The Golden Age of Walking Bots That Never Arrived

by jonbaer • 10 years ago

17 8 10 years ago

Animats• 10 years ago

It's not that bad. The DARPA humanoid competition reminds me of the first DARPA Grand Challenge for autonomous vehicles. Not the one in 2005, the first one in 2004. I was there, and it was embarrassing. I wrote a post-mortem. with pictures.[1] Here's how it went:

- CMU: Their vehicle was preprogrammed, not autonomous. Teams got a CD with the route, as GPS coordinates and a path width, 2 hours before the event. They had a trailer with workstations on site, and their large staff manually planned out the route using aerial photos. The vehicle had little or no automated obstacle avoidance. But DARPA's staff was too smart to fall for that. The competition manager (an active duty USMC colonel) had soldiers out the night before the event moving some of the obstacles. CMU's vehicle made it 7.4 miles and then plowed through a solid sheet metal fence, went off course, got stuck on a berm, and spun a wheel until a tire caught fire.

- Caltech: went off course and through a fence, and couldn't find a way back through the fence.

- SciAutonics II: went into an embankment and got stuck.

- Team DAD: stuck at a football-sized rock.

- TerraMax: in the preliminaries at the California Motor Speedway, plowed into a parked car set out as an obstacle and continued to push the car until remotely disabled.

- Virginia Tech, Axion, CajunBot, Palos Verdes High School, Blue Team, and Ensco: couldn't get out of the starting area.

- TerraHawk - didn't make it to the starting area.

That was 2004, which was so bad it was covered by the Comedy Channel. In 2005, there were 23 teams, and every vehicle worked better than any of the 2004 vehicles. Five vehicles finished the 132 mile course. Suddenly, autonomous vehicles were a real thing.

That was what I was expecting for the DARPA Humanoid Challenge. The first one was awful, and that was expected. But I expected to see improvement each year until there was full success.

Instead, Google bought most of the players and, it now seems, accomplished nothing with them. This does not speak well for Google. If you care about what you're doing, you probably don't want to sell your company to Alphabet.

[1] http://www.overbot.com/grandchallenge/note45.html

apalmer• 10 years ago

Two things:

1) of all the animals that move around freely on this planet, i think bipedal walking is amongst the rarest mode of transportation, i mean you got many animals that can go bipedal for short bursts but outside of that you got basically humans and birds... and for birds in general bipedalism is the secondary mode of locamotion... just seems illogical to spend so much effort to develop what is clearly a very limited and complex mode of operation.

also if as the article surmises google had any expectation that they were going to invest in robotics R&D and get some significant new product out of it in 5 years time... well then they deserved to get burned... some problems just always fall on the government and military in particular because of the timespans involved before something meaningful can sprout from the R&D efforts.

x1798DE• 10 years ago

I agree that there are easier modes of movement than bipedal motion, but to some extent bipedal motion is useful because human environments are designed for human-like motion.

That said, cats and dogs seem to do just fine.

marcosdumay• 10 years ago

Human environments are mostly designed for wheels.

mrfusion• 10 years ago

Don't forget ostriches who use walking as their only means of transportation.

Also remember our friend the kangaroo. Oh and the T. rex.

takatin• 10 years ago

....just seems illogical to spend so much effort to develop what is clearly a very limited and complex mode of operation

Quadrupedal motion is for locomotion. Bipedal motion is for transportation.

Bipedal motion frees up two limbs to carry stuff around, open locks, get into cars and drive, move into and out of crowded, people-packed spaces (elevators) where horizontal space is limited but vertical space is plenty and so on.

And since we’re taking about robots, the two freed up limbs need not be “normal” limbs but tools specialised for the job at hand that cannot be used for locomotion purposes (the "arms" can be guns, welding machines or forks for carrying weights)

Also, robots that look and act like humans are an easier sell and selling them is the hardest part about getting these robots mainstream.

Have no doubt, perfecting bipedal motion is a worthwhile endeavor.

mchahn• 10 years ago

perfecting bipedal motion is a worthwhile endeavor.

Especially since it is extremely efficient. The human legs act like pendulums and need only a small push every step to continue.

paulftw• 10 years ago

Sergey got bored with his new robot toy and chucked it into the same bin as google glass, wave, internet balloon and some other stuff that got cancelled before public announcement.

The Golden Age of Walking Bots That Never Arrived

Related Stories