Cross-Layer Approach to Wireless Collisions

Dina Katabi

802.11 devices are increasingly prevalent

distributed and bursty access

Home

office

City mesh

Two devices access medium together Collision

Job of the MAC:

Avoid Collisions!

And when they happen?

Be in denial!

In This Talk

• Collisions are not harmful– We can decode colliding packets as

efficiently as if they were sent separately

• Collisions are beneficial– We can exploit strategic collisions to

increase throughput Analog Network Coding

The Hidden Terminals Problem

Collision!

Alice Bob

The Hidden Terminals Problem

More Collisions!Retransmissions

Can’t get any useful traffic!

Alice Bob

Can we take two collisions and produce the two packets?

Pa

Pb

Pa

Pb

Yes, we can!

Yes, we can!

ZigZagExploits 802.11’s behavior• Retransmissions

Same packets collide again• Senders use random jitters

Collisions start with interference-free bits

∆1 ∆2Pa

Pb

Pa

Pb

Interference-free Bits

How Does ZigZag Work?

∆1 ∆2

Find a chunk that is interference-free in one collisions and has interference in the other

11

∆1 ≠∆2

Decode and subtract from the other collision

11

∆2

11

22

11

∆1

How Does ZigZag Work?

Find a chunk that is interference-free in one collisions and has interference in the other

∆1 ≠∆2

Decode and subtract from the other collision

∆211

2222∆1

How Does ZigZag Work?

33

Find a chunk that is interference-free in one collisions and has interference in the other

∆1 ≠∆2

Decode and subtract from the other collision

∆211

22 44∆1

How Does ZigZag Work?

33 33

Find a chunk that is interference-free in one collisions and has interference in the other

∆1 ≠∆2

Decode and subtract from the other collision

∆211

22 4444∆1

How Does ZigZag Work?

33 55

Find a chunk that is interference-free in one collisions and has interference in the other

∆1 ≠∆2

Decode and subtract from the other collision

∆211

66∆1

How Does ZigZag Work?

33 55 55

22 44

Find a chunk that is interference-free in one collisions and has interference in the other

∆1 ≠∆2

Decode and subtract from the other collision

∆211

6666∆1

How Does ZigZag Work?

22 44

33 55 77

Find a chunk that is interference-free in one collisions and has interference in the other

∆1 ≠∆2

Decode and subtract from the other collision

∆2

11

66 88∆1

How Does ZigZag Work?

22 44

33 55 7777

Find a chunk that is interference-free in one collisions and has interference in the other

∆1 ≠∆2

Decode and subtract from the other collision

• Delivered 2 packets in 2 timeslotsAs if packets didn’t collide

• Low-complexity linear decoder• No need for synchronization

How does the AP know it is a collision and where the second packet starts?

Time

AP received a collision signal

∆

Detecting Collisions and the Value of ∆

Time

AP received signal

Packets start with known preamble

AP correlates known preamble with signal

Correlation

Time

Correlate

∆

How Does the AP Subtract the Signal?

• Channel’s attenuation or phase may change between collisions

• Can’t simply subtract a chunk across collisions

Alice’ s signal in first collision

Alice’ s signal in second collision

Subtracting a Chunk

What if AP Makes a Mistake?

∆1 ∆211 11

2222

Bad News: Errors can propagate

33

Can we deal with these errors?

What if AP Makes a Mistake?

∆1 ∆2

What if AP Makes a Mistake?

Good News: Temporal DiversityA bit is unlikely to be affected by noise in both

collisions

Get two independent decodings

Errors propagate differently in the two decodings

For each bit, AP picks the decoding that has a higher PHY confidence [JB07, WKSK07]

Which decoded value should the AP pick?

∆1 ∆2

11 11

2222

33

AP Decodes Backwards as well as Forwards

ZigZag Generalizes

ZigZag Generalizes

∆1

∆211

22

11

22

• Flipped order

• Flipped order• Different packet sizes

ZigZag Generalizes

∆1 ∆2

11

22

11

22

ZigZag Generalizes

1

2

3

1

2

3

1

2

3

• Flipped order• Different packet sizes• Multiple colliding packets

11

22

11

22 22

11

333333

ZigZag Generalizes• Flipped order• Different packet sizes• Multiple colliding packets • Capture effect

Pa1

Pb

Pa2

Pb

3 packets in 2 timeslots better than no collisions

Performance

Implementation

• USRP Hardware• GNURadio software• Carrier Freq: 2.4-2.48GHz• BPSK modulation

USRPsTestbed

• 10% HT, 10% partial HT, 80% perfectly sense each other

• Each run randomly picks an AP and two clients

• Co-located 802.11a nodes that measure HTs. The USRPs use the same collision patterns as 802.11

802.11a

Throughput Comparison

Throughput

CD

F o

f co

ncur

rent

flo

w p

airs

Throughput Comparison

802.11

Throughput

CD

F o

f co

ncur

rent

flo

w p

airs

Hidden Terminals

Partial Hidden Terminals

Perfectly Sense

Throughput Comparison

ZigZag

Throughput

CD

F o

f co

ncur

rent

flo

w p

airs

802.11

Hidden Terminals get high throughput

Throughput Comparison

ZigZag

Throughput

CD

F o

f co

ncur

rent

flo

w p

airs

802.11

ZigZag Exploits Capture Effect

ZigZag improved average Throughput by 25%

Throughput Comparison

ZigZag

Throughput

CD

F o

f co

ncur

rent

flo

w p

airs

802.11

Improved hidden terminals loss rate from 72% to 0.7%

Hidden Terminals

Is ZigZag as efficient as if the colliding packets were sent in separate slots?

• For every SNR,

Check that ZigZag can match the BER of collision-free receptions

Is ZigZag as efficient as if packets were collision-free Receptions?

SNR in dB

Bit E

rror

Rat

e (B

ER)

Collision-Free Receptions

Is ZigZag as efficient as if packets were collision-free Receptions?

SNR in dB

Bit E

rror

Rat

e (B

ER)

Collision-Free Receptions

Is ZigZag as efficient as if packets were collision-free Receptions?

ZigZag-Decoded Collisions

SNR in dB

Bit E

rror

Rat

e (B

ER)

ZigZag is as efficient as if the colliding packets were sent separately

But can collisions be beneficial?

ZigZag makes collisions harmless

Current Wireless

Alice Bob

Current Wireless

Current approach requires 4 time slots

Can we do better?

Alice Bob

Naive Application of ZigZag

Collision!Alice Bob

Decode Collisions

Naïvely applying ZigZag requires 4 time slots

Decoding collisions at AP doesn’t reduce timeslots

But the AP doesn’t need to decode!

Alice Bob

Naive Application of ZigZag

What if the AP doesn’t decode?

Collision!

1) Alice and Bob transmit simultaneously

Alice Bob

What if the AP doesn’t decode?

Collision!

1) Alice and Bob transmit simultaneously2) AP amplifies and broadcasts the collision

Alice Bob

1) Alice and Bob transmit simultaneously2) AP amplifies and broadcasts the collision 3) Alice subtracts her packet from the collision

1fhj

What if the AP doesn’t decode?

1fhj11

Bob’s pkt11

Alice’s pkt

2 time slots instead of 4

Throughput Gain of 2x

Alice Bob

Extends Network Coding to Signals

Traditional network coding: nodes forward linear combinations of received packets

New approach: channel naturally creates linear combinations of signals

Analog Network Coding (ANC)!

Theoretical Limits

Theorem [Katti et al. ISIT’07]:Informal version: “For a symmetric two-way relay channel the

rate achieved by analog network coding at high SNR is double the rate achieved by pure forwarding”

Implemented in software radio

Evaluated in a similar testbed

Metric

Gain = Throughput in ANC /Current Throughput

Performance

Throughput Gain for the Alice-Bob Scenario

Throughput gain

CD

F

Median Throughput Gain is 1.7x

Related Work

Conclusion

Softcast: cross-layer wireless video without CSI