Adhoc Wireless Networks

1
Ad-hoc Wireless Networks

• With contributions from
• D. Patel, R. Shah, R. Jain, D. Estrin
• and many others

2
Ad-hoc Wireless Networking is hot

• Tu meeting at BWRC Ad-hoc wireless networking
  in Berkeley and beyond (or ha-hoo)
• 11 faculty in attendance V. Anantharam, D.
  Culler, R. Glaser (CE), J. Kahn, R. Katz, K.
  Pister, M. Potkonjak (UCLA), J. Rabaey, K.
  Ramchandran, P. Varaiya (del), J. Walrand
• Wide range of interests and topics applications,
  protocols, heterogeneous networks, energy,
  optical and RF
• Mailing list of over 40 people established!
• Will be repeated

3
Focus-2000 Conference in Berkeley

• Topic PicoRadio Networks
• Dates June 26-27
• Where Claremont Hotel
• Presentations on applications, networks,
  communications, architectures, and design
  methdology
• Check the BWRC web-site

4
Flow of presentation

• Traditional ad-hoc
• PicoRadio Network requirements
• Diffused routing
• Swarm intelligence
• Geographic routing

5
Multi-Hop Ad-Hoc Networks

• What does it mean to be Ad-Hoc?
• No pre-existing infrastructure (i.e. no wired BS)
• Network topology constantly changing
• What does Multi-Hop mean?
• Occurs when you use intermediate nodes as
  stepping stones to forward packets.
• Why? - Traditionally limited range.
• Node range is limited and sees only a limited
  number of other nodes.
• Require intermediates nodes to forward packets.
• Why? - Save ENERGY!!!

6
Proactive Protocol

• Autonomously tracks routes.
• Passes entire list of destinations to neighboring
  nodes only
• Couples easily with Locationing
• Nodes could send their calculated position in
  update which all nodes send anyway.
• Large amount of overhead, but optimal paths

7
Proactive Routing

• Network actively keeps routes up-to-date even if
  routes are not being used.
• But efficient if routes are used often
• Table Row size 10 bytes per entry
• Update size 6 bytes per entry

Table in node memory, 0(n)
Update broadcasted to neighbors
Updates grow in size as there are more
active destinations!!!
8
Reactive Protocol

• On demand
• Large initial cost, low maintenance cost
• Paths may grow suboptimal
• Positioning information useful if location of
  destination node known.
• Positioning a separate service. And more useful
  to protocol. (Help track neighbors)

9
In General

• Protocols are delay sensitive
• require immediate response for maximum energy
  savings
• need priority
• Finer granularity in cost metric will cause
  network oscillation.
• Find optimum hop/distance and fix power level to
  a constant

10
Proactive Periodic update.
Num of Nodes
40 - MTU 240 bytes 80 - MTU 480 bytes MTU
maximum transmission unit
Num of Nodes
11
DSDV

• Route settling problem due to unfortunate timing.
• Sequence number highest priority then cost
  metric.
• Happens more in well connected net.
• Fixes
• Timers, partial fix.
• WRP, more complex

12
Reactive

• Flooding
• Min (n-1)size
• assume 8 byte packet.
• Perfect Power Control

Packet Format
13
AODV
14
Comparison DSDV, AODV
Bytes. Startup costs
DSDV 6 bytes, AODV 8/12 bytes.
15
Route Maintenance
16
Picoradio Usage Scenarios

• Data Query Requests Give me the temperature
  distribution in the forum
• Request issued by monitor node and propagated
  through the network to the appropriate nodes
• Data is returned over discovered paths
• Spontaneous Data Broadcasts
• Data trigger exceeded
• Periodic emission (e.g. information in museum
  scenario)
• Programming or maintenance requests
• Set nodes to trigger at a particular threshold
• Query operation of the network

17
PicoRadio Addressing Modes

• Contents-based addressing
• id (type, range)
• Covers virtually all of the sensoring, actuator,
  and monitor activities
• Is annotated to specific function, not physical
  node
• Absolute addressing
• Unique identifier of physical network node
  necessary mostly for debugging and verification
  purposes

18
Alternative approaches(1) data-centric routing

• Exploits structure in the data since we cant rely
  on structure in node placement or addresses
• Sensor network is after all a data centric
  application.
• Based on reactive routing schemes

19
Directed Diffusion Concepts

• Application-aware communication primitives
• expressed in terms of named data (not in terms of
  the nodes generating or requesting data)
• Consumer of data initiates interest in data with
  certain attributes
• Nodes diffuse the interest towards producers via
  a sequence of local interactions
• This process sets up gradients in the network
  which channel the delivery of data
• Reinforcement and negative reinforcement used to
  converge to efficient distribution
• Intermediate nodes opportunistically fuse
  interests, aggregate, correlate or cache data

20
Local Behavior Choices

• 1. For propagating interests
• In our example, flood
• More sophisticated behaviors possible e.g. based
  on cached information, GPS
• 2. For setting up gradients
• Highest gradient towards neighbor from whom we
  first heard interest
• Others possible towards neighbor with highest
  energy

• 3. For data transmission
• Different local rules can result in single path
  delivery, striped multi-path delivery, single
  source to multiple sinks and so on.
• 4. For reinforcement
• reinforce one path, or part thereof, based on
  observed losses, delay variances etc.
• other variants inhibit certain paths because
  resource levels are low

21
Illustrating Directed Diffusion
Setting up gradients
Source
Sink
22
Swarm Intelligence (Sc. Am March 00)

• Shortest path algorithm based on observed
  behavior of ants
• Discovery of nodes (food) based on flooding
• Paths marked by scents (pheromones)
• First ant to arrive back found shortest path
• Nodes reinforced by frequent use
• Non-used paths evaporate
• Alternative paths available

23
Centroid Location

• Find edges
• Diffuse pheromone from the edges inward
• Find the lowest concentration using Min/Max
  sharing
• If you have the lowest concentration, turn yellow

Number Of Motes500 Communications Range.8
24
Mote Position Estimation

• Give GPS receivers to some motes and call them
  Basis Motes. Ask them to turn gray.
• Each Basis Mote diffuses its own pheromone
  throughout the group
• The position of any other mote can be estimated
  from the levels of basis pheromones present.

25
Geographical Routing (R. Jain) Motivation and
Main Result
500

• Routing with partial information
• Distributed, highly scalable and adaptive
  algorithm
• Small routing table sizes and communication
  overheads
• Quick adaptation to network changes
• Requirement Each station knows its position

Distance-vector or Link state routing
Table size
Geographical routing
500
Network size
26
Geographical Routing in Networks

• Each node knows its geographical
• position
• Each node knows its neighbors
• Each node learns of a few extra nodes
• Nodes dont know about the global
• topology
• Destination address is a geographical
• position to which the packet is to be
• delivered

A geographical network
Routing Problem To which neighbor does each
node forward the packet so that it gets to its
destination
27
Geographical Routing Algorithm
Routing Table Format

• Routing Algorithm
• Pkt arrives at S for node D at pos. P
• Node S finds the pos. in its routing table to
  which P is closest, and fwds to corres. neighbor
  
• If pos. P closest to node S, pkt is stuck
  initiate Route Discovery
• Route Discovery (BFS or DFS, etc.)
• P1 Finds acyclic path P(S,D)
• P2 Adds D as cell center at all nodes
  P(S,D)
• Initially, neighbors are discovered through hello
  messages

Position
Next hop
Node
Initial Routing Table for node S
28
Example 1 Geographical Routing
Pos(A) (1,1) Pos(B) (2,2) Pos(C)
(3,1) Links A ---- B B ---- C
B
Pos(C)
Pos(B)
---
Pos(C)
Pos(A)
A
C
A
Pos(C)
Pos(C)
C
Pos(C)
---
Pos(A)
---
Pos(B)
B
Pos(B)
B

• A gets a packet for Pos(C)
• A forwards it to B because pos(B) is closer to
  pos(C)
• B forwards it to C because pos(C) is closer to
  pos(C)

29
Example 2 Geographical Routing
Pos(D)
Pos(C)
---
Pos(A) (1,1) Pos(B) (2,2) Pos(C)
(3,1) Pos(D) (2.5,0) Links A ---- B B ----
C C ---- D
B
B
Pos(B)
Pos(D)
---
Pos(B)
Pos(D)
Pos(D)
D
Pos(D)
A
Pos(A)
C
A
C
Pos(C)
---
Pos(A)
---
Pos(D)
B
D
Pos(B)
Pos(C)
C
Pos(D)

• A gets a packet for Pos(D)
• Packet gets stuck at A because Pos(A) is closest
  to Pos(D)
• Initiate route discovery for D from A
• Update the routing tables and forward the packet

30
Geographical RoutingProperties of the Algorithm

• Main Results
• There are no cycles in the routing tables
• Every packet reaches its destination, if graph
  connected
• Average routing table size O(log n)
• Packets delivered even if errors in position
  information
• No cycles even in dynamic networks
• Communication overhead/ time

31
Summary

• Great opportunity offered by exploiting
  application properties
• Innovative distributed algorithms emerging
• Main targetsminimize routing tables and
  discovery and maintenance traffic
• Goal Operational simulation of network by next
  retreat