Virtual Localization for Mesh Network Routing Nick Moore / Ahmet S - PowerPoint PPT Presentation

Virtual Localization for Mesh Network Routing – 1/18 Virtual Localization for Mesh Network Routing Nick Moore / Ahmet S ¸ekercio˘ glu / Gregory K Egan Center for Telecommunications and Information Engineering, Monash University, Melbourne, Australia. Nick Moore / Ahmet S ¸ekercio˘ glu / Gregory K Egan

Virtual Localization for Mesh Network Routing – 2/18 Sensor Networks • Miniature sensors allow field measurements • Data must still be collected • Sensor networks allow sensors to communicate back to a central point Nick Moore / Ahmet S ¸ekercio˘ glu / Gregory K Egan

Virtual Localization for Mesh Network Routing – 3/18 Mesh Sensor Networks • All nodes are equal. • All routing computation is distributed. • Battery power is limited, and processing power and network usage are therefore expensive. Nick Moore / Ahmet S ¸ekercio˘ glu / Gregory K Egan

Virtual Localization for Mesh Network Routing – 4/18 Routing in a Mesh How can we route packets across the mesh? • hierarchical partitioning too inflexible • packet flooding too inefficient • route flooding • location based routing Nick Moore / Ahmet S ¸ekercio˘ glu / Gregory K Egan

Virtual Localization for Mesh Network Routing – 5/18 Greedy Forwarding Simplest algorithm for location based routing: forward packet to whichever neighbour is nearest the destination. A B D C G E F H M J K L • − HEFMDG is longer than − − − − − − − − → − − − − − → HEFCG , but F forwards to M as M is closer to G than C is. • − − − → JKL is blocked by a ‘void’. Nick Moore / Ahmet S ¸ekercio˘ glu / Gregory K Egan

Virtual Localization for Mesh Network Routing – 6/18 Determining Location • Na¨ ıve solution: GPS • ‘Anchor’ nodes (up to 20%) • Radio distance-finding Nick Moore / Ahmet S ¸ekercio˘ glu / Gregory K Egan

Virtual Localization for Mesh Network Routing – 7/18 Virtual Location • Location relative to other nodes • Axes do not correspond to real directions • Geometries may not correspond either • Internally consistent • Generally only useful for routing purposes Nick Moore / Ahmet S ¸ekercio˘ glu / Gregory K Egan

Virtual Localization for Mesh Network Routing – 8/18 n -neighbours A 2−neighbour 1 − n e i g h 1−neighbour b o u r B C 1−neighbour 1−neighbour D Nick Moore / Ahmet S ¸ekercio˘ glu / Gregory K Egan

Virtual Localization for Mesh Network Routing – 9/18 Spring Models Repulsion A C A t t r Attraction a c t i o n B Nick Moore / Ahmet S ¸ekercio˘ glu / Gregory K Egan

Virtual Localization for Mesh Network Routing – 10/18 Forces and Potentials - Equations • Springlike attraction F ∝ d to 1-neighbours U ij = k att · d 2 ; k att = 1 ij • Electrostatic-like repulsion F ∝ 1 /d 2 from 2-neighbours 1 k rep = 8 × 10 6 U ik = k rep · ; d ik + 1 • Node attempts to minimize total potential energy � U i = U ij j ∈ N Nick Moore / Ahmet S ¸ekercio˘ glu / Gregory K Egan

Virtual Localization for Mesh Network Routing – 11/18 Forces and Potentials - 1D U_C = U_CA + U_CB U_CB (attraction) U_CA (repulsion) A B min Nick Moore / Ahmet S ¸ekercio˘ glu / Gregory K Egan

Virtual Localization for Mesh Network Routing – 12/18 Forces and Potentials - 2D 100 0 0 100 −100 200 Nick Moore / Ahmet S ¸ekercio˘ glu / Gregory K Egan

K A I C H F X B G D J E

A C X B D

Virtual Localization for Mesh Network Routing – 15/18 200-node Mesh • 200 nodes • Each node is placed so that: ⋄ at least one existing node is in range ⋄ no nodes are within range/2 • similar to a rooftop network http://www.ctie.monash.edu.au/mesh/virt_loc/two.gif ./two.animated.gif Nick Moore / Ahmet S ¸ekercio˘ glu / Gregory K Egan

200-node Mesh Comparison: Actual Network Map 006000 : 200 / 200 / 200

Virtual Localization for Mesh Network Routing – 17/18 400-node Mesh • 400 nodes • Each node placed at random within a 1km x 1km grid • Node range 100m http://www.ctie.monash.edu.au/mesh/virt_loc/one.gif ./one.animated.gif Nick Moore / Ahmet S ¸ekercio˘ glu / Gregory K Egan

Virtual Localization for Mesh Network Routing – 18/18 Further Work • More sophisticated routing algorithms • 3D,4D virtual spaces (in submission to IEEE TPDS) • Node mobility / energy conservation • Multiple root nodes / anchors Questions? Nick Moore / Ahmet S ¸ekercio˘ glu / Gregory K Egan