Robots and Water and Whales, Oh my!

0df36316c97c9421b228b7208ae0cd57?s=47 marakana
August 23, 2012

Robots and Water and Whales, Oh my!

James Gosling, the Father of Java, talks about his new role as a software engineer for Liquid Robotics.

0df36316c97c9421b228b7208ae0cd57?s=128

marakana

August 23, 2012
Tweet

Transcript

  1. Robots and Water and Whales, Oh my! James Gosling http://liquidr.com

  2. Why? We’re critically dependent on the ocean But what data

    we have is sparse and expensive What if we could make the data dense and cheap? Eg. Predict hurricanes and tsunamis early and accurately?
  3. Genesis 2005: Roger Hine, Derek Hine and Joe Rizzi RC

    car for the ocean to listen to whales http://www.jupiterfoundation.org
  4. Not your typical robot ?

  5. Harvest Thrust from Waves

  6. Waves go Up

  7. Waves go Down

  8. Solar Cells

  9. Solar Cells The only actuator

  10. Exotic Materials Everywhere

  11. Wires

  12. Iridium satellite communication (GSM if close to shore)

  13. Iridium satellite communication (GSM if close to shore) 1$/kilobyte!

  14. Not a big data problem (1TB costs $109)

  15. Timeliness Continuous operation

  16. Free propulsion + Ruggedness + Solar = Long Duration

  17. What do you do with them?

  18. The weather buoy that installs itself

  19. Oil & Gas

  20. Pollution Monitoring

  21. Marine Mammal Monitoring

  22. Picket Fence around Marine Sanctuaries

  23. Global Warming Studies

  24. Fish Seafloor Sensors UAVs

  25. They’re really tough:

  26. High Seas

  27. Sharks!

  28. None
  29. None
  30. None
  31. None
  32. PacX challenge: 4 Wave Gliders, one ocean, public data. http://pacxdata.liquidr.com

    (over 140 gliders have been built)
  33. Demo

  34. Streaming Telemetry

  35. R<1 V>0 R>V

  36. No Rocket Science Roughly a lightweight, specialized, non-EE, JMS-ish thing

    Very Simple Tried-and-true techniques Optimized for small things that flow Every chunk individually authenticated Not trying to be general purpose The real key to simplicity.
  37. Not a Big Data Problem ($1/KB)

  38. But Reliability is Critical

  39. As is Authentication

  40. Data model is Simple: Bag of Bytes + TimeStamp

  41. There are nodes

  42. They contain data

  43. They contain data

  44. They publish

  45. They Subscribe

  46. They Subscribe

  47. They Subscribe Interests Authenticaton

  48. They Chat

  49. They Mesh

  50. They Mesh

  51. Viral Video Echo Cancellation

  52. Viral Video Echo Cancellation Seen That, Boring!

  53. Viral Video Echo Cancellation Seen That, Boring!

  54. Span Datacenters & ISPs JElastic Entic GoGrid Use no ISP

    special APIs
  55. Paranoid apps do multiple subscriptions

  56. App Paranoid apps do multiple subscriptions

  57. A Cluster is a Node

  58. A Cluster is a Node

  59. Cluster-to-Cluster links With filters Useful for • Sharding • Delivering

    to customer private datacenters • Jurisdictions (eg. Taxes)
  60. Cluster-to-Cluster links With filters Useful for • Sharding • Delivering

    to customer private datacenters • Jurisdictions (eg. Taxes)
  61. Retroactivity WTF? Screwups are always realized after the fact Never

    damage the data Computations are always a cache Metadata has timestamps too And they’re mutable!!
  62. Authentication OpenAM/OpenDS for users & groups Fine grained access requirements:

    every sensor of every craft by time interval for every sample Data delivered if subscriber meets requirements and is interested
  63. Thanks!