Class Introduction

COS 463: Wireless Networks

Lecture 1
Kyle Jamieson

[Parts adapted from H. Hassanieh, P. Steenkiste]

Course staff and office hours

Kyle Jamieson Longfei Shangguan

CS room 305 CS room 418C

Allen Welkie
CS room 418C

• Office hours: By appointment, synchronized with Lab

programming exercise deadlines on Piazza

2
 Wireless is increasingly prevalent
• Health and Fitness
Wireless Sensors
• Virtual Reality
etworks UAVs
• UAVs
• Internet of Things Sensors

Smart Home
Wireless VR Vehicular Networks
Wireless Vehicles
ata Centers

Cellular
Networks

3
Mobile connectivity for people:
Increased wireless demand

4
Next demand driver:
Billions of Wireless devices

5
Industrial Impact: Cisco Meraki
• Founders Biswas, Bicket, Aguayo PhD
candidates who left program
– Initial products: mesh networking
technology from grad school

• Pivot three years later

– Focus on cloud-managed Wi-Fi Roofnet Mesh Network

• Cisco acquisition, new Cisco

Wi-Fi product line

• Takeaway: Wireless
technology à industry impact
COS 463: Course Approach
• Gain deep knowledge of wireless networks by learning
across all layers of the stack
– Some “ELE” content, some “COS” content

• Taught from first principles: build up Computer

Systems, Signal Processing knowledge

• Test that knowledge and cement understanding by

hands-on programming lab assignments
– Build something >> “Know” something
– Labs on real software-defined radio hardware
• Real-world “surprises”

7
Goals of the Class
1. Understand wireless networks’ design and architecture
– From signals to bits to datagrams
– Understand design choices and tradeoffs

2. Understand how the design of wireless networks

interacts with the rest of the wired Internet

3. Gain proficiency in building real wireless networks

8
Course Contents
1. Wireless From the Transport Layer Downwards
– Transport over wireless, link layer, medium access

2. Overcoming Bit Errors

– Error Detection/correction, convolutional & “Rateless” codes

3. An Introduction to the Wireless Channel

– Noise, Multipath Propagation, radio spectrum

4. Wireless Physical Layer concepts

– OFDM, channel estimation, MIMO etc.

5. Boutique topics
– Visible light communication, low power, Wi-Fi localization

9
Prerequisites and Administrivia
• Assume basic familiarity with computer networking
concepts and programming

– COS 217 required

– Knowledge of C and one other programming language

helpful, but not required

– Not open to freshmen

• COS 463 is a COS systems-track course

10
Target audiences
• COS 463 is cross-registered with ELE (ELE 463)

1. COS students who want to extend their networking

background to wireless communications

2. ELE students who want to extend their wireless

communications knowledge to networks

11
Modes of delivery
• Lectures: Introduce concepts, gain background knowledge

• Precepts & Lab: hands-on training on wireless systems

– Learning by doing: building wireless systems with a
software defined radio platform
– Precepts and labs closely coupled

12
Lab logistics
• Lab location: 87 Prospect, Garden Level, Room 065

• Enrolled students as of Feb 1 are added to Salto room and

building access control list
– Visit a hotspot and tap your TigerCard to update
– If you weren’t enrolled Feb 1, send instructors a private
Piazza message requesting access
13
Lab: Building a spectrum analyzer
• Introduce you to basic signal
processing concepts
– Fast Fourier Transform (FFT)
– Low-pass filtering

• Learning how to use the

HackRF software-defined radio

14
Lab: Sharing the Wireless Medium
• We set up a transmitter in the room

• You listen for an empty time or

frequency slot and send in that slot

• We show the result in a live

“scoreboard”

15
Lab: Array Signal Processing
• Angle-of-arrival estimation using multiple antennas

• Transmit beamforming using multiple antennas

User 1

User 2

16
Readings
• Text book: Cory Beard and William Stallings,Wireless
Communication Networks and Systems
• Reference material: Eldad Perahia, Robert Stacey, Next
Generation Wireless LANs

17
Class Grading
• Midterm exam: 20%
• Final exam: 30%
• Five lab programming assignments: 40%
• Class participation: 10%
– Precept attendance and participation
– Activity on Piazza

• Midterm and final exam coverage: Everything mentioned

in lecture and precept, all readings, all labs
– Emphasis on the concepts discussed in lecture &
precepts, and lab content

18
Today
1. How do wireless and wired networks differ?

2. What makes wireless interesting?

3. What new services does wireless enable?

19
Wireless is less reliable

Alice Bob

• In wired networks, link bit error rate is 10-12 and less

• Wireless networks are far from that target
– Bit error rates of 10-6 and above are common!

• Why?

20
Today
1. How do wireless and wired networks differ?
– A shared wireless medium
– Less reliable links
– Mobility

2. What makes wireless interesting?

3. What new services does wireless enable?

21
Wireless is a shared medium
Cathy
• Wired networks: Alice Bob
Alice and Bob’s conversation is
independent of Cathy and
Eve’s conversation Eve

• Wireless networks:
Close by wireless
conversations share the same
wireless medium

22
Why is a point-to-point link the
wrong abstraction for building
wireless networks?

23
Reason #1: Interference
• Noise is naturally present Cathy
in the environment from Alice Bob
many sources

• Interference can be from

other users of the same
technology, other
technologies altogether

• Impacts the throughput

users can achieve Cathy
24
Reason #2: Can leverage broadcast

Router
Alice
Bob

• Want to exchange packets, but out of direct range

25
Solution using wired abstraction

1
2

Router 3
Alice
4 Bob

• Requires four transmissions in total

26
Idea: Router combines the packets

Router
Alice
Bob
= ⊕

27
Router broadcasts the combination

Router
Alice 3 Bob

• Requires just three transmissions in total

28
Sumary: Shared medium is very different
• Wireless’ shared medium is very different than point-to-
point wired links

• So need to think about wireless networks differently

• Interference is a major problem

• But also can leverage broadcast nature of wireless

– Four to three transmissions increases throughput

– Serve more users or increase app performance

• Better Skype calls

29
Today
1. How do wireless and wired networks differ?
– A shared wireless medium
– Less reliable links
– Mobility

2. What makes wireless interesting?

3. What new services does wireless enable?

30
Less reliable links:
Multipath propagation
Ceiling

S R
Floor
• Signal bounces off surface and interferes with itself

• Can be constructive or destructive, depending on the

respective path lengths

• Can be more than two paths

31
Today
1. How do wireless and wired networks differ?
– A shared wireless medium
– Less reliable links
– Mobility

2. What makes wireless interesting?

3. What new services does wireless enable?

32
Mobility affects link throughput
• Quality of transmission depends on distance, other factors

• Affects the throughput mobile users achieve

• Worst case: Outages, periods with no connectivity!

Alice

Throughput,
Alice to Bob
Bob
time 33
Mobility matters, even if stationary!
• Mobile people, devices affect wireless channel of
stationary nodes!

Alice
Throughput,
Alice to Bob

time
Bob

Cathy

34
Mobility matters, still!
Cathy
• Impact of mobility on
transmission is complex:
– Multipath effects Alice

Throughput,
Alice to Bob
Bob

time
35
Today
1. How do wireless and wired networks differ?

2. What makes wireless interesting?

– Deep intellectual challenges
– Cross-layer design

3. What new services does wireless enable?

36
Some things are well understood…
Q: What’s the capacity of a point-to-point link?
– Bits/second can ”reliably” communicate

• Before Shannon:
– Only way to make probability of bit error arbitrarily
small is to reduce the rate of communication.

• After Shannon (with some assumptions):

– Up to some rate C (Shannon Capacity), coding
can make chance of bit error arbitrary small!
37
…others aren’t understood well at all!
Q: What’s the capacity of a wireless network?

A [Information Theory]: “ ”
A [Computer Networks]: “Let’s build a better medium
access control protocol!”
38
Today
1. How do wireless and wired networks differ?

2. What makes wireless interesting?

– Deep Intellectual challenges
– Cross-layer design

3. What new services does wireless enable?

39
The argument for cross-layer design
• Traditional approach: • New Approach: Design
Optimize within layers and optimize across layers

Applications Applications
End-to-End
Transport Layer
(including Transport)
Communications
Network Layer
and Coding
Radio Hardware,
Physical Layer
Antennas
40
Why is layer separation sub-optimal?

why is point-to-point

41
 Scenario: Laptop in a Dead Spot
Scenario: Laptop in a “dead spot”
X
010101011111 X
011101011011
Loss
Loss
01010110111
With Layer Separation 01010110111
But access points are
a few bit errors à persistent loss
unlikely to have same bit
With layer separation: error
A few bit errors à
persistent loss

But two access points

unlikely to experience
error in the same bit

42
 Scenario: Laptop in a Dead Spot
Solution: A cross-layer Approach
X
010101011111 X
011101011011
High-speed Ethernet
Loss
Loss
01010110111
With Layer Separation 01010110111
But access points are
a few bit errors à persistent loss
unlikely to have same bit
error

• Physical layer delivers partially-correct packets

• Link/network layers combine correct bits across
different access points to correct errors
43
 Scenario: Laptop in a Dead Spot
A challenge for bit combining
X
010101011111 X
011101011011
High-speed Ethernet
Loss
Loss
“First
With Layer Separation
bitbitiserrors
1” à persistent loss But access points are
a few “First
unlikely to have same bit
error bit is 0”

Which access point

should we believe?

44
Idea: Network cooperates with PHY layer

• Physical layer already Confidence

estimates probability of
correctness (confidence) 1
in each 0/1 bit decision 0
01010110111
• Idea: Expose confidence
to the higher layers

Physical Layer

45
 Scenario: Laptop in a Dead Spot
Solution: Use confidences across layers
X
010101011111 X
011101011011
High-speed Ethernet
Loss
Loss
With Layer Separation
“First bit isà1 persistent
with loss But access points
“First bitare
is 0 with
a few bit errors
unlikely to have same bit
0.8 confidence” error 0.5 confidence”

Believe the access point

with higher confidence
46
Experiment: Packet
Experiment: delivery v. poor coverage
Packet Delivery vs. Poor Coverage

Fraction of Packets Delivered

Average
Average Bit Bit Errors
Error Rate
47
Experiment: Packet
Experiment: delivery v. poor coverage
Packet Delivery vs. Poor Coverage

Fraction of Packets Delivered

Layer Separation

Average Bit Bit

Average Error Rate
Errors
48
Experiment: Packet
Experiment: delivery v. poor coverage
Packet Delivery vs. Poor Coverage

Fraction of Packets Delivered

Cross-layer Approach

Layer separation

Average Bit Bit

Average Error Rate
Errors
49
Fundamental change in network architecture
• Traditional approach: • New Approach: Design
Optimize within layers and optimize across layers

Applications Applications
End-to-End
Transport Layer
(including Transport)
Communications
Network Layer
and Coding
Radio Hardware,
Physical Layer
Antennas
50
Today
1. How do wireless and wired networks differ?

2. What makes wireless interesting?

3. What new services does wireless enable?

51
New Services: Wireless localization
New Services: Wireless Localization
• GPS does not work indoorà
GPS does not work indoors Use WiFi
à use Wi-Fi to localize
to localize.

Indoor Navigation Business Analytics

WiFi Geofencing Indoor Robotic Navigation

52
How Do We Get Virtual Touch Screens?
How do we get virtual touch screens?

53
RFID motion tracking in the air

54
RFID technology: Smart homes
Smart Homes

• Localize everything and

anything in the home

RFID-tagged
Laptop Charger • RFID: Battery-free stickers
to tag any and every object
RFID-tagged
Laptop

RFID-tagged
Handbag

55
Low Power, Wide Area Networks
• Wireless, battery operated
sensors and devices

• Long Range: Kilometers

• Long lifetime: 10-20 years
• Low cost

• Applications:
– Smart electricity, water,
utility meters
– Asset tracking: e.g.
Vehicle fleet monitoring
– Agriculture monitoring [semtech.com]

56
Class Website
www.cs.princeton.edu/courses/archive/spring18/cos463

57
 Thursday
Systems & Networking Concepts: Layering,
End-to-End Arguments
Transport over Wireless I: TCP Split Connection

Friday Precept
Python Intro and Signal Processing Primer
Location: 87 Prospect Street, Room 065

58