Underwater Acoustic Sensor Networks: Medium Access Control, Routing and Reliable Transfer Peng Xie Dissertation Proposal Committee: Jun-Hong Cui, Reda A Ammar, Sanguthevar Rajasekaran, Bing Wang Computer Science & Engineering University of Connecticut
66
Embed
Underwater Acoustic Sensor Networks: Medium Access Control, Routing and Reliable Transfer Peng Xie Dissertation Proposal Committee: Jun-Hong Cui, Reda.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Underwater Acoustic Sensor Networks: Medium Access Control,
Routing and Reliable Transfer
Peng XieDissertation Proposal
Committee: Jun-Hong Cui, Reda A Ammar, Sanguthevar Rajasekaran, Bing Wang
Computer Science & Engineering
University of Connecticut
Outline
• Introduction– Motivation & challenges
• Three fundamental networking problems– Medium access control– Multi-hop routing– Reliable data transfer
• Conclusions and future work
Why Underwater?
• The Earth is a water planet– About 2/3 of the Earth covered by oceans
• Largely unexplored, huge amount resources to discover
• Many potential applications– Long-term aquatic monitoring
• Oceanography, seismic predictions, pollution detection, oil/gas field monitoring …
• Mobility & UW-A channel limitations open the door to very challenging networking issues
Objective & Contributions
• The final objective: – Build efficient, reliable, and scalable M-UWSNs
• This dissertation work address three fundamental networking issues:– Medium access control (resolving collision efficiently)– Multi-hop routing (routing data to sink efficiently)– Reliable data transfer (improving network reliability)
• This is the first Ph.D. proposal in the domain of underwater sensor networks at UCONN
Related PublicationsMedium Access Control• Peng Xie and Jun-Hong Cui, Exploring Random Access and Handshaking
Techniques in Large-Scale Underwater Wireless Acoustic Sensor Networks , Proceedings of IEEE/MTS OCEANS'06, Boston, Massachusetts, USA, September 18-21, 2006
• Peng Xie and Jun-Hong Cui, An Energy-Efficient MAC Protocol for Underwater Sensor Networks, to-be-submitted
Multi-hop Routing• Peng Xie and Jun-Hong Cui, SDRT: A Reliable Data Transport Protocol for
• Zheng Guo, Peng Xie, Jun-Hong Cui, and Bing Wang, On Applying Network Coding to Underwater Sensor Networks , Proceedings of ACM WUWNet'06 in conjunction with ACM MobiCom'06, Los Angeles, California, USA, September 25, 2006
Reliable Data Transfer• Peng Xie, Jun-Hong Cui, and Li Lao, VBF: Vector-Based Forwarding Protocol for
Underwater Sensor Networks , In Proceedings of IFIP Networking'06, Coimbra, Portugal, May 15 - 19, 2006
• Peng Xie, Jun-Hong Cui, and Li Lao, VBF: Vector-Based Forwarding Protocol for Underwater Sensor Networks , UCONN CSE Technical Report: UbiNet-TR05-03 , February 2005
Outline
• Introduction– Motivation & challenges
• Three fundamental networking problems– Medium access control– Multi-hop routing– Reliable data transfer
• Conclusions and future work
Medium Access Control
• General objectives: – Resolve collisions efficiently and effectively
• Evaluation metrics:– Channel utilization– Energy efficiency– Fairness– Delay– …(More depending on applications)
Challenges in M-UWSNs
• UW-A channel characteristics:– Long propagation delay
• Signal cannot reach dest. instantaneously
– Narrow communication bandwidth• Low data rate• Bandwidth must be shared by all nodes
• Passive sensor node mobility– Dynamic neighborhood makes coordination
very difficult if not impossible
Examine MAC Techniques
• Contention-free approach– TDMA, FDMA, CDMA
• Contention-based approach– Random access: ALOHA, slotted ALOHA– Collision avoidance with handshaking (RTS/CTS):
MACA, MACA-W
• We conducted a systematic study of random access and handshaking [Xie06:Oceans]
– Random access: sparse networks & low data traffic– RTS/CTS: dense networks & high data traffic
Existing MAC Protocols for Underwater Sensor Networks
• [Rodoplu05:Oceans]:– Network with ultra-low data traffic– Energy efficiency– Random access
• We propose R-MAC– A reservation-based MAC protocol
• Targeted networks– Traffic unevenly distributed & sporadic – Energy-efficiency is the highest priority – Channel utilization is not a critical concern
Basic Idea of R-MAC
• Each node works in cycles – Each node wakes/sleeps periodically
• A node sends data to another node– Sender reserves a time slot in receiver– Receiver informs all neighbors of reserved time slot – Sender sends data in reserved time slot
• How to make reservation? – Measuring propagation delays – Scheduling transmissions
The R-MAC Protocol
• Three phases– Latency detection
• Measure latencies between neighbors
– Period announcement• Collect period start times of neighbors
– Periodic operation• Reserve slot in intended node and send data
Phase I: Latency Detection
• Latency between A and B is: L= (T1-T2)/2
Node A
Node B
T1
T2L L
Phase II: Period Announcement
• Each node randomly selects period start time• Node B calculates difference of period start time
of node A with its own start time
LB-LA+LAB
LA
LAB
LB
LB-LA+LAB
A
B
Phase III: Periodic Operation (1)
• Each node powers on (listen window) and off (sleep window) periodically
• Data transmission is completed through REV/ACK-REV/DATA/ACK-DATA
• ACK-REV is treated with the highest priority – The first part of the listen window is reserved for
ACK-REV exclusively, called R-window– REV, DATA, ACK-DATA are scheduled to avoid the
R-windows of all nodes in the neighborhood
Phase III: Periodic Operation (2)
• The sender:– deliver REV to the target node in its listen window– specify the offset and duration of the reserved time
slot for data transmission in REV
• The receiver:– deliver ACK-REV to the sender in its R-window– reserve a timeslot for data transmission – deliver ACK-DATA after receiving data packets
• Other nodes: – Back off if receiving the ACK-REVs or sensing
collision in their R-windows
Sender in R-MAC
• Sender A schedules the transmission of REV to receiver B• Sender A specifies offset and duration of reserved time slot
Reserved time slot
STA
B
C
REV
REV
Receiver in R-MAC
• Receiver B schedules to send ACK-REVs to all neighbors• Sender A schedules the reserved time slot and Node C
keeps silence in this time period
time slot
A
B
C
Silence
ACK-REV
ACK-REV
Performance Evaluation
• Simulation settings:– Power consumption (UWM1000)
• Tx:2 Watts, Rx:0.75 Watts, idle:8 mW– Data rate
• 10kbps– Transmission range
• 90 m
• Performance metrics:– Goodput:
• Number of packets successfully received by receiver– Overhead:
• Energy consumption per data packet
Topology for Fairness
Node 3
Node 1
Node 4
Node 2
Node 0
80 m
20 m
60 m
20 m
Fairness
• All the nodes have almost equal goodputs
0 0.1 0.2 0.3 0.4 0.50
100
200
300
400
500
600
700
data rate (pkts/sec)
Go
od
pu
t
Node 1Node 2Node 3Node 4
Topology for Energy Efficiency
Node 3
Node 1
Node 4
Node 2
Node 0
30 m
20 m
40 m
20 m
Energy Efficiency
• R-MAC is more energy efficient than T-MAC
0 0.05 0.1 0.15 0.2 0.25
0.2
0.25
0.3
0.35
0.4
0.45
0.5
data rate (pkts/sec)
ove
rhe
ad
(Jo
ule
/pkt) R-MAC
T-MAC
Summary
• R-MAC – is energy-efficient– can achieve fairness– guarantees data packets collision-free (formal
proof)
Future Work
• Improve robustness of R-MAC against noisy channels
• Design efficient MAC solutions for mobile networks
Outline
• Introduction– Motivation & challenges
• Three fundamental networking problems– Medium access control– Multi-hop routing– Reliable data transfer
• Conclusions and future work
Challenges in M-UWSNs
• Hardest network environments for routing– Dynamic network topology– Large network scale– 3-dimensional space– High error probability– Energy constraint– Routing “voids”
Existing Routing Protocols for Terrestrial Sensor Networks
• Protocols for terrestrial sensor networks:– Directed Diffusion (DD)– GRADient Broadcast (GRAB)– Two-Tier Data Dissemination (TTDD)
• They are unsuitable for M-UWSNs– Dynamic network topology – 3-dimensional deployment
Our Solution
• We propose Vector-Based Forwarding (VBF) – A scalable, efficient and robust geo-routing approach
• The basic idea of VBF– Forwarding path represented by a vector– Node receiving packets
• Calculate its relative position• Forward packets if close to the vector
– Qualified nodes are in “routing pipe”• Controlled by pipe radius: W
VBF – An Illustration
VBF Enhancement
• Observations in dense networks– Too many nodes involved in data forwarding
• Solution: self-adaptation– Each node weighs the gain to forward a packet – Forwards packets adaptively
• Benefits of self-adaptation– Reduce energy consumption– Reduce packet collision– Can find optimal path (formal proof)
Self-Adaptation Algorithm
Pd
D Ad
F
Source(s1)
Sink(s0)
WW
R
A
W
p dd R B
Performance Evaluation
• Simulation settings:– 100×100×100 m3 cube– Transmission range: 20m– Source and sink are fixed– Other nodes are mobile
• Performance metrics:– Success rate (measure robustness)– Communication time (measure energy cost)
Impact of Density and Mobility
• VBF handles node mobility efficiently and effectively, and node density affects success rate and energy consumption
significantly
012345
500700
9001100
130015000
0.2
0.4
0.6
0.8
1
Number of nodesSpeed of nodes
Succ
ess
rate
(%)
01
23
45
500700
9001100
130015000
200
400
600
Number of nodesSpeed of nodes
com
mun
icat
ion
time
(sec
ond)
Impact of Pipe Radius
• When the pipe radius is large enough, VBF has the same success rate as naive flooding but with much less energy consumption
0 10 20 30 40 500
0.2
0.4
0.6
0.8
1
Radius (meter)
Succ
ess
rate
(%)
VBF
Naive Flooding
0 10 20 30 40 500
500
1000
1500
2000
Radius (meter)
Com
mun
icat
ion
time
(sec
ond)
VBFNaive Flooding
Robustness
• VBF is robust against packet losses and node failures
0 0.1 0.2 0.3 0.4 0.50
0.2
0.4
0.6
0.8
1
Error probability
Su
cce
ss r
ate
(%)
Robustness-packetlossRobustness-nodefailure
Summary
• VBF is– Energy efficient– Scalable– Robust (formal analysis)
Future Work
• Improve VBF– Adapt to non-uniformly distributed networks– Propose solutions to avoid routing “voids”
Outline
• Introduction– Motivation & challenges
• Three fundamental networking problems– Medium access control– Multi-hop routing– Reliable data transfer
• Conclusions and future work
Challenges in M-UWSNs
• Hardest network environments for RT– Highly error-prone communication channel– Long end-end propagation delay– Half-duplex acoustic channel– Dynamic network topology– Energy constraint
Examining Common Wisdoms• End-to-end approach
– not work well due to large RTT & high error probability
• Half-duplex channels limit complex ARQ– can only use Stop & Wait protocols – enhanced version to improve channel utilization
• S & W protocols with many feedbacks– have low energy efficiency
• Pure FEC approach– usually not energy efficient
Our Solution
• We propose segmented data reliable transport (SDRT) – A hybrid approach of FEC and ARQ
• The basic idea of SDRT– Data are first grouped into blocks at source– Each block encoded in simple & efficient codes– Source keeps pumping encoded data into
network till receiving a positive feedback in half-duplex channels
• The ratio of # of orig. data packet to the total time
– Inefficiency• The ratio of # of total packets sent to # of orig. data packets
Goodput
• SDRT improves the goodput significantly
0.1 0.2 0.3 0.4 0.50.01
0.1
1
10
Error probability
Go
od
pu
t (k
bp
s)
SDRTNaive ARQAccumulative-ARQ
Inefficiency
• SDRT reduces the number of packets sent
0.1 0.2 0.3 0.4 0.50
5
10
15
20
Error probability
Ine
ffic
ien
cy
Carousel-r3SDRT-r3Naive ARQ-r3Accumulative-ARQ-r3
SDRT Enhancement
• Observation: – Distance between sender and receiver: 30m RTT
(single hop) is 40ms time for trans. more than 60 packets (if packet size is 40bytes, data rate=500kbps)
– Too much overhead before receiving ACK
• Window size control– estimate # of packets for data reconstruction– send packets within window faster– send packets outside window slower– Thus save energy
• Critical to estimate the window size!
Simple Variant of Tornado Code
• Two-layer encoding scheme• Left degree is at least 3• Smaller maximum degree
dcb
dcba dca
a
b
c
d
dba
Model Validation (using SVT codes)
• Our model approximates the simulation results very well
0 0.1 0.2 0.3 0.4 0.5 0.61
1.5
2
2.5
3
3.5
4
4.5
Error probability
Ine
ffic
ien
cy
SimulationModel
Summary
• SDRT:– Improves channel utilization & energy efficiency– Relieves sender & receiver of manage burden – Well addresses dynamic network topology
Future Work
• Examine network coding for robustness
• Investigate congestion control
Outline
• Introduction– Motivation & challenges
• Three fundamental networking problems– Medium access control– Multi-hop Routing– Reliable data transfer
• Conclusions and future work
Medium Access Control• Current Status:
– Modeled and compared random access and handshaking (RTS/CTS) techniques
– Proposed an energy efficient protocol (R-MAC) for static networks
– Developed a simulation package for physical acoustic link and MAC in ns-2
• Future Work– Improve robustness of R-MAC in noisy channels – Design MAC solutions for mobile networks
Multi-hop Routing
• Current Status:
– Proposed a robust and energy-efficient routing protocol (VBF)
– Developed a self-adaptation algorithm to enable VBF to be adaptive to network density
– Implemented VBF in ns-2
• Future Work:– Enable VBF to handle non-uniform networks– Propose solutions to avoid routing voids
Reliable Data Transfer
• Current Status:– Proposed an efficient reliable protocol (SDRT)– Developed a model to estimate # of packets
needed
• Future Work:– Implement SDRT in ns-2– Examine network coding for robustness– Investigate congestion control and avoidance
Simulation Toolkit
• Current Status– Implemented acoustic physical link– Implemented R-MAC and VBF – Implemented MAC broadcast
• Future Work– Develop a complete package for all layers– Validate acoustic model with measurements
• Goal: release UWSN simulation package to the research community