1

Wireless Sensor Networks: In Search of Principles

Deborah Estrin

Director, NSF Science and Technology Center for Embedded Networked Sensing (CENS)

Professor, UCLA Computer Science Department

destrin@cs.ucla.edu

http://lecs.cs.ucla.edu/estrin

Contributors: Vlad Bychkovskiy, Alberto Cerpa, Jeremy Elson, Deepak Ganesan, Lew Girod, Ramesh Govindan, John Heidemann,

Bhaskar Krishnamachari, Fabio Silva, Wei Ye and members of CENS, LECS, and IPAM programsSponsors: DARPA, NSF, Intel, Sun, HS-SEAS

2

Roadmap

• Motivation– Driving applications– Need for new systems and algorithms research– Segue…Some structure: stack, taxonomy, load, metrics

• Recently-developed building blocks– Time synchronization– MAC– Adaptive Topology – Data centric routing and In-network processing

• Some emerging principles

3

Embedded Networked Sensing Potential

• Micro-sensors, on-board processing, and wireless interfaces all feasible at very small scale– can monitor

phenomena “up close”

• Will enable spatially and temporally dense environmental monitoring

• Embedded Networked Sensing will reveal previously unobservable phenomena

Seismic Structure response

Contaminant Transport

Marine Microorganisms

Ecosystems, Biocomplexity

4

Enabling Technologies

Embedded Networked

Sensing

Control system w/Small form factorUntethered nodes

ExploitcollaborativeSensing, action

Tightly coupled to physical world

Embed numerous distributed devices to monitor and interact with physical world

Network devices to coordinate and perform higher-level tasks

Exploit spatially and temporally dense, in situ, sensing and actuation

5

“The network is the sensor” (Oakridge National Labs)

Requires robust distributed systems of thousands of physically-embedded, unattended, and often untethered,

devices.

6

From Embedded Sensing to Embedded Control

• Embedded in unattended “control systems”

– Different from traditional Internet, PDA, Mobility applications

– More than control of the sensor network itself

• Critical applications extend beyond sensing to control and actuation

– Transportation, Precision Agriculture, Medical monitoring and drug delivery, Battlefied applications

– Concerns extend beyond traditional networked systems• Usability, Reliability, Safety

• Need systems architecture and an understanding for underlying algorithms to manage interactions– Current system development: one-off, incrementally tuned, stove-

piped

– Serious repercussions for piecemeal uncoordinated design: insufficient longevity, interoperability, safety, robustness, scalability...

7

New Design Themes

• Long-lived systems that can be untethered and unattended – Low-duty cycle operation with bounded latency– Exploit redundancy and heterogeneous tiered systems

• Leverage data processing inside the network– Thousands or millions of operations per second can be done

using energy of sending a bit over 10 or 100 meters (Pottie00)– Exploit computation near data to reduce communication

• Self configuring systems that can be deployed ad hoc– Un-modeled physical world dynamics makes systems appear ad hoc– Measure and adapt to unpredictable environment– Exploit spatial diversity and density of sensor/actuator nodes

• Achieve desired global behavior with adaptive localized algorithms– Cant afford to extract dynamic state information needed for

centralized control

8

Sample Layered Architecture

Resource constraints call for more tightly integrated layers

Can we define anInternet-like architecture for such application-specific systems??

In-network: Application processing, Data aggregation, Query processing

Adaptive topology, Geo-Routing

MAC, Time, Location

Phy: comm, sensing, actuation, SP

User Queries, External Database

Data dissemination, storage, caching

9

• Spatial and Temporal Scale

– Extent– Spatial Density (of

sensors relative to stimulus)

– Data rate of stimulii• Variability

– Ad hoc vs. engineered system structure

– System task variability– Mobility (variability in

space)• Autonomy

– Multiple sensor modalities

– Computational model complexity

• Resource constraints– Energy, BW– Storage, Computation

Systems Taxonomy

• Frequency – spatial and

temporal density of events

• Locality – spatial, temporal

correlation• Mobility

– Rate and pattern

Load/Event Models

Metrics

• Efficiency– System

lifetime/System resources

• Resolution/Fidelity– Detection,

Identification• Latency

– Response time• Robustness

– Vulnerability to node failure and environmental dynamics

• Scalability– Over space and

time

10

ENS Research

• Building blocks for experimental systems

– Fine grained time and location

– Adaptive MAC

– Adaptive topology

– Data centric routing

• Emerging principles…

These examples illustratenew combination ofconstraints and requirements

11

Fine Grained Time and Location(Elson, Girod, et al.)

• Unlike Internet, the location of nodes in time and space is essential for local and collaborative detection

• Fine-grained localization and time synchronization needed to detect events in three space and compare detections across nodes

• GPS provides solution where available (with differential GPS providing finer granularity)

• Acoustic or Ultrasound ranging and multi-lateration algorithms promising for non-GPS contexts (indoors, under foliage…)

• Fine grained time synchronization needed to support ranging and many other sensor network functions

12

Tiered System Design: IPAQs and UCB Motes

• Localization– IPAQs range to each other, create a

coordinate system– Mote periodically emits coded acoustic

“chirps” (511 bits)

– IPAQs listen for chirps (buffer time series - mote can’t do this)

– run matched filter and record time diff btwn emit- and receive-time of coded sequence

– Share ranges with each other via 802.11; trilaterate

• Time sync – Allows computation of acoustic time-of-flight– One IPAQ has a “MoteNIC” to sync mote

and IPAQ domains

14

Reference Broadcast Synchronization

• Distributed system of sensor nodes– Distinct nodes need “inter-node” synchronization

• Uses radio channel to relate local clocks of two nodes• “Multihop” synchronization: composition of time conversions.• Can be done post facto

• Eliminates effect of transmission variation • Receiver latency is low-variance

– Reception of broadcasts are closely correlated in real time– First bit arrives at receivers with small variations (and easy to filter)

010101001

CPU 1 Mote

CPU 2 Mote010101001

Mote Sender

50kb/S (20uS per bit)

15

Fine Grained Multi-hop Time Synch ResultsSnapshot from Running System that achieves

sec time synchronization relative to NTP ms over lossy wireless

Motes

IPAQ CPUsAnd Codecs

Linesannotatedwith offsetachieved betweenconnectedclocks

2 s

9 s

16

• Major sources of energy waste• Idle listening when no sensing events, Collisions, Control

overhead, Overhearing

• Major components in S-MAC• Message passing• Periodic listen and sleep

• Combine benefits of TDMA + contention protocols• Tradeoff latency and fairness for efficiency

Energy Efficient MAC design(Ye et al.)

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0 50 100 150 200 250 300

Ave

rage

Dis

sipa

ted

Ene

rgy

(Jou

les/

Nod

e/R

ecei

ved

Eve

nt)

Network Size

DiffusionDiffusion

Omniscient MulticastOmniscient MulticastFloodingFlooding

00.0020.0040.0060.0080.01

0.0120.0140.0160.018

0 50 100 150 200 250 300

Ave

rage

Dis

sipa

ted

Ene

rgy

(Jou

les/

Nod

e/R

ecei

ved

Eve

nt)

Network Size

DiffusionDiffusion

Omniscient MulticastOmniscient Multicast

FloodingFlooding

Over 802.11-like MAC Over energy-aware MAC

17

Message Passing

• Problem: In-network processing requires entire message

• Solution: Don’t interleave different messages– Long message is fragmented & sent in burst– RTS/CTS reserve medium for entire message– Fragment-level error recovery

— extend Tx time and re-transmit immediately• Other nodes sleep for whole message time

• Tradeoff fairness for energy and single-message level latency

18

Periodic Listen and Sleep

• Problem: Idle listening consumes significant energy• Solution: Periodic listen and sleep policy and mechanism to

coordinate• Turn off radio when sleeping; tradeoff latency for energy• Reduce duty cycle to ~ 10% (200ms on/2s off)• Schedules created using SYNCH

• Prefer neighboring nodes have same schedule for easy broadcast & low control overhead

Border nodes: two schedules requires two broadcasts

Node 1

Node 2

sleeplisten listen sleep

sleeplisten listen sleep

Schedule 2

Schedule 1

19

S-MAC Experimental results(implemented on UCB Mote over RFM radio)

• Topology and measured energy consumption on source nodes

Source 1

Source 2

Sink 1

Sink 2

• Each source node sends 10 messages — Each message has

10 fragments x 40B

• Measure total energy— Data + control +

idle

0 2 4 6 8 10

200

400

600

800

1000

1200

1400

1600

1800Average energy consumption in the source nodes

Message inter-arrival period (second)

Energy consumption (mJ)

802.11-like protocol Overhearing avoidanceS-MAC

Message Inter-arrival period

Energy consumed

20

Adaptive Topology: An example of Self-Organization with Localized Algorithms

• Self-configuration and reconfiguration essential to lifetime of unattended systems in dynamic, constrained energy, environment– Too many devices for manual configuration– Environmental conditions are unpredictable

• Example applications:– Efficient, multi-hop topology formation: node measures

neighborhood to determine participation, duty cycle, and/or power level

– Beacon placement: candidate beacon measures potential reduction in localization error

• Requires large solution space; not seeking unique optimal • Investigating applicability, convergence, role of selective global

information

21

Context for creating a topology: connectivity measurement study (Ganesan et al)

Packet reception over distance has a heavy tail. There is a non-zero probability of receiving packets at distances much greater than the average cell range

169 motes, 13x13 grid, 2 ft spacing, open area, RFM radio, simple CSMA

Can’t justdetermineConnectivityclusters thrugeographic Coordinates…

For the same reason you cant determine coordinates w/connectivity

22

Adaptive Topology Schemes

• Goal: exploit the redundancy in the system (high density) to save

energy while providing a topology that adapts to the application

needs

• Mechanism: empirical adaptation. Each node assesses its

connectivity and adapts participation in multi-hop topology based on the

measured operating region.

• Does not detect partitions, less efficient cases due to lack of global

knowledge

23

Example Performance Results (ASENT)(Cerpa et al., Simulations and Implementation)

Energy Savings (normalized to the Active case, all nodes turn on) as a function of density. ASCENT provides significant amount of energy savings, up to a factor of 5.5 for high density scenarios.

24

Directed Diffusion: Data Centric Routing

• Basic idea– name data (not nodes) with externally relevant attributes

• Data type, time, location of node, SNR, etc– diffuse requests and responses across network using application driven

routing (e.g., geo sensitive or not)– optimize path with gradient-based feedback– support in-network aggregation and processing

• Data sources publish data, Data clients subscribe to data– However, all nodes may play both roles

• A node that aggregates/combines/processes incoming sensor node data becomes a source of new data

• A sensor node that only publishes when a combination of conditions arise, is a client for the triggering event data

– True peer to peer system

• Implementation defines namespace and simple matching rules with filters– Linux (32 bit proc) and TinyOS (8 bit proc) implementations

25

Diffusion as a construct for in-network processing

• Nodes pull, push, and store named data (using tuple space) to create efficient processing points in the network– e.g. duplicate suppression, aggregation, correlation

• Nested queries reduce overhead relative to “edge processing”• Complex queries support

collaborative signal processing– propagate function

describing desired locations/nodes/data (e.g. ellipse for tracking (Zhao et al))

– Interesting analogs to emergingpeer-to-peer architectures

• Build on a data-centric architecturefor queries and storage

26

Nested Query Evaluation(A real experiment w/sub-optimal hardware)

(Heidemann et al.)

• Eevn simple nested queries greatly improve event delivery rate

• Specific results depend on experiment– placement– limited quality MAC

• General result: app-level info needed in sensor nets; diffusion is a good platform

• Concept of Data Centric vs. Address Centric more important than specific implementation

even

ts s

ucce

ssfu

lly r

ecei

ved

(%)

number of light sensors

flat

nested

60

1 2 3 4

80

40

20

27

A more general look at A more general look at Data Centric vs. Address Centric approachData Centric vs. Address Centric approach

(Krishnamachari et al.)(Krishnamachari et al.)

• Address Centric• Distinct paths from each source to sink.

• Data Centric• Support aggregation in the network where paths/trees overlap• Essential difference from traditional IP networking

• Building efficient trees for Data centric model• Aggregation tree: On a general graph if k nodes are sources and one is a sink, the aggregation tree that minimizes the number of transmissions is the minimum Steiner tree. NP-complete….Approximations:

• Center at Nearest Source (CNSDC): All sources send through source nearest to the sink.• Shortest Path Tree (SPTDC): Merge paths.• Greedy Incremental Tree (GITDC): Start with path from sink to nearest source. Successively add next nearest source to the existing tree.

28

Source placement: event-radius modelSource placement: event-radius model

29

Comparison of energy costsComparison of energy costs

Address CentricShortest path data centricGreedy tree data centricNearest source data centricLower Bound

Data centric has many fewer transmissions than does Address Centric; independent on the tree building algorithm.

Opportunism always pays;Greed pays only when things get very crowded

(Intanagowiwat et al. ns-2 more detailed simulations)

31

Programming Paradigm

• Move beyond simple query with in-network aggregation model

• How do we task a 1000+ node dynamic sensor network to conduct complex, long-lived queries and tasks ??– What constructs does the query language need to support?

• What sorts of mechanisms need to be “running in the background” in order to make tasking efficient?– Small databases scattered throughout the network, actively

collecting data of nearby nodes, as well as accepting messages from further away nodes?

– Active messages traveling the network to both train the network and identify anomalous conditions?

• Storage architecture

32

(Still hypothetical) Examples

• Map isotherms and other “contours”, gradients, regions– Record images wherever acoustic signatures indicate significantly

above-average species activity, and return with data on soil and air temperature and chemistry in vicinity of activity.

– Mobilize robotic sample collector to region where soil chemistry and air chemistry have followed a particular temporal pattern and where the region presents different data than neighboring regions.

• Raises requirements for some global context, e.g. “average” levels– Emerging role for distributed storage architecture

• Pattern identification: how much can and should we do in a distributed manner?– Similar to some vision/image analysis problems but distributed

noisy inputs

33

In search of Principles

…All we have thus far are heuristics/design themes

• Exploit density

• Use “localized” algorithms

• Procrastination Pays

34

Exploiting redundancy, density

• Design objectives

– Maximize system lifetime, coverage, accuracy, reliability

– … NOT to minimize nodes deployed

• Spatial and modal diversity can contribute to all objectives, e.g.:

– Adaptive topology/load sharing to increase system lifetime

– Spatial diversity to achieve coverage around obstacles

– Modal diversity to detect outliers in acoustic ranging

– Correlated measurements to calibrate

35

Localized algorithms

• Localized algorithms and in-network processing are mandated by energy

constraints and scale

– Challenge is to characterize and constrain global behavior that

result, e.g., designing for predictability in highly uncertain

environments

• Localized doesn’t mean flat fully-decentralized--Exploit self-configuring

“structure”

– Tiered and clustered systems

– Exploit some centralized resources and information

– Exploit built-in structures in Globally Ad hoc, Locally regular systems

(GALORE)

36

Lazy/Procrastinating/Just-in-time systems

• Post facto coordination

– Time synchronization

– Sensor calibration

• Don’t move a bit until needed: Leave data where it is detected until

needed

– Triggered systems

– Multi-resolution distributed storage architecture, Data centric

storage (DCS, (Ratnasamy (ICSI))

– Not really that simple: When and where is data needed to detect

patterns?

37

Some work in progress

• In network processing mechanisms and data, a few examples.

– Fine grained data collection, methods, tools, analysis, models (D. Muntz (UCLA), G. Pottie (UCLA), J. Reich (PARC))

– Collaborative, multi-modal, processing among clusters of nodes (e.g., F. Zhao (PARC), K. Yao (UCLA)

– Enable lossy to lossless multi-resolution data extraction (Ganesan (UCLA), (Ratnasamy (ICSI))

– Primitives for programming the “sensor network” (Estrin (UCLA), Database perspective: S. Madden (UCB))

– Modeling capacity and capability (M. Francischetti (Caltech), PR Kumar (Ill), M. Potkonjak (UCLA), S. Servetto (Cornell))

38

Towards a Unified Framework for ENS

• General theory of massively distributed systems that interface with the physical world

– low power/untethered systems, scaling, heterogeneity, unattended operation, adaptation to varying environments

• Understanding and designing for the collective

– Local-global (global properties that result…local behaviors that support)

– Programming model for instantiating local behavior and adaptation

– Abstractions and interfaces that do not preclude efficiency

• Cautionary questions

– Will we be able to generalize away from application-specific stove-pipe solutions?

– How to address social concerns about passive monitoring?

39

Pulling it all together

Collaborative Signal

Processing and Active

Databases

Adaptive Self-Configuration

Sensor Coordinated

Actuation

Environmental Microsensors

CENS Core Research Academic Disciplines

NetworkingCommunications

Signal ProcessingDatabases

Embedded SystemsControls

Optimization…

BiologyGeology

BiochemistryStructural Engineering

EducationEnvironmental Engineering

NetworkingCommunications

Signal ProcessingDatabases

Embedded SystemsControls

Optimization…

BiologyGeology

BiochemistryStructural Engineering

EducationEnvironmental Engineering

40

Follow up

• Embedded Everywhere: A Research Agenda for Networked Systems of Embedded Computers, Computer Science and Telecommunications Board, National Research Council - Washington, D.C., http://www.cstb.org/

• Related projects at UCLA and USC-ISI• http://cens.ucla.edu• http://lecs.cs.ucla.edu• http://rfab.cs.ucla.edu• http://www.isi.edu/scadds

• Many other emerging, active research programs, e.g., • UCB: Culler, Hellerstein, BWRC, Sensorwebs, CITRIS• MIT: Balakrishnan, Chandrakasan, Morris• Cornell: Gehrke, Wicker• Univ Washington: Boriello• Wisconsin: Ramanathan, Sayeed• UCSD: Cal-IT2

• DARPA Programs• http://dtsn.darpa.mil/ixo/sensit.asp• http://www.darpa.mil/ito/research/nest/