Mobile devices
Universität Stuttgart Institute of Parallel and Distributed Systems (IPVS) Universitätsstraße 38 D-70569 Stuttgart
PSense: Reducing Energy Consumption in Public Sensing Systems The 26th IEEE Conference on Advanced Information Networking and Applications (AINA-2012) Fukuoka, Japan 26.03.2012
Patrick Baier, Frank Dürr and Kurt Rothermel
Outline • Public Sensing Overview • Motivation • System Model • System Overview • Naive Sensing Approach • Optimized Sensing Approach
◦ Adaptive Positioning ◦ Ad-hoc Information Exchange • Evaluation • Related Work
• Conclusion & Outlook Research Group Distributed Systems
Universität Stuttgart 2
IPVS
Public Sensing • utilizes mobile devices to opportunistically collect sensor data. • extracts sensor data from devices of common people.
+ No costly sensor deployment
+ Large coverage area • can be used to record different type of sensor data, e.g. creating a noise map of a city.
Research Group Distributed Systems
Universität Stuttgart 3
IPVS
Motivation However, Public Sensing … • drains an extensive amount of energy on mobile devices
◦ Positioning (GPS) ◦ Sensing ◦ Cellular Communication
Coordination of sensing Acquiring sensor data Uploading sensor data
• needs the acceptance of the mobile device owners
◦ Less users results in decrease of effectiveness of sensing Challenge: High energy consumption reduces user acceptance
Goal: Design an energy efficient Public Sensing system Research Group Distributed Systems
Universität Stuttgart 4
IPVS
System Model • Server
◦ Unlimited battery supplies ◦ Knowledge of road graph • Mobile devices
◦ ◦ ◦ ◦ ◦
Limited battery supplies GPS device Sensor with limited sensing range
Mobile Device Sensor Range
Move according to road graph Neither direction nor speed of movement controllable
• Communication
◦ Server Mobile device: Cellular Radio (UMTS,GPRS) ◦ Mobile device Mobile device: Short-range Radio (WLAN) Research Group Distributed Systems
Universität Stuttgart 5
IPVS
PSense - System Overview • PSense takes queries for sensing locations (SL) defined by
◦ Geographical Position ◦ Type of sensor value (e.g. temperature, noise) • Mobile devices sense data when within sensing range to SL • Sensor data is delivered back to application Sensing Sensor Locations Data
Sensing Location Mobile device
Application Interface
Sensing Locations
PSense
Sensor Data Research Group Distributed Systems
Universität Stuttgart 6
IPVS
Naive Sensing Approach 1. Server forwards SLs to devices. 2. Each device fixes position every tGPS sec. 3. Device checks if it can sense a SL No: Continue with step 2 Yes: Upload data to server 4. Server sends update of SLs to devices 5. Server returns sensor data to application
Sensing Sensor Locations Data
PSense
tGPS
tGPS
Research Group Distributed Systems
Universität Stuttgart 7
IPVS
Overview Optimizations
1. Adaptive Positioning
◦ Goal: Reduce number of GPS fixes 2. Ad-hoc Information Exchange
◦ Goal: Reduce number of cellular messages Goal: Minimize energy consumption of mobile devices But:
Effectiveness of sensing should not decrease Number of recorded sensing locations should be the same Research Group Distributed Systems
Universität Stuttgart 8
IPVS
Adaptive Positioning (1) • Idea: Reduce number of GPS fixes by adapting positioning interval
1. Server determines graph-based distance to closest SL 2. Device calculates earliest time span tGPS when SL can be reached 3. Device delays next position for tGPS
dmin
Starting points for sensing
𝑡𝐺𝑃𝑆 =
time until next GPS fix
vmax
maximum speed of device
distance to closest sensing location
𝑑𝑚𝑖𝑛 𝑣𝑚𝑎𝑥
Research Group Distributed Systems
tGPS dmin
Universität Stuttgart 9
IPVS
Adaptive Positioning (2) Optimized Sensing Workflow: 1. Each time a device fixed its position, it queries server for :
◦ Distance dmin to closest sensing location SL ◦ Position (x,y)SL of SL 2. Device checks whether it can sense SL 3. Calculate tGPS and delay next position fix
m1
m2
Message
Content
m1
(x,y)Device
m2
dmin and (x,y)SL
SL dmin
𝑡𝐺𝑃𝑆 =
𝑑𝑚𝑖𝑛 𝑣𝑚𝑎𝑥
Number of GPS fixes decrease But: Number of cellular messages increase Research Group Distributed Systems
Universität Stuttgart 10
IPVS
Ad-hoc Information Exchange (1) Idea: Reduce cellular traffic using ad-hoc messages dmin SL
When node receives distance to closest SL from server, nearby nodes … • have almost the same distance to this sensing location • can recalculate time until the next GPS fix and possibly extend it Server communication can be reduced
Research Group Distributed Systems
Universität Stuttgart 11
IPVS
Ad-hoc Information Exchange (2) broadcast range
dmin
m1
dmin- rah Ad-hoc receiver
m2
Ad-hoc sender
• Upon receiving dmin from server, node sends ad-hoc broadcast • All broadcast receivers compute tGPS’ • If tGPS’ > trem , schedule next GPS fix to time tGPS’
𝑡𝐺𝑃𝑆′ =
𝑑𝑚𝑖𝑛− 𝑟𝑎ℎ 𝑣𝑚𝑎𝑥
Research Group Distributed Systems
trem
remaining time until next GPS fix
rah
maximum ad-hoc range
Universität Stuttgart 12
IPVS
Evaluation – Setup • Simulation in ns-2 • Mobility on road graph of Chicago (UDELModel, pedestrian speed) • Use energy model to compare results • Mean number of sensing locations: 50 • Simulation time: 6 hours • Maximum ad-hoc range: 100m
Operation
Energy [mJ]
GPS fix
75
Send Ad-hoc-Msg (802.11)
2
Recv. Ad-hoc-Msg (802.11)
1
Msg MD Server (GPRS)
80
Msg Server MD (GPRS)
40
Comparison of different approaches: Approach
Reference
1
Naïve sensing approach
naive
2
Using adaptive sensing approach
adapt
3
Using adaptive sensing + ad-hoc information exchange
adhoc
Research Group Distributed Systems
Universität Stuttgart 13
IPVS
Evaluation – Results (1) Energy Consumption
Effectiveness 60
Sensor Readings
Energy Consumption [MJ]
35 30 25 20 15 10
5
50 40 30 20 10 0
0
100
200
naive
300
400
adapt
500
adhoc
Number of Mobile Devices
100
200
300
400
500
→ Energy Consumption decreases with optimizations → Effectiveness of sensing is not affected Research Group Distributed Systems
Universität Stuttgart 14
IPVS
Evaluation – Results (2) Positioning
Cellular Communication 90
GPRS Messages
40
GPS fixes
35 30 25 20 15 10
80
70 60 50 40 30 20
5
10
0
0
100
200
naive
300
400
adapt
500
adhoc
Number of Mobile Devices
100
200
300
400
500
→ Number of position fixes decrease significantly → Cellular communication can be reduced using ad-hoc Research Group Distributed Systems
Universität Stuttgart 15
IPVS
Related Work • Optimizations in Public Sensing ◦ [Lu et al. 2010] Bubble-Sensing: Binding sensing tasks to the physical world ◦ [Philipp et al. 2011] A Sensor network abstraction for flexible public sensing systems ◦ [Jung et al. 2010] User-profile driven collaborative bandwidth sharing on mobile phones
• Efficient Positioning ◦ [Priyantha et al. 2011] Littlerock: Enabling energy-efficient continuous sensing on mobile phones
◦ [Farell et al. 2011] Processing continuous range queries with spatiotemporal tolerance
◦ [Weinschrott et al. 2009] Efficient capturing of environmental data with mobile RFID readers
Research Group Distributed Systems
Universität Stuttgart 16
IPVS
Conclusion and Outlook
• Conclusion ◦ Energy Consumption in Public Sensing can be reduced by ▪ Adaption of positioning interval ▪ Using short-range instead of cellular radio
◦ Effectiveness of sensing is not affected • Future Work ◦ Support of continuous sensing queries ◦ Redundancy avoidance by selectively sending sensing locations
Research Group Distributed Systems
Universität Stuttgart 17
IPVS
Thanks for your attention
Research Group Distributed Systems
Universität Stuttgart 18
IPVS
Backup: Ad-hoc Information Exchange • Why not broadcast graph based distance to nearest sensing location instead of Euclidean ones? dmin Ad-hoc receiver Ad-hoc sender
dmin
• Structure of road graph influences value of dmin
• Actual dmin for ad-hoc receiver is smaller than dmin from sender
◦ Receiver could miss sensing location if using too long dmin for calculation • Use of Euclidean distances avoids this scenario Research Group Distributed Systems
Universität Stuttgart 19
IPVS
PSense: Reducing Energy Consumption in Public Sensing Systems The 26th IEEE Conference on Advanced Information Networking and Applications (AINA-2012) Fukuoka, Japan 26.03.2012
Patrick Baier, Frank Dürr and Kurt Rothermel
Outline • Public Sensing Overview • Motivation • System Model • System Overview • Naive Sensing Approach • Optimized Sensing Approach
◦ Adaptive Positioning ◦ Ad-hoc Information Exchange • Evaluation • Related Work
• Conclusion & Outlook Research Group Distributed Systems
Universität Stuttgart 2
IPVS
Public Sensing • utilizes mobile devices to opportunistically collect sensor data. • extracts sensor data from devices of common people.
+ No costly sensor deployment
+ Large coverage area • can be used to record different type of sensor data, e.g. creating a noise map of a city.
Research Group Distributed Systems
Universität Stuttgart 3
IPVS
Motivation However, Public Sensing … • drains an extensive amount of energy on mobile devices
◦ Positioning (GPS) ◦ Sensing ◦ Cellular Communication
Coordination of sensing Acquiring sensor data Uploading sensor data
• needs the acceptance of the mobile device owners
◦ Less users results in decrease of effectiveness of sensing Challenge: High energy consumption reduces user acceptance
Goal: Design an energy efficient Public Sensing system Research Group Distributed Systems
Universität Stuttgart 4
IPVS
System Model • Server
◦ Unlimited battery supplies ◦ Knowledge of road graph • Mobile devices
◦ ◦ ◦ ◦ ◦
Limited battery supplies GPS device Sensor with limited sensing range
Mobile Device Sensor Range
Move according to road graph Neither direction nor speed of movement controllable
• Communication
◦ Server Mobile device: Cellular Radio (UMTS,GPRS) ◦ Mobile device Mobile device: Short-range Radio (WLAN) Research Group Distributed Systems
Universität Stuttgart 5
IPVS
PSense - System Overview • PSense takes queries for sensing locations (SL) defined by
◦ Geographical Position ◦ Type of sensor value (e.g. temperature, noise) • Mobile devices sense data when within sensing range to SL • Sensor data is delivered back to application Sensing Sensor Locations Data
Sensing Location Mobile device
Application Interface
Sensing Locations
PSense
Sensor Data Research Group Distributed Systems
Universität Stuttgart 6
IPVS
Naive Sensing Approach 1. Server forwards SLs to devices. 2. Each device fixes position every tGPS sec. 3. Device checks if it can sense a SL No: Continue with step 2 Yes: Upload data to server 4. Server sends update of SLs to devices 5. Server returns sensor data to application
Sensing Sensor Locations Data
PSense
tGPS
tGPS
Research Group Distributed Systems
Universität Stuttgart 7
IPVS
Overview Optimizations
1. Adaptive Positioning
◦ Goal: Reduce number of GPS fixes 2. Ad-hoc Information Exchange
◦ Goal: Reduce number of cellular messages Goal: Minimize energy consumption of mobile devices But:
Effectiveness of sensing should not decrease Number of recorded sensing locations should be the same Research Group Distributed Systems
Universität Stuttgart 8
IPVS
Adaptive Positioning (1) • Idea: Reduce number of GPS fixes by adapting positioning interval
1. Server determines graph-based distance to closest SL 2. Device calculates earliest time span tGPS when SL can be reached 3. Device delays next position for tGPS
dmin
Starting points for sensing
𝑡𝐺𝑃𝑆 =
time until next GPS fix
vmax
maximum speed of device
distance to closest sensing location
𝑑𝑚𝑖𝑛 𝑣𝑚𝑎𝑥
Research Group Distributed Systems
tGPS dmin
Universität Stuttgart 9
IPVS
Adaptive Positioning (2) Optimized Sensing Workflow: 1. Each time a device fixed its position, it queries server for :
◦ Distance dmin to closest sensing location SL ◦ Position (x,y)SL of SL 2. Device checks whether it can sense SL 3. Calculate tGPS and delay next position fix
m1
m2
Message
Content
m1
(x,y)Device
m2
dmin and (x,y)SL
SL dmin
𝑡𝐺𝑃𝑆 =
𝑑𝑚𝑖𝑛 𝑣𝑚𝑎𝑥
Number of GPS fixes decrease But: Number of cellular messages increase Research Group Distributed Systems
Universität Stuttgart 10
IPVS
Ad-hoc Information Exchange (1) Idea: Reduce cellular traffic using ad-hoc messages dmin SL
When node receives distance to closest SL from server, nearby nodes … • have almost the same distance to this sensing location • can recalculate time until the next GPS fix and possibly extend it Server communication can be reduced
Research Group Distributed Systems
Universität Stuttgart 11
IPVS
Ad-hoc Information Exchange (2) broadcast range
dmin
m1
dmin- rah Ad-hoc receiver
m2
Ad-hoc sender
• Upon receiving dmin from server, node sends ad-hoc broadcast • All broadcast receivers compute tGPS’ • If tGPS’ > trem , schedule next GPS fix to time tGPS’
𝑡𝐺𝑃𝑆′ =
𝑑𝑚𝑖𝑛− 𝑟𝑎ℎ 𝑣𝑚𝑎𝑥
Research Group Distributed Systems
trem
remaining time until next GPS fix
rah
maximum ad-hoc range
Universität Stuttgart 12
IPVS
Evaluation – Setup • Simulation in ns-2 • Mobility on road graph of Chicago (UDELModel, pedestrian speed) • Use energy model to compare results • Mean number of sensing locations: 50 • Simulation time: 6 hours • Maximum ad-hoc range: 100m
Operation
Energy [mJ]
GPS fix
75
Send Ad-hoc-Msg (802.11)
2
Recv. Ad-hoc-Msg (802.11)
1
Msg MD Server (GPRS)
80
Msg Server MD (GPRS)
40
Comparison of different approaches: Approach
Reference
1
Naïve sensing approach
naive
2
Using adaptive sensing approach
adapt
3
Using adaptive sensing + ad-hoc information exchange
adhoc
Research Group Distributed Systems
Universität Stuttgart 13
IPVS
Evaluation – Results (1) Energy Consumption
Effectiveness 60
Sensor Readings
Energy Consumption [MJ]
35 30 25 20 15 10
5
50 40 30 20 10 0
0
100
200
naive
300
400
adapt
500
adhoc
Number of Mobile Devices
100
200
300
400
500
→ Energy Consumption decreases with optimizations → Effectiveness of sensing is not affected Research Group Distributed Systems
Universität Stuttgart 14
IPVS
Evaluation – Results (2) Positioning
Cellular Communication 90
GPRS Messages
40
GPS fixes
35 30 25 20 15 10
80
70 60 50 40 30 20
5
10
0
0
100
200
naive
300
400
adapt
500
adhoc
Number of Mobile Devices
100
200
300
400
500
→ Number of position fixes decrease significantly → Cellular communication can be reduced using ad-hoc Research Group Distributed Systems
Universität Stuttgart 15
IPVS
Related Work • Optimizations in Public Sensing ◦ [Lu et al. 2010] Bubble-Sensing: Binding sensing tasks to the physical world ◦ [Philipp et al. 2011] A Sensor network abstraction for flexible public sensing systems ◦ [Jung et al. 2010] User-profile driven collaborative bandwidth sharing on mobile phones
• Efficient Positioning ◦ [Priyantha et al. 2011] Littlerock: Enabling energy-efficient continuous sensing on mobile phones
◦ [Farell et al. 2011] Processing continuous range queries with spatiotemporal tolerance
◦ [Weinschrott et al. 2009] Efficient capturing of environmental data with mobile RFID readers
Research Group Distributed Systems
Universität Stuttgart 16
IPVS
Conclusion and Outlook
• Conclusion ◦ Energy Consumption in Public Sensing can be reduced by ▪ Adaption of positioning interval ▪ Using short-range instead of cellular radio
◦ Effectiveness of sensing is not affected • Future Work ◦ Support of continuous sensing queries ◦ Redundancy avoidance by selectively sending sensing locations
Research Group Distributed Systems
Universität Stuttgart 17
IPVS
Thanks for your attention
Research Group Distributed Systems
Universität Stuttgart 18
IPVS
Backup: Ad-hoc Information Exchange • Why not broadcast graph based distance to nearest sensing location instead of Euclidean ones? dmin Ad-hoc receiver Ad-hoc sender
dmin
• Structure of road graph influences value of dmin
• Actual dmin for ad-hoc receiver is smaller than dmin from sender
◦ Receiver could miss sensing location if using too long dmin for calculation • Use of Euclidean distances avoids this scenario Research Group Distributed Systems
Universität Stuttgart 19
IPVS
