Title: Robust Gridbased Deployment Schemes for Underwater Optical Sensor Networks
1Robust Grid-based Deployment Schemes for
Underwater Optical Sensor Networks
- Abdullah Reza
- Janelle harms
- Department of Computing Science
- University of Alberta
- Edmonton, Alberta, Canada
2Organization
- Underwater Sensor Network
- Problems of Acoustic Communication and Motivation
- Problem Definition
- Formulation of Robust Grid-Based Deployment
Patterns - Performance Evaluation
- Conclusions and Future Work
3Underwater Sensor Networks (UWSN)
- Immense Research Potential
- Unmanned sea/ocean observation
- 70 of earth surface constitutes water
- A new and evolving field for research
4UWSN Architecture-2d Akyildiz et al. 1
5UWSN Architecture-3d Akyildiz et al. 1
6Difference from Terrestrial
- Nodes 1
- More expensive
- Sparse Deployment
- Uncorrelated reading
- Higher Energy consumption
7Difference from Terrestrial
- Physical Layer (Acoustic)
- Penetrates well in water
- Long range (several hundred meters)
- - Very low bandwidth (5 Kbps 2)
- - Very slow propagation (1500 m/s 3)
- - High error rate 4, 5
- - High cost of acoustic modems 2
8Optical Communication An Alternative of
Acoustic Communication
- Advantages
- High bandwidth
- experimented with 320 Kbps 2 and 4.4 Mbps 10
- Fast propagation
- Disadvantages
- Reduced range of communication
- 8m to 40m 2, 10
9Optical UWSN Target Application
- Stringent requirement for high bandwidth
- Smaller area of observation
- Observation area not close to the coast
10Optical UWSN Design Challenges
- Intelligent Deployment (expensive nodes)
- Connectivity is not inherent
- Cost of Deployment (directional transceivers)
11Optical UWSN Example 1 Vasilescu et al.2
- Hybrid network both optical and acoustic
interface - Acoustic communication is used for localization
- An AUV (Autonomous Underwater Vehicle) goes to
each node periodically and downloads sensed data
optically using short-range communication - A powerful and intelligent AUV is needed and high
amount of energy is expended in AUV movement
12Optical UWSN Example 2 Kedar et al. 16
- Oceanographic Contamination Detection System
- Distributed clusters of sensor nodes
- Each cluster consists of a sink and a number of
sensor nodes transmitting directly to the sink
optically - Inter-cluster connectivity to support wider area
of observation
13Our Approach An Optical UWSN with Static Sink(s)
- The VENUS Project at University of Victoria 11
14Problem Definition
- Place nodes on grid corners
- Select a set of point-to-point links from the
available pool - Placing nodes on the grid and setting their
interface directions can be done manually or
using AUVs
15Problem Definition Design Goals
- 1) Robustness
- Deterministic (2-edge-connected)
- Probabilistic
- 2) Path Quality
- 3) Interface-count
- Per node constraint (Maximum 1, 2 and 3
interfaces per node) - Minimum total interface in the network
16Optical Interface Model 2
- Maximum 1 interface per node
- 2 or more interfaces per node
17Robust Deployment Maximum 1 Interface per Node
- Cycles are needed to produce Connected Topology
- We have to span all nodes
- One Hamiltonian Cycle Always possible with even
number of rows and/or columns 12 - Robustness poor
- Path quality poor
18Robust Deployment Maximum 1 Interface per Node
- A Better Design 4 Hamiltonian Cycles
- Robustness Better
- Not 2-edge-connected yet
- Path Quality Better
- The sink has longer links and more than one
interface
19Robust Deployment Maximum 2 Interfaces per Node
- We have undirected edges now
- 2-degree constraint on each node
- We aim to build a 2-edge-connected topology
- Only way to build a 2-ec topology with 2-degree
constraint is to build a ring/cycle spanning all
nodes 13
20Robust Deployment Maximum 2 Interfaces per Node
- We use 4 undirected Hamiltonian Cycles
- Robustness
- 2-edge-connected
- Better than directed cycles
- Path Quality
- Better than directed cycles
21Robust Deployment Maximum 3 Interfaces per Node
- Design Approach
- 1) Generate a 3-degree constrained shortest path
spanning tree from the sink with minimum number
of 3-d nodes (optimal pattern) - 2) Add additional edges to make it
2-edge-connected - 3) Add additional edges to improve probabilistic
robustness - Introduce minimum number of 3-degree nodes in
each step
22Manhattan Distance Property 14
- Manhattan Distance between points (x1, y1) and
(x2, y2) in a grid is given by x1-
x2y1-y2
23Subproblem Quadrants
24Lower Bound for Number of 3-d nodes in a Quadrant
- Theorem Consider a quadrant with sides of size m
and n where mn. A sink is placed in a corner. A
3-degree constrained shortest-path tree rooted at
the sink and spanning all nodes inside and on the
boundaries of the quadrant require at least (m-2)
3-degree nodes.
25Optimal Pattern for a Quadrant
26Do Optimal Patterns for Quadrants Give Optimal
Pattern for the Grid?
- Depends on whether or not 3-degree constraint is
violated on the axes - In any case, (l1l2l3l4) remains the lower
bound - We call LB (l1l2l3l4) ? li
-
27A Pattern for a Quadrant that Avoids Conflict on
the Axes
- A pattern that draws edges only from the vertical
axis - Has minimum number of 3-d nodes (li ) of y x
but has minimum1 number of 3-d nodes (li1) if y
gt x
28Pattern Applied on the Entire Grid
- Shortest path spanning tree from the sink with
3-degree constraint - Number of 3-degree nodes in the worst case is
LB4 which can be shown to be actually LB3 - clockwise counterclockwise LB2 in worst case
29Next Step Add Edges to Make it 2-edge-connected
30Next Step Add 2 More Edges to Make it More Robust
31Summary of Proposed Deployments
TOP2
TOP3
TOP1
TOP4
TOP5
TOP6
32Summary of Proposed Deployments 12x12 Grid
33Static Evaluation of Deployment Topologies
- Two failure models
- Isolated Model 15
- Patterned Model 15
- Metrics Robustness Path Quality
- 12x12 grid with r 20m
34Results Isolated Failure
Path Quality
Robustness
35Results Patterned Failure
Path Quality
Robustness
b4m and ?3
36Dynamic Evaluation of Deployment Topologies
- Resilient Routing Protocols
- Memory-constrained Flooding (FLD)
- Dual Paths Protocol (DPP)
- Hop-by-Hop Acknowledgment with local update
protocol (HHA)
37HHA Protocol Routing Around Error Blobs
38Simulation Environment
- Single Packet generating source
- Ten packets generated per second, each with 1 Kb
payload - Three error blobs (20mx4m) moving at 15 cm/sec
speed in the grid in a random-walk fashion
39Simulation Metrics
- Delivery Ratio
- Average Delay Per Packet (ADPP)
- Average number of Payloads Transmitted per
Successful packet (APTS)
40Experiments
- 3 experiments for each protocol
- Vary hop distance of the source node from the
sink - Vary the size of obstructing error blobs (not
presented) - Vary the speed of obstructing error blobs (not
presented)
41Comparison of 3 protocols on TOP6
Hop Distance 10 A 20m b 4m Speed 15cm/sec
42Comparison of 3 protocols on TOP6
Hop Distance 10 A 30m b 4m Speed 15cm/sec
43Conclusions
- Robust deployment topologies for optical UWSN
with 1, 2 and 3 interfaces per node constraints - For 1 and 2 interfaces cases, topologies that
utilize four directed and undirected Hamiltonian
cycles in the grid, respectively. - For 3 interfaces case, formulation pattern for a
3-degree constrained shortest path spanning tree
in the grid with arbitrary root and arbitrary
dimension that results in (LB2) 3-degree nodes
in the worst case. - A series of deployment patterns built on the
3-degree constrained shortest path tree
44Conclusions
- Static evaluation of the proposed topologies show
that a very high degree of robustness (95) is
maintained by TOP6 even at reasonably harsh
conditions - Dynamic evaluation of the proposed topologies
with three simple routing protocols - FLD achieves a very high degree of delivery ratio
when applied on TOP6 but incurs excessive
communication - DPP fails to utilize the inherent redundancies of
the proposed topologies - HHA achieves delivery ratio as good as FLD with
very small communication overhead but slightly
higher average delay
45Future Work
- 3-degree constrained shortest path spanning tree
for a grid with LB number of 3-degree nodes (or
prove that such a tree cannot be formed) - Prove or disprove that our pattern is a pattern
with minimum number of 3-degree nodes in all
settings - Design similar patterns for 3D deployment
scenario - Consider non-grid deployments such as triangular
or hexagonal
46Future Work
- Experiment with multiple packet generating
sources in the network - Design more intelligent routing protocols that
perform load balancing - Use more intelligent multi-path routing protocols
in evaluation - Optimize HHA to avoid including path and link
information in the packet
47References
- 1 I.F. Akyildiz, D. Pompili, and T.
Melodia, Challenges for efficient communication
in underwater acoustic sensor networks, ACM
SIGBED Review, Volume 1, Issue 2, July, 2004,
pages 3-8. - 2 I. Vasilescu, K. Kotay, D. Rus, M.
Dunbabin, and P. Corke, Data collection,
storage, and retrieval with an underwater sensor
network, In Proc. of SenSys05, San Diego, CA,
USA, November, 2005. - 3 D. Pompili, T. Melodia, and I.F.
Akyildiz, Routing algorithms for
delay-insensitive and delay-sensitive
applications in underwater sensor networks, In
Proc. of MobiCom06, Los Angeles, CA, USA,
September, 2006. - 4 P. Sun, W.K.G. Seah, and P.W.Q. Lee,
Efficient data delivery with packet cloning for
underwater sensor networks, In Proc. of UT07 and
SSC07, Tokyo, Japan, April, 2007. - 5 P. Xie, J.H. Cui, and L. Lao, VBF
vector-based forwarding protocol for underwater
sensor networks, UCONN Technical Report
UbiNet-TR05-03, February, 2005. - 6 M. Molins, and M. Stojanovic,
Slotted FAMA a MAC protocol for underwater
acoustic networks, In Proc. of IEEE Oceans06,
Boston, MA, USA, September, 2006. - 7 P. Xie, and J.H. Cui, R-MAC An
energy-efficient MAC protocol for underwater
sensor networks, In Proc. of WASA 2007, Chicago,
IL, USA, August, 2007. - 8 W.K.G. Seah, and H.P. Tan, Multipath
virtual sink architecture for wireless sensor
networks in harsh environments, In Proc. of
InterSense06, Nice, France, May, 2006. - 9 J. H. Smart, Underwater optical
communications systems part 1 variability of
water optical parameters, In Proc. of IEEE
MILCOM 2005, Atlantic City, NJ, USA, October,
2005. - 10 J.W. Giles, and I.N. Bankman,
Underwater optical communications systems part
2 basic design considerations, In Proc. of
IEEE MILCOM 2005, Atlantic City, NJ, USA,
October, 2005. - 11 http//www.venus.uvic.ca/discover_venus
/mainpage.html - 12 S. Skiena, Implementing Discrete
Mathematics Combinatorics and Graph Theory with
Mathematica, Addison-Wesley, Redwood City, CA,
1990, pages 147-148. - 13 J. Llorca, A. Desai, and S. Milner,
Obscuration minimization in dynamic free space
optical networks through topology control, In
Proc. of IEEE MILCOM 2004, Monterey, CA, USA,
November, 2004 - 14 E.F. Krause, Taxicab Geometry, Dover,
New York, 1987. - 15 D. Ganesan, R. Govindan, S. Shenker,
and D. Estrin, Highly-resilient,
energy-efficient multipath routing in wireless
sensor networks, ACM SIGMOBILE Mobile Computing
and Communications Review, Volume 4, Issue 4,
October, 2001 - 16 D. Kedar and S. Arnon, A distributed
sensor system for detection of contaminants in
the ocean, In Proc of SPIE (Society of
Photographic Instrumentation Engineers), Vol.
6399, 639903(2006), Stockholm, Sweden, 2006.