Node Cooperation and Cognition in Dynamic Wireless Networks
Andrea Goldsmith
Stanford University
Joint with I. Maric, R. Dabora, N. Liu and D.C. Oneill
DAWN ARO MURI Program Review
U.C. Santa Cruz
September 5, 2007

Wireless Multimedia Networks
In Military Operations
•Command/Control
•Data, Images, Video
How to optimize QoS and end-to-end performance?

Challenges to meeting network performance requirements
 Wireless channels are a difficult and capacitylimited broadcast communications medium
 Interference severely degrades link performance
 Network dynamics require adaptive and flexible protocols as well as distributed control
 Wireless network protocols are generally ad-hoc and based on layering, but no single layer in the protocol stack can guarantee QoS

Interference in Wireless Networks
 Radio is a broadcast medium
 Radios in the same spectrum interfere
 Network capacity in unknown for all canonical networks with interference (even when exploited)




Z Channel
Interference Channel
Relay Channel
General wireless ad-hoc networks

 If treated as noise: Foe

SNR

N
P

I
Increases BER,
Reduces capacity
If decodable or precodable: Neutral
 Neither friend nor foe
Multiuser detecion (MUD) and precoding can completely remove interference
Common coding strategy to approach capacity

If exploited via coding, cooperation, and cognition

 Many possible cooperation strategies:
 Cooperative coding, virtual MIMO, interference forwarding, generalized relaying, and conferencing
“He that does good to another does good also to himself.”
Lucius Annaeus Seneca

Codebook Design
The Z Channel
 Capacity of Z channel unknown in general



Encoding strategy of X
1 impacts both receivers
We obtain capacity for a class of Z channels
Superposition encoding and partial decoding is capacity-achieving for these channels
Can show separation principle applies

Cooperation through Relaying
TX1
X
1
Y
3
=X
1
+X
2
+Z
3 relay
X
3
= f(Y
3
)
RX1
Y
4
=X
1
+X
2
+X
3
+Z
4
TX2
X
2
Y
5
=X
1
+X
2
+X
3
+Z
5
RX2

Relaying strategies:


Relay can forward all or part of the messages
 Much room for innovation
Relay can forward interference
 To help subtract it out

encoder 1 encoder 2 relay dest1 dest2
R
R
2
R
R
R
1
1
1
2

| X
2
,


I ( X
2
,
R
2

I
X
3
; Y
2
( X
1
, X
|
2
,
X
1
)
X
3
; Y
1
)

I ( X
1
; Y
1
R
2

I ( X
1
, X
2
X
3
)
, X
3
; Y
2
)

I ( X
2
; Y
3
| X
3
) for any distribution p(p(x
1
)p(x
2
,x
3
)p(y
1
•
The strategy to achieve these rates is:
,y
2
|x
1
,x
2
,x
3
)
- Single-user encoding at the encoder 1 to send W
1
- Decode/forward at encoder 2 and the relay to send message W
2
•
This region equals the capacity region when the interference is strong and the channel is degraded

 Scalability
 Increased capacity
 Reduced energy consumption
 Better end-to-end performance
We need more creative mechanisms for node cooperation in wireless networks

 Cognitive radios can support new wireless users in existing crowded spectrum
 Without degrading performance of existing users
 Utilize advanced communication and signal processing techniques
 Coupled with novel spectrum allocation policies
 Technology could
 Revolutionize the way spectrum is allocated worldwide
 Provide sufficient bandwidth to support higher quality and higher data rate products and services

Cognitive radios (CRs) intelligently exploit available side information about the
(a) Channel conditions
(b) Activity
(c) Codebooks
(d) Messages of other nodes with which they share the spectrum

 Underlay
 Cognitive radios constrained to cause minimal interference to noncognitive radios
 Interweave
 Cognitive radios find and exploit spectral holes to avoid interfering with noncognitive radios

Overlay
 Cognitive radios overhear and enhance noncognitive radio transmissions

 Cognitive radios determine the interference their transmission causes to noncognitive nodes
 Transmit if interference below a given threshold

I
P
NCR
NCR
CR CR
The interference constraint may be met


Via wideband signalling to maintain interference below the noise floor (spread spectrum or UWB)
Via multiple antennas and beamforming
 Challenges: measuring interference at RX and policy

 Measurements indicate that even crowded spectrum is not used across all time, space, and frequencies
 Original motivation for “cognitive” radios (Mitola’00)
 These holes can be used for communication



Detecting and avoiding active users is challenging
Hole location must be agreed upon between TX and RX
Common holes between TX and RX may be rare

 Cognitive user has knowledge of other user’s message and/or encoding strategy
 Used to help noncognitive transmission
 Used to presubtract noncognitive interference
RX1
CR
RX2
NCR

Proposed Transmission Strategy
Cooperation at CR TX
Precoding against interference at CR TX
Rate splitting
19
To allow each receiver to decode part of the other node’s message
 reduces interference
Removes the NCR interference at the CR RX
To help in sending NCR’s message to its RX
We optimally combine these approaches into one strategy

Transmission for Achievable Rates
The NCR uses single-user encoder
W
2
X
N
P
X
2
(.)
2
CR
RX1
The CR uses
NCR
RX2
Rate-splitting to allow receiver 2 to decode part of cognitive user’s message and thus reduce interference at that receiver
Precoding while treating the codebook for user 2 as interference to improve rate to its own receiver
Cooperation to increase rate to receiver 2
W
1
W
2
Rate split
CR
W c
W
1 a
P
U
1 c
(.)
X
2
N
NCR
U
1 c
N
X
2
N
P
U
1 a
| U
1 c
(.
| u
1 c
)
U
1 c
N
, U
1 a
N
X
2
N
X
1
N

• Follows from standard approach:
• Invoke Fano’s inequality
• Reduces to outer bound for full cooperation for R
2
=0
• Has to be evaluated for specific channels
How far are the achievable rates from the outer bound?

CR broadcast bound outer bound our scheme prior schemes

Introduction to Wireless
Network Utility Maximization
 Wireless networks operate over random time varying channels
 Fading distribution typically unknown
SNR
 Upper Layer performance is critical


Dictates application quality
Dictates user experience
Upper
Layers
Physical
Layer
 Application performance depends on multiple performance metrics
Rate
(R*,D*,O*)
 Rate


Delay
Outage
Delay ti me
Upper
Layers
Physical
Layer
Utility=f(Rate,Delay,Outage)
Outage

Wireless NUM Problem Statement
 Find network policies (control functions) that
 Optimize performance

At upper layers


Through optimal cross layer interaction
Utilizing information-theoretic coding strategies
 Meet constraints


Long term average: e.g. Power: E[S(·)]≤S
Instantaneous: e.g. Reliability: BER ≤  (·)
 Adapt gracefully to changing conditions

Network Utility Maximization (NUM)
 Model end-to-end performance as a utility function (typically a function of rate
Best effort
Diminishing returns
Contract with penalty
 NUM often applied to wireline/wireless networks
 Performs poorly in dynamic environments
 Dynamic NUM extends NUM to include dynamics in the links, interference, and network.

Interference and dynamics easily incorporated



Utility functions U(r)


Rate only
Does not “select” Rate-
Reliability operating point
Explicit Rate-Reliability tradeoff by sources
 U
B
(rate, reliability)

B controls tradeoff
 Sources select link code rate to meet reliability needs
Policies for


Link power S l
Link rates R l
(.) l=1,…,L
(.) l=1,…,L

U
1
( r
1
,

1
)
U
2
( r
2
,

2
Data
)
Data
U ( r
3
,
3
Data

3
)
Upper
Layers
Buffer
Physical
Layer
Upper
Layers
Buffer
Physical
Layer
Upper
Layers
Buffer
Physical
Layer
Data
Upper
Layers
Buffer
Physical
Layer
Upper
Layers
Buffer
Physical
Layer

Performance Improvement of Wireless NUM
Rate Benefits
BER (Reliability) Benefits
Beta controls tradeoff in
U
B
(rate, reliability)

 Interference can be exploited via cooperation and cognition to improve spectral utilization as well as end-to-end performance
 Much room for innovation
 WNUM can provide the bridge to incorporate novel coding methods into dynamic distributed networks.