1. Routing Considerations for Sensor Networks Lecture 12 October 12, 2004 EENG 460a / CPSC 436 / ENAS 960 Networked Embedded Systems & Sensor Networks Andreas Savvides andreas.savvides@yale.edu Office: AKW 212 Tel 432-1275 Course Website http://www.eng.yale.edu/enalab/courses/eeng460a

2. Announcements  Feng Zhao’s talk tomorrow 4:00pm @ AKW 500  Student session 3:20 – 4:00pm AKW 500  Reading for this lecture Zhao & Guibas Section 3.3 through 3.6  Reading for next lecture Directed Diffusion – paper posted on the class website  Today’s presentation IDSQ

3. Routing Considerations in Sensor Networks  Traditional TCP/IP routing not attractive for sensor networks Too much overhead and large routing tables  Sensor networks are more ad-hoc Each node acts as a router Still different than ad-hoc networks oProactive routing is too expensive oSome possibility for reactive routing such as –Fish-eye routing, AODV, DSR

4. Routing Goal  Focus on localized state-less routing Consider only local neighborhood  Classical separation of address and content does not hold Care about reaching the nodes rather than a particular address – what can be sensed by a node can most probably be sensed by neighboring nodes Interested in routing by attributes – data centric oNode’s location oNode’s type of sensors oRange of values in the sensed data  Notion of optimality can vary QoS routing – latency is important => shortest path Energy aware routing – longer paths are ok => avoid nodes with less energy

5. Geographic Routing  Aims to route based on very limited state information  Geographic routing protocols assume All nodes know their geographic location Each node knows its 1-hop neighbors Destination is a node with a given location Each packet can hold a limited amount of information as to where it has been in the network  Any issues with this? Needs to maintain information between node IDs and node location (referred to as location service)

6. Geographic Forwarding Approaches  Greedy distance routing: select the neighbor geographically closest to the destination and forward the data to that neighbor  Compass routing: pick the next node as the one that minimizes the angle to destination  What are the problems with the basic approaches Greedy distance routing – may get stuck in local minima Compass routing – may go in loops

7. Planarization of Routing Graph  To get protocols that guarantee data delivery, make graph planar  Remove some edges from your network graph G Aim: Keep the same connectivity but make the graph planar o no two edges in G should intersect each other In the planar subdivision of G each node is assumed to know the circular order of its neighbors Convex perimeter routing and other face routing protocols use this property

8. Common Planarization Methods  Relative Neighborhood Graph (RNG) The edge xy is introduced if the intersection of circles centered at x and y with radius the distance d(x,y) is free of other nodes  Grabriel Graph The edge xy is introduced if the diameter xy is free of other nodes  Both graphs RNG and Gabriel graphs can be found with distributed construction x y x y

9. Greedy Perimeter Stateless Routing(GPSR)  Geographic protocol based on the offline construction of planar graphs RDG, Gabriel, later on RDG suggested  Has 2 main phases forwarding and recovery  Forwarding is greedy  Recovery – uses a right-hand rule to recover from holes. It stops as soon as a node closer to the destination is found

10. Routing on a Curve  Specify a curve a packet should follow  Analytical description of a curve carried by the packet  Curves may correspond to natural features of the environment where the network is deployed  Can be implemented in a local greedy fashion that requires no global knowledge  Curve specified in parametric form C(t)=(x(t),y(t)) t – time parameter – could be just relative time  Each node makes use of nodes trajectory information and neighbor positions to decide the next hop for the packet.

11. Attribute Based Routing & Directed Diffusion  Nodes desire certain information and other nodes have some information. How do they find each other?  Use attribute value pairs to describe the data Attribute value recordInformation request record type = animaltype = animal instance = horseinstance = horse location = [89, 154]rect = [0,200,0,200] time = 2:45:23

12. Directed Diffusion  Each node names data with one or more attributes  Other nodes express interests based on these attributes  Network nodes propagate the interests and results back to the sink  Negative gradients inhibit the propagation of information & positive gradients encourage information propagation  Assumption: the sink will be interested in repeated measurements from a source for a period of time  Paa-Kwesi will give a detailed presentation of directed diffusion next time

13. Next Lecture  Geographic Hash Tables – Andreas  Directed Diffusion – Paa Kwesi