HyungJune Lee Martin Wicke Branislav Kusy Omprakash Gnawali and Leonidas Guibas Stanford University ACMIEEE IPSN10 April 15 2010 Traditional Data Delivery to Mobile Sinks ID: 402916
Download Presentation The PPT/PDF document "Data Stashing: Energy-Efficient Informat..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Data Stashing: Energy-Efficient Information Delivery to Mobile Sinks through Trajectory Prediction
HyungJune Lee
,
Martin
Wicke
,
Branislav
Kusy
,
Omprakash
Gnawali
, and
Leonidas
Guibas
Stanford University
ACM/IEEE IPSN’10
April 15, 2010Slide2
Traditional Data Delivery to Mobile Sinks in Wireless Ad-Hoc/Sensor Networks
Immediate
delivery from data source to mobile sinks
Proactive scheme: DSDV, OLSRReactive scheme: DSR, AODVPerformance degradesrapidly with increasing mobility
Data MULEs to collect data as it passes each of the sensor nodesWait until mobile sinks come to collectOften infeasible if we cannot control the movement
2
?
What’s a compromise between two extremes?
How to exploit the tolerated delay?
How to use regularity of mobility pattern?
How to select only a partial set of effective relays?Slide3
3
/34
Overview: Predictive
Mobile Routing
1. Trajectory PredictionAnticipated trajectory nodes2. Data request and trajectory announcement
3. Stashing node selection
To cover
the likely paths and minimize the routing cost
4. Data stashing
5. Data collection by mobile nodes Slide4
Summary of Contributions
Predictive Model of Users’ Trajectories
In the space of
wireless connectivityCapture Long-term behavior (in minutes) a set of the future connected relays
Predictive Data Delivery Propose an energy-efficient data delivery scheme to mobile sinksTurn even limited knowledge of future connectivity into networking benefit
A
4Slide5
Outline
[Off-line Learning Phase]
Mobile Trajectory Model
In the space of wireless connectivityFor packet delivery purpose[Routing]Prediction of Future Relay ConnectivityPredictive Data Delivery to Mobile Users[Evaluation]
5Slide6
Capturing Mobile Trajectory Patterns
Background
Trajectory: a sequence of node associations on a given spatial path
Trajectories from the same spatial trajectory are not necessarily identicalDue to imperfect links and radio signal strength fluctuationsGoal
To cluster similar mobile trajectories General trajectory pattern models explored by a number of spatial trajectories
a
l
q
o
r
t
z
b
p
y
u
i
x
s
T
=
a l o r t z b p y u
T’
=
a l q o r z s p i u z
T’’
=
a q r t z t s b y i x
6Slide7
Constructing trajectory clusters
Step I. Similarity measure
Step II. Hierarchical clustering
Step III. Compact representation
7Slide8
Step I: Similarity Measure
Similarity
measure (normalized)
Not a distance metric
8Slide9
Step II. Hierarchical Clustering
Hierarchical clustering :
Every point is its own cluster
Find most similar pair of clusters
Merge it into a parent cluster
Calculate the average similarity between objects in two clusters
Repeat
9Slide10
Step III: Probabilistic Representation
Execute multiple sequence
alignment
(using ClustalW
tool)- Computation complexity
Construct Profile
: A probabilistic representation
for efficient search in the usage phase
R T E A C E G I P D S
R E C E I G I P S D SY E C I R E C E I C G I G N G N D S
E D E C I G P D SR E C H C I G K D SR E C I G C
R I E C G S G D L D K SK E C G I G T D W D SR E C N I G D G T D S
R E P E C N I G I D G D K D S
10
-
RT-EACE-GIP----D--S
-R--E-CEIGIPS---D--S
--Y-E-C---I---------
REC-EICG--IGNG-ND--S
-ED-E-C---IGP---D--S
-R--E-CH-CIGK---D--S
-R--E-C---IGC-------
-RI-E-CG--SG-D-LDK-S
--K-E-CG--IGTD-WD--S
-R--E-CN--IG-DGTD--S
-REPE-CN--IGID-GDKDSSlide11
Summary: Mobility Trajectory Clustersin an off-line phase
Trajectory sequences
………………
……………………….
………………….………………………….……………11Slide12
Outline
[Off-line Learning Phase]
Mobile Trajectory Model
[Routing]Prediction of Future Relay ConnectivityPredictive Data Delivery to Mobile Users[Evaluation]
12Slide13
Prediction of Future Relay Connectivity
Given a partial test sequence,
1) First find the closest
cluster A variant of Smith-Waterman algorithm for local matchingWith the largest
F(*,*) among all profiles2) Find the highly overlapped region
Test sequence
:
Profile
:
R C E C N C
13
Mobility Profile Database
J
. . .
?Slide14
Prediction of Future Relay Connectivity
3) Obtain the most probable subsequences starting
from
J+1 through J
+W
J
W
14Slide15
Optimal Route Selection Using Predictive Knowledge
Data stashing:
Given a set of future trajectories of
multiple mobile users,Find the optimal stashing nodes for each data sourceConsidering Cover all possible future trajectoriesMinimize routing cost to the selected relay nodes
M1
M
2
A
T
3
T
1
T
2
T
4
T
5
T
6
N
15Slide16
Optimal Route Selection Using Predictive Knowledge
Optimization
problem
For sensor node A, Minimize total routing cost From sensor node itself To the selected stashing nodesSubject toStashing nodes cover all possible
future paths of multiple mobile usersSolved by LP/IP solvers such as CPLEX, Gurobi, GLPK, …
M1
M
2
A
T
3
T
1
T
2
T
4
T
5
T
6
N
16Slide17
Outline
[Off-line Phase]
Mobile Trajectory Model
[Routing]Prediction of Future Relay ConnectivityPredictive Data Delivery to Mobile Users
[Evaluation]Dynamic mobility modelPrediction AccuracyRouting performance ScalabilityTolerated Delay Load BalanceComputation for Selecting Stashing Nodes
17Slide18
Validated trajectory clustering using UMass
DieselNet
real-world dataset
: 34 buses, 4198 APs, 789 bus trips around UMass campus
Prediction method results in excellent stashing node selections for real-world dataPrediction Accuracy of Mobile Trajectory Model
18Slide19
Simulation Setup for Routing
TOSSIM under ‘
meyer
-light’ interference830x790 m2
716 nodes20 mobile trajectoriesVehicle moves at a random speed N(30, 52
) km/hVehicle sends a beacon every 1 sec
Each sensor node has data to deliver to mobile sinks19Slide20
Scalability depending on # of mobile sinks
Data stashing consumes less energy than
immediate point-to-point routingScalable with # of mobile sinks!Data stashing keeps high packet delivery even for network congestion
Data stashing performs closely to the upper bound by perfect prediction Even limited knowledge of future trajectories can significantly improve routing performance! (lower is better)
(higher is better)
20Slide21
W
: # of future trajectory
hops
Large W means more chance to exploit data stashing scheme
As W 1, data stashing should breakImplication
Trade-off:
Tolerated delay vs.
Network performance
Tolerated Delay W
(lower is better)
(higher is better)
21Slide22
Data stashing has a good
load balancing
performance compared to a point-to-point routing
immediately to mobile sinks
Load Balance
better
22
Immediate Routing
Data Stashing Slide23
PC: Dell Precision 390
(2.4 GHz Core 2 Duo)
Small Embedded: fit-PC2
(Intel Atom Z530 1.6GHz)Measured running time for solving the optimization problem - binary integer programFeasible even in a small embedded platform, taking less than 500ms
(lower is better)
23
Running time for a
source to compute stashing nodes Slide24
Conclusion
Dynamic mobile trajectory model in the space of
wireless connectivity, capturing wireless volatility
Mobile data delivery can be improved through mobility pattern learning and predictionEven limited knowledge of the future trajectory can improve networking performanceTake-home lesson:“If you know where someone is going (even uncertainly), you can deliver data to him more
efficiently and reliably.”24Slide25
Two problems
Current delivery scheme is “best-effort”
Current clustering method cannot share common pieces of
trajectories
More robust packet delivery:
When the system detects delivery would fail,
restashing
can significantly improve
robustness
Trajectory prediction and data stashing can be more intertwined
Multi-tier clustering:
Long trajectories can be
partitioned into short pieces
for efficient
clustering
On-line clustering
A
multi-tier
clustering approach can deal with extremely large complex networks
25
Limitations & Future
WorksSlide26
Questions?
HyungJune Lee
abbado@stanford.edu
26