Selecting Transmit Powers and Carrier Sense Thresholds in CSMA
Jason Fuemmeler, Nitin Vaidya, Venugopal Veeravalli
ECE Department & Coordinated Science Lab
University of Illinois at Urbana-Champaign
http://www.crhc.uiuc.edu/wireless/
WICON 2006 Boston, MA August 3, 2006
Funded in part by NSF and by a NSF Graduate Research Fellowship
Medium Access Control
The wireless channel All nodes share same medium Nodes can interfere with one another Channel can support multiple transmissions if
separated in space (spatial reuse)
Medium Access Control (MAC) is needed to use
the channel effectively
Question: How can we design MAC protocols to
maximize spatial reuse?
Power Control
Power control can be used to increase spatial reuse
MAC protocols utilizing power control must perform a balancing act Must maintain desired SINR at each receiver Need interference margin at each receiver to
maintain this SINR Increasing transmit power increases interference
margin But increasing transmit power increases
interference to other transmissions
Previous Research
PCMA [Monks01] Busy tones sent on out-of-band channel to
communicate current interference margins
PCDC, POWMAC [Muqattash03, Muqattash04] Control frames sent at maximum power to
communicate information about transmission powers and interference margins
Transmission power selection strategies in these protocols left unjustified
Previous Research
CS Threshold Selection in 802.11 [Zhu04] Does not address selection of transmit powers
In our work, we address both transmit power and carrier sense threshold selection in the IEEE 802.11 protocol
Physical Carrier Sensing
We primarily consider physical carrier sensing How it works:
Node is allowed to transmit only if channel is idle Channel assumed to be idle only if total power
seen at its location is less than carrier sense (CS) threshold
Idle channel should mean that transmitting will not cause a collision
A Two-Link Setup
A
BD
C
pow
er
distance
SCS Threshold
A Two-Link Setup
A
BD
C
pow
er
distance
S
I
Analytical Results
Collisions are doubly bad Waste channel resources now Waste channel resources upon retransmission
Intuitively, to prevent collisions large transmit power => small CS threshold
Analysis of collision prevention yields that the product of the transmit power and the CS threshold should remain constant throughout the network Bounds the amount of interference one link can
pose to another
Notation
pt: transmit power
pcs: carrier sense threshold
g: channel gain on the link γ: required SINR η: thermal noise β: the constant product k: number of worst-case interferers assumed
The Equations
pt pcs=
pcs=1k p tg −
The Role of k
Analysis uses collocation approximation A potential interferer sees same gain to both
transmitter and receiver
The value of k accounts for: The local topology around the link Any error introduced by the collocation
approximation
For k sufficiently large, collisions will be prevented on the link
ns-2 Simulation Setup
PHY layer was modified to be more accurate RTS/CTS disabled – physical carrier sensing
dominant η set to 0 to explore upper limit in spatial reuse UDP traffic, heavily loaded Topologies consisting of randomly placed links
Sample Topology
Schemes Considered
Fixed Rx Power Power at receiver held constant, CS threshold a
free parameter
Fixed Tx Power Transmit power held constant, CS threshold a free
parameter
Static k Our scheme with β held constant, k a free
parameter
Dynamic k (next slide)
Dynamic k Scheme
Each link adjusts its value of k dynamically Uses transmission failures as feedback Attempts to find minimum value of k such that
collisions are prevented on that link Minimum k <=> Minimum transmit power Algorithm used is heuristic
Throughput Comparisons
Fairness Issues
Our scheme does lead to some unfairness Links with high CS thresholds get to transmit more
often In general, short links are given preference
Could perhaps mitigate unfairness by having short links voluntarily lower CS threshold
Fairness Measure:
∑i=1
n
x i2
n∑i=1
n
xi2
Fairness Comparisons
Conclusions
Analyzed collision prevention conditions Concluded that product of transmit power and CS
threshold should remain constant throughout network
Simulation results indicate increased spatial reuse
Future Research
More detailed simulations Comparisons with non-802.11-based schemes Understand interactions with virtual carrier
sensing Better justified algorithm for adjustment of k Mitigation of unfairness
The End
Thanks for you attention!
Questions?