Transcript Sensors - University of Virginia, Department of Computer Science

Sensing and Hardware
CS 4501
Professor Jack Stankovic
Department of Computer Science
Fall 2010
HW - Mica2 and Mica2Dot
• ATMega 128L 8-bit, 8MHz, 4KB EEPROM, 4KB
RAM, 128KB flash
• Chipcon CC100 multi-channel radio (Manchester
encoding, FSK). From 10-20 ft. up to 500-1000ft.
Sensor Board
Sensor Board
Magnetometer-Compass
Ultrasonic Transceiver
Mica Weather Board
MicaDot Sensor Boards
Spec Mote
(3/6/2003)
• Size: 2x2.5mm, AVR RISC core, 3KB memory, FSK
radio (CC1000), encrypted communication hardware
support
Mica2
Rockwell WINS
• StrongARM SA 1100, 32-bit RISC
processor, 1MB SRAM, 4MB flash
• 900MHz spread spectrum radio,
with dedicated microcontroller:
32KB RAM, 1MB bootable flash
• 3.5”x3.5”x3” package size
•
•
•
•
acoustic sensor
magnetometer
accelerometer
seismic sensor
module
UCLA Medusa MK-2
• Radio-acoustic localization
• ATMega 128L 8-bit, 8MHz, 4KB flash,
4KB SRAM ( interface w/ sensors &
radio)
• ARM Thumb 32-bit,
40MHz, 1MB flash, 136KB
RAM (more demanding
processing)
• TR1000 radio Monolithics
(OOK, ASK modulation)
• Ultrasonic ranging system,
light & temperature
Medusa MK-2
• Can attach to infrastructure via a high
speed wire link
• Daisy chain motes
Acoustic Sensor
Magnetometer
Medusa MK-2
• Can power down various parts
independently to save power
–
–
–
–
Subsystems
Each sensor
Radio
CPU (might have multiple power saving
modes)
Specialized Hardware
• Environmental Motes (Berkeley, UVA)
• Medical Motes (Harvard/UVA)
– Wireless EKG
– Pulse Oximeter
• Robotic nodes
• New microprocessors/microcontrollers
– Use TI chips instead of Atmel
More Specialized HW
• CCDs
• Special logging mote (using camera
memory card)
• Stargates – heterogeneous WSNs
– Powerful
– Energy consumption is a problem
• New devices appearing continuously
Robo Mote
Trio Node
Solar Cells - Detecting Light
E-Tag Mote
SeeMote
Sensors
• Sensors must be small and low-power in
order to reduce energy and fit form
factor
• Packaging important
• Robustness to weather needed
Sensors
• Example of sensors
– Magnetic sensors
• Honeywell’s HMC/HMR magnetometers
– Photo sensors
• Clairex: CL9P4L
– Temperature sensors
• Panasonic ERT-J1VR103J
– Accelerometers
• Analog Devices: ADXL202JE
– Motion sensors
• Advantaca’s MIR sensors
– GPS
– Cameras
Actuators
• Examples of Actuators
–
–
–
–
Motor (for mobile nodes)
LEDs
Buzzer
Emit chemical
• In general, actuators may be powerful,
large, and complicated
– Can be outside of motes (e.g., turn on
lights, send a vehicle into system, …)
• What actuators should go on motes?
Properties of Sensors (14)
– Range
• Example
– HMC1053: +/-6 Gauss
– Accuracy
• Measure of error and uncertainty
– Repeatability
• HMC1002: 0.05%
– Linearity
• HMC1002: 0.1% (Best fit straight line +/- 1 Gauss)
Sensors
– Sensitivity
• How output reflects input?
– Efficiency
• Ratio of the output power to the input power
– Resolution
• Temperature within ½ degree
Sensors
• Response time
– How fast the output reaches a fraction of the expected
signal level
• Overshoot
– How much does the output signal go beyond the
expected signal level
• Drift and stability
– How the output signal varies slowly compared to time
• Offset
– The output when there is no input
Sensors
– Packaging
• Example – HMC1053: 16-PIN LCC packaging
– Property of the circuit
• Load of the circuit
• Power drain
– Initialization Time (important when
nodes are asleep and awakened
dynamically when an event occurs)
Sensors
• Signal Processing
– Process the sensor reading to make it useful
to the application
• Sensor fusion (heterogeneity possible)
• False alarm processing (false positives and false
negatives)
– The complexity varies from a simple threshold
algorithm to full-fledged signal processing and
pattern recognition
• New solutions needed on minimal capacity devices
Sensors
Indoor test, quiet environment without motion
500
400
300
7.IndoorQuiet
200
100
0
1
80
159 238 317 396 475 554 633 712 791 870 949 1028
• Raw reading of an MIR sensor in a
quiet environment
– The beginning period represents some
unknown noise, possibly due to the
positioning of the sensor
Sensors
39.64Hz.Milton.sb.MIR.DanWalk.3
120
100
80
60
39.64Hz.Milton.sb.MIR.DanWalk.3
40
20
0
1 13 25 37 49 61 73 85 97 109 121 133 145 157 169 181 193 205 217 229 241 253 265 277
• Raw reading of an MIR sensor as a person
walked by
– The all-zero period is due to unreliable UART
interface used to collect the reading and can be
ignored.
Acoustic Sensing
Three Cars
Initial
Calibration
No Detection
Detection when
Energy Crosses
Standard Deviation
Programming with Sensors
210
Sensor
Sensor
Voltage
AMP
Sensor
AMP
Voltage
ADC
ADC
ADC
MicroProc
MicroProc
MicroProc
ADC
2
10
2
28
• Resolution
• Sample Rate
V
Temp
0-100 C
Sensor
12
MicroProc
ADC
SPI
I2C
Resolution
ADC
• MAX1245
–
–
–
–
–
–
8 channels of analog input
Can sample up to 100,000 samples per sec
Resolution of 12 bits
Interfaces with SPI and I2C buses
Can enter low power mode
Interface to Processor: processor issues
commands to read channel
– Interfaces to sensors
ADC
• Sample rate
Too
slow
Nyquist
Sampling
Theorem
Temperature Sensor
• A22100
– Output voltage: 22.5mV/C over temperature
range of -50C to 150C
– Derive conversion equation (see spec sheet)
– Example: for 5 V power supply
• T = (V(out) – 1.375)/0.0225
• If V(out) = 1.94V then T = 25.1C
5V
A22100
GND
V(out)
Other Sensors
• Light
– Add power and ground
– Analog output voltage is proportional to incident
light
– May need an amp to detect full range
• Accelerometer
– Output voltage is proportional to acceleration and
power V(s)
– V(out) = V(s)/2 – (sensitivity * V(s)/5 *
acceleration)
– Sensitivity depends on particular accelerometer
RFID
• RFID
– Typical configuration
Plus
Microchip
With data
– Application: ID based intelligent control
• Such as access control, baggage ID, object tracking,
inventory management, …
RFID
– What makes RFID useful?
• Ubiquitous
• Low-cost (pennies)
– Compare RFID with motes
• Difference? Yes (today).
• Will they merge to be the same class
of hardware as motes?
– Active RFID tags exist (battery/sensors)
– Privacy and security issues
Intel WISP tag
• Essentially a batteryless sensor mote
– Light, temperature,
3d- accelerometer
– 10 feet range with
harvested RF power
• Requires RFID reader
and (large) antennas
Activity recognition using
WISP*
WISP tags on kitchen artifacts
Antenna layout in home
* Ubicomp 2009
WISP potential
• Battery-free solution to sensor
networks
• Great potential for elderly activity
inference and other smart home
applications
Sensor and Data Fusion
• Data Fusion – combine data from
multiple sources (not only sensors)
• Sensor Fusion – combine data from
multiple sensors
Signatures
• Objects/phenomena generate
signatures
• Type of energy (electromagnetic,
acoustic, ultrasonic, seismic, etc.
• Active or passive sensors
• Affected by weather, clutter,
countermeasures, etc.
Data Fusion
•
•
•
•
•
•
•
•
Ad hoc
Classical
Bayesian
Dempster-Shafer
Fuzzy Logic
Pattern Recognition
ANN
Etc.
Multi-Modal
• Robustness
• Act synergistically in high clutter and
inclement weather
• Example: Weather satellites use
microwave, millimeter wave, infrared
and cameras
• Example: Fog at an airport
• Example: Rain cools targets (PIR
sensors not as effective)
Fusion Architecture
ZigBee Coordinator
ZigBee Router/FFD
Raw Data to Knowledge
• Detection
• Classification
• Identification
Medical Care
Diabetes
Eating
Level
Toileting
Level
eating
toileting
Depression
Sleeping
Level
showering
Movement
Level
sleeping
Kitchen
visits
Light
Level
Weight
Level
Light
Weight
bathroom bedroom
visits
visits
Personal
location tracking
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
fron
t
floo
r
frid
ge
micr
owa
ve
pant
ry
cook
top
sink
flus
h
entr
ance
sink
sho
wer
moti
on
moti
on
moti
on
weig
ht
ligh
t
ligh
t
ligh
t
pres
sure
bed
roo
kitc
hen
kitc
hen
kitc
hen
kitc
hen
kitc
hen
bat
h
bat
h
bat
h
bat
h
bed
roo
kitc
hen
bat
h
bed
roo
bed
roo
kitc
hen
bat
h
bed
roo
Reference
• Sensor and Data Fusion, L. Klein, SPIE
Press, 2004.