Sitting in a lab on a swivel office chair is a robot worth a lot more than your car (probably more than your house) and it might one day save your life.
It's name is THOR, or Tactical Hazardous Operations Robot, and there are a couple of versions of it sitting around the Robotics and Mechanisms Laboratory (RoMeLa) at UCLA. They're the creation of Dr. Dennis Hong, who runs RoMeLa, and his team of graduate students.
Dr. Dennis Hong leads the Robotics and Mechanisms Laboratory at UCLA. Dennis Hong and his grad students have been working on versions of THOR for years. (Maya Sugarman/KPCC)
"I have a dream," says Hong. "In the future, when I become old, when I become a grandpa, I’m watching TV with my grandkids... And in the news there’s a burning building. And robots go into the burning building and rescue people. And I tell my grand kids… your grandpa developed that robot. How cool would that be?"
Notes fill a white board inside UCLA's Robotics and Mechanisms Laboratory on Friday morning, May 8, 2015. (Maya Sugarman/KPCC)
What's motivated Hong is the same thing that's driven the folks over at the Defense Department's Defense Advanced Research Projects Agency, or DARPA: they want to improve our ability to respond to catastrophes like the one at the Fukushima Daiichi Nuclear Power Plant in Japan following the earthquake and tsunami in 2011.
After the tsunami hit, explosions occurred at the power plant because of a buildup of hydrogen gas. Experts say that some of the damage could've been mitigated if humans had access to the area. It was too dangerous for people, so the thought is that if robots could do what humans do in a disaster situation, but go where humans can't, we'll be better off. And that's why they launched the DARPA Robotics Challenge.
The competition pits robot against robot to complete a series of tasks that they might experience if they were deployed in an emergency situation.
The robots need to drive a vehicle through a maze, get out of it, open a door, walk through the door, open a valve, walk over rubble, clear debris, climb stairs and use power tools to punch a hole in a wall. And all of this happens within mock rooms that DARPA built at the Pomona Fairplex.
Disney gave advice to obstacle course designers for DARPA's Robotics Challenge at the Pomona Fairplex. The challenge was first motivated by the 2011 explosion at the Fukushima Daiichi power plant in Japan. (Maya Sugarman/KPCC)
The winner of the competition will walk away with $2 million at the end of it all.
But even though these robots are competing now, it doesn't mean that the technology is ready to save lives.
Take Hong's THOR, for example:
"When you see these robots, they walk really, really slow and they’re really unstable," says Hong. "I consider these as like two-year-old babies. They’re just starting to learn how to walk. But in the next three, five years, you’ll see something very impressive soon."
This giant red button works as a "kill switch" if THOR is about to break itself while going through a test. (Maya Sugarman/KPCC)
THOR ambles along like a toddler when it's told to walk, but a toddler might not be able to control a power tool, drive a vehicle or put a hole in a wall.
"Our robot was able to drive a car, it climbed up ladders — just one step for the trials. It opens and close all valves. It can open doors and go through doors. And it can go over a rubble pile. And basically the competition is designed after these scenarios."
The robots in the competition, including THOR, are semi-autonomous, says Hong, which means they aren't being controlled by someone with a joystick: “A good way to explain this is you give high-level commands. For example, 'Robot, open that door.' Then the robot needs to figure out what do you mean by 'door,' where is it, how do I open the door. But the humans are always in control.”
In a building on the other side of the Fairplex, teams will be giving commands, much like they would in a disaster scenario. And also just like in a disaster scenario, they'll lose their line of communication with their robot at random intervals, which is planned by DARPA. Because in a disaster scenario the robots will need to be able to figure out how to finish tasks and stabilize themselves and wait for the next command.
THOR RD, left, the latest generation of the robot; a THOR shell; and THOR OP stand inside UCLA's Robotics and Mechanisms Laboratory on Friday morning, May 8, 2015. THOR was first developed at Virginia Tech with Dr. Dennis Hong. (Maya Sugarman/KPCC)
Compare real life robots to movie robots and you'll be disappointed. They're unstable, not completely autonomous and they still struggle with things like walking over uneven surfaces.
Getting in a car is another example of a normal everyday task that we take for granted. Because of the shift in weight and the movements required to squeeze into a seat, it's tough for scientists to program a robot to do the same. A real-world comparison? Think about the difficulty that an older person might have getting in and out of a car and compare that to what a robot has to do.
Still, the idea of failing doesn't phase Hong: "This is research. We don’t know if it is possible. It’s a challenge, but we need to do it so we’re doing it."
Battery life is also an issue. Most of the robots in the competition only have about a half hour of run time.
Even though you won't see these robots in a disaster situation tomorrow, maybe you will in the next five to 10 years, he says. Every little piece that works in a competition like this gets repurposed and moves into the next generation of machines.
And the ultimate hope is that they can be stationed in places like nuclear power plants so that, if disaster strikes, they can just undock and go help without human instruction.
That — and maybe they can make you dinner, Hong says:
"My dream is to have a butler robot, like Rosie the robot. Doing the dishes. Cooking the food. Let’s say we do have the technology to build a robot that can do all of that. We don’t, but let’s say that we have it. The cost is going to be more than the cost of a Ferrari... So I think we’ll see robots being used in real life where cost is not an issue. Like, for example, when human lives are at stake.
"You’ll probably see these robots being used in these type of scenarios first. And like cars and mass produced technology, the cost is going to go down and then it’s going to transfer to domestic use."
Inevitably one question kept coming up when I told people about this story: When are we going to have to bow down to our robot overlords?
Hong says that he gets this all the time and he's pretty tired of it. Dr. Gill Pratt, the program manager for the DARPA Robotics Competition, said something similar:
"I think that’s part of the difficulty that we have in this field, is that it’s very hard to get people to trust the machines even if the machines are in fact more trustworthy than a human being. They are not making their own decisions, and to be honest, a lot of human beings make pretty bad decisions too."
Don't project human intelligence into inanimate things, says Pratt. And Hong explains that we won't see a Chappie anytime soon.
"Walking with two legs, very difficult. You’re not going to see these robots used in real life in the next five years, maybe 10 years... [The artificial intelligence] that they describe in the movie Chappie, I think, is still science fiction... [The] arm manipulation. Those kind of things, I can easily see as being used in the next five years. Like an upper body on a tank tread for example... It sounds terrifying until it actually comes to you to save your life in an accident."
OK! OK! However, look at this thing from MIT.
If you'd like to see THOR and the other robots compete live, you can watch the stream here: