Sensor Networks

Pete Perlegos

2

Outline

Background Ad-hoc Wireless Networks Smart Dust – TinyOS PicoRadio

3

How are they possible? Moore’s Law:

The historical trend…

4

Lets look at the other side

Moore’s Law is also pushing a given functionality into a smaller, cheaper, lower-power unit.

486DX 1989 486DX 2001

5

Other Trends

Complete systems on a chip (SoC) Integrated low-power communication

RF, optical Integrated low-power transducers

power capacitor, solar cell, battery Integrated sensors

Detect light, heat, position, movement, chemical presence, etc.

6

Other Trends

7

Outline

Background Ad-hoc Wireless Networks Smart Dust – TinyOS PicoRadio

8

Ad-hoc Wireless Networks

No base stations or infrastructure required

Multi-hop wireless networks Each node can talk with a neighbor

Applications Sensor networks Intelligent control applications

9

Ad-hoc Wireless Networks

MAC schemes Addressing Routing

10

Geographical Routing Algorithm

Geographicalnetwork

Assumptions: Each node knows its own position and its neighbors’

position Nodes don’t know the global topology Destination address is a geographical position to

which the packet is to be delivered

11

A Simple Routing Algorithm

Routing Decision: Route to the neighbor which is nearest to the packet destination

Source

Destination

12

Problem with Simple Routing

Source

DestinationWall

Simple routing does not always work The Geographical routing algorithm is an

extension of the simple routing algorithm

13

Route Discovery

Packet gets “stuck” when a node does not have a neighbor to which it can forward the packet

When a packet is stuck, a Route Discovery is started to destination D

A path is found to D Entry [position(D), s(i+1)]

is added to the routing table of s(i) Source

DestinationWall

14

Routing Tables

Routing Table for Station n:

(x,y) position Neighbor

d(8,6)

b

Position of n -

Position of neighbor a a Routing Algorithm:

• Packet arrives for position p at node n• Node n finds the position to which p is closest and . forwards to the . corresponding neighbor

Position of neighbor b

Routing Tables: Routing tables contain some additional entries beside . neighbors

15

Example

Pos(A) = (1,1)Pos(B) = (2,2)Pos(C) = (3,1)

Links:A ---- BB ---- C

A

B

C

Pos(A) ---

Pos(B) B

Pos(B) ---

Pos(A) APos(C) C

Pos(C) ---

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)

Pos(C)

Pos(C)

Pos(C)

16

Route Discovery

Pos(A) = (1,1)Pos(B) = (2,2)Pos(C) = (3,1)Pos(D) = (2.5,0)Links:A ---- BB ---- CC ---- D

B

C

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

Pos(D)

Pos(D)

A

D

Pos(A) ---

Pos(B) B

Pos(D) ---

Pos(C) C

Pos(B) ---

Pos(A) A

Pos(C) C

Pos(C) ---

Pos(B) B

Pos(D) D

Pos(D) B

Pos(D) C

Pos(D)

Pos(D)

Pos(D)

17

Desirable Properties of Location Service

Spread load evenly over all nodes. Degrade gracefully as nodes fail. Queries for nearby nodes stay

local. Per-node storage and

communication costs grow slowly as the network size grows

18

Grid Location Service (GLS)

n

s

ss

s

s

s

s

s s

s is n’s successor in that square. (Successor is the node with “least ID greater than” n )

sibling level-0squares

sibling level-1squares

sibling level-2squares

19

GLS Updates

... 1

...

...

...

9

23, 2

11, 2

6

9

11

1623

6

17

4

26

21

5

19

25

7

3

292

...

...

...

...

...

...

......

...

1

8

1

location table content

location update

2

20

1

...

...

...

9

23, 2

11, 2

6

9

112

1623

6

17

4

26

21

5

19

25

7

3

292

...

...

...

...

...

...

......

...

1

8

... 1

location table content

query from 23 for 1

GLS Query

21

Outline

Background Ad-hoc Wireless Networks Smart Dust – TinyOS PicoRadio

22

Active Messages - Benefits

Event based model: Avoids busy waiting for data to arrive Allows the system overlap communication

with computation

Lightweight architecture: Balances the need for extensible

communication network while maintaining efficiency and agility

23

Active Messages – What are they?

Each Active Message contains: the name of a user-level handler to be

invoked on a target node upon arrival a data payload to pass in as arguments

The handler function serves a dual purpose: extracting the message from the network either integrating the data into the

computation or sending a response message

24

Active Messages – What are they?

The network is modeled as a pipeline with minimal buffering for messages. This eliminates many of the buffering

difficulties faced by communication schemes that use blocking protocols or special send/receive buffers.

To prevent network congestion and ensure adequate performance, message handlers must be able to execute quickly and asynchronously.

25

Event Based Programming Event handlers are invoked to deal with

hardware events (directly or indirectly). The lowest level components have handlers

connected directly to hardware interrupts: external interrupts, timer events, or counter

events Events propagate up through the component

hierarchy. To perform long-running computation,

components request to have tasks executed on their behalf.

Once executed by scheduler, tasks run to completion and execute autonomously with other tasks.

26

Event Based Programming

Events propagate up through the component hierarchy.

27

Processor Utilization

The vast majority of the power consumption occurs in the active state, with very little power used in the idle state

The system should embrace the philosophy of getting work done as quickly as possible and going to sleep

This is a great benefit of the event based model

28

Outline

Background Ad-hoc Wireless Networks Smart Dust – TinyOS PicoRadio

29

PicoRadio The ever-evolving scaling of the semiconductor

technology is enabling the co-integration of the interfacing, computation, position location, and communication functions into a single silicon circuit.

Benefits of the system-on-a-chip approach: Maximally reduces the size of the sensor node Allows the use of advanced circuit architectures

which provide the optimal trade-off between flexibility and energy-efficiency

The tight integration of communication and computation functions into a single chip will provide the desired functionality at the lowest possible cost and energy.

30

PicoRadio A range of technologies are still necessary for the

realization of ultra-low energy wireless sensor networks: The study of multi-hop networks, and MAC layers that

support low (but variable)-rate data transmission, while ensuring low energy-consumption.

Chip architectures that enable the implementation of these advanced algorithms. (A heterogeneous combination of programmable, configurable, and fixed components.)

Mapping the advanced networking and communication algorithms onto such an architecture is a real design methodology problem.

Ensuring and verifying that these distributed and embedded systems will behave correctly is especially hard.

An RF front-end that meets the demands of variable bit-rates and energy-efficiency.

31

Protocol Stack

Specification of protocol stack derived from: System requirements (top-down) Wireless channel properties (bottom-up)

UI

MAC

Transmit Receive

Synchronization

Filter

Tx_data Rx_data

Mulaw Mulaw

Transport

User Interface Layer

Transport Layer

Mac Layer

Data Link Layer

Voice samples

Tx/Rx

Service Requests

UI

MAC

Transmit - CRC Receive - CRC

Synchronization

Filter

Tx_data Rx_data

Mulaw Mulaw

Transport

User Interface Layer

Transport Layer

Mac Layer

Logical Link Layer

Voice samples

Tx/Rx

Service Requests

32

Energy Consumption

Multi-hop Networks and Low-Energy Consumption

Using several short hops to send a bit is more energy efficient than using one longer hop

33

Energy Consumption

Multi-hop Networks and Low-Energy Consumption

Energy optimal number of hops as a function of distance

34

Architecture

Conceptual PicoNode chip architecture

Allows for flexibility Trade-off between flexibility and energy

efficiency must be managed

35

Conclusion

We seem to be getting close to realizing networks of sensors. Lightweight, event-based

communication Shrinking die size (lower power and

cost) Advancement of system on a chip

development