For a human, reaching around an obstacle to pick that soda out of a crowded refrigerator is no big deal. However a robot with multi-jointed arms may find this challenging.
Tasks that involve this kind of movement are referred to as “motion planning,” which is actually a problem for robots—one that requires time-consuming computation. Simply picking up an object in an environment that has not been previously worked out for the robot may require several seconds of computation.
That’s about to change.
A team of researchers from Duke University has introduced a specially designed computer processor for motion planning that can plan up to 10,000 times faster than existing approaches while consuming a small fraction of the power.
The new processor is fast enough to plan and operate in real-time, and power-efficient enough to be used in large-scale manufacturing environments with thousands of robots.
"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."
Having quick motion-planning skills saves the task of modifying the environment around the robot.
Researchers have been working on motion planning techniques for nearly 30 years, which has resulted in recent advances reducing the time required to find a plan for a sophisticated robot to a few seconds. However most of these existing approaches rely on CPUs or faster (but more power-hungry) graphics processors (GPUs).
To make the skill even more efficient, the Duke team developed a new processor that’s specifically designed for motion planning.
"While a general-purpose CPU is good at many tasks, it cannot compete with a processor specially designed for just a single task," said Daniel Sorin, professor of electrical and computer engineering and computer science at Duke.
The new processor can perform collision detection, the most time-consuming aspect of motion planning. So for example, the processor can perform thousands of collision checks in parallel.
"We streamlined our design and focused our hardware and power budgets on just the specific tasks that matter for motion planning," said Sorin.
How It Works
The new technology works by breaking down the arm's operating space into thousands of 3-D volumes called voxels. From that point, the algorithm determines whether an object is present in one of the voxels contained within pre-programmed motion paths. With the help of specially designed hardware, the technology can check thousands of motion paths simultaneously, and then stitch together the shortest motion path possible using the "safe" options remaining.
"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."
"Previously, planning was done once per movement, because it was so slow," said Konidaris, "but now it is fast enough that it could be used as a component of a more complex planning algorithm, perhaps one that sequences several simpler motions or plans ahead to reason about the movement of several objects."
The technology opens the doors to new ways of using motion planning, and the team considers it a “game-changer for robotics.” The processor's speed and its power-efficient nature could pave the way for automation applications.
Konidaris, Sorin and their students are beginning to incorporate the technology into a spinoff company called Realtime Robotics, in order to commercialize it.