"Whither Robotics"

From Foocamp12

Revision as of 12:22, 13 June 2012 by Foocamp (Talk | contribs)
(diff) ← Older revision | Current revision (diff) | Newer revision → (diff)
Jump to: navigation, search


Whither Robotics

Gary Bradski

Robots until now

  • Automation has only penetrated about 9% of manufacturing and less of distribution
  • Robots are for the most part blind with scripted movements (welding robots in car plants are the prime example)
    • It takes as long to design the automation for making a car as it does to design the car
      • This is why cell phones are assembled by hand -- it takes longer to do the automation then the shelf life of the phone
  • There is a consistent demand among manufacturers and distributors to move to "Robotics 2.0"
    • These robots handle more variance using sensing, mostly visual (2D+3D) sensing. Thus, they see objects and determine their location and can move to interact or grasp them rather than just moving to exactly predetermined places to interact or grasp the part.

Robotics 2.0

  • Robotics is benefiting from advances in visual perception
    • Partly from algorithms and
    • partly from new, cheap 2D+3D sensing (Kinect, Asus Xion Pro).
  • Under FriendDA, I showed videos from my Willow Garage Spin off, Industrial Perception Inc.
    • These were in the areas of perception systems mounted on industrial robots that do recognition ("what") and 3D location determination ("where") of large and small objects for the purposes of robot manipulation (grasping, moving, placing).
    • Here's a video I can show (to be seen at ECCV 2012) 3D object recognition.
      • This video is "uncut" and shown at actual speed.
  • At the same time, robot arms are coming down. We'll soon see Heartland's robot, but other cheap, compliant arms are also coming on line
  • Much of this is supported by free, open source software, soon to be a foundation: OpenCV.org (or opencv)
  • We also see advances in robotic navigation:
    • In the home
    • Industrially (Industrial Perception Inc is selling an inexpensive bolt-on navigation system in factory and distribution settings) so far just to our partner
    • In cars.



  • Because of rising costs in Asia, transportation costs, and increasing productivity via automation, the US now has a chance to bring manufacturing back from China.
  • But, there's a cost:
    • The above robots are always judged against human labor, whether in driving, loading boxes or handling industrial parts.
      • This is a challenging standard: "unskilled labor" can be pretty skilled in rapidly and adaptability handling complex parts or driving in complex situations
        • But humans fatigue, get injured and work quality varies highly
      • Still, we are getting closer to another wave of automation replacing human labor due to the above improvements in algorithms and robot arms.
  • There's no choice
    • Manufacturing plants bring other employment around them: electricity, technical and human services
    • The technical expertise around manufacturing plants allows for easier cross-fertilization of ideas and skills
      • Un-conference in manufacturing might be cool

Transportation and Distribution

  • Self driving cars will also displace truck drivers
  • In door navigation and object recognition will bring about "pick to robots" supplanting "pick to man" (the thing that Kiva improved).
  • Clearly the above two points imply that robots will also start playing large roles in agriculture and mining.


  • As we improve perception and manipulation in manufacturing (first to pay), these skills will inevitably go across to domestic areas
    • We'll eventually have our robot butlers
      • First, we'll have to really nail human perception. You can't have a robot knocking into my frail mother approaching 90.
        • My 94 year old dad might enjoy a little robotic tousling however

Whither humanity

  • With robot manufacturing (and the fact that corporations are nearly complete legal persons in the US*), humans could be seen as just parasites on the corporate body: the most efficient corporations keep all their earnings to better themselves.
  • With robot cars and trucks, there may no longer be a reason to own these things. Dial them up on your smart phone, pay only for what you use. Idle cars are a waste.
  • With domestic robots and with most of your art, pictures, books, writings, and thoughts online, there might not be a reason to own a home anymore. Again, just pay for what you use, your roots are with your friends, families and your digitized content.
  • What might happen to society in such a case**? How could things work out?
    • Real communism: Workers/humans own the means of production. You could own stock in a mutual fund of productive assets of corporations.
    • You might not own your car or even your home. You might come and go, migrate with the seasons. Have a swappable ring of homes. Nothing but time shares, robotically cleaned, virtually customized.
      • Perhaps a mobile storage facility will hold your physical items, delivered on demand to where you are now or will be next.
    • Life might come to look like an un-conference: You co-house (house separate but shared food, workshops etc) with changing ring of friends and acquaintances. And move on with the tide or season. Contracting with various corporations as needed. Always having an default production income to fall back on.

  • * Note to libertarian wankers, early libertarian philosophers where adamantly opposed to this. Corporations were merely contractual agreements between individuals. They wanted no corporate veils. Only persons were persons.
    • We'll have to deal one day with the possibility of robot persons and with genetically altered machine-human "persons".
  • ** I'm a technical optimist "It will all work out". I'm a realist: "after millions die and suffer horribly".
Back to Session_Notes
Personal tools