Slashdot Mirror

← Back to Stories (view on slashdot.org)

Why So Many Robots Struggled With the DARPA Challenge

Posted by Soulskill on from the too-busy-trying-to-kill-all-humans dept.

stowie writes: The DARPA Robots Challenge concluded recently, and three teams were given prizes for completing all the tasks. The other robots in the competition struggled — not only were they unable to complete the required tasks, many of them were unable to even stay standing the entire time. So why did these robots have such a hard time? "DARPA deliberately degraded communications (low bandwidth, high latency, intermittent connection) during the challenge to truly see how a human-robot team could collaborate in a Fukushima-type disaster. And there was no standard set for how a human-robot interface would work. So, some worked better than others. The winning DRC-Hubo robot used custom software designed by Team KAIST that was engineered to perform in an environment with low bandwidth. It also used the Xenomai real-time operating system for Linux and a customized motion control framework. The second-place finisher, Team IHMC, used a sliding scale of autonomy that allowed a human operator to take control when the robot seemed stumped or if the robot knew it would run into problems." If nothing else, the competition's true legacy may lie in educating the public on the realistic capabilities of high-tech robots.

29 of 44 comments (clear)

  1. Sarah Connor? by Snufu · · Score: 2

    I am a robot sent from the future to ask you to open this bag of potato chips for me.

    1. Re:Sarah Connor? by ShanghaiBill · · Score: 1

      I am a robot sent from the future to ask you to open this bag of potato chips for me.

      Relevant Xkcd

    2. Re:Sarah Connor? by __aaclcg7560 · · Score: 1

      Did you bring dip with your chips?

    3. Re:Sarah Connor? by TWX · · Score: 1

      Ammunition supply is finite. Militaries of the world are always looking for ways to reduce the number of personnel, and they still have to manually perform maintenance tasks on all machinery including loading fuel and ordnance.

      Some day that might change, but right now there's no practical general-purpose war machine capable of servicing itself or being serviced by another machine without human involvement. Should a hostile AI want to do ill to humanity, it would either require support from some portion of humanity, or it would be limited to a small set of engagements until its machines-as-appendages ran out of supplies or broke down.

      --
      Do not look into laser with remaining eye.
    4. Re:Sarah Connor? by blue9steel · · Score: 1

      Should a hostile AI want to do ill to humanity, it would either require support from some portion of humanity, or it would be limited to a small set of engagements until its machines-as-appendages ran out of supplies or broke down.

      Fortunately humans have been shown to prioritize self over larger groupings so finding a small army of AI sympathizers and quislings shouldn't pose too much of a challenge for the budding AI overlords.

  2. Does it come down to hardware? by Iamthecheese · · Score: 1

    Honest question: In the end does it seem likely robots that succeed at these tasks mostly do so because, as in other DARPA challenges, they're the ones with high-end hardware?

    --
    If video games influenced behavior the Pac Man generation would be eating pills and running away from their problems.
    1. Re: Does it come down to hardware? by mediumcomputer · · Score: 1

      No they didn't. Teams such as Robosimian and Momaro we not bipedal. Just many did go bipedal because seven of the 25 were ATLAS variants and many were designed based on an open robotics forum (I forgot the name). Many of the total originals were unique body plans.

  3. They all struggled, at first. by 0100010001010011 · · Score: 5, Insightful

    Does everyone forget the DARPA Grand Challenge from 2004? The second year (2005) they had 5 vehicles finish. 4 of them within the 10 hour limit.

    In 2007 they had the 'urban challenge'.

    Now we have driverless cars and semis. Google and Uber are poaching a lot of of the grad students and professors from Stanford and Carnegie Mellon.

    1. Re:They all struggled, at first. by Jack+Griffin · · Score: 3, Insightful

      Now we have driverless cars and semis.

      Do we?
      Sure we have a whole bunch of cool stuff in controlled environments, but the nature of a public road is that it is uncontrolled. For some reason a lot of people underestimate how much of a difference there is between the two.

    2. Re:They all struggled, at first. by Anonymous Coward · · Score: 3, Informative

      Google's self-driving cars have logged plenty of time in traffic on public roads, so I don't know where you get the idea that nothing has happened outside of controlled conditions.

    3. Re:They all struggled, at first. by Necron69 · · Score: 1

      Perfect California weather on intensely pre-mapped out roads = 'controlled conditions'. When the cars can navigate I70 through the Rockies during a blizzard in heavy ski season traffic, let me know.

      Necron69

    4. Re:They all struggled, at first. by aardvarkjoe · · Score: 1

      When the cars can navigate I70 through the Rockies during a blizzard in heavy ski season traffic, let me know.

      When humans can navigate the I70 through the Rockies during a blizzard in heavy ski season traffic, let me know.

      --

      How can we continue to believe in a just universe and freedom to eat crackers if we have no ale?
    5. Re:They all struggled, at first. by jlar · · Score: 1

      Computers do not need to be able to navigate the car in all circumstances for self-driving cars to be useful. I would be happy if they could simply take over when driving on the freeway and warn me when it is time to leave the freeway again.

      The end goal is of course to let the AI control the car at all times. But a more modest start is just fine.

    6. Re:They all struggled, at first. by Jack+Griffin · · Score: 1

      Computers do not need to be able to navigate the car in all circumstances for self-driving cars to be useful. I would be happy if they could simply take over when driving on the freeway and warn me when it is time to leave the freeway again.

      So you would buy a robot car knowing it can't handle certain scenarios on the road, then trust it with your life to decide when that is? And you think this will be a big seller?

  4. Deep learning by Lennie · · Score: 4, Informative

    Yes, I'm sure the DARPA challenge is hard work, but I was much more impressed by how well they were able to apply deep learning for use with robots:
    http://newscenter.berkeley.edu...

    The fastest robot on the DARPA challenge took 45 minutes, look at how fast the robot is in the above video. It's much more close to how a human would do it.

    5 years ago from the same lab they took hours to do things and they were still using very little machine learning in comparison:
    https://www.youtube.com/watch?...

    And more importantly how close they are to using demonstrations (how about YouTube videos or from other people or robots doing similar tasks) to get robots to learn faster and many more tasks:
    https://www.youtube.com/watch?...

    I was also very much impressed the first time I saw what Deepmind had done:
    https://www.youtube.com/watch?...

    --
    New things are always on the horizon
    1. Re:Deep learning by Dr.+Spork · · Score: 2

      That's what's so great about the DARPA challenge: It reveals the difference between a lab's PR videos and the actual performance of robots they produce. It's not like Berkeley didn't try to score the two million from DARPA. IIRC, they failed. Comically.

    2. Re:Deep learning by PeDRoRist · · Score: 1

      Dude, that first video is accelerated 50 times!

      --

      Anything you do can get you slashdotted, including nothing.
    3. Re:Deep learning by PeDRoRist · · Score: 1

      Ah no, damn, I was talking about the second one, which you clearly stated was old.... stupid me.

      --

      Anything you do can get you slashdotted, including nothing.
    4. Re:Deep learning by samwichse · · Score: 1

      All this technology, but what I really want is a vacuum robot that can do stairs.

  5. origin by HiThereImBob · · Score: 2
    I had the opportunity to attend an award lecture early this year for one of the researchers working on the core mathematics / feedback control systems for bipedal walking robots. It's the basis for all of the DARPA robots and he covered many of the relevant topics. available here if anyone wants to know more:

    http://www.eecs.umich.edu/eecs...

  6. Detailed DARPA Challenge Videos... blocked! by Corporate+T00l · · Score: 2

    More detailed videos of how the challengers performed are available on the DARPAtv YouTube channel:

    https://www.youtube.com/user/D...

    However, it seems that the most interesting one, the Main Program Feed, is blocked in the US due to some kind of copyright issue.

    The DRC Finals Workshop is muted, also due to some kind of copyright issue (which makes it extra pointless since the video is just people talking on stage).

  7. Because its fucking hard! by Viol8 · · Score: 2

    The robots themselves arn't the problem, some of the hardware is fantastic. Its the software - moving a robot around and doing tasks in an unpredictable enviroment with obstacles is a monumentally hard task. Even the human brain takes a number of years to master it from birth so the chances of any one team of programmers suddenly mastering it with a robot is minimal. It'll be a gradual evolution of the software over the years.

  8. No by Viol8 · · Score: 2

    Its the software thats make or break for this task and software to accomplish this sort of thing is incredibly hard to write.

  9. Robotics club slogan "It is harder than it looks" by jgtg32a · · Score: 2

    I seem to remember hearing about a robotics club and their slogan was "it is harder than it looks"

  10. The robots are faking by PPH · · Score: 1

    The winner of the DARPA challenge gets sent inside of a hot, damaged reactor. No thanks. Send the meat bags instead. My maniputator motor was a bit sore this morning.

    --
    Have gnu, will travel.
  11. Re:Did they exclude EE for this challenge? by PPH · · Score: 1

    Robotics isn't a terribly difficult challenge for electrical engineering. Most of the sensor and motor control stuff is pretty much plug and play at this level.

    --
    Have gnu, will travel.
  12. That singularity... by gizmo2199 · · Score: 1

    Any day now...

    --
    This Sig does not Exist.
  13. Re:Apple by leonbev · · Score: 1

    Apple isn't stupid enough to invest a ton of R&D in this yet.

    They'll wait for someone to make a clunky but functional prototype that still has some usability issues first, and then make a pretty and easier to use version of that. Then they'll let their marketing machine convince their fan base that it's the most revolutionary thing since... well... the last Apple product that came out.

    Apple is really good at this. They did it laptops, then smartphones, then tablets, and most recently smartwatches (although that one still needs some work).

  14. Drunk robots? by jlar · · Score: 1

    From the video it seems like robots and drunk people behave somewhat similar. Or maybe the robots were in fact drunk. We will probably never know.