Robots That Jump

Robot Bodies Needed Before Robot Minds

DARPA Grand Challenge – For Humanoid Robots

Well, this is definitely interesting! In the spirit of the DARPA Grand Challenges for driverless vehicles, DARPA has apparently decided that those vehicles, do, indeed need drivers. The next challenge is to create humanoid robots that can drive cars. The fun begins on Oct. 1, 2012, the final demonstrations are due to be by the end of 2014.

Interestingly, Boston Dynamics – the company that has created agile 4-legged robots mules and a faster robot (admittedly running on bent metal knees)

This follows on the move by the US Navy to create humanoid robots which can fight fires, drawing on the Virginia Tech robots:

Once again, DARPA seems to have see where robots can go. In the 2005 challenge which I participated in (until semi-finals, at least) they had a lot of interest in innovate robotics that didn’t follow standard artificial intelligence engineering guidelines. The winner of the 2005 challenge, Stanley, didn’t used Ai to move -instead it ran stats on multiple, non-camera sensors and was “trained” to react to sensor patterns. DARPA’s willingness to support large numbers of entries indicates a flexibility of thinking that could move humanoid robotics beyond the run it has been stuck in ever since the Asimo debuted back in 2000.

What remains to be seen if the ambitious goals can actually be realized. Part of the DARPA spec is that the humanoid can sit in the car and drive, possibly even turning around to look through the windshield. This is not going to be easy. I don’t think I’ve ever seen a robot trying to look behind itself for any reason, much less one that involves grasping a seat, balancing, and swiftly moving vision between the front and back.This implies a highly flexible neck or upper body, again something I haven’t seen much of in humanoids.

That alone might be worth the $2 million grand prize – peanuts, really, considering the amount of money we spend on other things. The cost of a single Space Shuttle launch might sponsor 50 of these kinds of challenges.

So, instead of “robots that jump” we will have “robots that rubberneck”.


One response to “DARPA Grand Challenge – For Humanoid Robots

  1. John Nagle, Silicon Valley, CA June 9, 2012 at 8:55 pm

    The humanoid challenge is interesting. But driving isn’t the hard part. Getting into the vehicle is a tougher problem. Once in the vehicle, you can just treat the robot as a set of actuators and sensors for the vehicle controls. There’s no requirement to drive fast, so it’s not necessary to outdrive the visual stereo range. (In the 2005 Grand Challenge, the big breakthrough was using machine learning to determine that the far road, outside LIDAR profiling range, looks like the near road, for which profile data is available. That allowed speeds with stopping distances greater than LIDAR profiling range.)

    Of the other tasks, walking over rubble with a biped will be tough. It’s been done with quadrupeds, but never for a biped. Ideally, you’d like a robot that really could jump. More likely, we’ll see crawling in version 1.

    Opening a door with a key has been demonstrated in the DARPA ARM program. Climbing a ladder has been done in some Japanese hobbyist robot competitions. So those tasks are not out of reach.

    Turning a valve may be tough. Replacing a pump will be very tough. Using tools on threaded fasteners in less than benign environments hasn’t been done well. The DARPA ARM program effort uses a two-armed torso and can pick up a screwdriver. But it can’t tighten a screw with it. They need to demonstrate the pump replacement task in the ARM program. There’s some preliminary work to get done before doing this with the humanoid. Just for starters, the simulator DARPA is procuring from Willow Robotics isn’t up to handling bolt-on-nut or wrench-on-bolt situations. They hope to have that working in 2013, which puts the whole effort behind schedule.

    Tool using by robots with multiple arms and fingered hands, without fixtures, needs more development. There’s a lot that could be done there. If there was a market, I’d work on the problem.

    DARPA wants a limited man-machine interface for this project. If they’re lucky, they’ll get something like a video game, where you designate a target and menu-select a verb to apply to it. If not, they’re going to get a teleoperator.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: