Sensor Networks & TinyDB Author: Roman Kolcun Supervisor: - - PowerPoint PPT Presentation

Sensor Networks & TinyDB

Author: Roman Kolcun Supervisor: Julie A. McCann

2

Index

 Sensor Motes  Sensor Networks  Real world deployments  TinyDB

3

What a Sensor Can Sense?

4

What a Sensor Can Sense?

 Temperature  Humidity  Acceleration  Noise  Light  Magnetic Field  Gravity  Pressure  Heart Rate  Motion  Toxins  Nutrients  Glucose Level  Oxygen Level  Hormones  Proteins

5

Sensor Mote

 CPU ?

6

Sensor Mote

 CPU

 4 MHz  8 bit ATmega 128L, RISC

 Memory ?

7

Sensor Mote

 CPU

 4 MHz  8 bit ATmega 128L, RISC

 Memory

 128KB Program Flash Memory  4KB RAM  512KB Flash – serial access, max. 10-100k rewritten

 Wireless ?

8

Sensor Mote

 CPU

 4 MHz  8 bit ATmega 128L, RISC

 Memory

 128KB Program Flash Mem.  4KB RAM  512KB Flash – serial access, max. 10-100k rewritten

 Wireless

 IEEE 802.15.4 - 2.4GHz  250kbps  70 – 100m outdoor, 20 – 30m indoor

9

Beastie – Imperial

 8k bytes Flash program memory  1k byte SRAM  512 bytes EEPROM  4MHz  FM radio at 434.65MHz  10kbps maximum data rate (run at 5kbps as

standard)

 Maximum range 500m

10

1 mm3 Sensor

 Designed at Michigan University  Measures pressure in an eye every 15 minutes  Average Power Consumption 15 nW  To charge batteries:

 10 hours of

artificial light

 1.5 hours of

sun light

11

Power Consumption

 Typically supplied by small batteries

 1000 – 3000 mAh  1 mAh – 1 milliamp current for 1 hour  Power = Watts (W) = Amps (A) * Volts (V)  Energy = Joules (J) = W * time

 Power consumption

 Processor: 8mA active, <15μA sleep  Radio: 19.7mA receive, 11 – 17.4mA xmit,

~20ms/packet

 Sensor: 1μA – 100's mA, 1μs – 1s to sample

12

Example

 Battery: 1000mAh  How long a node can last and how much data

can a node receive?

13

Example

 Battery: 1000mAh  How long a node can last and how much data

can a node receive? 1000mAh / 19.7mA = ~50.7h 50.7h = 182 741s * 250kbps = ~5.7MB in real world approx. half of it = 2.85MB

14

Example

 Battery: 1000mAh  Sense 1 value every 30 seconds, receive 1

packet and send 1 packet. How long will battery last?

15

Example

 Battery: 1000mAh  Sense 1 value every 30 seconds, receive 1

packet and send 1 packet. How long will battery last? 8+29*0.015+1+19.7*0.02+13*0.02 = = ~10mA/30s = ~ 0.3363 mA/s 1000mAh = 3 600 000 mAs / 0.3363 = ~124 days

16

Sensor Networks - Problems

 Lossy, Ad-hoc radio communication  Really lossy radio communication  Node / link failures  Severe power constraints  Asymmetric links – if I can hear you, it does not

mean you can hear me

 Interference  Hidden node problem

17

Sensor Networks

18

Sensor Networks

19

Sensor Networks - Problems

 Duty cycling  Time synchronization  Node / link failure

20

Duty Cycling

… zzz … … zzz …

time wake-up period epoch time

21

Real World Deployments

 Great Duck Island

 temperature  relative humidity  infra-red termophile

22

Real World Deployments

 Golden Gate Bridge

 vibrations  temperature

23

Redwood

Humidity vs. Time

35 45 55 65 75 85 95

Rel Humidity (%)

101 104 109 110 111

Temperature vs. Time

8 13 18 23 28 33

7/7/03 9:40 7/7/03 13:11 7/7/03 16:43 7/7/03 20:15 7/7/03 23:46 7/8/03 3:18 7/8/03 6:50 7/8/03 10:21 7/8/03 13:53 7/8/03 17:25 7/8/03 20:56 7/9/03 0:28 7/9/03 4:00 7/9/03 7:31 7/9/03 11:03

Date

36m 33m: 111 32m: 110 30m: 109,108,107 20m: 106,105,104 10m: 103, 102, 101

24

Imperial Deployment

 Mobile Node (Mule) Experiment

25

Other Deployments

 Monitoring Space

 environmental and habitat monitoring (Duck Island,

Redwood Trees)

 precision agricultures  climate control  surveillance  intelligent alarms

26

Other Deployments (cont.)

 Things

 structural monitoring  ecophysiology  condition based maintenance (plane, bridges,

buildings, pipes)

 medical diagnostics  terrain mapping

27

Other Deployments (cont.)

 Interactions with things and encompassing

space

 monitoring wildlife habitats  disaster management  emergency response  ubicomp  process flow

28

Motivations for TinyDB

 Create an application which measures

• temperature. Make an average of temperatures
• ver 15°C.

29

Motivations for TinyDB

 Create an application which measures

• temperature. Make an average of temperatures
• ver 15°C.

 How would you change the application to make

an average of temperatures over 20°C ?

30

Motivations for TinyDB

 Create an application which measures

• temperature. Make an average of temperatures
• ver 15°C.

 How would you change the application to make

an average of temperatures over 20°C ?

 Recode the application and manually update every

node.

 Think about it while programming the application

and let it accept commands from the basestation

 Use TinyDB

31

TinyDB

 Supports a subset of Stream SQL  Whole network could be seen as ”sensor” table  Query syntax:

SELECT <aggregates>, <attributes> [FROM {sensors} | {buffer}] [WHERE <predicates>] [GROUP BY <expression>] [SAMPLE INTERVAL <const> | ONCE] [INTO buffer] [TRIGGER ACTION <command>]

32

TinyDB

 Example:

SELECT light, mag FROM sensors WHERE light > c1 AND mag > c2 SAMPLE INTERVAL 1s [FOR 3600s]

33

TinyDB

 Example:

SELECT light, mag FROM sensors WHERE light > c1 AND mag > c2 SAMPLE INTERVAL 1s

E(sampling mag) » E(sampling light) 1500 μJ vs 90 μJ In which order the predicates should be evaluated?

34

TinyDB

 Do we need to notify all sensors in the network?

SELECT light FROM sensors WHERE node_id > 20 SAMPLE INTERVAL 10s

35

What if the Result Depends on More than One Node?

36

What if the Result Depends on More than One Node? (cont.)

37

What We Do

 Adjust power transmission in order to minimise

interference using game theory

 Duty-cycling  Time synchronization  Mules  In-network data processing (joining data & data

filtering)

38

References

[1] Decentralised & Volatile Self-Adaptive,Self Organising WSNs by Julie A. McCann [2] Implementation and Research Issues in Query Processing for Wireless Sensor Networks by Wei Hong & Sam Madden [3] Modelling the Golden Gate Bridge using Wireless Sensor Networks by Guilherme Rocha, Shamim Pakzad and Bin Yu [4] http://www.coa.edu/greatduckisland.htm – College of the Atlantic – Great Duck Island Project

39

Where You Can Find Us

 Julie A. McCann: jamm@doc.ic.ac.uk  Roman Kolcun: rk1208@doc.ic.ac.uk  Lab: Huxley Building, 563