Hacker News new | ask | show | jobs
by Animats 3826 days ago
(http://www.phriends.eu/URAI_08.pdf)

That's a very good paper. There are many advantages to using two opposed springs driven by actuators to simulate muscles. You get muscle-like properties. You get energy storage and recovery. (Humans recover about 70% of energy from muscle springiness when running. Cheetahs, 90%. BigDog, 0%.)

As that paper points out, there are several ways to do this. The cleanest is a double-ended pneumatic cylinder with proportional spool valves at each end able to connect to pressure or exhaust. That was tried on a legged robot at CWRU some years ago. There are schemes with linear springs, string, two motors, and linkages, which tend to be bulky and complex.[1][2] Those work, but are more of a research design than a production mechanism. Somebody will do a better design, probably with rotational springs and no strings.

I once considered a design with two motors, rotational springs, and a differential. One motor controls impedance, the other controls position. If you don't need to change impedance rapidly, which you usually don't, the impedance motor can be much smaller and geared down.

[1] http://mech.vub.ac.be/multibody/topics/maccepa.htm [2] http://www.inacomm2013.ammindia.org/Papers/106-inacomm2013_s...