A Brief History of Communication

David Tse

Communication Systems What goes into the engineering of these systems?

Key Ingredients• Software• Hardware• Communication architecture, with

coding and signal processing algorithms

Communication channels can be very nasty!

Channel distortion, noise, interference……How do we communicate reliably over such channels?

Communication has a long history

• Smoke signals, telegraph, telephone…

• 1895: invention of the radio by Marconi

• 1901: trans-atlantic communication

State of affairs:Early 20th century

• Most communication systems are analog.

• Engineering designs are ad-hoc, tailored for each specific application.

Big Questions• Is there a general methodology for

designing communication systems?

• Is there a limit to how fast one can communicate?

Harry Nyquist (1928)• Analog signals of

bandwidth W can be represented by 2W samples/s

• Channels of bandwidth W support transmission of 2W symbols/s

From CT to DT Nyquist converted the continuous-time

problem to a discrete-time problem.

But has he really solved the communication problem?

No. You can communicate infinite number of bits in one continuous-valued symbol!

Claude Shannon (1948) His information theory

addressed all the big questions in a single stroke.

Randomness Shannon thought of both

information sources and channels as random and used probability models for them.

encodersource channel decoder

Everything is bits Shannon showed the universality

of a digital interface between the source and the channel.

source SourceEncoder

Channel encoder channel

Bit stream

Information is like fluid• Every source has an entropy rate H

bits per second.• Every channel has a capacity C

bits per second• Reliable communication is possible

if and only if H < C.

Simple example: binary symmetric channel

0

1

0

1

1-p

p1-p

p

1

p

C

10.50

C = 1+ p log p + (1-p) log (1-p)

Initial Reactions Engineers didn’t understand what he was

talking about.• People were still stuck in the analog world.• Complexity way too high for

implementation technology of the day.• He didn’t really tell people exactly how to

design optimal communication systems.

50 years later….• Our communication infrastructure

is going fully digital.• Most modern communication

systems are designed according to the principles laid down by Shannon.

Internet

S

D

Lessons for Us• Think different• Think big• Think simple

Mobility in Embedded Networks

People and their stuff

Transportation systems(e.g., cars)

Environmental and wildlife monitoring(e.g., Princeton ZebraNet Wildlife Tracker)

Inventory and supply chainmanagement (e.g., RFID tags)

Ubiquitous Mobility• All nodes potentially move

– Network topology changes with time• Efficient routes require knowledge of

topology– Control traffic: distance vector or link state

updates, flooded discovery packets, …

Shanno• Scaling to large networks?– How costly is the dissemination of enough

information to allow for “reasonably good routes”?

– Does control traffic grow more quickly than capacity of the network?

Position-based Routing• Position-based routing:

– Geographic coordinates rather than graph to make routing decisions

• Local routing decisions based on positions of destination and neighbors

• Separation into– Location service: where is the destination?– Local routing protocol: select next hop

towards destination

Bla• Bullet 1• Bullet

– Bullet 2.0– Bullet 2.1

• Bullet 3

Location Services• Challenge: construct a distributed

database out of mobile nodes• Approaches:

– Virtual Home Region: hash destination id to geographic region -> rendez-vous point for source and dest (Giordano & Hamdi, EPFL tech. report, 1999)

– Grid Location Service: quad-tree hierarchy, proximity in hashed id space (Li et al., Mobicom 2000)

– DREAM: Distance Routing Effect Algorithm (Basagni & Chlamtac & Syrotiuk, Mobicom 1998)

Last Encounter History• Question:

– Do we really need a location service?• Answer:

– No (well, at least not always)• Observation:

– Only information on network topology available for free at a node is local connectivity to neighboring nodes

– But there is more: history of this local connectivity!• Claim:

– Collection of last encounter histories at network nodes contain enough information about current topology to efficiently route packets

Last Encounter Routing• Can we efficiently route a packet from a source to

a destination based only on LE information, in a large network with n nodes?

• Assumptions:– Dense encounters: O(n^2) pairs of nodes have

encountered each other at least once– Time-scale separation: packet transmission (ms) <<

topology change (minutes, hours, days)– Memory is cheap (O(n) per node)

• Basic idea:– Packet carries with it: location and age of best (most

recent) encounter it has seen so far– Routing: packet consults entries for its destination along

the way, “zeroes in” on destination

Definition: Last Encounter Table

A

B

encounter at Xbetween A and B at t=10

B: loc=X, time=10C: ...

A: loc=X, time=10C: ...D: ...

X

Fixed Destination

A

Moving Destination

A

A

A

AA

A

Exponential Age Search (EASE)

time

-T

0

?

source destination

EASE: Messenger Nodes

time

-T

0

-T/2

EASE: Searching for Messenger Node

time

-T

0

-T/2

Search: who has seendest at most T/2 secs ago?

EASE: Forwarding the Packet

time

-T

0

-T/2

Forwarding towards new positionwith T:=new encounter age

EASE: Sample RouteDef:

anchor point of age T = pos. of dest. T sec ago

EASE:- ring search nodes

until new anchor point of age less than T/2 is found

- go there and repeat with T=new age

src

dst

Performance of EASE• Length of routes clearly depends on

mobility process– Cannot work without locality– Counterexample: i.i.d. node positions every

time step• Model:

– 2-D lattice, N points, fixed density of nodes– Each node knows its own position– Independent random walks of nodes on

lattice• Cost = forwarding cost + search cost

Cost of EASE Routes• Claim:

– The asymptotic expected cost for large N of EASE routes is on the order of shortest route, i.e., total forwarding cost is O(shortest path):

• Forwarding cost:– Geometric series of ages -

> geometric series of EASE segments

– Total length = O(shortest path)

Search Cost

• Single step search cost is small compared to forwarding cost:– Show that density of messenger nodes

around current anchor point is high– Depends on:

• Number of unique messenger nodes encountered by destination = O(log T)

• Distance traveled by messenger nodes= same order as destination

Interpretation: Distance Effect and Mobility Diffusion

• Observation: required precision of destination’s location can decrease with distance– DREAM algorithm: exploit distance effect to decrease state

maintenance overhead– When a node moves by d meters, inform other nodes in disk of

radius c*d meters– Relax separation of location service and routing service

• Basic idea behind last encounter routing:– Exploit mobility of other nodes to diffuse estimate of

destination’s location “for free”– Concurrently for all nodes

destination

Improvement: Greedy EASE

Simulation: Random Walk Model

•N nodes•Positions i.i.d.•Increments i.i.d.

Simulation: Random Walk Model

Heterogeneous Speeds: Slow Dest

Heterogeneous Speeds: Fast Dest

Heterogeneous Speeds

Simulation: Pareto Random Walk

•N nodes•Positions i.i.d.•Increments i.i.d.,heavy-tailed distancedistribution

Simulation: Random Waypoint

•N nodes•Positions i.i.d.•Every node has awaypoint•Moves straight towardswaypoint at constantspeed•When reached, newwaypoint selecteduniformly over area

Pareto RW and Random Waypoint

Related Idea: Last Encounter Flooding

• With coordinate system– Last-encounter information: time + place– EASE/GREASE algorithms

• Blind, no coordinate system– Last-encounter information: time only– FRESH algorithm: flood to next anchor

point– Henri Dubois-Ferrière & MG & Martin

Vetterli, MOBIHOC 03

FRESH: Last Encounter Flooding

Summary: Last Encounter Routing• Last Encounter Routing uses position information that is

diffused for free by node mobility– Last encounter history: noisy view of network topology– Packet successively refines position estimate as it moves

towards destination– Mobility creates uncertainty, but also provides the means to

diffuse new information• No explicit location service, no transmission overhead

to update state!– Only control traffic is local “hello” messages– At least for some classes of node mobility, routes are efficient!– Key ingredients: locality, homogeneity, mixing of trajectories

• Rich area for more research:– Prediction– Integration with explicit location services & routing protocols– More realistic mobility models

• Ref: MG & Martin Vetterli, IEEE INFOCOM 03