Bounding the Interconnection Gap for Vehicular Internet Access Presented by Prasun Sinha Authors Zizhan Zheng Prasun Sinha and Santosh Kumar ID: 573515
Download Presentation The PPT/PDF document "Alpha Coverage:" 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
Alpha Coverage: Bounding the Interconnection Gap for Vehicular Internet Access
Presented by:
Prasun
Sinha
Authors:
Zizhan
Zheng
†
,
Prasun
Sinha
†
and
Santosh
Kumar
*
†
The Ohio State University,
*
University
of MemphisSlide2
Internet Access for Mobile VehiclesApplications
Infotainment
Cargo tracking
Burglar tracking
Road surface monitoring
Current Approaches
Full Coverage
Wireless
Wide-Area Networking (WWAN)
Fully Covered
WiFi
Mesh
Opportunistic Service
Roadside
WiFi
Slide3
Current Approach I (of II): Full CoverageWireless Wide-Area Networking
3G Cellular Network
3GPP LTE (Long Term Evolution)
WiMAX
Either long range coverage (30 miles) or high data rates (75 Mbps per 20 MHz channel)
3 Mbps downlink bandwidth reported in one of the first deployments in US Google WiFi for Mountain View 12 square miles, 400+ APs1 Mbps upload and download rate Not very practical for large scale deployment due to the prohibitive cost of deployment and management
3
Google
Wifi
Coverage Map
http://wifi.google.com/city/mv/apmap.htmlSlide4
Current Approach II (of II): Opportunistic Service via In-Situ APs
Prototype
Drive-Thru Internet (Infocom’04,05)
In-Situ Evaluation
DieselNet
(Sigcomm’08, Mobicom’08)Interactive WiFi connectivity (Sigcomm’08)Cost-performance trade-offs of three infrastructure enhancement alternatives (Mobicom’08) MobiSteer (Mobisys’07)Handoff optimization for a single mobile user in the context of directional antenna and beam steering Cabernet (Mobicom’08)Fast connection setup (QuickWiFi) and end-to-end throughput improvement (CTP)ProblemsOpportunistic service, no guaranteeUnpredictable interconnection gap
4
Our solution: an intermittent coverage model that provides predictable data service to mobile users at low cost
Internet
AP
AP
APSlide5
Roadmap Alpha Coverage – An Intermittent Coverage ModelA general definition – intuitive but intractable Two simplifications
Alpha Network Coverage (
N
-Coverage)Applies when route information is unknown Ex: Burglar trackingAllows a factor log (n) approximationAlpha Path Coverage (P -Coverage) Applies when route information is given Ex: bus trace in DieselNet, cached model in MobisteerAllows a more efficient factor log (n) approximationEvaluationFuture WorkSlide6
Road Network Model and Problem Statement Model
Model a road network
R
as an undirected graph
G
R with edge length at most (by inserting artificial intersections if needed). Model a movement as a path on GR (not necessarily ending at intersections). Model access points as points on GR (modeling the worst case of communication range). Given GR and A
0 µ
V [GR] that models a set of APs previously deployed
Determine if the deployment provides the desired coverage (to be defined), and if not
Find a minimum set of
points
A
in
GR so that when new APs are deployed at these locations, A0 [ A provides the desired coverage.6
v5
v
1
v
2
v
3
v
4
v
6
v
7
v
8
s
t
v
9
v
4
v
5
v
1
v
2
v
3
v
6
v
7
v
8
v
9Slide7
Alpha Coverage: an Intermittent Coverage Model
A
deployment provides
-Coverage
to a road network R if any path of length on GR touches at least one point representing an access-point. FeaturesProvides a guarantee on the worst case inter-contact gapProvides an estimation of the cumulative data serviceChallengesEven verifying -Coverage is NP-complete since there is a reduction from HAMILTONIAN PATH to itSimplified models are needed
7Slide8
Alpha Coverage w/o Route InformationA deployment provides Network Coverage of distance (
N
-
Coverage for short) if any path f(a,b)
with dist(a,
b) (graph distance) at least is covered by at least one AP
–
Coverage
implies
N –Coverage, but not vice versa 8
s
t
v
9
v
4
v
5
v
1
v
2
v
3
v
6
v
7
v
8
= 5
s
t
v
9
v
4
v
5
v
1
v
2
v
3
v
6
v
7
v
8
-Coverage
N
-
CoverageSlide9
Alpha Coverage w/o Route Information (Cont.)Polynomial time verifiable
The optimization problem (
N
-Cover) is NP-hard
Reduction from VERTEX COVER restricted to triangle-free, 3-connected, cubic planar graphsO(log |V|) approximationAssumption: New APs are deployed only at the vertices of GR (real or artificial road intersections) Introducing a factor of 2Reduce
N -Cover to node version low diameter graph decomposition
GVY algorithmHigh computation time complexity for large networks
9
v
1
v
2
v
3
v
4
v
5
v
6
v
7
v
8
v
9
v
1
v
3
v
4
v
6
v
8
= 2Slide10
Alpha Coverage with Route InformationMotivation: use route information to design a more efficient algorithm
Assumption: a set of paths
F
is given where |
F
| = O(p(|V|))Ex 1) a set of shortest paths obtained from a road network databaseEx 2) a set of most frequently traveled paths learned from historical traffic data
Decompose each given path into -paths
A deployment provides Path Coverage of distance (
P
-Coverage
for short) if any -path in
F
is covered by at least one AP.Polynomial time verifiable, the optimization problem is still NP-hardO(log |V|) approximation: reduce P -Cover to Minimum Set Cover
10Slide11
Simulation Setting Road network
A
4
km x
4
km region around the center of Franklin County, OHAbout 1000 intersections, 1300 road segments Obtained from 2007 Tiger/Line Shapefiles + Mercator projection
Moving scenariosRestricted random way point: each movement follows a shortest path and has length at least
5
mobile nodes, moving
1
hour each, 10 scenarios
Various speed limits
Ns-2 simulation
The transmission range of each AP is
100m11Slide12
Deployment methodsP –Coverage
Rand-1: a set of randomly selected vertices of
G
R
Rand-2: a set of points on randomly selected edges of
GRRand-3: the region is divided into 50m x 50m cells; APs are deployed at the centers of a set of randomly selected cells.12
An instance of
P -Cover, = 3000
m
Simulation Setting (Cont.)Slide13
Simulation Results
13
21 APs are used
The maximum gap for
P -Coverage
is about 214 sec, bounded by the time spent on two adjacent moves
The maximum gap for a random deployment can be larger than 2000 sec
Inter-contact gap (sec
)
= 3000m
(m)
CDF
Standard deviation (sec)Slide14
Future WorkImprove the efficiency of
N
-
Coverage
Combinatorial algorithms for fractional vertex
multicut Connected -CoverageConnect each AP to at least one of the gateways with Internet backhaulJoint Coverage and connectivity optimizationA bound on the number of hops to gateways(,)-Coverage: Enabling Assured Data Service Guarantees that each user moving through a path of length has access to at least units of data.
Challenges: variable data rates, traffic density, and contact durations; unknown association schedules
14Slide15
Alpha Coverage w/o Route Information (Cont.)Polynomial time verifiable
The optimization problem, called
N
-Cover, is NP-hard
There is a reduction from VERTEX COVER restricted to triangle-free,
3-connected, cubic planar graphsO(log |V|) approximation: reduce N-Cover to Minimum Vertex Multicut Assumption: New APs are deployed only at the vertices of
GR
(real or artificial road intersections) => introducing a factor 2Step1: Find the set of -pairs, treat their midpoints as terminals
Step2: Solving the fractional vertex
multicut
problem -- the dual of node version maximum
multicommodity
flow problem
Step 3: Rounding the solution by low diameter graph decomposition (GVY).
15