This will be an exciting weekend for the robotics world. Starting today, the DARPA Grand Robotics Challenge takes place at the Homestead-Miami Speedway. Seventeen teams will compete on a grueling eight stage course that will test the limits of robotics as we know it. In order to win, humanoid robots will have to drive a vehicle, climb a ladder, move debris, manipulate tools, just to name a few tasks. This weekend is just the trials though and the tasks will be taken in stages, but by next year they will have to complete all eight stages consecutively. Though the anticipation is high, we do have to recognize the monumental task that was put down upon these teams. As with the DARPA autonomous vehicle competition back in 2004, the first year of the challenge, no team made it past the 8 mile mark. In other words they all failed, a year later however, multiple teams crossed the finish line. The leader of the winning team from Stanford, Sebastian Thrun, went on to work with Google’s Autonomous car group.
I thought I’d review a few of the robots that you will likely hear about over the next few days and years. You might have already seen Atlas, the Humanoid robot from Boston Dynamics. More Terminator looking than human , he weighs in at about 330lbs and over 6 ft tall. Atlas, originally known at PETMAN was originally conceived to test the design and effectiveness of chemical suits for the military.
More recently, NASA’s Johnson Space Center unveiled Valkyrie. The design feels like you’re looking at a character from Star Wars. Several team members came from the team that worked on the Robonaut, so there will certainly be life for Valkyrie, even after the challenge.
Then of course there’s THOR, (Tactical Hazardous Operations Robot). From RoMeLa, the lab that created CHARLI, DARWIN, and SAFFIR. With Dennis Hong on the team, THOR is certainly one of the robots I would be watching out for.
However the results play out, the competition both this weekend and at next year’s finals will really let the world know where the cutting edge of robotics lies. There is no doping class and Hollywood will not be able to add in any CGI to improve performance. Personally, I am not expecting any of these robots to actually complete all the tasks. Most robots, I am anticipating, will actually be remotely operated at some point in an effort to accelerate machine learning. Essentially showing the robot what to do, then allowing it to perform on its own at a later time. This will not be much different that teaching a child how to use its body and learn. Situations like this are actually be very useful in telepresence and remote presence applications, allowing the possibility for some semi-autonomy in future platforms, while maintaining a highly anthromorphoric platform to work with. While pieces of this already exist in consumer and enterprise platforms, expect it to become much more ubiquitous and likely a standard option on some models.
In the meantime while I wait for 2014, Robots, start your CPU’s, and may the best team err, qualify.