A new chip could bring robots out of the lab and into the real world, according to Phys.
A team of researchers led by George Konidaris and Daniel Sorin claim that they will be able to speed up robots to the point where they can leave the lab and specially designed factories and deal with the messy real world.
“When you think about a car assembly line, the entire environment is carefully controlled so that the robots can blindly repeat the same movements over and over again,” said George Konidaris, assistant professor of computer science and electrical and computer engineering at Duke. “The car parts are in exactly the same place every time, and the robots are contained within cages so that humans don’t wander past. But if your robot is using motion planning in real time and a part is in a different place, or there’s some unexpected clutter, or a human walks by, it’ll do the right thing.”
The Duke Robotics group at Duke University will present the paper “Robot Motion Planning on a Chip” at the RSS 2016 conference. The researchers describe how they can speed up motion planning by three orders of magnitude while using 20 times less power. They achieved this by using a custom processor they developed, rather than using general purpose CPUs and GPUs.
“Sometimes described as the ‘piano mover’s problem,’ motion planning refers to the challenge of dealing with scenarios by working out which movements to make. It’s something that humans pick up instinctively at a very young age, but which is extremely tough for a robot.”
— The Real BrianPiercy (@brianpiercy) June 24, 2016
Current robotics technology uses chips much like the one found in your computer to plan the motion of robots. This is very slow, even when done on high end computer chips.
“The state of the art prior to our work used high-performance, commodity graphics processors that consume 200 to 300 watts,” said Konidaris. “And even then, it was taking hundreds of milliseconds, or even as much as a second, to find a plan. We’re at less than a millisecond, and less than 10 watts. Even if we weren’t faster, the power savings alone will add up in factories with thousands, or even millions, of robots.”
The reduction in power demands also means that the technology is ripe for miniaturization.
Counter-intuitively, it is the simple problems in robotics that are hard, while the hard problems are easy. The first computer system to do integral calculus was already completely designed and working in 1969, as described in a paper by University of California researcher Maxwell Rosenlicht. However, when it comes to mastering movement in 3D space, robots can still be outperformed by an average three year old, according to Digital Trends.
— SciTech Digest (@SciTechDigest) June 26, 2016
“Human brains have evolved to be very good at some hard computational problems. Planning may seem like a simple 3D problem, but actually the robot must find a sequence of joint positions for every joint in its arm — so if it has an arm with 7 joints it is searching for a 7-dimensional path.”
The new chip solves this nearly-intractable problem by separating space into voxels, 3D versions of the more familiar pixels. The chip then marks some voxels as forbidden if there is an object in them and stops the robot from moving into those voxels. Finally, the chip picks the most efficient path from the remaining safe options.
The chip is designed such that the path can be recalculated just as easily if an obstacle, like a human, moves into the newly planned path of the robot. It can quite easily recalculate the path to avoid anything moving at everyday speeds.
“While a general-purpose CPU is good at many tasks, it cannot compete with a processor specially designed for just a single task.”
— Automaton (@AutomatonBlog) June 21, 2016
[Photo by Matt Cardy/Getty Images]