1-1 cmpe 259 sensor networks katia obraczka winter 2005 deployment, organization, localization

1-1

CMPE 259 Sensor Networks

Katia Obraczka

Winter 2005

Deployment, Organization, Localization

1-2

Announcements

Homework due on 02.14. Submission: e-mail to katia, cintia,

kumarv@soe. Plain text or pdf.

Final project presentations. March 15th. from 4-7pm.

Venkatesh Rajendra’s MAC presentation. Wed, Feb 16th.

Homework 2. Exam.

1-3

Node localization

1-4

Node Localization

For some sensor network applications, exact location is critical. Tracking. Monitoring.

For most applications, having location information enhance value of information.

Also needed in geographic routing.

1-5

How to determine node location?How to determine node location? May be trivially available if:

Satellite based GPS is feasible and available on all nodes.

All nodes are hand-placed and pre-configured with location coordinates.

Otherwise it is quite challenging (even with a fraction of known reference/beacon nodes).

Typically, one assumes some nodes have position information (e.g., through GPS), but not all.

1-6

Basic Localization ApproachesBasic Localization Approaches Proximity:

Near/far. Connectivity information.

Lateration/ranging (based on distance estimate): Received signal strength. Time difference of arrival. Time of arrival.

Angulation. Location service.

1-7

Ad Hoc Positioning System

1-8

APS

Provides (approximate) location information for all nodes given location information from subset of nodes.

Positioning mechanism requirements: Distributed. Energy-efficient.

• Minimize communication and processing. Robustness.

• In the face of partitions.

1-9

APS basics

Employ same principles as GPS for computing positions. I.e., triangulation.

Landmarks: nodes that know their position.

Distance to landmarks propagate hop-by-hop. Distance vector approach. Once node has distance to 3 landmarks, it

can compute its position.

1-10

Hop-by-hop propagation

DV-hop. DV-distance. Euclidean.

1-11

DV-Hop

Nodes get distance in hops to landmarks. Landmarks compute distance of a hop.

Landmarks get distance to other landmarks in hops. Landmarks know euclidean distance to other

landmarks.

Landmark broadcasts hop distance. Controlled flooding. Once nodes gets and forwards hop distance, it will

drop subsequent ones.

Nodes use triangulation to compute their position based on the position of the landmarks.

1-12

DV-distance

Distances measured using received signal strength.

1-13

Euclidean propagation

Euclidean distance to landmark is propagated.

Node needs at least 2 neighbors with known estimates to landmark.

Also need distance from node to these neighbors and distance between neighbors.

1-14

Evaluation

Simulations using ns-2. 100 nodes. 2 topologies: isotropic and anisotropic. Metrics:

Location error. Coverage. Overhead.

Use of APS-estimated locations in routing.

1-15

Results

From paper… DV-based perform relatively well with

low overhead. Euclidean-based exhibits better

accuracy at the expense of signaling.

1-16

Organization

1-17

Tracking

Given a sensor network, use the sensors to determine the motion of one or more targets

Typically requires more cooperation among entities than other examples we have seen Compare: “is there an elephant out there?” vs.

“where has that particular elephant been?”

1-18

Tracking challenges Data dissemination and storage Resource allocation and control Operating under uncertainty Real-time constraints Data fusion (measurement interpretation)

Multiple target disambiguation Track modeling, continuity and prediction Target identification and classification

1-19

Tracking domains

Appropriate strategy depends on the sensors’ capabilities, domain goals and environment Requires multiple measurements? Bounded communication? Target movement characteristics? No single solution for all problems

For example… Limited bandwidth encourages local processing Limited sensors requires cooperation

1-20

Why not centralized?

Scale! Data processing combinatorics Resource bottleneck (communication,

processing) Single point of failure Ignores benefits of locality

1-21

Why not (fully) distributed?(i.e. everyone tracks)

Redundant information and computation Can increase uncertainty Lack of unified view High communication costs

(exception: overhearing [Fitzpatrick 2003])

1-22

Organization-based tracking

Use structure, roles to control data and action flow

Can be static, or dynamically evolved [Brooks 2003]: Spontaneous coalition formation [Horling 2003]: Partitions, mediated clustering [Li 2002]: Hierarchical information fusion [Yadgar 2003]: Hierarchical teams [Wang 2003]: Roles and group formation [Zhao 2002]:

Geographic groups

1-23

Using and Maintaining Organization in a Large-Scale Sensor Network

1-24

Problem Domain

Fixed doppler radars Requires multiple,

coordinated measurements

Multiple targets Shared 8-channel RF

communication

1-25

Sensor Characteristics

Hardware Fixed location,

orientation Three 120° radar

heads Agent controller

Doppler radar Amplitude and

frequency data One (asynchronous)

measurement at a time

1-27

Organizational Control

Use organization to address scaling issues Environment is partitioned

Constrains information propagation Reduces information load Exploits locality

Agents take on one or more roles Limits sources of information Facilitates data retrieval

Other techniques also built into negotiation protocol and individual role behaviors

1-29

Typical Node Layout

•Nodes are arranged or scattered, and have varied orientations.•One agent is assigned to each node.

1-30

Partitioning of nodes

•The environment is first partitioned into sectors.•Sector managers are then assigned.

1-31

Sector manager

Generate scanning plans. Assign track managers. Keep local sensor information.

1-32

Distribution of scan schedule

•Sector members send their capabilities to their managers.•Each manager then generates and disseminates a scan schedule.

1-33

Track Manager Selection

•Nodes in the scan schedule perform scanning actions.•Detections reported to manager, and a track manager selected.

1-34

Track manager

Organizes tracking task. Discovers sensors capable of tracking

target. Determines track schedule.

When to perform scan. Fidelity, timeliness, etc.

1-35

Managing limited resources

•Track manager discovers and coordinates with tracking nodes.•New tracking tasks may conflict with existing tasks at the node.

1-36

Data fusion (track generation)

•Tracking data sent to an agent which performs the fusion.•Results sent back to track manager for path prediction.

1-38

Sector size Sector manager load.

Smaller sector –› smaller manager directory. Larger sector –› better sector coverage.

Track manager actions. Smaller sector –› fewer update messages. Larger sector –› fewer directory queries.

Depends on sensor density, sensor range, target speed, etc.

Empirical evaluation of how sector size affects performance.

1-39

Experimental setup

Radsim simulator 36 sensors 1-36 equal sized

sectors 4 mobile targets 10 runs per

configuration

Hypothesis: sector size of 6-10 agents is best

1-40

Communication characteristics

Larger sectors with more agents leads to less messaging overall.

Agents per Sector0 5 10 15 20 25 30 35 40

0

1

2

3

4

5

1-41

Load disparity

Large sectors increase SM comm. Load.

Greater disparity in activity load.

Agents per Sector0 5 10 15 20 25 30 35 40

0

50

100

150

200

250

300

350

1-42

Domain metrics

Communication distance increases with larger sectors Track migration

triggered by boundaries

…but better accuracy. More measurements

due to lower control overhead

Agents per Sector0 5 10 15 20 25 30 35 40

0

10

20

30

40

50

60

Agents per Sector0 5 10 15 20 25 30 35 40

0

1

2

3

4

5

1-43

What’s best?

Find inflection point in graphs’ intersection

Empirical evidence supports sector size from 5-10 sensors

This would vary, depending on sensor and environmental characteristics

Agents per Sector0 5 10 15 20 25 30 35 40

0

10

20

30

40

50

60

Agents per Sector0 5 10 15 20 25 30 35 40

0

1

2

3

4

5

Agents per Sector0 5 10 15 20 25 30 35 40

0

1

2

3

4

5

1-44

Conclusions

Specific results are domain-specific. However, this demonstrates that

organizational controls can affect performance.

1-45

Deployment

1-46

DeploymentDeploymentDepend application.Two main classes:

Structured placement.Random deployment.

In both classes the two main goals are network connectivity and sensor coverage.

Costs have to do primarily with equipment and energy.

1-47

Network connectivityNetwork connectivity

Idealized* geometric model for wireless links: perfect connectivity within radio range R.

Network graph G formed by nodes as vertices and these links as edges.

Basic notion of connectivity: there exists at least one multihop path between any pair of nodes in the network; could be generalized to k-connectivity, existence of Hamiltonian cycle, etc.

*Caveat: Perhaps good for preliminary analysis, but known to be unrealistic

R

1-48

Random deploymentRandom deployment

E.g., scattered from an aircraft/robot, mixed into concrete.

Issues of average density and range settings are important.

Connectivity issues can be explored using the Theory of Random Graphs and Percolation Theory.

1-49

Random graphsRandom graphs Bernoulli Random Graphs G(n,p): edge between any pair of

the n nodes independently with probability p. Geometric Random Graphs G(n,R): n nodes placed with a

uniform random distribution in a finite region; edge between any pair of nodes within range R.

1-50

Sensor coverageSensor coverage Application specific. Some possible notions of coverage:

Density of placement (average/max distance between nodes).

Percentage of desired (known a-priori) measurement points covered.

Percentage of the operational area that is covered with “k” sensors: k-coverage.

Others?

1-51

Sensor Placement for Effective Coverage

1-52

Goals

Minimum number of sensors for adequate coverage. Adequate coverage: every grid point is

covered with minimum confidence level.

1-53

Sensor placement

Greedy approach. Place one sensor at a time. Algorithm terminates when:

No more sensors or Sufficient coverage achieved.

1-54

Approaches

Place sensor to reduce miss probability. Maximize average coverage.

Place sensor to increase coverage. Maximize coverage of grid point that is

covered the least. Relax grid structure.

1-55

Evaluation

Simulation of case studies. Main metric: number of sensors.

1-56

Results

From paper…