OMC community member Duncan Haldane has built the world’s greatest jumping robot. No, seriously. Haldane’s bot is named Salto (short for “saltatorial locomotion on terrain obstacles”), weighs 100 grams, is 26 centimeters tall when fully extended, and can jump to 1.007 meters. Impressive, to say the least.
Salto is appearing on the cover of Science Robotics, the newest journal from Science, the peer-reviewed academic journal of the American Association for the Advancement of Science (AAAS), which has been in production since 1880 and is one of the most respected science journals in the world.
Haldane is no beginner to robotics. He’s currently a working on his PhD in robotics at UC Berkeley and is a National Science Foundation (NSF) fellow as well as a National Defense Science and Engineering Graduate (NDSEG) awardee. For the Salto design, Haldane drew inspiration from nature, specifically a small nocturnal primate called the lesser galago (Galago senegalensis), otherwise known as a bushbaby, known for being particularly adept at leaping. (This Smithsonian video does a great job of explaining just how their bodies are built specifically for jumping.)
What's special about Salto is that it can perform huge jumps continuously, which enables it to do new things like dynamically jump off walls to push itself higher. Haldane’s bot is almost entirely carbon fiber, but he made the joints out of small Igus polymer bushings and used molded polyurethane pieces to connect everything together.
He employed the Othermill to cut the honeycomb carbon-fiber composite plates for the body and leg structure, mill wax positives for molding the polyurethane pieces, and make a little breakout board that the motor encoder IC (an AS5047P from Ams) is mounted on.
We spoke with Haldane to learn more about the extensive process that went into creating such a remarkable and potentially game-changing build. On this project, he collaborated with Justin Yim, Mark Plecnik, and their advisor Ron Fearing.
What was your R&D process?
Lengthy. I started off by nailing down a performance metric for robot and animal jumpers to figure out which systems we should be looking to for inspiration. We found that the galago jumps twice as well as any robot that came before ours. We looked to some biomechanical analyses and extracted the general principle that lets the galago jump that well. Turns out they have this super-crouch posture that lets them load a whole bunch of energy into their stretchy tendons and then release that energy explosively.
We took that idea and made a leg mechanism that can perform that behavior using a single motor. That was hard. We actually came up with a new way to synthesize linkages in the process. We basically specified that we want a straight-line motion, created a million-degree polynomial function whose solutions are linkages that accomplish that motion, and optimized those solutions for things like compactness, low internal forces, and inertial balancing. We have a detailed paper on our linkages being published in an upcoming issue of the Journal of Mechanisms and Robotics.
Then came the prototyping process. I started with a laser-cut acrylic version of the mechanism, banged together with rivets for joints. That validated our ideas, so the next version was made out of prefabricated carbon-fiber components with Igus bushings, as well as some aluminum tube and molded polyurethane. The final version replaced the aluminum tube with a carbon-fiber tube made for high-performance kites. I also made my own geared brushless motor using a custom-wound drone motor and some quality time in the machine shop.
What issues did you run into?
Mostly the robot exploding itself. It's packing a whole lot of force for something this small. For reference, the final leg mechanism weighs 11 grams and has a peak internal force of 200N, which is like 185,000% of its own weight.
Another hard part was getting the right spring. It has to store a whole lot of energy in basically no mass and be packed very tightly. I ended up going with a torsional latex spring, like the one that HEBI Robotics uses.
We also had to make our own motor driver for the brushless motor because nothing that small existed.
How did you troubleshoot the issues?
We've nailed down most of the issues and are up and jumping. It was just trial and error for a few months as we iterated mechanism designs (in simulation mostly) and the material system used to build the robot. The final version of the leg mechanism was “42-j-u-12,” with each of those being an iterator on the design (as in version number 22.214.171.124), to give you an idea of the number of iterations we tried.
What did you learn?
With the robot build, we've proven to ourselves that a jumper like this is possible. We also found that you can use it to do new things, like the wall-jump.
What are you going to do with the project next?
Next up, we'll be getting the robot to chain together longer series of jumps, bouncing around the environment to get to spots that couldn't be reached before.
Special thanks to Duncan Haldane, Science Robotics, UC Berkeley, the National Science Foundation (NSF), and the National Defense Science and Engineering Graduate (NDSEG) program for sharing this work with the world.
If you're interested in checking out parts made on the Othermill and learning more about how the Othermill is being used to accelerate research and engineering in other universities and businesses, request a sample.