Research Profile

1
Research Profile

• Guoliang Xing
• Assistant Professor
• Department of Computer Science and Engineering
  Michigan State University

2
Background

• Education
• Washington University in St. Louis, MO
• Master of Science in Computer Science, 2003
• Doctor of Science in Computer Science, 2006,
  Advisor Chenyang Lu
• Xian JiaoTong University, Xian, China
• Master of Science in Computer Science, 2001
• Bachelor of Science in Electrical Engineering,
  1998
• Work Experience
• Assistant Professor, 8/2008 , Department of
  Computer Science and Engineering, Michigan State
  University
• Assistant Professor, 8/2006 8/2008, Department
  of Computer Science, City University of Hong Kong
• Summer Research Intern, May July 2004, System
  Practice Laboratory, Palo Alto Research Center
  (PARC), Palo Alto, CA

3
Research Summary

• Mobility-assisted data collection and target
  detection
• Holistic radio power management
• Data-fusion based network design
• Publications
• 6 IEEE/ACM Transactions papers since 2005
• 20 conference/workshop papers
• First-tier conference papers MobiHoc (3), RTSS
  (2), ICDCS (2), INFOCOM (1), SenSys (1), IPSN
  (3), IWQoS (2)
• The paper "Integrated Coverage and Connectivity
  Configuration in Wireless Sensor Networks" was
  ranked the 23rd most cited articles among all
  papers of Computer Science published in 2003
• Total 780 citations (Google Scholar, 2009 Jan.)

4
Methodology

• Explore fundamental network design issues
• Address multi-dimensional performance
  requirements by a holistic approach
• High-throughput and power-efficiency
• Sensing coverage and comm. performance
• Exploit realistic system platform models
• Combine theory and system design

5
Selected Projects on Sensor Networks

• Integrated Coverage and Connectivity
  Configuration
• Holistic power configuration
• Rendezvous-based data collection

6
Coverage Connectivity

• Select a subset of sensors to achieve
• K-coverage every point is monitored by at least
  K active sensors
• N-connectivity network is still connected if N-1
  active nodes fail

Active nodes
Sensing range
Sleeping node
Communicating nodes
A network with 1-coverage and 1-connectivity
7
Coverage Connectivity

• Select a set of nodes to achieve
• K-coverage every point is monitored by at least
  K active sensors
• N-connectivity network is still connected if N-1
  active nodes fail

Active nodes
Sensing range
Sleeping node
Communicating nodes
A network with 1-coverage and 1-connectivity
8
Connectivity vs. Coverage Analytical Results

• Network connectivity does not guarantee coverage
• Connectivity only concerns with node locations
• Coverage concerns with all locations in a region
• If Rc/Rs ? 2
• K-coverage ? K-connectivity
• Implication given requirements of K-coverage and
  N-connectivity, only needs to satisfy max(K,
  N)-coverage
• Solution Coverage Configuration Protocol (CCP)
• If Rc/Rs lt 2
• CCP connectivity mountainous protocols

ACM Transactions on Sensor Networks, Vol. 1 (1),
2005. First ACM Conference on Embedded Networked
Sensor Systems (SenSys), 2003
9
Understanding Radio Power Cost
Radio States Transmission Ptx Reception Prx Idle Pidle Sleeping Psleep
Power consumption (mw) 21.2106.8 32 32 0.001
Power consumption of CC1000 Radio in different
states

• Sleeping consumes much less power than idle
  listening
• Motivate sleep scheduling Polastre et al. 04, Ye
  et al. 04
• Transmission consumes most power
• Motivate transmission power control Singh et al.
  98,Li et al. 01,Li and Hou 03
• None of existing schemes minimizes the total
  energy consumption in all radio states

10
Example of Min-power Backbone
c

• a sends to c at normalized rate of r
  Data Rate / Bandwidth
• Nodes on backbone remain active
• Backbone 1 a ? b ? c
• Backbone 2 a ?c, b sleeps

b
a
11
Power Control vs. Sleep Scheduling
Transmission power dominates use low
transmission power
Power Consumption
3Pidle
2PidlePsleep
1
r0
Idle power dominates use high transmission power
since more nodes can sleep
12
Problem Formulation

• Given comm. demands I( si , ti , ri ) and
  G(V,E), find a sub-graph G(V, E) minimizing

sum of edge cost from si to ti in G
independent of data rate!
node cost

• Sleep scheduling

• Sleep scheduling
• Power control

• Sleep scheduling
• Power control
• Finding min-power backbone is NP-Hard

13
Two Online Algorithms

• Incremental Shortest-path Tree Heuristic
• Known approx. ratio is O(k)
• Adapt to dynamic network workloads and different
  radio characteristics
• Minimum Steiner Tree Heuristic
• Approx. ratio is 1.5(PrxPtx-Pidle)/Pidle ( 5
  on Mica2 motes)

ACM International Symposium on Mobile Ad Hoc
Networking and Computing (MobiHoc), 2005
14
Data Transport using Mobiles
Base Station
5 mins
150K bytes
Robomote _at_ USC
10 mins
500K bytes
5 mins
100K bytes
100K bytes

• Analogy
• What's best way to send 100 G data from HK to DC?

Networked Infomechanical Systems (NIMS) _at_ UCLA
15
Rendezvous-based Data Transport

• Some nodes serve as rendezvous points (RPs)
• Other nodes send data to the closest RP
• Mobiles visit RPs and transport data to base
  station
• Advantages
• Combine In-network caching and controlled
  mobility
• Mobiles can collect a large volume of data at a
  time
• Minimize disruptions due to mobility
• Achieve desirable balance between latency and
  network power consumption

16
Summary of Solutions

• Fixed mobile trails
• Without data aggregation, an optimal algorithm
• With data aggregation, NP-Hard, a constant-ratio
  approx. algorithm
• Free mobile trails w/o data aggregation
• Without data aggregation, NP-Hard, an efficient
  greedy heuristic
• With data aggregation, NP-Hard, a constant-ratio
  approx. algorithm
• Mobility-assisted data transport protocol
• Robust to unexpected comm./movement delays

ACM International Symposium on Mobile Ad Hoc
Networking and Computing (MobiHoc), 2008 IEEE
Real-Time Systems Symposium (RTSS), 2007
17
Impact of Data Fusion on Network Performance

• Data fusion in sensor networks
• Combine data from multiple sources to achieve
  inferences
• Value fusion, decision fusion, hybrid fusion
• Enable collaboration among resource-limited
  sensors
• Fusion architecture in wireless sensor networks
• Sensors close to each other participate in fusion
  
• Fusion is confined to geographic proximity
• Impact on network-wide performance
• Capability of sensors is limited to local fusion
  groups
• Complicate system behavior
• Modeling, calibration, mobility etc. becomes
  challenging

18
Our Work on Data Fusion

• Virtual fusion grids
• Dynamic fusion groups for effective sensor
  collaboration
• Sensor deployment
• Controlled mobility in fusion-based target
  detection
• System-level calibration in fusion-based
  sensornet
• Project ideas
• Focus on fundamental impact of data fusion

19
(No Transcript)
20
Problem Formulation
base station

• Constraint
• Mobiles must visit all RPs within a delay bound
• Objective
• Minimize energy of transmitting data from sources
  to RPs
• Approach
• Joint optimization of positions of RPs, mobile
  motion paths and data routes

mobile
rendezvous point
source node
21
(No Transcript)