Transcript Transmission Power Control introduction

Dynamic Transmission Power Control
in Wireless Ad-Hoc Networks
EE194 Wireless Sensor Networks
Stuart Peloquin & Joe Cerra
Outline






Goals
Reasons for changing transmit strength
Hardware
Distributed algorithms
Proposed basic algorithm
JiST / SWANS
Goals
• Reduce single node and entire network
power consumption


Define the link metric as a function of
transmit power
Efficiently update link information for use in
effective routing
• Simulate an adaptive transmitter power
algorithm using JiST/SWANS
Project considerations
• Highly mobile environment



Distributed algorithms, no central server
Distributed routing tables
Increased loss model
• Hardware considerations

Crossbow Mica2 Motes
• RSSI capable
• Programmable transmit power
• Simulation considerations

JiST/SWANS
Why use dynamic power
model?
• Free space signal attenuation:
theoretical

Loss ~ d2
• Loss model ~ d4:



Mobile nodes
Large d: d > 4πhthr/λ
Multi-path fading due to:
• Urban/congested environment
• Indoors
Why use dynamic power model?
●
●
●
Save energy at each node with reduced transmission cost
Easily accessed and updated accurate link metric for routing
Can eliminate some channel assignment issues:
Overlap.
This channel cannot
be used here.
4 Channels:
Black
Red
Blue
Yellow
Able to change
power needs
All nodes operate at
full transmission
power.
Single-hop vs. Multi-hop

As seen before:
• Transmitted power ~ d2 – d4
• 2 nodes 50m away could drastically
decrease the power needed to
communicate if relay nodes are used
50m
Power ~
504
~
6.25x106
50m
10m
Power ~ 104 * 5 ~ 5x104
Single-hop vs. Multi-hop

When that model fails
• From link metrics shown, B is the obvious choice for routing
• How to establish these routes



Polling: Very wasteful
Random inquiries: Can be shown to perform better than constant polling
Include link status messages in some/all data packets
A
A.
B.
3
7
2
1
4
Information propagation

Need for an effective routing protocol
• Must be able to adapt quickly to changing
network topology and link status
• Cannot be over-cumbersome


Each node does not have unlimited memory to store
this data
Need for an effective method to update
link metrics
• RSSI is only so good on its own


Each node can reduce/increase the cost to send to its
neighbor
Each node should also forward that information so
routing tables can change
RSSI

Received Signal Strength Indicator
• Available on most Transceivers. RSS
can be interrogated from receiver.
• RSS to dB conversion rates available

How to use RSS?
• RSS does not directly translate to
distance
• Sending node should include the
transmitted signal strength

Roughly, loss ~ RSS – Transmitted power
Hardware

Mica2, Mica2 DOT
• CPU


Active: 8ma
Sleep: <15uA
• Transceiver – data rate: 38.4 Kbaud





Send: 25 – 27 mA (maximum power)
Receive: 8 – 10 mA
Sleep: <1uA
RF Power: -20 - 10 dB (programmable)
Received Sensitivity: -98 - -101dB (RSSI
capable)
Hardware – Mica2
• Typical Battery life:

1000 milliamp - hours
• CPU:

1000/8mA ~ 125 hours at full CPU consumption
• Transmitter:


25mA * (38000)-1 sec/bit * (60)-1 hour/sec ~ 1.1x105 mA – hour /bit
1000/1.1x10-5 ~ 9.12x107 bits
• Doesn't consider MAC encoding scheme, collision error
checking
• Simple 8-bit CDMA: 9.12x107/8 = 1.14x107 bits =
1.425x106 bytes
• 1.425x106 ~ 1.36MB: 1 floppy disk
Distributed Algorithms
Determine appropriate transmission
power levels per node
Reduce or increase the amount of
neighbors a node has
Reduce average RF interference



•
Outline of 5 distributed algorithms…
Fixed Transmission Power
Distributed Algorithm # 1


Simplest solution
Arbitrarily assign a fixed transmission
power level to all nodes.
• Does not adjust transmission power at all.
http://bwrc.eecs.berkeley.edu/People/Grad_Students/czhong/documents/kubischtxCtrl.pdf
Local Mean Algorithm (LMA)
Distributed Algorithm # 2
1.
2.
3.
All nodes start with same initial transmission power.
Every node periodically broadcasts a LifeMsg.
These nodes than count the number of responses
(LifeAckMsg) they receive.

4.
5.
6.
Called NodeResp
If NodeResp < NodeMinThresh, than node increases
transmission power
If NodeResp > NodeMaxThresh, than node decreases
transmission power
If NodeResp is between these bounds, than the node
does not change its transmission power.
http://bwrc.eecs.berkeley.edu/People/Grad_Students/czhong/documents/kubischtxCtrl.pdf
Threshold in Mean Number of
Neighbors (LMN)
Distributed Algorithm # 3

Similar to previous algorithm.
• LifeAckMsg also contains its own number of
neighbors.

Node receiving the LifeAckMsg computes a
mean value from this
• The new NodeResp
http://bwrc.eecs.berkeley.edu/People/Grad_Students/czhong/documents/kubischtxCtrl.pdf
Global Solution with Equal
Transmission Power
Distributed Algorithm # 4

Uses the Equal Transmission Power (ETP)
Algorithm.
• Assigns a uniform transmission power to all nodes
• Chooses the minimal value to ensure a fully connected
network.

ETP Algorithm on next page…
http://bwrc.eecs.berkeley.edu/People/Grad_Students/czhong/documents/kubischtxCtrl.pdf
Equal Transmission Power (ETP)
Algorithm
1.
2.
3.
4.
Among the node pairs that are not yet
connected, choose the one with the
smallest distance.
Set transmission power of all nodes to a
value sufficient to connect these two
nodes.
Check connectivity of the resulting
network. If not connected, loop.
When network is connected, minimum
power level is found.
http://bwrc.eecs.berkeley.edu/People/Grad_Students/czhong/documents/kubischtxCtrl.pdf
Global Solution with Diverse
Transmission Power
Distributed Algorithm # 5

Uses the Diverse Transmission Power (DTP)
Algorithm.
• Creates a connected network
• Does not set all transmission ranges to the same value.
• Tries to find a minimal power level for every node.

DTP Algorithm on next page…
http://bwrc.eecs.berkeley.edu/People/Grad_Students/czhong/documents/kubischtxCtrl.pdf
Diverse Transmission Power (DTP)
Algorithm
1.
2.
3.
4.
Among the node pairs that are not yet
connected, choose the one with the
smallest distance.
Set transmission power of these two
nodes to a value sufficient to connect
them.
Check connectivity of the resulting
network. Loop if not connected.
When network is fully connected,
minimum power level is found.
http://bwrc.eecs.berkeley.edu/People/Grad_Students/czhong/documents/kubischtxCtrl.pdf
A B
Proposed Basic Algorithms
Before initiating communication with A

1.
2.
Make sure there will be no communication interference
Put node into FCL if above conditions apply
A sends RTS(FCL, Ld) at maximum power
When B receives RTS


1.
2.
Check whether there is some channel in the FCL
Ensure some channel is a free channel after CTS and
Data transmission durations, and that it does not
interfere with any other channel
B replies with CTS(Dj, NAVCTS, PCTS)
Each mobile host keeps power[…] array


•
For each host Id, power[Id] is the power to transmit
http://pages.cpsc.ucalgary.ca/~caox/papers/multichannel02.pdf
Java Simulation Environment

JiST
• New and rapidly growing simulation
environment


Simulates using virtual machines
SWANS
• Runs on top of JiST
• Functionality similar to ns2 and GloMoSim


Component based architecture
Parallelized
JiST
Java in Simulation Time


High performance, discrete event simulation
engine that runs over standard java virtual
machine
Advantages:
• Efficient
• Transparent
• Standard
http://jist.ece.cornell.edu/docs/040325-yorku.pdf
SWANS
Scalable Wireless Ad hoc Network Simulation
•Newest technology in wireless sensor networking simulation
•Simulation program features include:
• Routing protocol, field dimensions, number of nodes, client/server
pairs, transmissions, packet loss probability, node movement rate
•Control Flow: Route request, route reply
•Data Structures: Buffers, route cache, route tables – scalable
•Layers: Routing, network, MAC
•Routing: DSR, ZRP
•Detail: Approximate physical level, packet level
http://jist.ece.cornell.edu/docs/040108-swans-dsr.pdf
Future Plan

Simulate a mobile wireless network in
JiST/SWANS.
1. Without power control
2. With power control
•
Develop a refined power control algorithm.
References








http://bwrc.eecs.berkeley.edu/People/Grad_Students/czho
ng/documents/kubischtxCtrl.pdf
http://pages.cpsc.ucalgary.ca/~caox/papers/multichannel0
2.pdf
http://jist.ece.cornell.edu/docs/040325-yorku.pdf
http://jist.ece.cornell.edu/docs/040108-swans-dsr.pdf
http://computer.howstuffworks.com/mote4.htm
http://www.xbow.com/Products/Product_pdf_files/Wireless
_pdf/6020-0042-06_B_MICA2.pdf
http://www.engineering.uiowa.edu/~ece195/2005/lectures
/lecture03.ppt
http://groups.csail.mit.edu/robotics/journal_club/papers/na
na.dankwa.ee.pdf