Energy Efficient Routing

1
Energy Efficient Routing
LAMAR UNIVERISTY Computer Science Department
By Rui Luo Supervisor Dr. Lawrence J.
Osborne Committee Member Dr. Chung-Chih
Li Committee Member Dr. Bo Sun Fall 2005
2
Agenda

• Background
• Sensor, WSNs, TinyOS, Applications
• Current Routing Protocols In WSNs
• Reviews, Compare
• Algorithm Design
• Target Model, Principle, Algorithm
• Implementation
• Platform, Issues, Architecture
• Testing
• Methodology, Simulator, Chat
• Conclusion and Future Work

3
Sensor and WSNs

• Poor powered
• Battery, Ambient Energy (solar cell)
• Constrained computing resource
• Memory space
• CPU
• Limited Communication abilities
• Low bandwidth
• Transmission range

Agenda
4
Applications

• Habitat Monitoring
• (Event-driven, Randomized Deployment)
• Environment Observation and Forecasting (EOFS)
• (Time-driven, Query-driven)
• Health Applications
• (Event-driven, Time-driven)
• Structure Health Monitoring (SHM)
• (Deterministic Deployment)
• Home, Office Applications, and Other

Agenda
5
WSNs vs. MANETs
Application Dependent Routing is known Low
Bandwidth Weak Nodes Usually Stationary
WSNs
Limited Transmission Range
MANETs
Powerful Nodes, High Bandwidth, Application
Independent, Routing created on Demand, Nodes
come and go
Agenda
6
WSNs Design Issues

• Challenges
• Global address scheme
• Data need to be routed to a particular BS
• Constrained abilities of nodes
• Stationary Mobility
• Application dependent
• Position awareness
• Redundancy data

Agenda
7
Tiny OS

• Component Based
• Rapid development, Small size binary
• Event Driven
• Save more energy
• Multi-Hardware platform
• Berkeley/Crossbow mica2 3rd generation, wireless
  reprogramming

Agenda
8
Routing Protocol Review

• By network structure
• Flat Network, Hierarchical Networks, Geographic
  Information Based
• By protocol operation
• Negotiation based, Multi-Path based, QoS based,
  Coherent based

Agenda
9
Routing Protocol Comparison

• Small Minimum Energy Communication Network (MECN)
• GPS, has to compute relay region
• Our protocol
• GPS free, no relay region computation

Agenda
10
Target Model

• Over densed deployment
• Stationary
• Nodes send data to the root node
• Battery changing is reasonable
• Demand lifetime of the whole network is much
  longer than a single life time of each sensor.

Agenda
11
Principle

• Radio Propagation Model
• Near field zone
• Signal strength is strong, but very short
• Free space path loss zone
• 20 dB/decade, radio travel through the air
• Excess path loss zone.
• 20-50 dB/decade, affected by the ground

Agenda
12
Principle Cont.

• Friis equation
• GTx transmitter antenna gain
• GRx receiver antenna gain
• ? wavelength (same units as d)
• d distance separating Tx and Rx antennas
• L system loss factor ( 1)

Agenda
13
Principle Cont.

• Key idea
• To reach the distance two times far away, we need
  four times transmission power.
• Note
• In read world, the index is not constant and
  usually between 2 to 4.

Agenda
14
Principle Cont.

• Shown by graph
• Two ways to get D from S
• Send to D directly
• Less transmission delay
• Use more energy
• More chance collision
• Hop by R
• More transmission delay
• Use less energy
• Less chance collision

D
R
S
Agenda
15
Principle Cont.

• Most Energy Efficient Region (MEER)
• Definition A region in which a relay node is
  preferred if total energy is the concerned
  metric.
• Shape Draw the shape according to the definition
  in Matlab
• Note Not necessary to be in 2-D space.

Agenda
16
Algorithm Principle Cont.

• Shape of MEER for attenuation rate index equal to
  3

Agenda
17
Algorithm

• Problem
• Limited transmission range
• GPS is expensive.
• In real world, space is twisted.

Agenda
18
Protocol

• Approximate by considering one hop

Agenda
19
Implementation Platform

• Red Hat Linux 9.0
• Tiny OS 1.1.0
• ncc 1.1.1 (source)
• 1.1.2 is not compatible with gcc-3.2.2
• gcc 3.2.2
• IBM-JDK 1.3 for Linux
• Tiny OS cannot be installed normally for 1.4
• Atemu 0.4 (source)

Agenda
20
Implementation Issues

• Memory Constrain
• Mica2 4k RAM, 512K ROM
• Ram usage (byte)
• TinyDB less than 3100, Moté 849
• Our program about 1k ram, most of them are used
  for outgoing buffer
• Transmission Power
• Network density is controllable
• Max150, reaches 100m in Atemu
• PM start1, increment10

Agenda
21
Implementation Issues Cont.

• Packet loss
• Causes exposed node, Outgoing buffer full
• Solution
• Avoid packet loss by a big buffer in test
• In real application, the buffer can be small
• Dynamic memory management?
• Advantages more adaptive, flexible
• Disadvantages more CPU load
• Choose Static memory allocation

Agenda
22
Implementation Issues Cont.

• AM message or Low level Comm.?
• Goal
• Make the program easy to be used
• Allow other protocols run in one application
• Choose
• AM message with direct control to radio mode.

Agenda
23
Implementation Architecture
Agenda
24
Implementation Debugging

• Problem
• Atemu logs the low level events
• Good for simple program debugging
• Not good for a distributed protocol debugging

Agenda
25
Implementation Log File

• Sample
• 000 -- system timer 0 4
• 000 ltpmackgt to 7
• 006 ltpmackgt to 1
• 007 parent ETOR power 0 21 21
• 001 parent ETOR power 6 32 21
• 006 ltpmackgt to 7
• 007 drop pmack from 6
• 008 ltpmgt 31
• 007 ltpmgt 31

Agenda
26
Implementation Log File Analysis

• Unix text tool
• Example Show final topology
• !/bin/bash
• for i in cut -b1-3 1 sort -u
• do
• grep i 1 grep parent tail -n 1
• done
• Eegraph
• Will be told later

Agenda
27
Implementation Auto ID

• Problem
• Node ID is hard coded into programs
• TOSSIM does substitution automatically
• Atemu does not.
• Solution
• Modify the source code of Tiny OS
• Modify the default Makefile
• A bash script is used to iterate the node id

Agenda
28
Testing Methodology

• Simulator
• TOSSIM
• Simple, no high fidelity
• Ns2
• Classical, not feasible for low level simulation
• Atemu
• High fidelity, slow, no future support
• Avrora
• New, high fidelity, fast
• Current not support IBM-JAVA

Agenda
29
Testing Methodology Cont.

• Visual tool eegraph
• Functionalities
• Visualize the log file
• Output data for chart

Agenda
30
Testing Metrics

• Topology
• Total ETOR
• Cost for data transmission
• Number of the node in the network (N)
• Total energy consumption (TEC)
• Cost for this protocol

Agenda
31
Testing Topology Starting Power

• Node 14,2,6
• Rough detection of 2,6
• Node 15,13,17
• Direct reachable to 17
• Conclusion
• Height of the tree
• Transmission delay

S1
S40
Agenda
32
Testing Constructing Starting Po...

• Conclusion
• No obvious relationship between constructing time
  and starting power

Agenda
33
Testing total ETOR-Starting Power

• Conclusion
• Higher starting power results in higher total
  ETOR

Agenda
34
Testing TEC Starting power

• Conclusion
• A higher starting power results in less total
  energy consumption. However, this effect is not
  obvious at the beginning.

Agenda
35
Testing Topology - Increment

• Conclusion
• Height of the tree
• Transmission delay

i10
i60
Agenda
36
Testing Constructing Increment

• Conclusion
• Higher increment results in fast network
  constructing.

Agenda
37
Testing Total ETOR-Increment

• Conclusion
• A higher increment results in higher total ETOR

Agenda
38
Testing TEC - Increment

• Conclusion
• Higher increment results in lower TEC

Agenda
39
Testing Topology - Fixed

• Fixed transmission power is equal to the view of
  network density
• Topology for different densities

Agenda
40
Testing Constructing - Fix

• Conclusion
• Higher transmission power results in fast network
  construction.
• Network density is one of the most important
  factors.

Agenda
41
Testing total ETOR - fix

• Conclusion
• Either blue or green is not good
• Yellow is preferred
• Network density is one of the most important
  factors

Agenda
42
Testing TEC - Fix

• Conclusion
• Higher trans. power results in higher TEC
• A modified protocol can be used to stop using
  energy after the network setting up, hence trans.
  power has less relationship with TEC.

Agenda
43
Future Work

• How to use the protocol in a hierarchical network
• Adaptive TPC period
• More testing in real world applications

Agenda
44
Conclusion

• The algorithm is quite applicable in an
  over-densed network.
• The algorithm can behave differently to meet the
  transmission delay metric.
• Compare with the shortest path algorithm, the new
  algorithm saves more energy and introduces more
  transmission delay.

45
References

• Mica2 Mote Datasheet
• Jamal N. Al-Karaki, Ahmed E. Kamal. Routing
  Techniques in Wireless Sensor Networks A Survey
• Ning Xu, A Survey of Sensor Network
  Applications
• H.T. Friis, Introduction to radio and radio
  antennas
• Jonathan Polley, Dionysys Blazakis, Jonathan
  McGee, Dan Rusk, John S. Baras, ATEMU A
  Fine-grained Sensor Network Simulator
• V. Rodoplu and T. H. Meng, Minimum Energy Mobile
  Wireless Networks"

Agenda
46
Questions?
Agenda