Presented by Anwar Saipulla Umass Lowell Slides are adopted
from original authors presentation 1
Slide 2
Papers An Optimization Framework for Joint Sensor Deployment,
Link Scheduling and Routing in Underwater Sensor Networks Leonardo
Badia, Michele Mastrogiovanni, Chiara Petrioli, Stamatis
Stefanakos, Michele Zorzi, WUWNet06 Surface-Level Gateway
Deployment For Underwater Sensor Networks Saleh Ibrahim, Jun-Hong
Cui, Reda Ammar, Milcom07 Deployment Analysis in Underwater
Acoustic Wireless Sensor Networks D. Pompili, T. Melodia, I.F.
Akyildiz, WUWNet06 2
Slide 3
Leonardo Badia*, Michele Mastrogiovanni +, Chiara Petrioli +,
Stamatis Stefanakos +, Michele Zorzi *IMT Lucca, Italy University
of Padova, Italy + University of Rome La Sapienza, Italy 3
Slide 4
Outline Underwater sensor networks Joint Optimization of Link
Scheduling, Routing and Sensor Placement Underwater challenges
(propagation, delay, interference) Proposed solution Conclusions
and future works 4
Slide 5
Underwater Sensor Networks Sensor networks are generally used
for surveillance, monitoring and low-cost communication. In
underwater scenarios, they can be employed to detect earthquakes,
incoming tsunamis, water pollution, global warming, and other
important phenomena. 5
Slide 6
Underwater Sensor Networks The network can comprise columns of
submarine acoustic sensors, e.g. linked to moored buoys and hence
to the land. 6
Slide 7
Underwater Sensor Networks For the modeling purpose, the
network can be often represented as a directed graph, where the
sensor nodes need to deliver their measurements to a sink. If nodes
position are assumed to be stationary and link gains are
sufficiently stable, it can be thought of optimizing this delivery.
7
Slide 8
Underwater Sensor Networks This optimization involves an
underlying TDMA access, where transmission slots of a periodic
frame are assigned in a centralized and coordinated manner. The
optimization must account for: Power expenditure of the nodes (the
traditional energy saving needs assumed for RF sensors is even more
important for underwater networks) Transmission and interference
constraints. 8
Slide 9
Joint Optimization of Instead of optimizing each network layer
separately, we go for a cross-layer solution. We seek to optimize
jointly: Link scheduling (including Power Control) Routing Node
placement This is obtained through a proper ILP framework. 9
Slide 10
Joint Optimization of This can be done by representing with
binary variables: The positioning of a node in a candidate position
i, represented by y i The activation of a link (i,j) at time t,
represented by X ij (t) The allocation of these variables must
respect several constraints. 10
Slide 11
Underwater challenges Such optimization procedures have been
studied for traditional wireless networks. However, the underwater
medium has special characteristics that need to be addressed. In
particular, the most important differences are in the propagation
and delay aspects. 11
Slide 12
Underwater challenges Usually, in wireless radio networks the
signal received power is a strongly decreasing function of the
distance d (e.g., ~ d -4 ) For underwater scenarios: where in
particular:
Slide 13
Underwater challenges For RF networks, physical propagation
delay is negligible (when it is not zero, it is mainly because of
the processing delay). Underwater, it is not. It can be comparable
to or even larger than the packet transmission time. bb time Node a
cc time Node b aa cc cc
Slide 14
ILP Formulation main constraints: For any case where t e and t
f overlap 14
Slide 15
Proposed solution For our preliminary investigations: Power
Control is dealt with by considering a single link for any
available power level. Only links with a gain high enough are
considered. Interference constraints are modeled by means of a
simple two-step approach: the ILP considers the protocol
interference model and check the SIR a posteriori, if it is
violated adds a constraint and re-runs. 15
Proposed solution We used GLPK version 4.8, an exact solver of
ILP problems. The numerical data reproduce UWM1000 LinkQuest
underwater acoustic modem. The grid size is 600 600 200 meters. Two
power levels are available (2W and 8W). 17
Slide 18
Proposed solution The higher the parallelism of the solution,
the better. Hence, the optimal solution has a high degree of
parallelism. Shown links are not activated at the same time, but
this does not guarantee they do not interfere!! Links 2 18 and 4 18
are activated at the same time but do not interfere, thanks to the
different delays! 18
Slide 19
Conclusions and Future Work Optimal link scheduling and routing
can obtain the highest possible performance in severely constrained
scenarios such as underwater acoustic networks. However, their
evaluation is subject to many design constraints. Optimized
solution are not easy to find. 19
Slide 20
Conclusions and Future Work Capturing the interference
phenomena into a viable mathematical model can be an interesting
topic for future research. Possible ways to further improve the
optimization framework are under study. 20
Slide 21
Questions? 21
Slide 22
Papers An Optimization Framework for Joint Sensor Deployment,
Link Scheduling and Routing in Underwater Sensor Networks Leonardo
Badia, Michele Mastrogiovanni, Chiara Petrioli, Stamatis
Stefanakos, Michele Zorzi, WUWNet06 Surface-Level Gateway
Deployment For Underwater Sensor Networks Saleh Ibrahim, Jun-Hong
Cui, Reda Ammar, Milcom07 Deployment Analysis in Underwater
Acoustic Wireless Sensor Networks D. Pompili, T. Melodia, I.F.
Akyildiz, WUWNet06 22
Slide 23
Saleh Ibrahim, Jun-Hong Cui, Reda Ammar Computer Science &
Engineering Dept. University of Connecticut 23
Slide 24
Underwater Sensor Networks Radio Does Not Work Well in Water
Acoustic Communication More Practical High Propagation Delay
Propagation speed 1500 m/s for sound vs. 3x10 8 m/s for EM waves
Low Available Bandwidth Heavily depending on transmission range
& frequency Most acoustic systems operate below 30kHz Range x
Rate product less than 40 km x kbps 24
Slide 25
Motivation Use Surface Gateways to Improve Performance of UWSN:
Delay, Energy Consumption, etc. Multiple Surface Gateway Nodes
Relay Traffic between Underwater Nodes and the Control Center
Slide 26
Surface Gateways Deployment Given : UWSN, Node Locations Data
Gen. Rates Candidate Locations for Surface Gateways Find : Optimal
Deployment Locations Minimizing : Expected End-to-end Delay
Expected Per Packet Energy Consumption Required Number of Surface
Gateways 26
Slide 27
Plan Integer Linear Programming Problem Solve Sample Problems
to Investigate Effect of: Number of Surface Gateways Network Load
Underwater Deployment Pattern 27
Slide 28
ILP Constraints (1) For each candidate location t i Limit
number of surface gateways 28
Slide 29
ILP Constraints (2) No flow f in edge e i going into a
candidate location t i, unless a node is deployed there 29
Slide 30
ILP Constraints (3) Flow conservation at each node End-to-End
Flow conservation
Minimizing Expected Delay Delay t of Edge e L message length, B
bit-rate, l(e) distance, v p propagation velocity. Minimize
Expected End-to-End Delay Minimize 32
Slide 33
Minimizing Expected Energy Energy Consumption at Edge e s (e)
is the transmission power at e Minimize Expected End-to-End
Per-Packet Energy Consumption Minimize 33
Slide 34
Sample Uniform UW Problem Underwater 7x7 2D-Mesh of sensors at
100m deep. 600x600m horizontal area Comm. range R=150m Data
generation g=1pkt/sec Candidate locations 5x5 mesh Fixed packet
length, propagation velocity, transmission power
Slide 35
Results and Observations (1) End to End Delay vs. # of Gateways
with Varying Bit-rate Due to Transmission time Minimum # Gateways
that Makes the Problem Feasible.
Slide 36
Results and Observations (2) Energy Per Packet vs. # of
Gateways with Varying Bit-rate Due to Transmission time
Slide 37
Sample Random UW Problem Underwater 49 Randomly-located UW
nodes Same horizontal area Multiple problem instances Candidate
Locations Same 5x5 mesh Same communication range, data generation
rate, packet length, propagation velocity, transmission power.
Slide 38
Results and Observations (3) Random U.W. deployment results
very similar to Uniform case.
Slide 39
Results and Observations (4) 39
Slide 40
Conclusion and Future Work Conclusion Benefits of Multiple
Surface Gateways Effect of Network Load Effect of Underwater
Deployment Pattern Future Work Developing Heuristic Solutions
Candidate Deployment Location Schemes Joint Optimization of Surface
and UW 40
Slide 41
Questions? Discussions [1] Key Assumption: known Candidate
Deployment Locations [2] Other dangerous assumption: symmetric
communication links; A virtue sink: ignoring communication issues
between sink nodes.. My solution: grid point approximation 41
Slide 42
Papers An Optimization Framework for Joint Sensor Deployment,
Link Scheduling and Routing in Underwater Sensor Networks Leonardo
Badia, Michele Mastrogiovanni, Chiara Petrioli, Stamatis
Stefanakos, Michele Zorzi, WUWNet06 Surface-Level Gateway
Deployment For Underwater Sensor Networks Saleh Ibrahim, Jun-Hong
Cui, Reda Ammar, Milcom07 Deployment Analysis in Underwater
Acoustic Wireless Sensor Networks D. Pompili, T. Melodia, I.F.
Akyildiz, WUWNet06 42
Slide 43
Deployment Analysis in Underwater Acoustic Wireless Sensor
Networks D. Pompili, T. Melodia, I.F. Akyildiz Broadband and
Wireless Networking Laboratory School of Electrical and Computer
Engineering Georgia Institute of Technology 43/29
Slide 44
Outline Propose 2D and 3D architectures for UW-ASNs State
objectives of the paper Study graph properties of 2D ocean-bottom
UW-ASNs Propose and evaluate 3D deployment strategies Conclusions
and future work 44/29
Slide 45
45/29 Two-dimensional Architecture
Slide 46
Three-dimensional Architecture 46
Slide 47
Objectives 2D Architecture: Determine the minimum number of
sensors and uw- gateways to achieve communication and sensing
coverage Provide guidelines on how to choose the optimal deployment
surface area, given a target region 3D Architecture: Evaluate
different deployment strategies Determine the minimum number of
sensors needed to achieve the target sensing coverage 47
Slide 48
Graph Properties of Bottom UW-ASNs We analyze the graph
properties of devices (sensors and uw-gateways) when they are
deployed on the ocean surface, sink, and reach the bottom We study
the trajectory of sinking devices deployed on the ocean surface
when: Sensors are randomly deployed on the ocean surface (e.g.,
scattered from an airplane), or Sensors are accurately positioned
on the surface (e.g., released from a vessel) 48/29
Slide 49
Triangular-grid Surface Deployment Sensors with same sensing
range r Optimal deployment to cover a 2D area with minimum number
of sensors: Center sensors at the vertex of a grid of equilateral
triangles 49/29 d A FE D d d r r r *d = (3)/2*r d/r = 1.732*r
Slide 50
Triangular-grid Coverage 50/29 Coverage=0.95 Ratio of sensor
distance and sensing range=d/r=1.95
Slide 51
Minimum No. of Sensors in Tri-grid A 1 =100x100m 2 A 2
=300x200m 2 r in[10,35]m d*/r=1.95 N =
{100*100/[(3/4)*(1.95*15)^2]} = 26.99N =
*{300*200/[(3/4)*(1.95*15)^2] }= 161.96