a survey on sensor networks ian f. akyildiz, weillian su, yogesh sankarasubramaniam, erdal cayirci...
TRANSCRIPT
A Survey on Sensor Networks
Ian F. Akyildiz, Weillian Su, Yogesh Sankarasubramaniam, Erdal Cayirci
Presented by Alexandra Czarlinska
IEEE Communications Magazine, August 2002
Goals and Structure of this Paper
Survey Communication
Protocols for the 5 Layers
Special
constraints of
Sensor NetworksCommunication
Topology
Communication in Sensor Networks
Do these protocols work for Sensors? Encourage
new research
Analysis and
Conclusions
Introduction
Communication Topology
Sink / Base Station
Task Manager Node
Internet or Satellite
Self-organizing, non-homogenous Sensor Network
End User
Multi-hop wireless
Cluster-Head or Aggregator
Density of nodes μ(R) = N πR2/A
N = # of nodes in area AR is radio range
area A
Tiny sensor node with 7 sensing devices ; Photoresistor, Temperature, Barometric pressure, Barometric Pressure and Temperature, Humidity, Thermopile and Thermistor
UC- Berkeley mote
Inside a Sensor Node
Location Finder Mobilizer
Power Unit
Sensor ADCProcessor
StorageTX/RX
Actuator
Power Generator
Special Constraints for Communication in Sensor Networks
Fault Tolerance
Handle loss of nodes
ScalabilityHandle high density of nodes
CostsNodes die, make them low cost
Hardware LimitationsNodes are tiny
Changing Topology
Nodes moving, new nodes, loss of nodes
Hostile Environment
Survive and maintain communication
Transmission Media
wireless: RF, optical, infrared
PowerLimited Tx, computation, and lifetime
Security ?Security ?Confidentiality, Authentication
etc
Protocol Stack and Sensor Network Management
5. Application Layer5. Application Layer
4. Transport Layer 4. Transport Layer
3. Network Layer3. Network Layer
2. Data Link Layer 2. Data Link Layer
1. Layer Physical1. Layer Physical
Po
wer
Po
wer
Mo
ving
Mo
ving
Co
llabo
ration
C
ollab
oratio
n
Sensor Network Manage-ment
1. Physical Layer
Responsible for frequency selection, modulation and dataEncryption. The big issue here is Power!
encryption
???modulation
freq 915 MHz?
0 1 1 0 1 0 1 . . .
source
Transmit
Sink
d distance
Low-lying antenna
Power to transmit
≈ dn where 2 ≤ n ≤ 4
High antenna
1. Physical Layer
• More research on smaller hardware that uses less power (battery!)
• Protocols that exploit node density and multi-hop redundancy
Have/Know
Need
• Binary Modulation needs less energy than M-ary modulation
• Direct Sequence Spread Spectrum is low power (transmission and security)
•Ultra wideband (UWB) uses low power, does well under multi-path and has simple Tx/Rx
2. Data Link Layer
Responsible for multiplexing of data streams, MediumAccess control (MAC) and Error Control
Problem: MAC ensures QoS and Bandwidth, not Power conservation, central base
Cellular MAC
Base Basewired
wireless
Problem: MAC ensures QoS under mobility. Not enough nodes, Tx power needed is too much, central base
Bluetooth and MANET
Master
slave
2. Data Link Layer
MAC Protocol
Method
SMACS and EAR
Fixed allocation of duplex time slots at fixed freq.
Hybrid TDMA/ FDMA
Sensors
Centralized freq. and time division
CSMA-based for Sensors
Contention-based random access
Power Conservation
Random wake up during setup and turning radio off while idle. Exploits large bandwidth available compared to sensor data rate.
Min energy use in Hardware. Picks the mix of TDMA/FDMA to min energy use
Uses constant “listening time” to min energy.
What we still need to improve
Not great for very mobile networks (good for mostly stationary nodes).
Should use more computation over handshaking. Use more Power-saving modes (sleeping). Need to derive bounds on energy needed by sensors.
3. Network Layer
Scheme Description
Flooding Broadcasts data to all neighbor nodes
Gossiping Sends data to one randomly selected neighbor
LEACH Forms a clusters to minimize energy loss
SPINSends data to sensor nodes only if they are “interested”, has 3 types of messages (ADV, REQ, DATA)
Directed Diffusion
Sets up gradients for data to flow from source to sink during interest dissemination
Have/Know
Also have: Pegasis, GEAR
Routes data supplied by the Transport Layer
3. Network Layer
Scheme Description
Power Efficiency Routing
Pick a route based on: max Power Available (PA) or, min Energy (ME), or Min Hop (MH) or Max Min PA.
SMECNCreates a sub-graph of the sensor network that contains the minimum energy path
SARCreates multiple trees where the root of each tree is one hop neighbor from the sink
Have/Know
• allow higher topology changes• Allow higher scalability
Need
3. Network LayerPower Efficiency Routing
1- Max Power Available (PA) route, 2- Min Energy (ME) route, 3- Min Hop (MH) route and 4 - Max Min PA route
A1- Max Power Available route
P = 2
P = 2
P = 2B
P = 3
P = 1
P = 4
Total P = 6
4 -Max Min
Route along which the min PA is larger than the min PA of other routes
4. Transport Layer
Helps to maintain the flow of data if the Application Layerrequires it. Needed if End-User accesses the SensorNetwork through the Internet
Have/Know
Current TCP:
• has a window mechanism that may not suit Sensor Networks
• uses end-to-end Global Addressing (nodes may have Attribute-based or Location-based addressing)
• TCP and UDP not based on power conservation and scalability
4. Transport Layer
More research is needed to see if we need new protocols and what they might be. Suggestion by authors:
Need
TCP Splitting: TCP on one side, modified UDP on the other
UDPSink
Task Manager Node
Internet
TCP
UDP
End User
5. Application Layer
Task Manager Node
Internet Sink
End User
Makes the hardware and software of the lower layerstransparent to the Task Manager Node (and End-User)
5. Application Layer
SMPSMP TADAPTADAP SQDDPSQDDP
Sensor Management Protocol
Rules for Data Aggregation, Time sync.
Moving, turning nodes on/off
Task Assignment and Data Advertisement
Protocols
Sensor Query and Data Dissemination Protocol
Interest Dissemination from:
- user to nodes- nodes to user
Issue queries, collect replies
Allows for Attribute-based and Location-
based addressing
5. Application Layer
Attribute-Based Addressing:
The locations of nodes that sense temperature higher than 70 degrees
Location-Based Addressing:
What are the temperatures read by sensors in region A
In general, more research is needed in all the areas mentioned
Need
Data Centric Routing (not Address Based Routing):
Strengths and Contributions
Great overview of what exists, why it’s good or not good for Sensor Networks
Identifies and motivates new areas of research that are needed
Identifies certain general “design principles”
Gives a table of on-going Sensor Network research projects
“A Survey on Sensor Networks”
Analysis: Weaker side
• Does not mention security enough (should be designed right into the system, not after)
• Does not mention Asymmetric Links
• Does not mention Distributed Protocols sufficiently (could have no Base Station)
Keep in Mind
• Paper assumes that nodes do not have much energy but there is new research in:
– solar cells, temperature gradients, vibrations, RF Transfer, Micro Heat Engines (using MEMS)
• Assumes high density of cheap nodes vs. a few high quality nodes (keep in mind NASA)
• Does not always assume the most General Topology (ie: there could be no Base Station, there could be multiple sensing devices on one node etc)
References
• I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, E. Cyirci, “ A survey on Sensor Networks”, Computer Networks, 38(4):393-422, March 2002
• http://www.greatduckisland.net
• Chee-Yong Chong, S. P. Kumar, “Sensor networks: evolution, opportunities, and challenges”, Proceedings of IEEE, pp 1247-1256, August 2003
Appendix
Additional slides with more detail
Sensor Network Communication Architecture
area A
Fault Tolerance rk(t) = e – λk t
λ k is failure rate of node k
t is time period
Density of nodes μ(R) = N πR2/A
N = number of nodes in area A
R is radio range
Ian F. Akyildiz
Yogesh
Sankarasu-
bramaniam
Weilian Su
Erdal
Cayirci
Authors of this Survey
Georgia Institute of Technology
Prof: wireless and satellite networks, next-gen Internet
Phd: Timing recovery, ad hoc routing and Sensor Networks
Phd: Sensor Networks and next-generation wireless
Sensor Networks, mobile comm., tactical and military
School of Electrical and Computer Engineering