autonomous prototype sensors (motes) 4 MHz, 8bit MCU, 4 KB RAM, 8KB ROM short-range (1-10ft.) radio light and other sensors LED and serial port outputs running TinyOS: flexible, component- based, multithreaded OS with small memory footprint PC-based simulator (up to 10 3 motes) Scalable Routing in Networks of Embedded Sensors Motivation advances in technology enable sensing, computing and communication in devices as small as 1 mm 3 individual devices are severely resource constrained potential need for rapid-deployment of large-scale (10 3 – 10 5 nodes) of such devices in hostile environments. need for multi-hop point-to-point routing. Objective scalable routing protocol in resource (memory, power, and bandwidth) constrained ad-hoc wireless sensor networks Future Research incorporate real-time message delivery constraints, multicasting, multi-path routing explore fault-tolerance aspects of routing: massive and simple node failures, anticipated (deaths) and unanticipated faults explore application-specific knowledge for data aggregation, energy-aware routing advantages  caching on routing info from overheard messages and data messages possible  source routing on short trips, next-(vicinity?)-hop on long  fixed message size  controlled per-node memory usage  route changes contained to vicinity Existing protocols DSR AODV source routing – message carries routing info - aggressive route caching possible - message size and per-node memory consumption grows uncontrolled with network size - changes in remote areas make node’s routing info obsolete next-hop routing – message carries little routing info - message size fixed - per-node memory consumption controlled advantages non-gratuitous HELLO messages neighbors discovered as needed, stale info kept to minimum possible simple implementation of reliable next-hop delivery Routing requirements scalable up to 10 3 motes (20 – 50 hops) on-demand, self-configuring, fault-tolerant low overhead  few messages, controlled message length  low per-node memory requirements Proposed Performance Measurement compare performance of VSV vs. AODV and DSR field test (real motes) - 5x5 grid of sensors periodically taking light readings and sending data to several destinations - measure overhead, latency, message propagation, signal attenuation, error rate, use results in simulation simulated test motes in arbitrary locations on a plane - 10% simultaneous senders destinations implemented DSR, AODV, and VSV in TinyOS, done preliminary testing in simulator and the field measured mote’s radio and computation performance Work Completed Goals and Motivation Requirements and Existing Protocols Platform Expected performance overhead (memory+msg.) avg. number of hops DSR VSV AODV VSV should behave as DSR when short trips dominate AODV when long trips dominate Vicinity Source Routing (VSV) Design Ideas ABC data ABC ? ABC ALIVE? ABC ACK ? ABC RERR A sends data to C via B B rebroadcasts to C A overhears? link to B is ok A sends ALIVE? to B B returns ACK A gets ACK? link to B is bad yes no yes no Vicinity routing message carries routing info to a fixed distance (3 hops in the example) F learns its 4-hop vicinity AEDBCF G AABABCBCDCDE source dest. message gathers routing info as it travels both protocols assume the existence of the link layer that continuously maintains node’s neighborhood (IEEE ) – not efficient for sensor networks On-demand neighbor awareness node learns about neighbors when requested link verified on data trans- mission Regi Oommen Hassan Gobjuka Mahesh TamboliMikhail Nesterenko