Routing Protocols
for
Sensor Networks
Agenda
General Properties
Architectures and Requirements
Routing Protocols Classification
10 Suggested Routing Protocols:
LEACH DD
PEGASIS MCF
TEEN TTDD
APTEEN RW
SPIN RR
Acknowledgements
E. Magistretti (U. Bologna Italy)
J. Kulik (MIT; BBN Co.)
R. R. Choudhury, P. Kyasanur & N. Vaidya (UIUC)
P. Desai (UFL)
D. Braginsky and D. Estrin (UCLA)
S. Hazarika, W. Chen, Y. Gong & X. Liu (UMASS)
T. Kwon & Mjnam (SNU Korea)
R. Peterson & D. Rus (Dartmouth C.)
H.C. Chung, K. Ghoshal & J. Krishna (TAMU)
C. Tavoularis (Cornell )
G. Dong (Virginia U.)
WSN
Dartmouth College
Concepts
Application: Military
From UMASS
Environmental
From UMASS
Future Health
Circulatory Net
Agenda
General Properties
Architectures and Requirements
Routing Protocols Classification
10 Suggested Routing Protocols:
LEACH DD
PEGASIS MCF
TEEN TTDD
APTEEN RW
SPIN RR
General Properties (1)
Mainly for Information Collection
Single Owner
Up to Hundreds of Thousands of Nodes
Disposable Nodes
Cheap Nodes
Security Concerns
General Properties (2)
Bounded Directed Stream (from/to Sink)
Somewhat Limited Computation Capability
Limited Communication Capability
Limited Power Resources
Node may not have Unique ID
Common case - Stationary Nodes
Agenda
General Properties
Architectures and Requirements
Routing Protocols Classification
10 Suggested Routing Protocols:
LEACH DD
PEGASIS MCF
TEEN TTDD
APTEEN RW
SPIN RR
General Architecture (1)
Sensor Network Node Main Components
Sensor Unit
ADC – Analog Digital Converter
CPU – Central Processing Unit
Power Unit
Communication Unit
General Architecture (2)
General Requirements (1)
Varying Network Size
Inexpensive Nodes Equipment
Long Lifetime (Power)
Load-Balancing
Self-Organization
Re-tasking and Querying Capability
General Requirements (2)
Sensible Data Aggregation
Consolidation of Redundant Data
Application Awareness
Tradeoff
Communication for Computation
Possible Mobility
Agenda
General Properties
Architectures and Requirements
Routing Protocols Classification
10 Suggested Routing Protocols:
LEACH DD
PEGASIS MCF
TEEN TTDD
APTEEN RW
SPIN RR
Protocol Classification (1)
Proactive –
First Compute all Routes;
Then Route
Reactive –
Compute Routes On-Demand
Hybrid –
First Compute all Routes;
Then Improve While Routing
Protocol Classification (2)
Direct
–
Node and Sink Communicate Directly
(Fast Drainage; Small Scale)
Flat(Equal) –
Random Indirect Route
(Fast Drainage Around Sink; Medium Scale)
Clustering(Hierarchical) –
Route Thru Distinguished Nodes
Protocol Classification (3)
Location Aware –
Nodes knows where they are
Location-Less –
Nodes location is unimportant
Mobility Aware –
Nodes may move –
Sources; Sinks; All
Protocol Classification (4)
Unicast –
One-to-One Message Passing
Multicast (actually Local Broadcast) –
Node-to-Neighbors Message Passing
Broadcast –
Full-Mesh – Source to Everyone
Protocol Classification (5)
Query Models:
Historical Queries: Analysis of historical data
“What was the watermark 2h ago in the southeast?”
One-time Queries: Snapshot view
“What is the watermark in the southeast?”
Persistent Queries: Monitoring over time
“Report the watermark in the southeast for the next 4h”
Protocol Classification (6)
Agenda
General Properties
Architectures and Requirements
Routing Protocols Classification
10 Suggested Routing Protocols:
LEACH DD
PEGASIS MCF
TEEN TTDD
APTEEN RW
SPIN RR
Low Energy Adaptive Clustering Hierarchy
1 - LEACH – Discussed …
Self-Organizing – Adaptive Clustering
Protocol Highlights
Cluster-Heads elect themselves –
Now – “Random Round-Robin”
Future – Power-Based Probability
Nodes die in random
Stationary Sink
Localized Coordination
Data Fusion
Low Energy Adaptive Clustering Hierarchy
1 - LEACH (2)
“Hot Spot” Problem
Main Drawbacks
(Nodes on a path from an event-congested area
to the sink may drain)
Inadequate for Time-Critical Applications
Stationary Sink – Maybe Unpractical
Basic Algorithm assumes any node can
communicate with sink – limited scale
Low Energy Adaptive Clustering Hierarchy
1 - LEACH (3)
Works in Rounds, each with
Main Procedures
Set-Up (Short) and Steady-State (Long)
Set-Up Phase - subdivided:
– Advertisement (I am a Cluster-Head)
– Cluster Set-Up (I am in your Cluster)
– Schedule Creation (This is your slot)
Steady-State Phase:
– Data Transmission using TDMA
Low Energy Adaptive Clustering Hierarchy
1 - LEACH (4)
Everyone uses the same channel
Main Procedures
Different clusters use different CDMA codes
Code chosen in random
Cluster-Head communicate with Sink
Can be extended to Hierarchical Clustering
Low Energy Adaptive Clustering Hierarchy
1 - LEACH (5)
Illustrations
Low Energy Adaptive Clustering Hierarchy
1 - LEACH (6)
Illustrations
Power-Efficient Gathering in Sensor Information Systems
2 - PEGASIS (1)
Token-Passing Chain-Based
Protocol Highlights
Considered Near-Optimal (in a sense)
Nodes die in random
Stationary Nodes and Sink
Every node have a global network map
Data Fusion
Greedy chain construction
Power-Efficient Gathering in Sensor Information Systems
2 - PEGASIS (2)
Stationary Nodes
Main Drawbacks
Global Information
Limited Scale:
Information travels many nodes
Assumes any node can communicate
with sink
Power-Efficient Gathering in Sensor Information Systems
2 - PEGASIS (3)
Greedy Algorithm Construct Chain –
Main Procedures
Start at a node far from sink and
gather everyone neighbor by neighbor
Node i (mod N) is the leader in round i
Nodes passes token thru the chain to leader
from both sides
Each node fuse its data with the rest
Leader transmit to sink
Power-Efficient Gathering in Sensor Information Systems
2 - PEGASIS (4)
Illustrations
Power-Efficient Gathering in Sensor Information Systems
2 - PEGASIS (5)
Illustrations
Rounds Until Death
Threshold sensitive Energy Efficient Sensor Network
3 - TEEN (1)
LEACH based Clustering
Protocol Highlights
Smart data transmission (Saves Power)
Nodes dynamic reconfiguration ability
Suits for Time-Critical applications
Threshold sensitive Energy Efficient Sensor Network
3 - TEEN (2)
“Hot Spot” Problem
Main Drawbacks
Cluster-Heads need to listen constantly
Wasted time-slots
Can’t distinguish dead nodes
Other LEACH problems…
Threshold sensitive Energy Efficient Sensor Network
3 - TEEN (3)
LEACH Proactive Clustering
Main Procedures
Node transmit in timeslot only if both:
– Value greater then a Hard Threshold (HT)
– Value differs from last transmitted value
(SV ) by more then a Soft Threshold (ST)
After transmission SV is reset
Threshold sensitive Energy Efficient Sensor Network
3 - TEEN (4)
Illustrations
Adaptive Periodic Threshold-sensitive Energy Efficient Sensor Network
4 - APTEEN (1)
Improved (Adaptive - Hybrid) TEEN
Protocol Highlights
All TEEN Features
More flexible logic and timeslots
Multi-type Queries:
– Historical (What was the temp. then?)
– One-time (What’s the temp. now?)
– Persistent (Tell me the temp for 2 hours)
Can distinguish dead nodes
Adaptive Periodic Threshold-sensitive Energy Efficient Sensor Network
4 - APTEEN (2)
LEACH problems…
Main Drawbacks
Complex logic
Adaptive Periodic Threshold-sensitive Energy Efficient Sensor Network
4 - APTEEN (3)
LEACH Proactive Clustering
Main Procedures
Node transmit in timeslot only if both:
– Value greater then a Hard Threshold (HT)
– Value differs from last transmitted value
(SV ) by more then a Soft Threshold (ST)
Or If did not transmit for a max time (TC )
Or if queried by some sink
After transmission SV is reset
Adaptive Periodic Threshold-sensitive Energy Efficient Sensor Network
4 - APTEEN (4)
Illustrations
Power Consumption:
As could be expected –
APTEEN is better the LEACH
but not as good as TEEN
Sensor Protocol for Information via Negotiation
5 - SPIN (1)
Network-wide Broadcast Limited by
Protocol Highlights
Negotiation and using Local Communication
Flooding problems solved:
Implosion – same data from many neighbors
Detection of overlapping regions
Excessive resources consumption (Blindness)
Needs only Localized Information
DataFusion
Two main protocols SPIN-PP & SPIN-BC
Sensor Protocol for Information via Negotiation
5 - SPIN (2)
Broadcast - Limited Scale –
Main Drawbacks
every node handles O(n) messages
Data is updated throughout network –
unnecessary in many cases
Network lifetime - not clear
High degree nodes = High power needs
Sensor Protocol for Information via Negotiation
5 - SPIN (3)
SPIN-PP (Point-to-Point Communication)
Main Procedures
Data is described by meta-data ADV msg.
Node has data sends ADV to neighbors
If neighbor do not have data sends REQ
Node responds by sending the DATA
This process continues around the network
Nodes may aggregate their data to ADV
In a Lossy Network ADV may be repeated
periodically and REQ if not answered
Sensor Protocol for Information via Negotiation
5 - SPIN (4)
SPIN-BC (Local Broadcast Communication)
Main Procedures
ADV and DATA sending like PP (but in B.C.)
Since only one REQ answer is needed, any
node waits a random interval and B.C. REQ
only if none was received yet.
The rest – like SPIN-PP
Sensor Protocol for Information via Negotiation
5 - SPIN (5)
Node with data
Illustrations
ADV
Node with data advertises to all its neighbors
Sensor Protocol for Information via Negotiation
5 - SPIN (5)
Node with data
Illustrations
REQ
Neighbor requests for data and it is sent
Sensor Protocol for Information via Negotiation
5 - SPIN (5)
Node with data
Illustrations
DATA
Node with data advertises to all its neighbors
Sensor Protocol for Information via Negotiation
5 - SPIN (5)
Node with data
Illustrations
ADV
Receiving node sends ADV to neighbors
Sensor Protocol for Information via Negotiation
5 - SPIN (5)
Node with data
Illustrations
Already
has data
(or dead)
REQ
Receiving neighbors requests for data.
Sensor Protocol for Information via Negotiation
5 - SPIN (5)
Node with data
Illustrations
DATA
Receiving node sends ADV to neighbors
Directed Diffusion
6 - DD (1)
Hybrid Data Centric Routing –
Protocol Highlights
Looking for Named Data
Query–Response Model
Performs Better than Flooding
Robust and Fault Tolerant (bypass faults)
Localized Interactions
Data Fusion - Application Specific Filters
Directed Diffusion
6 - DD (2)
“Hot Spot” Problem near sink
Main Drawbacks
Periodic Broadcasts of “Interest”
Reduces Network Lifetime
Trade-off: Energy Efficiency vs.
Robustness and Scalability
Complex Data Aggregation -
may Lead to Expensive Node
Directed Diffusion
6 - DD (3)
A Query (Interest) is Broadcasted by a node
Main Procedures
(sink)
Query Reaches Relevant Sensor Sources
This Sets-Up Exploratory Gradients
Once Data is Available in a Source
it is Sent Back via Reinforced Path
Failing Links / Nodes are being Gradually
Bypassed
Directed Diffusion
6 - DD (4) CLASS_KEY IS INTEREST_CLASS
LONGITUDE_KEY GE 10
LONGITUDE_KEY LE 50
LATITUDE_KEY GE 100
Illustrations
LATITUDE_KEY LE 120
SENSOR EQ MOVEMENT
Source INTENSITY GE 0.6
CONFIDENCE GE 0.7
INTERVAL IS 10
EXPIRE_TIME IS 100
Sink
Interest = Interrogation
Gradient = Who is interested
Directed Diffusion
6 - DD (4) FilterAttrVec
CLASS_KEY EQ DATA_CLASS
Illustrations
SENSOR EQ MOVEMENT
INTENSITY GE 0.7
Source 3. addFilter (FilAttrVec, FilterCallback)
1. subscribe (InterestAttrVec, Callback) 2. subscribe (AttrVec, ApplCallback)
InterestAttrVec
CLASS_KEY EQ INTEREST_CLASS
LONGITUDE_KEY IS 35
LATITUDE_KEY IS 110
SENSOR IS MOVEMENT Sink
Interest = Interrogation
Gradient = Who is interested
Directed Diffusion
6 - DD (4) Interests Setting up
gradients
Illustrations
Source
Sink
Interest = Interrogation
Gradient = Who is interested
Directed Diffusion
6 - DD (4)
Sending data …
Illustrations
Source
4. h = publish (SensedAttrVec)
5. send (h, SensedAttrVec)
SensedAttrVec Sink
CLASS_KEY IS DATA_CLASS
LONGITUDE_KEY IS 35
LATITUDE_KEY IS 110
SENSOR IS MOVEMENT
INTENSITY IS 0.8
CONFIDENCE IS 0.7
Low rate event
Directed Diffusion
6 - DD (4)
Illustrations
Source m1a 6. FilterCallback.recv (Message
m1)
m1b m2 m2
CLASS_KEY IS DATA_CLASS
LONGITUDE_KEY IS 35
m2
LATITUDE_KEY IS 110
SENSOR IS MOVEMENT
INTENSITY IS 0.8
CONFIDENCE IS 0.8
7. sendMessage (Message new)
Low rate event
Directed Diffusion
6 - DD (4)
Illustrations
Source 8. ApplCallback.recv (NRAttrVec)
Sink
Low rate event
Directed Diffusion
6 - DD (4) … and Reinforcing the
best path
Illustrations
CLASS_KEY IS INTEREST_CLASS
LONGITUDE_KEY GE 10
LONGITUDE_KEY LE 50
LATITUDE_KEY GE 100
Source LATITUDE_KEY LE 120
SENSOR EQ MOVEMENT
INTENSITY GE 0.6
CONFIDENCE GE 0.7
INTERVAL IS 1
EXPIRE_TIME IS 90
Sink
Low rate event Reinforcement = Increased interest
Directed Diffusion
6 - DD (5)
Illustrations
Source
Sink
Recovering
from node failure
Low rate event
Reinforcement
High rate event
Directed Diffusion
6 - DD (5)
Illustrations
Source
Sink
Stable path
Low rate event
High rate event
Directed Diffusion
6 - DD (6)
Illustrations
Source
Sink
Recovering
from link failure
Low rate event
Reinforcement
High rate event
Directed Diffusion
6 - DD (6)
Illustrations
Source
Sink
Stable path
Low rate event
Reinforcement
High rate event
Use: “Interests set up gradients drawing down data”
Minimum Cost Forwarding
7 - MCF (1)
Cost-Field min Cost from Node to Sink on
Protocol Highlights
Optimal Path
Slop-Down the Cost-Fields to Get to Sink
Minimize Multiple Transmissions using
Back-Off Algorithm Based on Node Cost
Localized Communication
Minimum Cost Forwarding
7 - MCF (2)
High Time Complexity (due to back-off)
Main Drawbacks
Many Sinks – Large Cost Tables
Cost Field Set-Up Time O(N)
No Load-Balancing
Minimum Cost Forwarding
7 - MCF (3)
Broadcast ADV msg. and get Answers from
Main Procedures
all Sinks Create Cost-Fields
Calculate Back-Off Timer Proportional to
Cost per each Sink
Needed Information Sent thru Slop
If no ACK until Timer Expires – Resend ADV
Minimum Cost Forwarding
7 - MCF (4)
Illustrations
Cost
A
B
C
Timeline
Minimum Cost Forwarding
7 - MCF (5)
Illustrations
A=150
110
S = 200 B = 120
C = 90
130 50
100
60
90
Sink = 0
Two-Tier Data Dissemination
8 - TTDD (1)
Grid Structure Clustering
Protocol Highlights
Stationary Location-Aware Nodes
Mission Aware – Infrequent Changes
Greedy Geographical Forwarding –
Building Grid
Localized Communication
Two-Tier Data Dissemination
8 - TTDD (2)
No Mobile Sensors
Main Drawbacks
Requires Location Information
Grid Nodes may Drain
Two-Tier Data Dissemination
8 - TTDD (3)
Grid Build using Greedy Algorithm and
Main Procedures
Location Awerness
Node Floods Messages to Dissemination
Nodes
Dissemination Nodes Forward to Sink
If a Node Fails – Grid is Fixed
Two-Tier Data Dissemination
8 - TTDD (4)
Illustrations
Dissemination Node
Data Announcement
Source
Data
Sink
Immediate Query
Dissemination
Node
TTDD Basics
Two-Tier Data Dissemination
8 - TTDD (5)
Illustrations
Dissemination Node
Trajectory
Data Announcement Forwarding
Source
Data Immediate
Dissemination
Node
Sink
Immediate
Dissemination Trajectory
Node Forwarding
TTDD Mobile Sinks
Two-Tier Data Dissemination
8 - TTDD (6)
Illustrations
Dissemination Node
Trajectory
Data Announcement Forwarding
Source
Data Immediate
Dissemination
Node
TTDD Multiple Mobile Sinks
Random Walks
9 - RW (1)
Finding a Random Walk over a Grid
Protocol Highlights
Multi-path Routing Different Routes
Load Balancing at Different Times
Large Scale Networks
Nodes Assumed to be Mostly Stationary
No Location Information Needed
Little State Information
Localized Communication
Random Walks
9 - RW (2)
Topology may not be Practical
Main Drawbacks
(Nodes are Assumed to be Located at
Cubic Grid Junctions)
Random Walks
9 - RW (3) - RSG
Regular Static Graphs
Main Procedures
Find coordinates differences (Dx, Dy) using
Distributed Bellman Ford (local comm.)
For every node compute probability of
moving on X and Y
(By the diagonal to the destination)
On each node move to a adjacent one on X
or Y using that probability. Adjust near end.
All Paths together draws a straight “Banana”
Random Walks
9 - RW (4) - ISG
Irregular Static Graphs (Some dead nodes)
Main Procedures
Same as RSG but…
If one adjacent node is missing – go to the
other (with p=1).
If both are missing – go to a neighbor
whose B-F distance to the destination is
strictly smaller than the current node
(This will create a detour).
(Could optimize by not getting to that node…).
Random Walks
9 - RW (5) - DG
Dynamic Graphs (Nodes may sleep and wake)
Main Procedures
Same as ISG but…
When a node changes state: the one-hop
neighbors change B-F labels and possibly
trigger further label (distances) changes
Concerns:
– Delays in propagating updates
– Sensitivity to inaccuracies in labels
Random Walks
9 - RW (6) - RSG
…...
Illustrations
0 1 2 N-1
0 S
de[3,2]=2
1 u4
P4
2 u3 P3 v P1 u1
[3,2]
P2
...
u2
N-1 R
Random Walks
9 - RW (7) – RSG vs. ISG
Illustrations
0 1 2 3 0 1 2 3
0 1/2 2/3 3/4 1/2
1,20 1,10 1,4 1,1 0 1,7 1,3 0,0 0,1
1/2
1/3
1/4
1/2
1
1
1/3 1/2 1/3 1/3 1 1/2
1 1,10 2,6 3,3 4,1 1 1,4 2,3 2,3 2,1
2/3
1/2
2/3
2/3
1/2
1
1
1/4 2/3 1/2 1/2
2 1,4 3,3 6,2 10,1
2 1,1 0,0 2,2 4,1
3/4
1/3
1/2
1/2
1
1
1 1 1 1 1 1
3 1,1 4,1 10,1 20,1
3 1,1 1,1 3,1 7,1
ISG ISG
Random Walks
9 - RW (8) – RSG vs. ISG
Illustrations
A Random walk by flipping a fair coin
RSG (DG Similar) ISG
Load Distribution - Narrow
Random Walks
9 - RW (9) – RSG vs. ISG
Illustrations
A Random walk by RSG/ISG algorithms
RSG (DG Similar) ISG
Load Distribution - Flat
Rumor Routing
10 - RR (1)
Observation: for many application any
Protocol Highlights
arbitrary path will do –
No Need for a Shortest Path
Nodes are Densely Distributed
Bidirectional Links
Localized Communication
Stationary Nodes
Meet Trails of Queries and Events
Rumor Routing
10 - RR (2)
Attractive only when the ratio between
Main Drawbacks
events and queries is inside a threshold
where it is not attractive to flood
neither.
Optimal parameters depend heavily on
topology (but can be adaptively tuned)
Does not guarantee delivery
Rumor Routing
10 - RR (3)
Movement on the net is done by several
Main Procedures
agents, trying (randomly) to walk straight.
Every node maintains lists of neighbors and
events (how to get to the reporting node).
An agent coming from and event is
updating nodes it visits.
An agent coming from a query is searching
for ways to the reporting nodes.
High probability the lines will intersect.
Rumor Routing
10 - RR (4)
Illustrations
Event 1 Knows Event 1 Agent
Event 2 Knows Event 2 Knows Both Event
Rumor Routing
10 - RR (5)
Illustrations
Event
Source
Query
Source
Agenda
General Properties
Architectures and Requirements
Routing Protocols Classification
10 Suggested Routing Protocols:
LEACH DD
PEGASIS MCF
TEEN TTDD
APTEEN RW
SPIN RR
Conclusions
WSN will spread to many applications
Properties and Requirements are both
Unique and Diversified
Routing Protocol choice
is and probably will continue to be
Application Driven
More Analysis, Simulations and new
Ideas are needed for every category
References (1)
Q. Jiang, D. Manivannan, Routing Protocols for Sensor Networks,
IEEE Consumer Communications and Networking Conference
(CCNC'04), 2004.
R. Jurdak, C. V. Lopes, P. Baldiy, A Framework for Modeling Sensor
Networks, 19th Annual ACM Conference on Object-Oriented
Programming, Systems, Languages, and Applications (OOPSLA'04),
2004.
W. Heinzelman, A. Chandrakasan, and H. Balakrishnan, Energy-
Efficient Communication Protocol for Wireless Microsensor
Networks, IEEE Proceedings of the IEEE International Conference
on System Sciences, 2000.
S. Lindsey, C. S. Raghavendra, PEGASIS: Power Efficient GAthering
in Sensor Information Systems, IEEE Aerospace Conference, 2002.
References (2)
A. Manjeshwar and D. P. Agrawal, TEEN: A Protocol for Enhanced
Efficiency in Wireless Sensor Networks, Proceedings of the 1st
International Workshop on Parallel and Distributed Computing
Issues in Wireless Networks and Mobile Computing (with
IPDPS'01), 2001.
A. Manjeshwar and D. P. Agrawal, APTEEN: a hybrid protocol for
efficient routing and comprehensive information retrieval in
wireless sensor networks, Proceedings of the International Parallel
and Distributed Processing Symposium (IPDPS'02), 2002.
J. Kulik, W. Heinzelman, and H. Balakrishnan, Negotiation-Based
Protocols for Disseminating Information in Wireless Sensor
Networks, Wireless Networks, Vol. 8, pp. 169-185, 2002.
C. Intanagonwiwat, R. Govindan, D. Estrin, J. S. Heidemann, and
F. Silva, Directed Diffusion for Wireless Sensor Networking,
IEEE/ACM Transactions on Networking, vol. 11, no. 1, pp. 2-16,
2003.
References (3)
F. Ye, A. Chen, S. Lu, L. Zhang, A Scalable Solution to Minimum
Cost Forwarding in Large Sensor Networks, Proceedings of the
10th IEEE International Conference on Computer Communications
and Networks (ICCCN'01), 2001.
F. Ye, H. Luo, J. Cheng, S. Lu, and L. Zhang, A Two-Tier Data
Dissemination Model for Large-scale Wireless Sensor Networks,
ACM International Conference on Mobile Computing and
Networking (MOBICOM'02), 2002.
S. D. Servetto, G. Barrenechea, Constrained Random Walks on
Random Graphs: Routing Algorithms for Large Scale Wireless
Sensor Networks, In the Proceedings of the 1st ACM International
Workshop on Wireless Sensor Networks and Applications
(WSNA'02), 2002.
D. Braginsky, D. Estrin, Rumor Routing Algorithm For Sensor
Networks, In the Proceedings of the 1st ACM International
Workshop on Wireless Sensor Networks and Applications
(WSNA'02), 2002.
SenseYour
Network
Dude
Karl
Friedrich
Hieronymus
Baron of
Munchausen
(1720-1797)