1. Wireless Sensor Networks Craig Ulmer

2. Background: Sensor Networks n Array of Sensor Probes (10-1000) n Collect In-Situ Data about Environment n Wireless Links –Relay Data –Collaboration

3. NASA Applications n Primary –In-Situ Data Collection –Precision Landing Guidance –Vehicle Health Sensors n Secondary –Trail Markers –Relay Networks

4. Motivating Application: Exploration of Mars n Scientific Phenomena: –Thermal Currents –Dust Storms –Seismology n Engineering Challenge: –No GPS –No Communication Infrastructure –Size, Mass, & Power Constraints

5. Modern Sensor Nodes UC Berkeley: COTS Dust UC Berkeley: Smart Dust UCLA: WINS Rockwell: WINS JPL: Sensor Webs

6. Node Hardware Embedded Processor Transceiver Memory Sensors Battery Limited Lifetime 8-bit, 10 MHz Slow Computations 1Kbps - 1Mbps, 3-100 Meters, Lossy Transmissions 128KB-1MB Limited Storage 66% of Total Cost Requires Supervision

7. Networking n Multi-Hop Routing –Limited Transmission Range n Routing Issues: –Irregular Topologies – Data Transport Aware –Power Aware– Fault Tolerant

8. Scientific Value n Multiple Data Points: Time and Position n Temporal Synchronization –Hierarchical Schemes n Position Estimation –Digital Ranging –Offline Triangulation d1d1 d2d2 d3d3 

9. Sensor Network Initialization Deploy Wake/Diagnosis Organize into ClustersRoute

10. SensorSim n Sensor Network Simulator –How well do Algorithms Perform? –Algorithms as State Machines n Configurable Modules for Flexibility –Simulation at Different Levels n Java Based –Platform Independent

11. Simulator Node Layers Application Routing Link Sensor Triggers Clustering Algorithms, Reliable Routing Data Fusion Clock Synchronization Medium Access, Commercial Chipsets Node

12. Example: Election Clustering n Distributed Algorithm n Nodes Elect Leaders, Form Groups n Limited Knowledge Trial Member Trial Leader Member Undecided Join Nearest

13. Example: Fixed Leader Clustering n Predefined Cluster Leaders n Find Nearest Leader n “Mutiny” if Leader too Far Away Sleep LeaderUndecided Member

14. Other Simulators n ns –CMU Monarch Extensions for Ad Hoc Wireless –WiNS: Wireless Network Simulator –LEACH/PEGASIS Extensions to WiNS n GloMoSim / UCLA n Opnet

15. Why Another Simulator? n Previous Sims: LAN-Biased –Assume Thick Layers (802.11,TCP, Telnet) –End-to-End Networking n Sensor Nets: Different Architecture –How Can We Network w/ Minimal Hardware? –Interested in Node Behavior –Adapting Other Sims is Same Job

16. Ongoing Work: Network Algorithms n Given Clusters, How do we Route? –Limited Route Table Storage –Traffic Often Directed –Loop-Free –Minimal Route Updates n How does Node know Location in Network? –“Identifying ID” Number

17. Sunrise Synchronization n Use Sunrise as Synchronization Point –Earlier Risers are More Eastern –Smooth with Cluster Values, Neighbor Clusters n Gross Estimate of East-West Dimension

18. Conclusions n Sensor Networks Valuable Collection Agents n Minimal Hardware, Adapt Algorithms to Match n Use Scientific Observations in Routing n SensorSim Ongoing Work for Analysis