Robots Will Drive, Bust Through A Wall, And Make Repairs In DARPA’s New Robotics Challenge
Five years after the DARPA Grand Challenge robotic cars are already hitting the roads and states are preparing for their eventual arrival. Now DARPA is launching a new Robotics Challenge that will test the most advanced robotics solutions in a simulated disaster obstacle course. The challenge – and the cash prize – will almost certainly prove a major catalyst for the near future of robotics technologies.
In the event of a natural or man-made disaster, it’s always better to put robots in harm’s way instead of humans. With this in mind, the Robotics Challenge is a wide open, task-driven test. Instead of specifying specific technologies, DARPA is saying get to the finish line anyway you can.
What DARPA wants to see is a robot that has human-like “mobility and manipulation” abilities. At a disaster scene robots will have to make use of the same machinery and tools that human rescue teams have to use. The different stages of the challenge are meant to simulate an emergency response to a natural or manmade disaster. The robot will enter an open-frame vehicle like a John Deere Gator or Polaris Ranger, turn it on, and drive it – steering, throttle, brakes and all – to the disaster scene. Once it’s pulled up to the pile of rubble, it will exit the vehicle and climb over the sloped terrain littered with loose rocks, trees, ditches, and other obstacles it has to negotiate or avoid. Eventually the robot will reach an entryway blocked with debris that it will have to remove. Once the debris is cleared, it has to operate a door handle and push the door open. Inside, it will have to climb a ladder to reach a catwalk. After crossing the catwalk it will reach a concrete panel or a framed wall that it has to bust through using something like an electric hammer or chisel. Waiting for it on the other side of the panel will be a series of pipes, only one of which will be leaking. The robot has to spot the smoke or hear the hissing sound to locate the faulty pipe and then close the pipe’s turn valve. Lastly, the rescue robot’s day will end after locating and replacing a cooling pump.
They somehow forgot to include pulling small children from a burning building.
DARPA hasn’t yet decided by what criteria exactly the robots will be judged except that the robots get points for operating autonomously and using less energy.
If you’re a big fan of humanoid robots like Petman, you may be biased towards imagining a band of Superman-like – or Terminator-like – robots coming to the rescue. But DARPA emphasizes that the winning robot will get the job done, humanoid or not. They want to make it clear that a team should go ahead if they think an arachnoid robot would do the job better than a spiderman robot.
The team that builds the winning robot pockets $2 million.
The robots won't be required to be completely autonomous but will operate with "supervised autonomy." Under supervised autonomy, the controller gives general commands without having to carry out basic functions of perception and action.
I expect that Boston Dynamics will be one of the competing teams. But it will be interesting to see which direction they go in. Will they go straight humanoid and try to develop Petman, or will Alpha Dog be the more sensible choice? Maybe we’ll see something entirely new.
DARPA realizes that the challenge is really, really tough, calling it “DARPA hard” but not impossible. For them it’s a longterm commitment. The competition will be held once a year, at the endpoint of two separate phases. Phase 1 will last 15 months, beginning October 1, 2012 to December 31, 2013. There’s no cash prize for the Phase 1 winners. But they qualify to move onto the second phase, which will last 12 moths from January 1, 2014 to December 31, 2014. Any company or research team, anywhere in the world, can compete.
Needless to say, the humanoid robot envisioned by DARPA will require some seriously advanced technologies, and lots of them. To maximize their chances for success, the program is broken up into different tracks so that entrants can tackle a problem that is suited to their strength. Track A is for teams developing both hardware and software, while Track B is for software developers only. Any proposal that is chosen for Tracks A and B will be funded – $3 million for Track A, $375,000 for Track B. Tracks C and D will compete in the same categories, except they’ll have to build their robot and software at their own expense.
DARPA will build its own robot so that teams focused solely on software development will have a test bed. The so-called Government Furnished Equipment (GFE) platform will have arms with two or three fingers and 7 degrees of freedom, legs with 6 degrees of freedom, and a head equipped with stereo vision and laser radar.
The challenge was inspired by the Fukushima disaster last year in Japan. Gill Pratt, the DARPA program manager in charge of the challenge, said in the first 24 hours following the disaster crucial tasks couldn’t be performed because it was too dangerous for people to go into the ruined reactor. DARPA is certainly serious about making sure the US can take care of business. The $2 million cash prize is just a small part of their investment. Including the funded research, DARPA will spend up to $34 million for its rescue robot of the future.
Each of the different tasks could probably be performed separately by different robots, but DARPA's effort to bring the different technologies into a single robot will undoubtedly create something great. I'm excited for our competitors. May the best robot win! What it will look like is anybody's guess.