Soft Robots

I’ve been working on this project for a few months, focused on changing how soft robots get designed and made. Traditionally these robots are complex to design and build, and they require and unexpectedly large amount of hand labor to stitch together. This ends up with parts being produced slowly, with small deviations from known working designs. I’ve been trying to come up with a method that allows you to design a robot in CAD, queue up the design on a powder printer, cast silicone into the printed mold, and pull out a working robot. The idea is to allow for a huge variety of geometry, experimentation, and prototypes that are quick and inexpensive to produce. I want to make the process a whole lot more like a scientific experiment, where you test and observe multiple samples while adjusting a single variable.

I’ve been focusing on these tapered tentacle shapes because they’re pretty easy to model and mold, but there’s some more complex and exciting geometry coming out of the machines all the time. Here’s a list of posts on the project, including information on the methods, techniques, and materials involved:

You can also find a running feed of photo documentation from the project here. I’ve also uploaded a number of the major revisions to Thingiverse.

12 thoughts on “Soft Robots

  1. Hi Matt,

    Nice website and documentation. I have been very interested in soft robots using silicone and SMA or pneumatics and am really excited you’re sharing your work. I was wondering if you’ve worked on position sensing and control of the tentacles? I am interested in how these designs can be used for manipulation and locomotion.


    1. Howdy,
      Thanks for the complements. I haven’t explored any sensors or closed loop controls. I’ve only been doing control from the pneumatics end, using solenoid valves to meter out how much air gets into the robots. I do have some ideas for control schemes but am mainly focusing on how to build these kinds of robots quickly and cheaply rather than things like controlling them and making them more durable. It’s a lot of work, but the advantage is the field is wide open for people who want to make big breakthroughs in the research.

      I wrote a little on how you could possibly control a soft robot at the bottom of this post –


      1. Hi Matt,
        Idea: instead of thinking of pressures as being binary, on or off, why not use a base three pressure grade? Something like parking the needle of a hard disc in the middle… you would have pressure, neutral, vacuum. That would let you both contract AND expand your muscles with the same controller. Or? For instance, if you pressurize that suction cup *before* gripping the object, then simply by releasing the pressure you would activate it, and hold it until upping the pressure to release. Just an idea… I enjoy your work and blog!

        1. Jim,
          It’s a neat idea. I’ve been playing with geometries that might collapse elegantly when vacuum is applied. Right now I’m aimed at getting some reliable mechanisms out of the manufacturing methods I’ve got available. This has meant a lot of modeling, but not too much experimentation with the control systems. If you see any experiments with vacuum pressure as a method for moving soft actuators, or examples from nature that do the same, please send them my way.

          Thanks for the kind words,

Leave a Reply to Mary Wu Cancel reply

Your email address will not be published. Required fields are marked *