Revenge of the (Robot) Nerds

August 2013
By Bryan Bergeron

In the world of commercial robotics, brawn seems to have it over brains. The huge mechanical arms that bend, bolt, and bond metal don’t have much in the way of computational capacity. Why supply a robot designed for dull, dirty, and often dangerous work with more than a smidgen of a brain? The first reason that comes to mind is safety.

There are useful reflex loops programmed into robotic TIG welders in some automobile assembly lines – including the ability to sense when contact has been made with a soft, easily damaged human. These safety mechanisms can be lifesaving.

Having worked with a desktop 3D printer for several months, I’ve come to appreciate the computational overhead of slicing and dicing a 3D model in such a way that it can be quickly rendered in 3D. Although this computation is largely performed on a desktop before porting the rendering files over to my 3D printer, the onboard processing is still magnitudes beyond the computational abilities of the CNC router that it replaced.

For robotics to really blossom, there has to be more computational intelligence available for navigation, sensing, human interaction, and selfrepair. There’s a good chance there will be more of this intelligence available in the next few years.

One reason for my optimism is the never-ending march of progress on the microcontroller front. Compare the Raspberry PI to the popular Arduino, introduced only a few years ago. The price/performance point for the Pi simply blows away the Arduino — at least as far as raw computational ability goes. Then, there’s the new Propeller chip and C compiler from Parallax which creates some interesting synergies with parallel processing hardware and a conventional programming language. There’s probably a 1 GHz Arduino on the drawing board somewhere.

However, these are examples of evolutionary progress — the same sort of progress that we’ve seen in robotics for the past few decades. What about revolutionary, game-changing progress? When do we get even a glimpse of the robotic intelligence portrayed by Data in Star Trek or David in Prometheus?

There’s a good chance we’ll see it before the decade is out, thanks in part to projects such as the BRAIN (Brain Research through Advancing Innovative Neurotechnologies) initiative. One goal of this initiative is to develop new technologies that can record the activity of the brain at the level of individual neurons.

In contrast, much of our current understanding of the way the brain works is at a gross level, using tools such as functional magnetic resonance imaging (fMRI) that can map oxygen consumption by different regions of the brain.

Assuming the BRAIN initiative is funded, the AI and robotics communities should have much more data to work with. I expect the research findings to provide the basis for new algorithms and computational methods for intelligent robotics. Eventually, robots will become known for their computational prowess, and not simply as fancy R/C drones or replacements for semi-skilled human labor. SV