1. Control over Networks

2. The Coming Revolution  Electronics advances Moore’s law Computing/memory  Metrology advances MEMS  Communication advances Layered abstractions, MAC protocols, wireless standards  Devices are cheap, small form factor, low power, intelligent, networked  Sensing will become ubiquitous  What can we do with them?

3. Applications  Event detection Did something happen? Avalanches, fires, security breach, chemical leaks  Event Localization Where did it happen?  Surveillance/Mapping Typically fixed topography identified by co-operating mobile agents  State Estimation  Feedback Control  Understanding biological networks Swarms, flocks, etc How does complex collection motion emerge naturally from simple comm?  Bio-mimetic networks  Robotic networks

4. The revolution has not come yet …  Many papers from academics  A few start-ups (mainly selling motes and OS) Arch Rock Dust Networks  Some wireless protocols  Some competing OS for WSN  But no home run! No compelling broad-based consumer application Main successes are in security  Why? Is power/lifetime the bottleneck? Cost? Do we need motes to be sub $10? $1? Cheaper alternate solutions?

5. Sources  Conferences  Journals  Some survey articles I like

6. Research Ontology  WSN OS  WSN Hardware integration  Applications  Cooperation/Coordination/consensus  Fundamental studies Limits on achievable performance Understanding the interplay between comm and control How does the reliability of the intervening network degrade performance?

7. Sensor Nodes  Power Lifetime often the limiting factor  Intelligence, i.e. local computation Can imbed processors with reasonable capability  10 x 10 matrix inversion, products, etc  Memory Cheap, light, so it isn’t an issue  Communication with neighbor nodes Dedicated link to a specific neighbor Broadcast to neighbors  Mobility Optional – needed in some applications Would argue that mobility consumes far more power than communication But can program Mote to dock at recharge station when needed  Operating System Manages all resources Don’t want to custom program any mote

8. Scalability  No custom programming of any mote  Universal OS  Are all motes identical?  Fault detection capability  Programming the objective seamlessly  Change objective by reprogramming any mote  Objective change is sent to all motes  OS updates from any mote  Re-configurability  Layered abstractions  Virtual Co-location  Middle-ware  Adaptation in the WSN framework

9. Applications  Monitoring asynchronous events  Continuous environmental monitoring  State estimation  Feedback control  Distributed computing

10. Decentralized Control

11. Distributed Control

12. Wireless Actuators?  I am skeptical Actuators (opening closing valves, changing rotor position, etc) simply need too much power So there has to be a substantial power source The penalty of actuator failure is too steep In fact, we should likely have redundant actuation Need large batteries or AC power supplies This power supply can also be used for communication Literature survey?