Ad Hoc Networks:#

Architectures & Protocols

Dr Ljiljana Simić
iNETS, RWTH Aachen University
SS2016

Important  Note:  
These   course   notes   may   contain   some   copyrighted  
material.   The   copyright   of   this   material   covers   its  
use   in   class   and   for   educa7onal   purposes,   but   you  
are   not   allowed   to   distribute   this   course   material  
freely.   Under   the   code   of   appropriate   use,   please  
refrain   from   uploading   the   provided   source   files   or  
documents  to  any  publicly  accessible  system  outside  
RWTH  Aachen  University  without  prior  permission.  

1
 #
   
WELCOME TO THE COURSE!
 
The  course  should  be  fun  …  
 
 
 
 
 
 
 
 
 
 
 
 
 

#
   
WELCOME TO THE COURSE!
 
…  and  hopefully  you  will  learn  something  during  it.  
 
 
 
 
 
 
 
 
 
 
 
 
 

2
 Teaching Staff
Lecturer:    
   Dr  Ljiljana  Simić  
           [email protected]­‐aachen.de  
 
 
 

TA:    
   Mr  Fei  Liu,  M.Sc.    
 (          [email protected]­‐aachen.de  
             
             
               

Logistics

Lectures:    Thu  15.15  –  16.45    

                 
Exercises:    Thu    17:00  –  17:45  
     (NOT  today,  21.04.2015)  
 
       
 
 
 

3
 Exam
   
§  oral  exam    
-  Tue  2nd  August  
-  Wed  3rd  August  
-  Thu  4th  August  

§  per-­‐student  schedule  announced  towards  end  of  semester  

§  if  you  have  a  CLASH  with  another  exam,  contact  TA  NOW  
       
 
 
 

L2P
§  we  are  using  L2P      
§  big  announcements  are  also  sent  by  email  
 
§  in  principle  course  pages  include:  
§  announcements  
§  notes  on  the  homework  
§  possibility  for  discussion  forum  
§  lecture,  homework  &  extra  material  

§  you  must  register  on  L2P!  

4
Check your RWTH email#
regularly!

How do I get 1.0 or A+?

5
 How do I get 1.0 or A+?
§  prerequisites:  no  strict  ones    
§  but  you  should  know  something  about  computer  
networks  and  something  about  programming  
 
§  we  expect  a  reasonable  amount  of  work,  but  also  credit  it  
§  homework  assignments  (typically  w.r.t.  research  papers)    
§  exam  (a.k.a.  “reality  check”)  

§  in  the  homework  collabora've  learning/discussion  allowed  

§  BUT  copying/cheaRng  is  NOT!  
§  you  copy  =  you  get  ZERO  for  that  week’s  homework  
§  plus  it  is  pointless:  we  will  find  what  you  know  in  the  exam!  

“The dog ate my homework…”

6
 Course material
 
§  no  textbook  –topic  is  combina7on  of  several  different  fields  
and  no  perfect  textbook  exists  
 
§  also  s7ll  in  the  R&D  phase  
§  s7ll  emerging  area  –  no  “right  solu7ons”  
§  we  encourage  open  discussion!  
   
§  we  will  rely  on  lecture  slides  &  original  research  papers  

Relevant textbooks
 
C.K.  Toh  (2002)  Ad  Hoc  Mobile  Wireless  Networks:  Protocols  and  Systems,  Pren7ce-­‐Hall  
§  nice  general  introduc7on  to  ad  hoc  networks,  but  does  not  cover  everything  we  want,  
also  somewhat  “old”  
 
C.  Perkins,  Ed.  (2001)  Ad  Hoc  Networks,  Addison-­‐Wesley    
§  essen7ally  ‘just’  a  combina7on  of  ar7cles  on  rou7ng.  

C.  Siva  Ram  Murthy  &  B.S.  Manoj  (2004)  Ad  Hoc  Wireless  Networks:  Architectures  and  Protocols,  
Pren7ce-­‐Hall  
§  850  pages.  probably  best  of  the  bunch…but  it  is  a  bit  heavy  
 
 
Feng  Zhao  &  Leonidas  Guibas  (2004)  Wireless  Sensor  Netoworks:  Informa'on  Processing  
Approach,  Morgan  Kaufman  
 
H.  Karl  &  A.  Willig  (2007)  Protocols  and  Architectures  for  Wireless  Sensor  Networks,  Wiley  
 

7
 Material and orientation
§  working  somewhere  in  the  middle  of  Electrical  Engineering,  
Computer  Science  and  Computer  Engineering  
 
§  slides  contain  the  main  points  and  some  addi7onal  material  
§  also  some  handouts  that  will  give  useful  background  informa7on  
 
a  very  important  and  key  part  of  the  course  is  formed  by    
original  research  arRcles  

Goals
§  course  is  planned  to  make  sure  that  everyone  has  necessary  
background  informa7on  on  this  field  
§  but  it  will  not  make  you  an  expert,  yet  J  

§  also  designed  to  teach  and  train  you  to  criRcally  read    
research  papers  and  evaluate  ideas  
§  transi7on  to  becoming  a  professional  engineer  
§  this  also  means  reading  efficiently  and  quickly  
 
§  in  small  part  it  will  also  teach  you  how  to  express  your  
ideas  and  present  them  to  others  

8
 Strange devices

SMEAGOL lab @ iNETS

Rough Contents – Part I

§  IntroducRon   hello
§  organiza7on,  goals,  material  
§  history  of  ubiquitous  compu7ng  
§  vision  of  the  field  
§  what  is  an  ad  hoc  network?  
§  how  to  read  research  (scien7fic)  papers  
 
§  Graph  Theory  &  RouRng  Fundamentals  
§  graph  theory  &  basic  rou7ng  review  
§  stable  marriage  problem  
 
§  RouRng  for  Ad  Hoc  Networks  
§  rou7ng:  classical  
which way
§  rou7ng:  DSR,  AODV,  ZRP,  taxonomy   do I go?
§  broadcast  storms  
§  improved  metrics,  energy  awareness  
§  link  reversal  algorithm(s)  
 

9
 Rough Contents – Part II
 
§  MAC  Layer  for  Wireless  Networks  
§  historical  evolu7on:  MACA,  MACAW,  CSMA/CA  
§  hidden  terminal  problem  
§  power-­‐aware  MAC  
§  MAC  with  direc7onal  antennas  
 
§  ImplemenRng  Ad  Hoc  Networks  over  Wireless  
§  radio  possibili7es  for  ad  hoc  networks    
§  IEEE  802.11,  IEEE  802.15.4,  Bluetooth  

§  Wireless  Sensor  Networks  

 
 

hello
– Introduction

10
(R)evolution road to ad hoc networks
 
1.  compu7ng  revolu7on  
2.  mobile  wireless  revolu7on  (cellular,  Wi-­‐Fi)  
3.  Internet  revolu7on  

§  next  wave:  intelligently  connecRng  these    

 è  ad  hoc  wireless  networks,  “Internet  of  Things”  
 
§  technical  challenges  for  communica7ons  engineers:  
-­‐  low  power  
-­‐  no  infrastructure  (self-­‐organizing)  
-­‐  mobility  

The “Future” …

- telemedicine
- video phone

- smart watch

11
 “Cool” Internet appliances
world’s smallest#
web server
http://www-ccs.cs.umass.edu/
shri/iPic.html

web-connected printer#
http://littleprinter.com/

Kurose & Ross 2003

IP picture frame
http://www.ceiva.com/

Wi-Fi enabled smart toaster#

http://legrandours.com/3924/647021/gallery/jamy-smart-toaster

Convergence? ...
First  Color  TV  
Broadcast,  1953  
HBO  Launched,     InteracRve  TV,    
1972   1990  

Telephone,  1876  

Early  Wireless   Handheld  Portable  

Phones,  1978   Phones,  1990  

ENIAC,  1947   WinTel  

PenRum  
PC,  1993  
Computer   First  PC   IBM   Apple   Apple   IBM  
+  Modem   Altair,   PC,   Mac,   Powerbook,   Thinkpad,  
1957   1974   1981   1984   1990   1992  
Apple  
Newton,  
1993  
HP  Palmtop,  
1991  
 
Red  Herring,  10/99

12
 … divergence and competition
Atari  Home  
Pong,  1972  

PenRum   Network   Free   Sega   Internet-­‐enabled  

PC,  1993   Computer,   PC,  1999   Dreamcast,     Smart  Phones,    
1996   1999   1999  
Game  Consoles  
PenRum  II   Personal  Digital  Assistants  
PC,  1997   Digital  VCRs  (TiVo,  ReplayTV)  
Apple   E-­‐Toys  (Furby,  Aibo)  
iMac,  1998   Smartphones  
Tablets  
Palm  VII   E-­‐book  Readers  
PDA,  1999    

PROLIFERATION OF DIVERSE#
END-DEVICES & ACCESS NETWORKS

 
Red  Herring,  10/99

Information appliances
 
§  different  design  constraints  based  on  
intended  use,  enhances  ease  of  use  
§  desktop  PC,  mobile  PC  
§  desktop  “smart”  phone  
§  mobile  telephone  
§  tablet  
§  set-­‐top  box,  digital  VCR  
§  in-­‐car  naviga7on  
 
§  implicaRons:    
§  shio  from  computer  design  to  consumer  design  
§  heterogeneous  “standards”  &  hybrid  networking  
§  interacRve  networking,  access  on  demand,  QoS  

13
 New network beyond the PC
Internet  computers   500+  
million   today’s  Internet  
Internet  users   2+  billion  

automobiles  
663  million  

telephones   embedded  
ca.  5.2  billion  (4  billion  mobile)  
networks  
electronic  chips  
30+  Billion  

Internet of Things
§  by  2020,  predicRons  of  20-­‐50  billion  devices  connected  to  Internet  

14
 Wireless data transmission
Fibre  op7c  
Wired   IR  
Mbps   100  
MBR  -­‐  Mobile  
  Broadband  Network  
DSL   10     (LTE)  
WLAN  
   
1  
UMTS  –  3G  
Cordless  Phone  
0.1   EDGE  –  2.5G  

GSM,  IS-­‐95,...  –  2G  

0.01  
Room   Building   Fixed   Walking   ’car’  
Indoor   Outdoor  
plenty of room at the bottom!

A new computer class emerging

log  (people  per  computer)  

Mainframe  

Minicomputer  

WorkstaRon   “smart  dust”  

PC   ad  hoc  wireless  
sensor  networks  
Laptop  
smartphone   ???  

year  

15
 Ubiquitous computing
§  Mark  Weiser’s  1996  hallmark  paper:  
“Coming  Age  of  Calm  Technology”  

§  basic  no7on  could  be  summarized  as  the  observa7on:  

§  mainframe  –  many  people  share  a  computer  
§  personal  computer  –  one  computer  per  person  
             …..internet,  distributed  compu7ng….  
§  ubiquitous  compuRng  –  many  computers  share  us  

§  “calm  technology”  –  technology  moves  from  our  “visible”  

reality  to  the  background,  and  becomes  almost  invisible  –  and  
bounces  between  our  aren7on  and  non-­‐aren7on  

aside: Mark Weiser

§  chief  technologist  at  Xerox  PARC      
 

§  conceived  the  idea  of  Ubiquitous  Compu7ng  in  1988  

 

§  Netscape  (1994),  Gopher  (1991),  Windows  3.1  (1992)  

16
 aside: Xerox PARC
§  almost  mythical  laboratory  and  quality  in  the  systems    
research  (see  e.g.  book  “Dealers  of  Lightning”)  
§  Bob  Taylor  lead  CSL  (leo  PARC  later)  

§  PARC  researchers  invented:  

§  personal  computers  –  Alto  
§  mouse  
§  windows  –  Star  
§  bitmapped  terminals  
§  icons  
§  Ethernet  
§  Smalltalk  
§  Bravo  –  first  WYSIWYG  program  
§  laser  printer  
 
 

Weiser: “Future Technologies”

§  screens:  80  dpi  (10  million  pixels)  
§  technologically  doable  in  2003  (we  have  those)  
§  now  cheap  enough  for  everyday  consumer  use  
§  2048  x  1900  ~  3,9M  pixels    (ca.  250  €)  

§  processors  (embedded)  

§  1,000  MOPS  @  low  power  
§  200  MOPS  (200MHz)  available  and  scaling  @low  power  

§  hard  disks  for  size  of  “matchbox”  with  60MB  

§  smaller  ones  available  at  4+  GB  

17
 Internet of Things
§  by  2020,  predicRons  of  20-­‐50  billion  devices  connected  to  Internet  

Internet of Everything?

18
 Internet of Everything?

§  smart  home  

§  smart  city  
§  healthcare  
§  logis7cs  
§  environment  
§  industrial  automa7on  (Industry  4.0)  

Design for embedded is tough!

19
Miniaturization and microelectronics
               Itanium2  (241M)  

nearly  a  thousand  8086’s  

would  fit  in  a  modern  
microprocessor  

MEMS  
on  chip  

Actuation Sensing CommunicaRon  

Processing  &  
     Storage  
I  SD  Q  SD  
PLL   baseband  
filters  
mixer  
LNA  

Radios at present

§  433,  866  MHz  ISM  

§  2.4  GHz  WLANs  
§  5  GHz  WLANs  
§  Bluetooth  
§  RFID  tags  
§  <40  MHz  FM-­‐radios  
 

20
 What is an ad hoc wireless network?
§  ad  hoc  (La7n)    =    created  for  a  parRcular  purpose  
§  wireless            =    mobile  hosts  (&  unreliable  links)  
§  network            =    informaRon  sharing  

§  “without  infrastructure”,  “self-­‐organizing”  

 
 
 
  ad  hoc  wireless  network:  informa7on  sharing  
between  mobile  hosts  for  a  par7cular  purpose  
 
 
 
 
§  also  “MANET”  (mobile  ad  hoc  network)  

§  latest  big  rush  towards  IoT,  “wireless  sensor  networks”  

How a MANET works

shared  data  
connec7on   user  

two-­‐way  
communicaRon  
mulR-­‐hop  
communicaRon  

radio  
signals  

resource  sharing  
network  cloud   physical  wireless  local  connecRvity  

21
 Explanation of a MANET
§  no  need  for  fixed  infrastructure  
§  each  node  equipped  with  one  or  more  radios  
§  radios  can  be  heterogeneous    
§  each  node  free  to  move  about  while  communica7ng  
§  paths  between  nodes  can  be  mulR-­‐hop  

Applications of MANETs
           current  applicaRons              emerging  applicaRons  
§  shared  whiteboard   §  electronic  payments  from  
applica7on  (office   anywhere  
workgroup)   §  habitat  monitoring  
§  mul7-­‐user  games   §  high-­‐precision  agriculture  
§  robo7c  pets   §  biomedical  
§  PAN/home/office  wireless   §  earthquake  monitoring  
network   §  industrial  automa7on  
§  communica7ons  for   §  smart  home  
emergency/disaster  relief  
§  security  
§  what you invent!

22
MANET Advantages & Disadvantages
§  dynamic  topologies  
§  bandwidth  
§  restricted  7me-­‐bounded  service  (voice,  video)  
§  loose  7me-­‐bounded  service  (data)  
§  energy-­‐constrained  opera7on  
§  limited  physical  security  

Goals of MANET designer

§  discover  links  

§  build  paths  
§  maintain  shared  data  
§  provide  required  Quality  of  Service  (QoS)  

23
 And mobility…

What is mobility in MANETs?

§  a  user/device  that  moves:  
§  between  different  geographical  loca7ons  
§  between  different  networks  
§  between  different  communica7on  devices  
§  between  different  applica7ons  

24
 Mobility means changes in …
§  hardware  
-  lighter  
-  more  robust  
-  lower  power  
§  wireless  communicaRon  
-  can’t  tune  the  network  for  sta7onary  access  
§  network  protocols  
-  name/address  changes  
-  delay  changes  
-  error  rate  changes  

Mobility means changes in …

§  fidelity/data  consistency  
-  strong  consistency  no  always  possible  
§  locaRon/transparency  awareness  
-  transparency  not  always  possible  
§  security  
-  lighter-­‐weight  algorithms  
-  end-­‐point  authen7ca7on  harder  
-  devices  more  vulnerable  

 
 

25
 Mobility, in conclusion
generally,  mobility  stresses  all  resources  further:  
-  CPU  
-  power  
-  bandwidth  
-  delay  tolerance  
-  radio  spectrum  
-  physical  size  
-  loca7ons  for  device  placement  
-  constraints  on  peripherals  and  GUIs  
-  human  aren7on  
 

Design issues in Ad Hoc networks#

example: wireless sensor networks (WSN)
§  design  constrains  of  WSN:  highly  energy-­‐constrained,  
low-­‐complexity  collabora7ve  nodes  
§  unlike  conven7onal  wireless  networks,  the  protocols  
designed  for  WSN  must  allow  closer  collaboraRon  or  
awareness  among  the  layers  of  the  protocol  stack,  
especially  L1-­‐L3  

Application
Transport
Network
Data Link
Physical

26
 Design issues in Ad Hoc networks#
example: wireless sensor networks (WSN)
§  e.g.  WSN  MAC  protocol  puts  radio  transceivers  in  
sleeping  mode  as  much  as  possilbe  to  save  energy  
§   but  if  MAC  not  jointly  designed  with  rou7ng  algorithms,  
poor  overall  performance  e.g.  excessive  packet  delay  

Application
Transport
Network
Data Link
Physical

Design issues in Ad Hoc networks#

example: wireless sensor networks (WSN)
§  conversely,  e.g.  data-­‐centric  rouRng  algorithms  with  data  
aggregaRon  create  special  requirements  on  MAC    

Application
Transport
Network
Data Link
Physical

27
 Design issues in Ad Hoc networks#

MANETs: new networking paradigms

Application
Transport
Network
Data Link
Physical

Reading homework
historical  perspecRve  on  visions  of  future  technology  (on  L2P):  

§  M.  Weisser  and  J.  Seely  Brown,  “The  coming  age  of  calm  technology,”  
Xerox  PARC,  1996.  
 
§  M.  Weisser,  “The  computer  for  the  21st  century,”  Scien'fic  American,  
1991.  

§  J.  C.  R.  Licklider,  “Man-­‐Computer  Symbiosis,”  1960.  

§  J.  C.  R.  Licklider,  “The  Computer  as  a  Communica7on  Device,”  Science  
and  Technology,  1968.  

§  V.  Bush,  “As  we  may  think,”  The  Atlan'c  Monthly,  1945.  

28
 HP CoolTown
§  People,  Places,  Things:  Web  Presence  for  the  Real  World.  
Cooltown  Project  at  HP.  In  WMCSA  2001  

§  CoolTown  Project  

§  what:  HP  wanted  to  build  some  sort  of  bridge  between  electronic  and  
real  physical  world  
§  ra7onale:    
§  web  is  transparent  access  because  of  standardiza7on  of  hrp,  TCP/IP,…  
§  web  access  will  be  ubiquitous  
§  a  lot  of  useful  informa7on  and  content  is  already  in  Web  
§  where:  homes,  offices,  shops  
§  people:  interact  with  visitors  etc.  in  different  places  
§  things/services:  we  find  those  and  are  able  to  control  them  

§  hnp://www.youtube.com/watch?v=U2AkkuIVV-­‐I  

Remember …
§  register  on  L2P  
§  if  you  cannot,  email  Fei  Liu  and  he  will  add  you  
§  reading  for  your  homework  will  be  posted  here  
 

29
– How to NOT plagiarise

Plagiarism: DON’T!

30
  
Plagiarism: DON’T!
§  in  your  homework  assignments,  you  will  ooen  be  asked  ques7ons  
on  exis7ng  original  work  
§  to  help  you  answer,  you  are  also  encouraged  to  consult  other  
sources  (i.e.  from  the  scien7fic  literature  &  in  some  (rare)  cases  
white  papers,  company  websites,  blogs,  etc.)  

§  BUT  you  have  to  answer  in  ALL  YOUR  OWN  WORDS  
 
 

§  copy/paste  from  any  source  =  PLAGIARISM  

                       =  ZERO  FOR  ASSIGNMENT  

 
Plagiarism: DON’T!
 

§  in  general,  any  work  you  submit  in  your  name  must  be  in  ALL  
YOUR  OWN  WORDS  
 

§  copy/paste  from  any  source  =  PLAGIARISM!  

         
§  YOU  MUST  BOTH  CITE  YOUR  SOURCE  &  PARAPHRASE  
§  hrp://web.mit.edu/academicintegrity/wri7ng/paraphrasing.html  
 
§  no  citaRon/insufficient  paraphrasing  
         ALSO  =  PLAGIARISM!  
 

31
 In-text citations: HOW-TO
§  in  general,  provide  a  list  of  References  at  the  end  of  your  
technical  wri7ng/engineering  report/thesis,  accompanied  by  
explicit  in-­‐text  citaRons  of  those  sources,  for  the  purpose  of:    
§  credibility  (support  for  facts  you  are  giving,  how  reader  can  trust  your  
report:  "says  who?")    
§  acknowledgement  (of  source  of  idea  or  proposal  -­‐  avoiding  plagiarism)    
§  further  reference  point  for  reading  about  or  checking  a  given  fact  or  
proposal.    
§  e.g.  "This  is  the  best  prospect  of  allevia7ng  the  spectrum  
                 crunch  [4,  15-­‐17].”  
§  note  that  in-­‐text  cita7on  is  NOT  only  used  for  direct  quota7ons  or  
defini7ons  of  terms!  
§  whenever  an  idea  is  not  your  own,  you  must  clearly  iden7fy  the  source  of  
the  idea/informa7on  

Plagiarism in your h/w assignment

submission: consequences
 

§  1st  7me  =  ZERO  points  for  that  assignment  

 
§  2nd  7me  =  ZERO  points  for  all  assignments,  
                     your  assignments  will  no  longer  be  marked  

PLEASE  DO  NOT  WASTE  OUR  TIME!  

32
 – How to read a #
research paper

Anatomy of a research paper

§  typical  paper  includes:  
§  Abstract  
§  IntroducRon  
§  mo7va7on,  problem  descrip7on  
§  research  ques7ons  that  are  being  addressed  by  this  paper  
§  experiment  setup/  system  model  
§  results  &  analysis  
§  discussion  
§  Conclusions  (incl.  future  work)  
 
§  apart  from  these  some  “white  papers”  and  reports  might  have  
§  Execu7ve  summary  (=  Abstract)  
§  Appendices    

33
 Why do you read a paper?
§  understand  and  learn  new  contribu7ons  
§  BUT  
-­‐  not  all  papers  are  “good”  
-­‐  not  all  papers  are  “interes7ng”  
-­‐  not  all  papers  are  “worthwhile”  for  you  
§  also  only  a  small  frac7on  of  papers  are  “easy  reading”  
-­‐  and  one  should  not  take  them  as  exemplars  of  English  prose  
 

§  you  have  to  learn  to  iden7fy  a  good  paper  &  spend  your  7me  
wisely  
1.  Breadth  
2.  Depth  
3.  React  

How to read a research paper: breadth 1st

§  ask  yourself,  what  is  this  paper  about?  (breadth)  

§  read  the  Title  and  the  Abstract  
-­‐  if  you  s7ll  don’t  know  what  this  paper  is  about,  
then  this  might  be  a    bad  paper.  
§   read  the  Conclusions  
-­‐  are  you  now  sure  you  know  what  this  paper  is  about?    
-­‐  if  not  it  is  a  BAD  paper.    
-­‐  in  this  course  we  MOSTLY  avoid  such  papers.    
§  read  the  IntroducRon  
§  read  the  secRon  headings  
§  read  tables  and  graphs  and  capRons  -­‐  see  what  they  say  

34
How to read a research paper: breadth 1st

§  see  who  wrote  it,  where  it  was  published,  when  was  it  wriren  
(credibility)  
-­‐  note  that  this  is  a  problema7c  thing,  but  credibility  is  there  to  
take  in  account.  
 
§  skim  References  to  see  if  the  authors  are  aware  of  relevant  
related  work;  see  if  you  know  the  relevant  work;  see  if  you  
know  a  relevant  work  that  they  didn’t  refer  to  

How to read a research paper: depth 2nd

§  approach  with  scienRfic  skepRcism  

-­‐  but  use  healthy  skep7cism!  
-­‐  skep7cism  do  not  mean  rejec7ng  everything  
 

35
How to read a research paper: depth 2nd

§  examine  the  assumpRons:  

§ do  their  results  rely  on  any  assump7ons  about  trends  in  
environments?  
§ are  these  assump7ons  reasonable?  
-  e.g.  “Lets  assume  that  there  are  billions  of  powerful  
computers,  connected  by  a  high  speed  network,  spread  
across  the  world,  our  system  will  …”  
-  e.g.  “Our  system  can  enable  you  to  run  Windows  Vista  
on  a  33MHz  Intel  386  with  640K  main  memory”  
-  e.g.  “We  assume  that  40Tbit/s  wireless  links  become  
commonly  available  next  year  for  this  1€  appliance”  

How to read a research paper: depth 2nd

§  examine  the  methods:  

§  did  they  measure  what  they  claim?  

§  can  they  explain  what  they  observed?  

§  it  is  easy  to  dump  your  experimental  results  on  the  
paper;  as  a  reader  you  want  an  analysis  of  why  the  
system  behaves  a  certain  way,  not  the  raw  data.  
§  did  they  have  adequate  controls?  
§  were  tests  carried  out  in  a  standard  way?  were  the  
performance  metrics  standard?  if  not,  do  they  explain  their  
metrics  clearly?  

36
How to read a research paper: depth 2nd

§  examine  the  staRsRcs:  (there  is  truth,  lies,    and  then  there  is  
sta7s7cs!)  
§  were  appropriate  staRsRcal  tests  applied  properly?  
§  did  they  do  proper  error  analysis?  
§  are  the  results  staRsRcally  significant?  
§  common  mistake:  “We  performed  our  experiment  once  
at  4  am  and  no'ced  a  ten  fold  improvement.  Thus  we  
conclude  that  our  system  is  beZer”  
§  be  very  careful  with  percentages  
§  Method  A:  0.01  seconds,  our  Method:  0.005  seconds  
§  “our  method  shows  100%  improvement  over  method  A!”  

How to read a research paper: depth 2nd

§  examine  the  conclusions:  

§  do  the  conclusions  follow  logically  from  the  results?  
§  “We  performed  our  experiments  with  8  palm  pilots  and  
saw  a  10  fold  improvement.  Hence  we  conclude  that  our  
system  will  scale  to  millions  of  palm  pilots”  
§  what  other  explanaRons  are  there  for  the  observed  effects?    
§  what  other  conclusions  or  correlaRons  are  there  in  the  data  
that  the  authors  did  not  point  out?  
§  “Earlier  work  performed  experiments  using  a  2  Mbit  
wireless  network.  Our  system  (incidentally)  used  a  11  
Mbit  network  and  saw  a  5  fold  improvement.  So  our  
technique  works!”  

37
 How to read a research paper: react 3rd

§  take  notes  

§  highlight  major  points  
§  react  to  the  points  in  the  paper  
§  place  this  work  with  your  own  experience  
§  if  you  doubt  a  statement,  note  your  objecRon  
§  if  you  find  a  pleasing  quotaRon,  write  it  down  
§  construct  your  own  example  
§  summarize  what  you  read  
§  maintain  your  own  bibliography  of  all  papers  that  you  ever  
read  

How to summarize/review a paper

§  how  to  summarize:  do  not  rewrite  the  abstract!  
§  it  is  useless  to  rewrite  the  abstract:  your  summary  should  go  
beyond  that  and  also  include  your  “reac7ons”  

§  how  to  classify:  

§  find  suitable  keywords,  phrases  to  classify  main  topic(s)  of  the  
paper  

§  how  to  review  (referee  process):  

§  is  this  seminal  (trailblazing)  paper?  
§  one  step-­‐further  on  XYZ  
§  yet-­‐another-­‐paper  on  XYZ  
§  visionary  paper:  but  is  it  real  vision  or  “hot  air”  
§  cri7cism  should  be  put  into  the  form  of  sugges7ons:    
§  “how  to  make  the  paper  berer”  
§  readability  (not  only  a  marer  of  taste)  

38
 How to write a research paper
§  write  it  such  that  anyone  who  reads  it  using  the  method  we  
discussed  understands  your  ideas.  

§  clearly  explain  what  problem  you  are  solving,  why  it  is  
interes7ng  and  how  your  solu7on  solves  this  interes7ng  
problem    

§  be  crisp  -­‐  explain  what  your  contribu7ons  are,  what  your  ideas  
are  and  what  are  others’  ideas.  

39