smartphone-based sensor networks and some statistical
TRANSCRIPT
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Smartphone-based sensor networks and somestatistical challenges: the Earthquake Network
Android application
Francesco Finazzi
University of Bergamo
29 November 2013 - University of Glasgow
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Outline
Outline
Monitoring networks
Wireless sensor networks
Smartphone-based sensor networks (SSN)
Analysis of SSN data
Earthquake Network Android application
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Classic monitoring networks
Classic (wired) monitoring networks
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Monitoring network node
Monitoring network node
Byres Road - Scottish air quality network
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Monitoring network node
Air quality monitoring network
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Wireless sensor network
Wireless sensor network
Volcano monitoring - Harvard Sensor Networks Lab
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Smartphone-based sensor networks
Smartphone-based sensor network (SSN)
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Smartphones as nodes
Can we use smartphones?
Data transferGSM signal - not easy to have in remote areas
PowerExternal batteries, solar panels
Sensing capabilitiesAcceleration, geomagnetic �eld, temperature, illuminance,sound pressure, atm. pressure and humidityVisible light (camera)Position
Sensor accuracy and calibrationNot always good
GeolocationGPS or GSM signal
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Smartphones as nodes
Can we use smartphones?
Data transferGSM signal - not easy to have in remote areas
PowerExternal batteries, solar panels
Sensing capabilitiesAcceleration, geomagnetic �eld, temperature, illuminance,sound pressure, atm. pressure and humidityVisible light (camera)Position
Sensor accuracy and calibrationNot always good
GeolocationGPS or GSM signal
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Smartphones as nodes
Can we use smartphones?
Data transferGSM signal - not easy to have in remote areas
PowerExternal batteries, solar panels
Sensing capabilitiesAcceleration, geomagnetic �eld, temperature, illuminance,sound pressure, atm. pressure and humidityVisible light (camera)Position
Sensor accuracy and calibrationNot always good
GeolocationGPS or GSM signal
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Smartphones as nodes
Can we use smartphones?
Data transferGSM signal - not easy to have in remote areas
PowerExternal batteries, solar panels
Sensing capabilitiesAcceleration, geomagnetic �eld, temperature, illuminance,sound pressure, atm. pressure and humidityVisible light (camera)Position
Sensor accuracy and calibrationNot always good
GeolocationGPS or GSM signal
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Smartphones as nodes
Can we use smartphones?
Data transferGSM signal - not easy to have in remote areas
PowerExternal batteries, solar panels
Sensing capabilitiesAcceleration, geomagnetic �eld, temperature, illuminance,sound pressure, atm. pressure and humidityVisible light (camera)Position
Sensor accuracy and calibrationNot always good
GeolocationGPS or GSM signal
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Smartphones as nodes
Can we use smartphones?
Data transferGSM signal - not easy to have in remote areas
PowerExternal batteries, solar panels
Sensing capabilitiesAcceleration, geomagnetic �eld, temperature, illuminance,sound pressure, atm. pressure and humidityVisible light (camera)Position
Sensor accuracy and calibrationNot always good
GeolocationGPS or GSM signal
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Smartphones as nodes
Can we use smartphones?
Data transferGSM signal - not easy to have in remote areas
PowerExternal batteries, solar panels
Sensing capabilitiesAcceleration, geomagnetic �eld, temperature, illuminance,sound pressure, atm. pressure and humidityVisible light (camera)Position
Sensor accuracy and calibrationNot always good
GeolocationGPS or GSM signal
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Smartphones as nodes
Can we use smartphones?
Data transferGSM signal - not easy to have in remote areas
PowerExternal batteries, solar panels
Sensing capabilitiesAcceleration, geomagnetic �eld, temperature, illuminance,sound pressure, atm. pressure and humidityVisible light (camera)Position
Sensor accuracy and calibrationNot always good
GeolocationGPS or GSM signal
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Smartphones as nodes
Can we use smartphones?
Data transferGSM signal - not easy to have in remote areas
PowerExternal batteries, solar panels
Sensing capabilitiesAcceleration, geomagnetic �eld, temperature, illuminance,sound pressure, atm. pressure and humidityVisible light (camera)Position
Sensor accuracy and calibrationNot always good
GeolocationGPS or GSM signal
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Smartphones as nodes
Can we use smartphones?
Data transferGSM signal - not easy to have in remote areas
PowerExternal batteries, solar panels
Sensing capabilitiesAcceleration, geomagnetic �eld, temperature, illuminance,sound pressure, atm. pressure and humidityVisible light (camera)Position
Sensor accuracy and calibrationNot always good
GeolocationGPS or GSM signal
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Smartphones as nodes
Can we use smartphones?
Data transferGSM signal - not easy to have in remote areas
PowerExternal batteries, solar panels
Sensing capabilitiesAcceleration, geomagnetic �eld, temperature, illuminance,sound pressure, atm. pressure and humidityVisible light (camera)Position
Sensor accuracy and calibrationNot always good
GeolocationGPS or GSM signal
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Smartphones as nodes
Can we use smartphones?
Data transferGSM signal - not easy to have in remote areas
PowerExternal batteries, solar panels
Sensing capabilitiesAcceleration, geomagnetic �eld, temperature, illuminance,sound pressure, atm. pressure and humidityVisible light (camera)Position
Sensor accuracy and calibrationNot always good
GeolocationGPS or GSM signal
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Smartphones as nodes
SSN applications
Environmental �eld estimation
Temperature, pressure, humidity,...
Earthquakes (accelerometer)
Avalanches/landslides (microphone)
Air quality (camera+�lter)
Fire/smoke detection (camera)
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Smartphones as nodes
SSN applications
Environmental �eld estimation
Temperature, pressure, humidity,...
Earthquakes (accelerometer)
Avalanches/landslides (microphone)
Air quality (camera+�lter)
Fire/smoke detection (camera)
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Smartphones as nodes
SSN applications
Environmental �eld estimation
Temperature, pressure, humidity,...
Earthquakes (accelerometer)
Avalanches/landslides (microphone)
Air quality (camera+�lter)
Fire/smoke detection (camera)
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Smartphones as nodes
SSN applications
Environmental �eld estimation
Temperature, pressure, humidity,...
Earthquakes (accelerometer)
Avalanches/landslides (microphone)
Air quality (camera+�lter)
Fire/smoke detection (camera)
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Smartphones as nodes
SSN applications
Environmental �eld estimation
Temperature, pressure, humidity,...
Earthquakes (accelerometer)
Avalanches/landslides (microphone)
Air quality (camera+�lter)
Fire/smoke detection (camera)
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Smartphones as nodes
SSN applications
Environmental �eld estimation
Temperature, pressure, humidity,...
Earthquakes (accelerometer)
Avalanches/landslides (microphone)
Air quality (camera+�lter)
Fire/smoke detection (camera)
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
SSN types
SSN types
Private SSNs - Sensor network with smartphones as nodes.The network is operated by a network manager which isresponsible for buying, installing and maintaining the nodes.
Public SSNs - Sensor network based on smartphonesbelonging to their single owners. The smartphone is part ofthe network only if the owner agrees to be part of it.
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
SSN types
SSN types
Private SSNs - Sensor network with smartphones as nodes.The network is operated by a network manager which isresponsible for buying, installing and maintaining the nodes.
Public SSNs - Sensor network based on smartphonesbelonging to their single owners. The smartphone is part ofthe network only if the owner agrees to be part of it.
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Public SSNs
Public SSNs architecture
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Public SSNs
Public SSNs
Dynamic network: the total number of nodes, the number ofactive nodes and the location of the nodes changecontinuously;
Location is not always known precisely (can we usenon-localized smartphones?);
Heterogeneous nodes: di¤erent vendors and models;
Sensors are often uncalibrated;
Asynchronous data acquisition;
Preferential sampling (remote areas?);
Someone else pays device and energy.
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Public SSNs
Public SSNs
Dynamic network: the total number of nodes, the number ofactive nodes and the location of the nodes changecontinuously;
Location is not always known precisely (can we usenon-localized smartphones?);
Heterogeneous nodes: di¤erent vendors and models;
Sensors are often uncalibrated;
Asynchronous data acquisition;
Preferential sampling (remote areas?);
Someone else pays device and energy.
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Public SSNs
Public SSNs
Dynamic network: the total number of nodes, the number ofactive nodes and the location of the nodes changecontinuously;
Location is not always known precisely (can we usenon-localized smartphones?);
Heterogeneous nodes: di¤erent vendors and models;
Sensors are often uncalibrated;
Asynchronous data acquisition;
Preferential sampling (remote areas?);
Someone else pays device and energy.
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Public SSNs
Public SSNs
Dynamic network: the total number of nodes, the number ofactive nodes and the location of the nodes changecontinuously;
Location is not always known precisely (can we usenon-localized smartphones?);
Heterogeneous nodes: di¤erent vendors and models;
Sensors are often uncalibrated;
Asynchronous data acquisition;
Preferential sampling (remote areas?);
Someone else pays device and energy.
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Public SSNs
Public SSNs
Dynamic network: the total number of nodes, the number ofactive nodes and the location of the nodes changecontinuously;
Location is not always known precisely (can we usenon-localized smartphones?);
Heterogeneous nodes: di¤erent vendors and models;
Sensors are often uncalibrated;
Asynchronous data acquisition;
Preferential sampling (remote areas?);
Someone else pays device and energy.
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Public SSNs
Public SSNs
Dynamic network: the total number of nodes, the number ofactive nodes and the location of the nodes changecontinuously;
Location is not always known precisely (can we usenon-localized smartphones?);
Heterogeneous nodes: di¤erent vendors and models;
Sensors are often uncalibrated;
Asynchronous data acquisition;
Preferential sampling (remote areas?);
Someone else pays device and energy.
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Public SSNs
Public SSNs
Dynamic network: the total number of nodes, the number ofactive nodes and the location of the nodes changecontinuously;
Location is not always known precisely (can we usenon-localized smartphones?);
Heterogeneous nodes: di¤erent vendors and models;
Sensors are often uncalibrated;
Asynchronous data acquisition;
Preferential sampling (remote areas?);
Someone else pays device and energy.
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
SSN data analysis
SSN data analysis
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Classic monitoring network - Data matrix
Classic monitoring network - Data matrix
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Classic monitoring network - Data matrix
SSN - Geolocation
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Classic monitoring network - Data matrix
Dynamic network
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Classic monitoring network - Data matrix
Dynamic network
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Classic monitoring network - Data matrix
SSN - Data table
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Calibration
Sensor calibration
Smartphone sensors are often uncalibrated or...
they require periodic calibration
Manual calibration is not easy/practicable
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Calibration
"On the �y" calibration based on geofence
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Calibration
Calibration curve
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Introduction
Earthquake Network Android app
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Earthquake Network as SSN
Earthquake Network
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Earthquake Network as SSN
Earthquake Network
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Why and what
Why and what
Why detect quakes?
O¢ cial networks take between 15 and 60 minutes to notifyPopulation noti�cationIntensity (shake map)
What can be detected?
Time of the eventIntensityEpicenter (not easy)
What cannot be detected?
Magnitude (energy)
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Why and what
Why and what
Why detect quakes?
O¢ cial networks take between 15 and 60 minutes to notifyPopulation noti�cationIntensity (shake map)
What can be detected?
Time of the eventIntensityEpicenter (not easy)
What cannot be detected?
Magnitude (energy)
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Why and what
Why and what
Why detect quakes?
O¢ cial networks take between 15 and 60 minutes to notifyPopulation noti�cationIntensity (shake map)
What can be detected?
Time of the eventIntensityEpicenter (not easy)
What cannot be detected?
Magnitude (energy)
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Why and what
Why and what
Why detect quakes?
O¢ cial networks take between 15 and 60 minutes to notifyPopulation noti�cationIntensity (shake map)
What can be detected?
Time of the eventIntensityEpicenter (not easy)
What cannot be detected?
Magnitude (energy)
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Why and what
Why and what
Why detect quakes?
O¢ cial networks take between 15 and 60 minutes to notifyPopulation noti�cationIntensity (shake map)
What can be detected?
Time of the eventIntensityEpicenter (not easy)
What cannot be detected?
Magnitude (energy)
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Why and what
Why and what
Why detect quakes?
O¢ cial networks take between 15 and 60 minutes to notifyPopulation noti�cationIntensity (shake map)
What can be detected?
Time of the eventIntensityEpicenter (not easy)
What cannot be detected?
Magnitude (energy)
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Why and what
Why and what
Why detect quakes?
O¢ cial networks take between 15 and 60 minutes to notifyPopulation noti�cationIntensity (shake map)
What can be detected?
Time of the eventIntensityEpicenter (not easy)
What cannot be detected?
Magnitude (energy)
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Why and what
Why and what
Why detect quakes?
O¢ cial networks take between 15 and 60 minutes to notifyPopulation noti�cationIntensity (shake map)
What can be detected?
Time of the eventIntensityEpicenter (not easy)
What cannot be detected?
Magnitude (energy)
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Why and what
Why and what
Why detect quakes?
O¢ cial networks take between 15 and 60 minutes to notifyPopulation noti�cationIntensity (shake map)
What can be detected?
Time of the eventIntensityEpicenter (not easy)
What cannot be detected?
Magnitude (energy)
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Why and what
Why and what
Why detect quakes?
O¢ cial networks take between 15 and 60 minutes to notifyPopulation noti�cationIntensity (shake map)
What can be detected?
Time of the eventIntensityEpicenter (not easy)
What cannot be detected?
Magnitude (energy)
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
HW/SW architecture
HW/SW architecture
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
HW/SW architecture
HW/SW architecture
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
SW architecture
App dashboard
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Vibrations and quakes
Vibration vs earthquake
Vibration: smartphone movement
Earthquake: waves of energy travelling through the Earth�slayers
Smartphones measure vibrations, the network measuresearthquakes
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Vibrations and quakes
Vibration vs earthquake
Vibration: smartphone movement
Earthquake: waves of energy travelling through the Earth�slayers
Smartphones measure vibrations, the network measuresearthquakes
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Vibrations and quakes
Vibration vs earthquake
Vibration: smartphone movement
Earthquake: waves of energy travelling through the Earth�slayers
Smartphones measure vibrations, the network measuresearthquakes
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Vibration detection
Vibration detection
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Accelerometer calibration
Accelerometer calibration
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Vibration detection
Vibration detection - Battery charging
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Earthquake detection
Earthquake detection
(non geolocalized smartphones?)
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Dynamic network
Active nodes
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Dynamic network
Active nodes
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Earthquake Network
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Dynamic network
Active nodes
Downloads: � 270000 (11 months)Total active users: � 30000Active nodes at any time: � 10000Enabled nodes at any time: 100� 250 (3%� 8%)3% of 1.4 billion (potential) smartphones is a large number
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Dynamic network
Active nodes
Downloads: � 270000 (11 months)Total active users: � 30000Active nodes at any time: � 10000Enabled nodes at any time: 100� 250 (3%� 8%)3% of 1.4 billion (potential) smartphones is a large number
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Dynamic network
Active nodes
Downloads: � 270000 (11 months)Total active users: � 30000Active nodes at any time: � 10000Enabled nodes at any time: 100� 250 (3%� 8%)3% of 1.4 billion (potential) smartphones is a large number
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Dynamic network
Active nodes
Downloads: � 270000 (11 months)Total active users: � 30000Active nodes at any time: � 10000Enabled nodes at any time: 100� 250 (3%� 8%)3% of 1.4 billion (potential) smartphones is a large number
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Dynamic network
Active nodes
Downloads: � 270000 (11 months)Total active users: � 30000Active nodes at any time: � 10000Enabled nodes at any time: 100� 250 (3%� 8%)3% of 1.4 billion (potential) smartphones is a large number
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Conclusions
Conclusions
Environmental sensors are becoming available for o¤-the-shelfproducts
Smartphones are the natural candidates to carry aroundsensors
However:
A SSN should be based on a large number of smartphones(not easy)Calibrations of heterogeneous sensors can be an issueStatistical methods/software for SSN space-time data may notbe immediately available
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Conclusions
Conclusions
Environmental sensors are becoming available for o¤-the-shelfproducts
Smartphones are the natural candidates to carry aroundsensors
However:
A SSN should be based on a large number of smartphones(not easy)Calibrations of heterogeneous sensors can be an issueStatistical methods/software for SSN space-time data may notbe immediately available
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Conclusions
Conclusions
Environmental sensors are becoming available for o¤-the-shelfproducts
Smartphones are the natural candidates to carry aroundsensors
However:
A SSN should be based on a large number of smartphones(not easy)Calibrations of heterogeneous sensors can be an issueStatistical methods/software for SSN space-time data may notbe immediately available
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Conclusions
Conclusions
Environmental sensors are becoming available for o¤-the-shelfproducts
Smartphones are the natural candidates to carry aroundsensors
However:
A SSN should be based on a large number of smartphones(not easy)Calibrations of heterogeneous sensors can be an issueStatistical methods/software for SSN space-time data may notbe immediately available
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Conclusions
Conclusions
Environmental sensors are becoming available for o¤-the-shelfproducts
Smartphones are the natural candidates to carry aroundsensors
However:
A SSN should be based on a large number of smartphones(not easy)Calibrations of heterogeneous sensors can be an issueStatistical methods/software for SSN space-time data may notbe immediately available
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Conclusions
Conclusions
Environmental sensors are becoming available for o¤-the-shelfproducts
Smartphones are the natural candidates to carry aroundsensors
However:
A SSN should be based on a large number of smartphones(not easy)Calibrations of heterogeneous sensors can be an issueStatistical methods/software for SSN space-time data may notbe immediately available
Introduction Monitoring networks WSN SSN SSN data analysis Earthquake Network Conclusions
Pro version