Transcript Lecture Slides

Announcements

Project progress reports due today.

Homework 2 ready later today –
due 6/2 (next Friday)

Graded HW 1 and solutions ready
shortly.

Third paper summary on ad-hoc
networks due next Wednesday.
Ad-Hoc Wireless
Networks

Main Characteristics







Each node generates independent data
Any node can communicate with any other.
No centralized controller (self-configuring)
Data transmitted in (short) packets
Links typically symmetric.
Nodes may be mobile and/or power constrained.
Typically a large number of nodes
Applications

Battlefield communications

Wireless LANs

Emergency infrastructures

Short-term networks (e.g. convention)

Sensor networks



Medical applications (on-body)
Buildings
Wide area

Cellular phone evolution

Communication infrastructure for
automated vehicles


Automobiles
Airplanes
Widely different channel characteristics,
distances, mobility, and rate requirements.
Design Issues

Link Layer design

Channel sharing (MAC/reuse)

Reliability/QOS

Routing

Network topology

Network management/control
Must exploit synergies
between design layers
Link Layer Issues

Modulation and Coding




Robustness
Rate requirements
Performance
Adaptive techniques


Bandwidth requirements


Typically distributed
Antenna design




Control and communication requirements
Power control


Rate, power, BER, code, framing.
Smart antennas
Multipath mitigation
Multiuser detection
Connectivity

Binary or adaptive.
Channel Access

Frequency-Division

Time-Division

DS Spread Spectrum

FH Spread Spectrum

Frequency reuse
 Bandwidth efficient
 Distributed allocation
 Dynamic channel allocation
hard for packet data
Frequency Division

Fixed allocation inefficient
 Hard
to implement when node
locations dynamically change

Distributed dynamic channel
allocation hard to do

FD typically only used to
create hierarchical networks
Time-Division

Fixed allocation inefficient and
impractical (as in FD)

Aloha



Inefficient
No capture
Carrier sensing


Hidden nodes degrade performance
Busy tone may interfere with
transmission to other nodes.
Busy Tone
Spread Spectrum
Code Assignment

Common spreading code for all nodes
 Collisions occur whenever receiver
can “hear” two or more transmissions.
 Near-far effect improves capture.
 Broadcasting easy

Receiver-oriented
 Each receiver assigned a spreading
sequence.

All transmissions to that receiver use
the sequence.

Collisions occur if 2 signals destined
for same receiver arrive at same time.

Can randomize transmission time.

Little time needed to synchronize.

Transmitters must know code of
destination receiver


Complicates route discovery.
Multiple transmissions for broadcasting.

Transmitter-oriented



Each transmitter uses a unique
spreading sequence
No collisions
Receiver must determine sequence of
incoming packet




Complicates route discovery.
Good broadcasting properties
Poor acquisition performance
Preamble vs. Data assignment
Preamble may use common code that
contains information about data code
 Data may use specific code
 Advantages of common and specific
codes:




Easy acquisition of preamble
Few collisions on short preamble
New transmissions don’t interfere with
the data block
Data link control

Packet acknowledgements needed
 May be lost on reverse link
 Should negative ACKs be used.

Combined ARQ and coding
 Retransmissions cause delay
 Coding may reduce data rate
 Balance may be adaptive

Hop-by-hop acknowledgements
 Explicit acknowledgements
 Echo acknowledgements





Transmitter listens for forwarded packet
Not possible with directive antennas.
Large delays in FIFO queues.
More likely to experience collisions than a
short acknowledgement.
Hop-by-hop or end-to-end or both.
Connectivity

Determining connectivity
 SNR measurements
 Bit/Packet error rate

Connectivity control
 Link
can adapt to maintain
connectivity (adapt rate, power,…)

Interaction with routing protocol.
 Power
increase may affect other
nodes (Bambos technique).

How many connected nodes
constitute a network
 Or,
take what you can get.
Routing (1987)

Flooding





Broadcast packet to all neighbors
Inefficient
Robust for fast changing topologies.
Little explicit overhead
Point-to-point routing


Routes follow a sequence of links
Connection-oriented




Explicit end-to-end connection
Less overhead/less randomness
Hard to maintain under rapid dynamics.
Connectionless


Packets forwarded towards destination
Local adaptation
Route dessemination

Route computed at centralized node



Distributed route computation




Most efficient route computation.
Can’t adapt to fast topology changes.
Each node transmits connectivity
information to other nodes.
Nodes determine end-to-end route
based on this local information.
Adapts locally but not globally.
Nodes exchange local routing tables



Node determines next hop based on
some metric.
Deals well with connectivity dynamics.
Routing loops common.
Routing (1999*)

Table-driven




Destination-sequenced distance-vector
Clusterhead gateway switch routing
Wireless routing protocol
On-Demand Routing





On-demand distance vector routing
Dynamic source routing
Temporally ordered routing
Associativity-based routing
Signal stability routing
*”A review of current routing protocols for ad hoc mobile
wireless networks,” Royer and Toh, IEEE Personal
Communications Magzine, April 1999.
Packet Forwarding

Overhead information
 Routing information
 Packet identifiers
 Priority/delay information
 Tradeoffs in overhead size

Synergies of routing and
packet forwarding with link
layer.
Other Network
Issues

Network Capacity

Admission Control

Interface with wired networks

Security

Upgrades
 Software
changes
 Software
radios
Network Capacity

Capacity limits of ad-hoc 3D
networks.



Data rates per node
Number of nodes
Assumptions






N users uniformly distributed over the interior
of a sphere.
Each user communicates with another user
randomly chosen among all users.
Signal power decays based on free space path
loss.
All users transmit at the same power.
No channel separation or diversity.
Interference acts as additive white Gaussian
noise
Capacity Bounds

The total number of bits that may be
transmitted by all users, per second,
is approximately
C K N
3
Proportional to the cube root of N

Lower Bound


Based on deterministic routing scheme.
Upper Bound


Similar formula
Uses convexity
Lower Bound Proof
Sketch





Estimate the effects of interference in the
limit of large N.
Construct a series of cell tessellations with
useful properties.
Use the weak law of large numbers to
prove the existence of one user in each
cell.
Specify a routing and transmitting scheme
using time sharing.
Determine the capacity of this scheme,
which lower bounds the capacity of the
best scheme.
What has changed
since 1985?

Signal processing is better,
cheaper, and lower power.

More powerful channel codes.

Multiuser detection and smart
antennas.

Signal strength measuring
techniques available in radios.

How would we leverage these
developments to make better
ad-hoc networks?
Sensor Networks

Sensor Networks








Data highly correlated in time and space.
Low homogeneous rates.
Links typically asymmetric.
Data flows to centralized location.
Energy is the driving constraint.
1000-100,000 Nodes
Have a common mission.
Very different from typical ad-hoc networks