The upcoming DARPA Robotics Challenge isn’t just starring Boston Dynamics’ PETMAN robot. A small selection of teams from around the world will be funded to the tune of $2M USD over several years to build their own custom platforms for the challenge. The deadline for proposals was in late May, and the teams are now learning whether or not they made the cut.
Unfortunately the European team led by Icarus Technology‘s Davide Faconti wasn’t selected. Having already completed two impressive full-sized humanoid robots called REEM-A and REEM-B, Davide was in a much better position than many teams from the United States to build a working robot for the challenge. I had full confidence in him and wanted to lend a hand if I could. With his permission, I am now able to share the conceptual work I did for his team’s proposal.
As made clear by the challenge breakdown, the robots are required to perform a variety of complex tasks including driving a vehicle, opening doors, climbing ladders, and using tools. Initially the thinking was to use the REEM-C, which I had previously helped conceptualize. Using a combination of my 3D model and some photographs, I did the following renders to show what it would look like doing parts of the challenge.
However, after doing these renderings Davide hypothesized that an ape-like robot might be more capable. This led to the design of a new robot called Silverback. It would have a basic humanoid structure but would be capable of walking on all fours for added stability. Davide provided some simple renderings showing the proportions of the robot, which I then fleshed out.
Davide’s ape-like robot design
The change came as something of a surprise, so I had to reuse some components of my design for the REEM-C in order to make the deadline. However, one area where I put in extra effort was Silverback’s head. I did a number of conceptual sketches (a sample of which you can see in the image below) before I came up with something we were both happy with.
Unlike other humanoids, the head design didn’t have to look friendly – and it required that I fit a Kinect or Prosense sensor, and two stereoscopic camera systems. In hindsight, I thought it might be better to use a set-up similar to the HRP-2 Promet (which hides some of its cameras in helmet protrusions), but time had run out. If I were to return to it now, I would probably play with it looking much more gorilla-like.
Due to the relatively fragile nature of robot manipulators, it would have arm guards extending beyond the hands to take the brunt of the robot’s weight. This could have a spring-loaded suspension mechanism inside it to lessen ground impact. Furthermore, it would have hooks on its arms that could be used to climb ladders.
Another change to the standard humanoid design came in its knee and ankle-foot structures. The knees would have to bend in both directions, allowing it to crawl or kneel. And it would not need as many actuators in its ankles because its foot would be made of flexible (but very strong) carbon fiber planks.
Think something along the lines of the under-actuated springy prosthetics used by amputee sprinters (image right).
Here is a video showing the robot on a turntable:
Sadly this robot will never make the journey from design to reality, but I am still hoping that PAL Robotics will utilize my design for REEM-C (see it in my Illustrated Guide to Designing Humanoid Robots part 1 and part 2). I wish Davide and Icarus Technology the best in the future, and look forward to seeing what the other teams come up with.
I am still open to working with other teams if they would like some help finalizing the look of their robots (contact me using the form here).