Post on 15-Jan-2016
uSense: A Unified Asymmetric Sensing Architecture for Wireless Sensor Networks
Yu Gu, Joengmin Hwang, Tian He and David Du
Minnesota Embedded Sensor System (MESS)Department of Computer Science & Engineering
http://mess.cs.umn.edu
Outline
Motivation Overview of uSense Architecture Global Scheduling Algorithms System Implementation and
Evaluation Conclusion
Motivation (Different Services)
Coverage
Life Time
Detection Delay
Stealth Distance
A Flexsible Solution?
…
C. Gui et al., Mobicom’04S. Ren et al., MC2R 2005
X. Wang et al. Sensys’03Q. Cao et al. IPSN’05C. Chiasserini et al. Infocom’04
T. Yan et al., Sensys‘03M. Cardei et al., Infocom’05
S. Kumar et al. Mobicom’04S. Slijepcevic et al. , ICC’01D. Tian et al. , Wireless Communications and Mobile Computing Journal 2003
New Design Philosophy
Sensor Functions
Essential functions Non-Essential functions
uSense Asymmetric Architecture
uSense Asymmetric Architecture
Sensor i
Sensor j
Sensor k
GenericSwitching Algorithm
Sensor Network
Algorithm 1
Algorithm 2
Algorithm n
…..
Para
met
er T
rans
latin
g
ComputationalEntity
Connectivity
Parameters
Generic Switching Algorithm
Scheduling Bits
Sensor i
Sensor j
Sensor k
GenericSwitching Algorithm
Sensor Network
Switching Rate
10110101
On
0.5HZ 16s round time
Algorithm2
Algorithm1
Algorithm2
Algorithm1
Algorithm2
Algorithm2
Algorithm1
Have We Solved Some Problems?
Flexibility issue of existing protocols?
Before:
Algorithm1
Complete code for
Algorithm2
Parameters2
Parameters2
Parameters2
Parameters2
Generic Switching
Generic Switching
Generic Switching
Generic Switching
Have We Solved Some Problems?
Flexibility issue of existing protocols?
uSense:Parameters1
Parameters1
Parameters1
Parameters1Schedule BitsSwitching
Rate
•Only need to disseminate two parameters
•Generic Switching unaffected
Outline
Motivation Motivation uSense ArchitectureuSense Architecture Global Scheduling Algorithms System Implementation and
Evaluation Conclusion
Sensor Network
Sensor i
Sensor j
Sensor j
Generic Switching Algorithm
Para
mete
r Tra
nsl
ati
ng
CCP
…
uScan
Virtual Patrol
Computational Entity
DiffSurv
Algorithm n
Algorithm 1
Algorithm 2
Algorithm 3Connectivity
Parameters
uScan Overview
Sensor Network
Sensor i
Sensor j
Sensor j
Generic Switching Algorithm
Para
mete
r Tra
nsl
ati
ng
CCP
…
uScan
Virtual Patrol
Computational Entity
DiffSurv
Connectivity
Parameters
Tile Level Scheduling
Line Scanning
Systolic Scanning
Node Level Scheduling
Para
mete
r Tra
nsl
ati
ng
CCP
…
uScan
Virtual Patrol
Computational Entity
DiffSurv
uScan Overview
Tile Level SchedulingTessellations
Line ScanEnergy saving
Systolic ScanMinimal Worst-case Breach
(10000)*
(00100)*
(00001)*
Node Scheduling {N1,N5} {N2,N3} {N1,N4}N2
N1
N3
N5
N4
T1
T2
T3
T4
T5
Schedule(N1)=(00100 00000 00100)*
Schedule(N2)=(00000 00100 00000)*
Schedule(N3)=(00000 00100 00000)*
Schedule(N4)=(00000 00000 00100)*
Schedule(N5)=(00100 00000 00000)*
(00100)*
Polynomial solution for set-cover
Physical Coverage
Bipartite Graph
Set Cover
Outline
MotivationMotivation Overview of uSense ArchitectureOverview of uSense Architecture Global Scheduling AlgorithmsGlobal Scheduling Algorithms System Implementation and
Evaluation Conclusion
System Implementation
Life Time under Full Coverage
At Node Density of 4, uSense outperforms DiffSurv 1.6 times
DiffSurv: T. Yan et al. “Differentiated Surveillance service for sensor networks”, Sensys 2003
Life Time under Partial Coverage
At Node Density of 10, uSense outperforms Virtual Patrol 27 timesVirtual Patrol: C. Gui and P. Mohapatra, “Virtual Patrol: a new power conservation design for surveillance using sensor
networks”, IPSN 2005
Future Work
Incorporating more sensing protocols to further strengthen the flexibility of uSense
Consider communication issues in uScan scenarios
Resilience to node failures
Conclusion (1)
uSense, provides a unified sensing architecture Simple generic switching algorithm in
sensor nodes Allows changing sensing algorithms
with only two parameters
Conclusion (2)
uScan, a two-level global scheduling algorithm Seamlessly supported by the uSense Generic node scheduling for different
sensing coverage requirements Significantly more energy efficient
than the localized algorithms