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.
“Motion planning seems easy for humans because we do it so fast,” George Konidaris, Assistant Professor of Computer Science and Electrical and Computer Engineering at Duke University, tells Digital Trends. “But that’s deceptive; like a lot of problems in AI and robotics, just because it’s easy for us doesn’t mean it’s easy for computers. 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.”
Fortunately, Konidaris and a Duke colleague named Dan Sorin think they’ve come up with an answer — and they’ve turned it into a new company called Realtime Robotics. What Konidaris and Sorin have created is a dedicated computer processor designed entirely for motion planning. It’s a chip that not only consumes a tiny fraction of the power of comparable processors, but can also carry out the task up to 10,000 times as fast as regular general purpose processors.
Related: Russian robot escapes from its lab
“Our special-purpose processor is designed just to do motion planning,” Sorin tells Digital Trends. “It doesn’t run software; it doesn’t read instructions from memory and execute them. Instead our processor has only the hardware needed to do motion planning, and it only expends power to do motion planning.”
Such an advance could be revolutionary in a world in which robots are moving beyond being specific purpose machines, designed to operate in a single domain — and becoming general purpose robots, capable of dealing with a complex world which may not be the same each time.
“This technology could liberate robotics from highly structured environments, with applications ranging from much more flexible and inexpensive robot manufacturing to lightning-fast decision-making for autonomous cars,” Konidaris continues. “Any robot that must interact with an environment that has not been carefully engineering for it specifically — in other words, the sorts of environments we all deal with every day — will have to do motion planning, and do it fast.”