GLIDER: Gradient Landmark-Based Distributed Routing for Sensor Networks
description
Transcript of GLIDER: Gradient Landmark-Based Distributed Routing for Sensor Networks
GLIDER: Gradient Landmark-Based Distributed Routing for Sensor Networks
Qing Fang, Jie Gao, Leonidas J. Guibas, Vin de Silva, Li ZhangDepartment of Electrical Engineering, Computer Science, Mathematics, Stanford University
Information Dynamics Lab, HP Labs
INFOCOM 2005
SpeakerSpeaker:: Shih-Yun HsuShih-Yun Hsu
OutlineOutlineIntroductionGLIDER
Landmark selectionLocal landmark coordinateNamingRouting
SimulationConclusions
IntroductionIntroductionRouting algorithm are classified
ProactiveRouting table
Proactively maintained and take advantage of the hierarchical structure of IP addresses to enable route discovery
Hard to maintain routing table when topology changes frequently
ReactiveAODV or DSR
Flooding the network in order to discover the desired route
IntroductionIntroductionRouting algorithm in WSNs
Power conservation becomes a serious concern and flooding is undesirable
IntroductionIntroductionGeographical routing
Compute routes that are often close to the best possible
Do so with very little overhead in maintaining auxiliary routing structures
GPS receivers can be costly and lead to cumbersome node form factors
IntroductionIntroductionLocalization algorithms
Only some nodes, called anchor nodes, know their location by manually or GPS
Other nodes determine their location by estimating their distances to three or more of these anchors and then become anchors themselves
Localization algorithms are still quite expensive in terms of computation or communication, and often insufficiently accurate
IntroductionIntroductionGLIDER
Location-unawareTopology-enabled routing
The whole networks are connectivity
GLIDERGLIDERLandmark selectionLocal landmark coordinateNamingRouting
Landmark selectionLandmark selectionDesire to have several landmarks lying close
to topological features, such as hole boundariesArrange for nodes near the boundary to be selected
as landmarks with higher probability than interior nodes
Expect the number of landmarks to be proportional to the number of holes (or topological features) of the sensor domain
Landmark selectionLandmark selection
GLIDERGLIDERLandmark Voronoi Complex (LVC)
GLIDERGLIDERCombinatorial Delaunay Triangulation (CDT)
GLIDERGLIDEREach node in the same tile is knowing their
home landmark and the shortest path to the home landmark
GLIDERGLIDEREach landmark also flooding to neighbor tile
Each node is knowing the shortest path to neighboring landmark
Each landmark is knowing other landmark positionsEach landmark builds its shortest path tree by CDT
Local landmark coordinateLocal landmark coordinateEach node calculate the distances or hops
between home landmark and neighboring landmark to itself
1
2
3
(1, 2, 3)
NamingNamingEach node has ID and Name
ID is uniqueName is not necessary unique
Name isHome landmark ID For local landmark coordinate
Ex: (Name, 1, 2, 3)
RoutingRoutingGlobal routingLocal routing
Local routingLocal routingLocal routing (intra-routing) is done by
gradient descent using the local landmark coordinatesGreedily routing
Global routingGlobal routing
SD
SimulationSimulationC++ programming
Network-level simulations using ns-2 will be undertaken in the near future
2000 sensor nodes23 landmarks
5 landmarks are near the hole boundary18 are chosen randomly
Using a Gaussian random variable with standard deviation equal to 50%
SimulationSimulation
20 pairs of sources and destinations selected at random, with path length about 40 hops in each case
The density of network is important
SimulationSimulation
GLIDER GPSR
SD
S
S
S
D
D
D
SimulationSimulation
S
D
S SS
SS
D
D D
DD
GLIDER GPSR
SimulationSimulation
GLIDER GPSR
45 pairs of randomly chosen source and destination
ConclusionsConclusionsThis paper propose a topology-based and
location-unaware routingThis paper combines LVC and CDT to
maintain routingIt will more power balance than GPSR
Thank You!!Thank You!!