Slashdot Mirror
US Robots Win Big Down Under
An anonymous reader writes "US teams dominated the MAGIC 2010 autonomous robotics competition, mapping and neutralizing simulated bombs at the 250,000 sq. meter Royal Showgrounds in Adelaide, Australia. Leading the pack with a team of fourteen robots was Team Michigan, principally from the University of Michigan, followed by the University of Pennsylvania, and RASR. This contest marks the beginning of practical robots that not only think for themselves, but also actively coordinate with a human commander."
Congrats to the teams that did well. I know a bunch of Australian teams that looked into entering and decided not to because:
a) It was an engineering challenge more than a research challenge,
b) It was closer to that ethical line of making killer robots than, say, the DARPA Grand Challenge autonomous vehicle competition,
c) There was an extremely compressed timeline to actually make anything, and
d) The prize is mostly prestige. i.e. It wouldn't come anywhere near the development costs even for the teams that won.
So, it was a less than perfect competition. But that also means that the teams that did well in it did well under difficult conditions, so good for them. :)
They used the Hokuyo UTM-30LX Laser RangeFinder (LIDAR) which has a MSRP of $5,600 and a 30m range (270 degree FOV). I wonder if the Kinect would be a low-cost/low-resolution alternative in some environments (e.g., urban)? And at $150 each, one could use three or four Kinects for a wide field of view.
What one fool can do, another can. (Ancient Simian Proverb)
I think robots are nice and have loads of practical uses, but honestly I'm just waiting for something like LCARS to be practical. Integrated compute control of all the major systems in the house, etc.
The only thing that never really made sense to me were typing things out. In Enterprise they had a keyboard of sorts but there weren't nearly enough keys to cover most of the major symbols.
I suppose I'm just in love with the general concept of it.
I just hope robots don't become cheap soldiers that any rich guy can own his personal army.
I imagine eventually the UN is going to draw a line between remote-controlled drones (UAVs like the Predator) and AI bots and forbid AI bots from being used, at the very least, in direct combat. Besides, there are a lot of issues at hand with bots; EMPS, for one. Robots won't be nearly as agile and fast as a human running for his life can be, so I imagine they would be far more vulnerable to specialized weaponry designed to counteract them (or hell, even conventional "big bang" weaponry like grenade launchers, rockets, missiles, etc.) Robots can be hacked and reprogrammed, soldiers cannot so easily. It would be a P.R. disaster if an Army Combat bot is seized by an enemy combatant with off-the-shelf gear and turned on its own soldiers.
I don't believe that robots will be practical enough (cost-wise) to be used as soldiers for at least 20-30 years (if we and/or the international community would even allow such a thing to happen).
Random Thoughts From A Diseased Mind (Not For Dummies)
Interesting idea, however, using a lower resolution sensor leads to a more complicated model. SLAM and other mapping techniques are generally probabilistic based. It depends whether or not they have the processing power and energy to find a viable solution using the Kinect or other visual senors.
There is a large subset of the SLAM community devoted to this, Visual-SLAM; check it out.
Turkey showed up and put on quite a show. However, there was a Japanese team that did not show.
Odds that the Kinect will work outdoors should be quite low, as it relies on an array-based infrared system. Alternatively, a laser range finder uses a highly focused pulse of light at (nearly) a single point, which performs better in natural sunlight. It seems quite likely that Kinect will be popular in the near future for indoor robotics and robotics education, but indoor/outdoor robustness is strongly desired these days and scanning LIDARs won't disappear until robust Flash LADAR becomes common
Laser range finders are a must for accurate mapping and localization. I work with the UTM and other LIDARS on my robots, and the maps the produce are extremely accurate. Vision based navigation is possible, but it takes a lot of computation, and a lot of work to account for the uncertainty introduced. I'd say if you have the money, use both. Kinect might work well in a crunch, but as of now vision based SLAM is still in its infancy.