Here’s another example of some nice springy robots research to follow on from my little rant the other day. As well as looking cool and running fast, one of the vague notions I have about springy robots is you should be able to use that springyness to increase your energy efficiency. Springs can store and release energy, and they do it pretty damn well when they are extending and compressing at the right frequency – i.e. the resonant frequency. If you push a spring at this frequency you won’t be fighting against the energy being released from your previous pushes, so you tend to get an ever growing bouncing motion. It should be possible to make use of this in a robot, though it might be a bit of a challenge.
So here’s a nice recent attempt at something like this: EduBot is a robot that is designed to have controllable springyness in its legs. The nice thing about it is that it can adapt the amount of compliance for difference terrains.
It’s got arc shaped legs with adjustable sliders that set the compliance. It then, apparently, has various adaptive strategies to set the compliance and gait for the best efficiency on a given terrain. The details are in a conference paper from the recent International Conference on Robotics and Automation, but I don’t think the proceedings have been published yet and I wasn’t at the conference, so I haven’t read it. I’m getting my info from the brief writeup over at the excellent IEEE robotics blog.
So, again without having read the paper, I can tell you that apparently the robot was generally most efficient with stiffer legs except on certain terrains. To be honest, this isn’t that surprising. The robot’s legs are pretty “wheel” shaped, and we know that stiff wheels are generally more efficient, so long as the running surface is relatively smooth. That’s why train wheels are made of steel (because that’s great if you’re on a nice flat piece of metal), but a road car needs to have a bit of softness in the tires to deal with the bumps. It’s called rolling resistance, but if you need proof, try deflating your bicycle tyres a bit and comparing how much harder it is to pedal.
Maybe I’ll look at this again once I find the paper. But I wonder if what’s going on is that the robot is suffering from too much rolling resistance and isn’t able to make use of the resonant bouncing effect. There is an example of a nice bouncy looking motion at about 1:50 into the video, but a lot of the other gaits look less smooth. Anyway, nice robot.