Regular followers of the Critical Link blog will know that I’ve been interested in robots since my grad school days at UMass, when I got to work with a robot named Harvey. So, naturally, I enjoyed a recent post by Altera’s Ron Wilson on how robots are taking over. He’s not really worried about robots seizing control and aiming for world domination – at least not yet. No, Ron’s focus is on how robots are influencing development, and how they can impact both your hardware budget and your schedule.
“The issue is that concepts—and with them, design requirements—from the world of robotics are filtering into other kinds of embedded systems. The infiltration seems to follow an identifiable sequence. First, low-cost sensors, often from smart-phone technology, and actuators, such as servo motors from radio-controlled (RC) models, increase the number and complexity of control loops in new designs. Then demands for increasing autonomy gradually pass control of the system from human operators to the system itself: first automating sequences of related actions, then shifting the human-machine interface from actions to goals, then moving toward full autonomy. Think, for example, of a car evolving from a manual to a fully-automatic transmission, to an automated driver-assist system, and on to full self-driving capability. Embedded systems are becoming robots.”
“Obviously, as you move along this path the computing load on the system increases. But by how much? In what algorithms? And how do you provision for these new computing loads?” (Source: Altera)
Ron walks, in considerable detail, through an example of a toy robot: a six-legged apparatus, or hexapod, that resembles a drone. The legs can rotate and bend at their two leg joints. While the example is pretty simple,
“It also conceals a remarkable complexity, which will provide us a rich example of how computing intensity scales with seemingly reasonable changes in requirements.”
The requirements will include the ability to calculate the angles that will let the hexapod take a walk, and I/O to support multiple pulse trains. Factor in that, in real life, the hexapod will need to navigate uneven terrains and step over things, which will add to the computing load and require the ability to construct your own 3D map of the terrain. Enter cameras and machine-vision algorithms.
And so on…More power. A “wireless link to offload much of the computation.”
Ron’s point is that, as robotics become more prevalent in applications, complexity will increase, as well. New architectural models will be required, and assumptions of the past will go out the window. He ends with a warning:
“As robotics infiltrates the rest of the embedded design world, the wise will plan well.”
Well, you can’t argue with that, other than to say that embedded design has always required thoughtful planning.
Anyway, I probably haven’t done justice to Ron’s piece, so I encourage you to go over and read it in its entirety.
Just wanted to point out that Altera is a Critical Link technology partner. Our MitySOM-5CSx combines the Altera Cyclone V System on Chip (SoC), memory subsystems and onboard power supplies. This SoM provides a complete and flexible CPU and FPGA infrastructure for highly-integrated embedded systems.