Title: Localized Routing and Coordination in Wireless Sensor Networks
1Localized Routing and Coordination in Wireless
Sensor Networks
Tutorial ASI WSN Hong Kong, Dec. 7, 2006 pm
Ivan Stojmenovic Ivan_at_site.uottawa.ca www.site.uot
tawa.ca/ivan
2www.site.uottawa.ca/ivan/adhoc.html
- T0 Introduction (42 slides no time)
- T1 Routing (60 slides, 2pm-3pm)
- T2 Coordination Topology control for sensor area
and communication coverage (36 slides,
330pm-430pm) - T3 Data Gathering (54 slides) Design guidelines
for reporting, anycasting, multicasting,
location service(no time) - T4 Broadcasting (48 slides, no time Dec. 11 at
HK Polytechnic Univ.) - T5 Construction and application of sparse
connected wireless ad hoc and sensor networks
(57 slides no time Dec. 12, 230pm at HK
Baptist Univ.) - T6 Partitioning and connectivity (18 slides, no
time) - T7 Challenging issues to the applications of
large scale sensor networks (16 slides, no
time)
3Introduction
- Applications
- physical properties
- Research problems in sensor networks
- MAC, TCP, Data centric operations
- Network layer data communication and
coordination - Localized algorithms
- Simulations
- How to write research articles
4Sensor may measure
- Distance, Direction, Speed
- Humidity, Soil makeup
- Temperature, Chemicals
- Light, Vibrations, Motion
- Seismic data, Acoustic data
- strain, torque, load, pressure
- Self-configure into wireless multi-hop network
5Sensors Physical layer
- Sensing hardware
- Processor CPU (lt20 Mhz),
- Memory lt 4KB RAM
- (low) Power supply
- Transceiver (lt20 kbps)
- Receivers (0/1/2)
- Low-cost
- Miniature cm
- Multi-functional
- Hundreds/thousands sensors spread
GPS receiver ?? 7mm x 7mm x 2mm accuracy ?
6Physical layer continued
- Wireless communication
- RF noise and multipath fading causes severe
packet loss - Easy to eavesdrop and to launch spoofing or
Denial-of-Service attacks - Mobility requires rerouting of packets
- Infrared and Optical line of sight alternatives
- Physical risk - nodes are defective, lost,
damaged, compromised, or expired - Limited bandwidth and power (battery)
- One-to-all communication by omni-directional
antennas
7Micro Sensor Hardware Platform
- Berkeley/Crossbow MICA2 Mote
- 250 (MICA2 sensor board)
- www.xbow.com
- ATMEGA microcontroller
- 4 KB RAM, 4 KB EEPROM, 128 KB Flash
- CPU 8 MHZ
- Chipcon CC1000 radio, lt20 kbps
- Multifrequency
- Two AA Batteries
- 1 Duty Cycle can extend lifetime from 5 days to
3-6 months - Extension boards for sensors
8More on physical layer
- Routing tables size ? Data replication ?
- Security ? Processing time ?
- Data compression, error control
- Same frequency ? two frequencies ? Frequency
hopping ? - Sensors in active state spend considerably more
energy than sensors in sleep state - Reliability of individual sensor measurement
? Reliability of sensor network measurement? - There are many types of sensors !
9Journals on sensor networks
- Several journals including mobile computing
- Ad Hoc Networks (Elsevier), Akyildiz, 2003
- ACM Transactions on Sensor Networks 2005
- Int.J.Distributed Sensor Networks, Iyengar,(TF)
2005 - Int.J. Sensor Networks, Y. Xiao, InderScience,
2006. - Ad Hoc Sensor Wireless Networks, An Int. J.
Stojmenovic, OCP, 2005. - Handbook of Sensor Networks Algorithms and
Architectures (Stojmenovic, ed.), Wiley, 2005. - The number of researchers, publications, journals
and conferences grows exponentially since 2000
10Research problems
- Physical,MAC,Network,TCP,Application layers, and
cross-layering - Energy scavenging (chapter)
- Authentication, key management, security
(chapter) - Operating systems (chapter), databases
- Path exposure, target location, classification,
tracking (chapter) - Data gathering and fusion (chapter)
- Localization (position determination) (chapter)
- Time synchronization and calibration (chapter)
is global synchronization needed? Local
synchronization (like temperature across a
country)? - Chapters in Handbook of Sensor Networks
Algorithms and Architectures (Stojmenovic, ed.),
Wiley, 2005.
11Physical layer issues
- Distributed signal processing (chapter)
- All models are unrealistic (easy to criticize),
but some of them are useful - Unit disk graph? two nodes can communicate if
and only if the distance between them is at most
R (transmission range) - Sensing model ? Event can be sensed if and only
if it is up to distance S from sensor (sensing
range)? - Fixed or variable R/S ranges ?
- (Omni)directional antenna? One or more channels?
- More realistic physical layer ? Indoor-outdoor?
Air/underwater? Position information? - Are sensors static or could be mobile?
- One or more sinks? Static or mobile?
12Medium access for sensors
- Ideal MAC (for initial simulations), or
- IEEE 802.15.4 Zigbee standard for low rate
wireless networks (chapter) - Two network topologies are allowed by the
standard, both relying on the presence of central
coordinator. - peer-to-peer topology sensors may
communicate directly, - star-shaped topology they must communicate
through coordinator. - Sensors are time synchronized,
- and follow a joint sleep-active schedule.
- they are active at the same time, followed by
longer sleep periods. At the beginning of active
periods, they compete for upcoming slots to send
messages
13Embedded Sensor Lightweight Operating Systems
- Embedded OS memory footprint too large, not
truly open source, too limited - Berkeleys TinyOS
- Program in NesC
- Event-driven OS run-to-completion
- Run-to-completion operating systems are very
small, simple, and efficient, but because most of
the scheduling and synchronization burden is
pushed to the individual tasks, they are only
applicable to very simple uses. NRC Report - Mote sensor board programming board
Simple multithreading (MANTIS OS, open source)
14Transport layer in sensornets
- Reduced traffic ? reduced congestion
- error rate is increased due to MAC problems,
disconnection is possible due to mobility or
power failure ? Wireless TCP ? TCP - Traditional end-to-end reliability does not apply
- Acknowledgements are power consuming
- Buffering abilities limited
- QoS issues are of different type reliability of
little information rather than quantity/delay
15Reliability
- Multiple correlated data flow from event to sink
- Spatial correlation among data
- Several reports arrive at sink, or
- Several reports are combined at intermediate
nodes to reduce communication (data fusion) - Collective reliability
- Transport problem configure the reporting rate
to achieve the required event detection
reliability at the sink with minimum resource
utilization.
16Basic scenario for area monitoringone sink,
thousands sensors
event
17Deployment strategies
- Embedded sensors to objects in factories
- Deterministic placement by humans, robots
- Randomized placement by plane, artillery, by
humans or robots - Initial deployment and redeployments
18Query types
- Event-driven when sensor decides that it has
something to report (e.g. high temperature) - On-demand by request from monitoring station
- On-demand whole sensing region, or
- On-demand geocasting region (only sensors inside
a geographic region to report) - On-demand multicasting regions (sensors inside
few regions)
19Addressing queries
- Address centric query to an individual node
(e.g. IP routing) - Data-centric addressing query to a geographic
region (position information essential) - Sensors may not have IDs to reduce overhead
- Route based on contents of data
- Send me all sensor readings with Temp gt 40 C
(data-driven routing)
20Data fusion in sensor networks
- New datum important to forward?
- Combine it with other received data
- Minimize of bits to forward
- Coding
- Reliability of new datum sensing distance and
malfunctioning ?
21Data aggregation
- Some sensors may aggregate data by doing some
computation average, summation, highest etc - Collaborative signal processing fuse data
from multiple sensors - Sensors may divide jobs some are sensing and
forwarding, some are receiving and
forwarding some are aggregating data and
forwarding
22Network layer Data communication
- Routing find a path from a source to
destination - Broadcasting send from source to all nodes
data dissemination - Multicasting send from source to several nodes
- Geocasting send a packet from source to all
nodes inside a region - Design guidelines for efficient data
communication
23Network layer coordination
- Sensor area coverage decide which sensors should
sleep and which ones should be active - Backbone creation which active sensors should be
in backbone for more efficient routing,
broadcasting etc. - backbones clusters, connected dominating sets
- Choose links, transmission radii, for desired
protocol operation
24Localized algorithms
- Scalability algorithms work well (or still work)
on ad hoc networks with large number of nodes - Globalized algorithms global network information
or global structure required (e.g. for shortest
path) - Localized algorithms Decisions made based only
on information from neighbors and natural
additional information (e.g. destination for
routing) - Local localized Maintenance remains local
- Quazi-local localized Local changes may trigger
global updates - Mobility or changes between active and sleep
periods in ad hoc and sensor networks require
localized algorithms, preferably local localized
25Memorization and message count ?
- Avoid/reduce memorization at nodes, because that
node may not be active or at expected place when
the stored information is needed - Number of messages between neighbors few ?
O(degree) ? More ? - Number of messages between neighbors to run a
protocol should be very limited (e.g. under
five), possibly even zero after hello messages
for backbone construction, since - Bandwidth and power are limited, and
- Impact of realistic physical layer unreliable
receptions
26Community differences
- Engineers dominated by 2000 with almost full
control - Computer scientists algorithms are more
important than simulations simplified
simulations to extract major properties if
routing A is better than routing B on a single
routing task with ideal MAC layer and home-made
simulator, it is highly likely that 1000 routing
tasks A will be better than 1000 routing tasks B
with 802.11 and NS-2, Glomosym.. - Mathematicians very good theoretical results,
with lot of theorems/proofs, not much practical
relevance, not much support to computer
scientists either (your algorithm is too
simple, there are no theorems..) worst case
performance more important than average case
27Simulations and writing traps to avoid
- I am best approach compare with something
expected to be worse, ignore existing better
solutions - Use simulator that was used by others hide many
detail, but add transport and medium access layer
immediately to network layer - Literature review incomplete
- Simulation diagrams to be impressive, choose
parameter values showing good performance - Algorithmic description incomplete, vague, or
given by pseudo-code - Discuss the impact of ten parameters at once, not
isolating one and making conclusions on it before
introducing the next one. - Assumptions are mostly realistic, but often
forgetting to measure something, e.g.
communication overhead for mobile ad hoc networks
when applying centralized solution
28How to write research articles
- See manuscript at www.site.uottawa.ca/ivan
- Repeat contribution in four sizes title,
abstract, introduction, full text, as if each was
stand-alone - Problem statement, including assumptions and
limitations - Existing solutions why they are (not) competing
- New solution(s)
- Comparison with competing solutions analytical,
simulation - Advantages and drawbacks
- References and Conclusions
29Broadcasting Data dissemination
- Blind flooding each node retransmits when first
copy received, ignores other copies - For dense networks, blind flooding is unreliable
at MAC layer contentions, collisions,
redundancy, increased latency - Blind flooding works well on area dominating set,
when sensing range communication range, since
it is not dense ?20 packets sent without need - Sensing range lt comm. range ? blind flooding
becomes less efficient - Improved solution connected dominating sets
(CDS) - DSeach node is either in DS or has a neighbor in
DS - Dozens solutions recently proposed
30Wireless sensor and actor networks
Actors active nodes, higher energy and
computation, action possible, may be mobile
Task Manager Node
Sink
Sensor
Actor
31Coordinated actuator movement- move/place
sensors to improve area coverage- move to help
sensors determine positions- move to create
fault tolerant network
32Data communication in sensor actuator networks
- Sensor-sensor, sensor-actuator,
actuator-actuator data communication - Anycasting sensor to report to any actuator
- Anycasting with guaranteed delivery
- Multicasting sending report to certain actuators
- Routing, broadcasting minimal hop count, minimal
power consumption, bounded delay - Controlled mobility to improve data communication
- Currently hot subject with small number of
seed articles
33Info at www.site.uottawa.ca/ivan
- articles book chapters, papers
- WWASN June 2007 (General co-chair) Dec. 15
(deadline) - IEEE MASS Oct. 2007 (Program co-chair) March 31
(deadline) - Four books edited
- EIC for journals AHSWN, IJPEDS
- Advice for writing theses and papers
34(No Transcript)
35Who is the father of radio broadcasting and
wireless communications ?
36Nikola Tesla 1856-1943
The Father of wireless communication
- The Serbian-American
- inventor,
- electrical engineer,
- and scientist
- www.teslasociety.com
37Radio/wireless transmission
- was invented by Nikola Tesla in 1893 - Tesla
demonstrated it to the public in 1898 in New
York - Radio remote controlled submarine by
logic gate and three separate radio frequencies
38The US Supreme Court awarded the patent for
radio communications to Tesla in 1945, taking it
away from Marconi
The "World's radio station, Long Island, build
in 1900 for remote wireless control transmission
throughout the world news, music, photographs
and even electricity! However, that great plan
could not be carried out because Tesla did not
have sufficient resources finish it.
39Tesla the genius who lit the world
- Transformers for long distance transfers of
alternate current electricity - Polyphase motors to use the current
- Built the worlds first hydroelectric plant at
Niagara Falls, 1895
Winning over Edisons direct current proposal
40Teslas other inventions
- Tesla experimented and published on x rays in
1895-6, independently in parallel with Roentgen - Neon and fluorescent lighting
- Wireless lighting
- NASAs robot on Mars, Shanghais fast train etc.
based on Teslas inventions
41Lost inventions
- When Nikola Tesla announced he was working on the
transmission of free wireless electric energy, - His laboratory was destroyed.
- And he lived in poverty
- Among other achievements
- Made a ship invisible to a live audience!
- Worked on time/space object transfer (USA army
project including Einstein) - His documentation is still classified by the US
government
After his withdrawal from the army project, there
is evidence that he was assassinated
42 Celebrating the 150th anniversary of Nikola
Tesla's birth in 2006