Topological Hole Detection in Wireless Sensor Networks and its Applications
Stefan FunkeDepartment of Computer Science, Stanford University, U.S.A.
DIAL-M-POMC 2005DIAL-M-POMC 2005
SpeakerSpeaker:: Shih-Yun HsuShih-Yun Hsu
DIAL-M-POMCDIAL-M-POMC
Discrete Algorithms and Methods for Mobile Computing and CommunicationsWorkshop in conjunction with ACM/SIGMOBILE
MobiCom (1997~ 2004)
Principles of Mobile ComputingWorkshop in conjunction with
ACM/SIGACT and SIGOPS PODC (2001)ACM/DISC (2002)
OutlineOutlineIntroduction
Related works
Main methodsTopology hole findingCoarse Boundary Sampling and Pruning
ApplicationsExperiment evaluationConclusions
IntroductionIntroductionDue to cost restrictions and to achieve the
maximum life-time by energy savingsThe characteristics of sensors
Low-capability devicesTemperatureHumidity
Small radio device that allows for communication between nearby sensor nodes
Easy to be deployed by airplanes
IntroductionIntroductionTo achieve the maximum life-time
It is impossible to equip energy-hungry GPS unitNone of the sensor nodes is aware of its geographic loca
tion
IntroductionIntroductionThere are many holes in the monitoring region
Fall right into the flames and be destroyedPlunge into a lake or pond and be unable to
perform their monitoring taskFall from airplane on the grand then break
Detecting the boundaries of such holes in the monitored space created by fire or other phenomena
Related worksRelated worksGLIDER: Gradient Landmark-Based
Distributed Routing for Sensor NetworksGeographic Routing without Location
InformationMAP: Medial Axis Based Geometric Routing
in Sensor Networks
Main methodsMain methodsTopology hole findingCoarse Boundary Sampling and Pruning
Topology hole findingTopology hole findingBasic concept
Beacon
Euclidean lengthhole
Unit Disk Graph (UDG)
Topology hole findingTopology hole finding Monitoring (connected) region Beacon Any points dp(x) denotes the minimum Euclidean
length from p to x The isolevel (contour of level) of k
The sub-graph of UDG induced by I(k) might be disconnected
2R Rp R
x R
( ) { ( ) }pI k x R d x k
1 2( ) { ( ), ( ), ...}I k C k C k
p
xdp(x)
I(k)
C1(k)
C2(k)
Topology hole findingTopology hole finding Pick a local beacon q Compute hop-distances h(v’) to q Mark all nodes v which do not ha
ve a 2-hop neighbor v’ with h(v’) > h(v)
C1(k)
q
v
Topology hole findingTopology hole findingbeacon
Border nodes
Topology hole findingTopology hole finding
Topology hole findingTopology hole finding
Topology hole findingTopology hole finding The first beacon was chosen rand
omly Maintain a variable CBD(v) (Clos
est Beacon Distance) storing the (hop-)distance and choose the last 3 beacons as far as possible1
2
3
4
Coarse Boundary Sampling and Coarse Boundary Sampling and PruningPruningA natural way to reduce this number is to com
pute a maximal independent set (MIS) within all the marked nodesMaximal independent sets in radio networks
Thomas Moscibroda, Roger WattenhoferDepartment of Computer Engineering and Networks Laborato
ry, ETH Zurich, Switzerland
ACM Symp. on PODC 2005
Coarse Boundary Sampling and Coarse Boundary Sampling and PruningPruning
Coarse Boundary Sampling and Coarse Boundary Sampling and PruningPruning
Density
ApplicationsApplicationsGLIDER: Gradient Landmark-Based Distri
buted Routing for Sensor NetworksQing Fang, Jie Gao, Leonidas J. Guibas, Vin de Sil
va, Li ZhangDepartment of Electrical Engineering, Computer Scienc
e, Mathematics, Stanford UniversityInformation Dynamics Lab, HP Labs
INFOCOM 2005
ApplicationsApplications -GLIDER--GLIDER-
S
D
ApplicationsApplications -GLIDER--GLIDER-
Paths that share the same subsequence of tiles are kept apartLoad-balance
ApplicationsApplications -GLIDER--GLIDER-
GLIDER for random landmark selection
GLIDER for topology-aware landmark selection
ApplicationsApplications -GLIDER--GLIDER-
In inter-tile, the GLIDER protocol is also load-balance
ApplicationsApplications -GLIDER--GLIDER-
In intra-tile, the GLIDER protocol could not be load-balance
Near Far
ApplicationsApplications -GLIDER--GLIDER-
Load imbalance due to Landmarks being too close to boundaries
ApplicationsApplications -GLIDER--GLIDER-
ApplicationsApplications -GLIDER--GLIDER-
Landmarks sends a HELLO message with distance counter 0 which increases at every hop
The value △(v) is then the minimum counter value over all messages received
dlocal(p)=min(d(p, qi)) New position of landmark p’=dloca
l(p)/3 p still in the tile of p’ Any tile will not contain a whole
hole If d(p’, q’)<dlocal(p) (p and q are c
loser) Removed q’
p
q1
q2
q3
q4
P’
ApplicationsApplications -GLIDER--GLIDER-
ApplicationsApplicationsGeographic Routing without Location Infor
mationAnanth Rao, Sylvia Ratnasamy, Christos Papadimi
triou, Scott Shenker and Ion StoicaUniversity of California, Berkeley
INFOCOM 2003
ApplicationsApplications - - Geographic Routing Geographic Routing --
ApplicationsApplications - - Geographic Routing Geographic Routing --
ApplicationsApplications - - Geographic Routing Geographic Routing --
Holes might obstruct the shortest paths between nodes of the network and hence their lengths are not a good estimate of the true geometric distance
ApplicationsApplications - - Geographic Routing Geographic Routing --
Truthful distances
Not truthfuldistances
ApplicationsApplications - - Geographic Routing Geographic Routing --
P is the set of boundary nodesThe distance measured between a pair
is truthful, if the respective shortest path in the communication graph from p to q providing this estimate does not contain any as intermediate node
( , )p q P P
r P
ApplicationsApplications - - Geographic Routing Geographic Routing --
ApplicationsApplicationsMAP: Medial Axis Based Geometric Routin
g in Sensor NetworksJehoshua Bruck, Jie Gao, Anxiao (Andrew) Jiang
California Institute of Technology, USCaltech, US
MobiCom 2005
ApplicationsApplications -MAP--MAP-
ApplicationsApplications -MAP--MAP-
ApplicationsApplications -MAP--MAP-
ApplicationsApplications -MAP--MAP-
ApplicationsApplications -MAP--MAP-
ApplicationsApplications -MAP--MAP-
ApplicationsApplications -MAP--MAP-
Near Far to the border
Experiment evaluationExperiment evaluation4900 nodes800×800 square regionCommunication range is 15(average degree 5),
20(10), 27(18), 40(39)The degree is rcommunication/rsense
Unit disk graphs (UDG)Random Uniform DistributionsRandomly perturbed Grid
Non-UDG
UDG with Random Uniform UDG with Random Uniform DistributionsDistributions
15(5) 20(10)
27(18) 40(39)
Communication Range (Ave. degree)
UDG with Randomly perturbed UDG with Randomly perturbed GridGrid
15(5) 20(10)
27(18) 40(39)
Communication Range (Ave. degree)
Non-UDGNon-UDG
With UDG With Non-UDG
Non-UDGNon-UDG
Degree 8 Degree 16 Degree 20
ConclusionsConclusionsThis paper we have presented a rather simple
and straightforward algorithm for detecting holes in a wireless communication networkLocation-unawareHigher density is better
This paper also sketched further applications of hole finding routine, where the knowledge about holes in the network provides for better performance of existing topology-based, location-free protocols
Thank You!!Thank You!!