previous result is that edges exist arbitrarily.


 Graphs used in previous proof unlikely to arise in practice.
 In realistic networks, edges are more likely to exist between close
nodes, and do not exist between distant nodes.
 Unit Disk Graphs: edge present if distance between nodes less than
parameter r.
 Therefore: if edge absent, distance between nodes is greater than r.
 Does this information help us solve the localization problem?

3 2

Red edge would exist in unit disk graph, so unit disk


graph localization would not solve Set Partition.
4

0
36

Complexity of Localizing Unit Disk Graphs


 Theorem: Localization for sparse sensor networks is NP-hard.
 Method:
 Reduction from Circuit Satisfiability to Unit Disk Graph Reconstruction.

 Reduction is by construction of a family of graphs that represent Boolean circuits.


 Rigid bodies in the graph represent wires.
 Relative position of rigid bodies in the graph represent signals on wires.
 NOT and AND gates built out of constraints between these bodies expressed in the
graph structure.
 There is a polynomial-time reduction from Circuit Satisfiability to Unit Disk Graph
Reconstruction, in which there is a one-to-one correspondence between satisfying
assignments to the circuit and solutions to the resulting localization problem.

Unit Disk Graph Reconstruction (decision problem)


Circuit Satisfiability (NP-hard): Input: Graph G along with a parameter r, and the square of
Input: A boolean combinatorial circuit. each edge length (luv)2 (to avoid irrational edge lengths).
Composed of AND, OR, and NOT gates
Output: YES iff the circuit is satisfiable. Output: YES iff there exists a set of points in R2 such that
distance from u to v is luv if uv is an edge in G and greater
than r otherwise.
37

Localization in Trilateration Graphs


 As one adds more edges, localization becomes easier: There are classes of
globally rigid graph which are easy to localize.
 Trilateration graphs are localizable in polynomial time.
 Remember: One gets a trilateration graph from a connected network by tripling
the sensing radius.

 Algorithm:
 If initial 3 vertices known, localize
vertices one at a time until all vertices
localized.
 Else starting with each triangle in the
graph, proceed as above until all
localized.
 O(|V|2) or O(|V|5).
38

Connectivity in Random Networks


 The random geometric graph Gn(r) is the random The following guarantees Gn(r) is k-
graph associated with formations with n vertices connected with high probability for
with all links of length less than r, where the some constant c large enough and
vertices are points in [0,1]2 generated by a two constant k:
dimensional Poisson point process of intensity n. 2
limn   nr  c
log n
r Penrose, ‘99

 Note: Need nr2/(log n) > c, for some c, to


guarantee even connectivity.
 Theorem: If nr2/(log n) > 8, with high
probability, Gn(r) is a trilateration graph.

 This identifies conditions under which a


simple iterated trilateration algorithm will
succeed in localization.
39

Trilateration in Random Networks


Iterative Trilateration
Localized mode:
Broadcast position.

Unlocalized mode:
Listen for broadcast.
if broadcast from (x,y) heard,
Determine distance to (x,y).
if three broadcasts heard
Determine position
Switch to localized mode

 Sensors have 2 modes.


 Sensors determine distance from heard transmitter.
 All sensors are pre-placed and plugged in

But how
fast?
40

Asymptotics of Trilateration in Random Networks

Beacons Sensing radius E[tloc]


n )
O(1) O( log
n
n) O(
logn

n
O( )
logn O( log
n
n) O( logn )

O(n) O( log
n
n) O(1)

Running times to complete localization using trilateration for different beacon densities.
41

NP-hardness of Localization
 Fine-grained localization is NP-hard due to NP-hardness
of realizing globally rigid graphs.

 This means that localization of networks in complete


generality is unlikely to be efficiently solvable.

 Motivates search for reasonable special cases and


heuristics. Explains hit-or-miss character of previous
approaches.
 Changing sensing radius can predictably convert connectedness
to global rigidity and trilateration.
42

Outline
 Introduction to Localization

 Conditions for Unique Localization

 Computational Complexity of Localization

 Localization in Sparse Networks


43

Motivation
 Being able to precisely localize only trilateration
networks is unsatisfying.
 Trilaterationnetworks contain significantly more
constraints than necessary for unique localizability.
 Can we localize networks with closer to the minimal
number of constraints?
5 5

1 2 1 2
Red edges unnecessary 7 6 7 6
for unique localizability. 8 8

3 4 3 4
Trilateration graph Globally rigid subgraph
44

Bilateration Graphs
 A bilateration graph G is one with a bilateration ordering:
an ordering of the vertices 1,...,n such that the complete
graph on the initial 3 vertices is in G and from every vertex j
> 3, there are at least 2 edges to vertices earlier in the
sequence.

 Theorem: Bilateration graphs are rigid (but not globally rigid).

 Theorem: Let G = (V,E) be a connected graph.


Then G2 is a bilateration graph.
6’
6’

 Bilateration graphs are finitely localizable in O(2|V|) time. 5’


2 5
6
 Algorithm: 6’
0
 If initial 3 vertices known, finitely localize
vertices one at a time by computing all
possible positions consistent with neighbor
positions until all vertices finitely localized. 4’’’ 1 4
 Else starting with each triangle in the graph, 3
proceed as above until all finitely localized. 3’

4’ 4’’
45

Localization in Doubled Cycles


 Based on finite localization of bilateration graphs,
localization is uniquely computable for globally
rigid doubled cycles.
 Completes in O(2|V|) time.
 Assumes nodes in general position.

6’
6’

 “Sweep” Algorithm: 5’ 5
2
 Fix the position of three vertices. 6
6’
 Until no progress made: 0
 Finitely localize each vertex connected
to two finitely localized vertices.
 Remove possibilities with no consistent
descendants. 4’’’ 1 4
3
3’

4’ 4’’
46

Localization in Doubled 2-connected Graphs


 2-connected graphs are a union of cycles (they have an Ear Decomposition).
 The ear decomposition gives a ordering in which cycles may be localized
using previous algorithm.
 Note: This means if we have angles, we can localize 2-connected networks.

Biconnected network with its ear decomposition. Doubled biconnected network.


47

Localization on General Sparse Networks


 Worst-case exponential time algorithm for localization
in sparse networks:
5 6 5
6
4’
2 2
3 3’
3 0 0

6’
1 5’ 1

4 7 4
7

 For which types of network does sweep localization work?


 Theorem: Shell sweep finitely localizes bilateration networks.
 Theorem: Shell sweep uniquely localizes globally rigid bilateration
networks.
 If G is connected, when run on G2, shell sweep produces all possible
positions for each node. If G2 globally rigid, gives the unique positions.
 Question: How many globally rigid networks are also bilaterations?
48

Shell Sweep on
Random Network
 Typical random graph.
 Starting nodes randomly
chosen.
 Shell sweep uniquely
localizes localizable
portion.
 Also non-uniquely
localizes nodes rigidly
connected to localized
region.
49

Performance
on Large
Network
 500 node graph with
considerable anisotropy
and 4.5 average degree.
 Shell sweep computes
in <5 seconds* with no
intermediate position set
exceeding 128.

* As a JAVA applet on a zoo


node with a dual 2.8GHz CPU
and 2GB RAM
50

Failing Case

 Globally rigid network.


 Connection between
clusters unbridgeable
by bilateration.
51

Extent of Sweep Localization

Sweeps in Random Network Sweeps in Regular Network

 Sweeps localizes more nodes than


trilateration, and almost all localizable nodes!
 In regular networks, sweeps localizes
significantly more nodes than trilateration.
 Most incremental localization algorithms are
trilateration based.
 Key point: Many globally rigid random
geometric graphs are bilateration graphs.
52
Summary of Localization Density
Spectrum
 Localization is NP-hard in general, but there are classes of graphs that
are easy to localize.
 Complete graphs.
 Trilateration graphs.
 Graphs that we know how to localize in worst-case exponential time:
 Doubled biconnected graphs.
 Basic idea: more edges make localization easier.
 Goal: to understand which networks can be localized and which are
problematic.

Consider all possible networks on n sensors

Some networks can be Some networks can be Some networks can be Unlocalizable
localized in O(|V|2) time: localized in O(|V|5): localized in exponential
Trilateration graphs with known Trilateration graphs with unknown time:
ordering ordering Doubled biconnected graphs
Globally rigid bilateration graphs
53

When Does Localization Become Easy?


Dense 1 Easy

Complete Graph

Trilateration Graph 3r1 Polynomial time

Bilateration Graph 2r2 Exponential


Globally Rigid NP-hard

3-connected r3

Sparse 0 Unsolvable
Number of edges Sensing radius in Gn(r) Complexity of
realization
54

Conclusion and Future Work


 Formalized the localization problem and its solvability.
 Showed that the problem is fundamentally
computationally hard.
 Constructively characterized easily localizable
networks.
 Provided algorithm that localizes more nodes than
previous incremental algorithms.

 Next:
 Localization using maps.
 Localization using angular order information.
 Localization in networks of mobile nodes.
 Localization in 3D or on 3D surfaces.
 Full system from deployment to localization.
55

Our Work in the Field


 “Rigidity, Computation, and Randomization in Network Localization” - [Infocom 2004]
 Conditions for unique fine-grained localization.
 Initial computational complexity results.
 “On the Computational Complexity of Sensor Network Localization” - [Algosensors 2004]
 Computational complexity results.
 “A Theory of Network Localization” - [Transactions on Mobile Computing 2006]

 “Graphical Properties of Easily Localizable Sensor Networks” - [under review]


 Characterizing easily localizable ad-hoc networks.
 “Precise Localization in Sparse Sensor Networks” - [Accepted to Mobicom 2006]
 Algorithm for localization in sparse ad-hoc networks.

 “Localization in Partially Localizable Networks” - [Infocom 2005]


 Investigation of partially localizable networks.
 Localizability-aware network deployment.

 “Towards Mobility as a Network Control Primitive” - [Mobihoc 2004]


 Location-aware controlled node-mobility algorithm for sensor network optimization.
56

Acknowledgements
 I would like to thank all my collaborators, without whom this work would not have
been possible.

 Brian D.O. Anderson (Australia National University and NICTA)


 James Aspnes
 P.N. Belhumeur (Columbia University)
 Pascal Bihler
 Ming Cao
 Tolga Eren
 Jia Fang THANK YOU FOR LISTENING
 Arvind Krishnamurthy
ANY QUESTIONS?
 Jie (Archer) Lin
 Wesley Maness
 A. Stephen Morse
 Brad Rosen
 Andreas Savvides
 Walter Whiteley (York University)
 Y. Richard Yang
 Anthony Young

You might also like