The Glaucus is a soft robotic quadruped composed of a single seamless silicone part. It has a complex network of interior channels, created via a lost wax process, that turn into actuators when pressurized with air. It’s able to walk with a diagonal gait, similar to a gecko or Glaucus Atlanticus sea slug, using only two input channels.

The Glaucus was created to demonstrate a method for fabricating soft robots of nearly any geometry with arbitrary interior structures. It’s been my goal, since beginning my research into soft robotics, to simplify the process of prototyping and refining designs. Often the barrier between an interesting bench prototype and practical application is how it scales into production. If methods for experimenting with the core concepts, evaluating them in a context that represents their final manufactured state, and refining them for mass production don’t exist, the idea is very likely to languish on the bench.

Many of our most transformative technologies were enabled by mass manufacture. Computation was transformed by the microchip, not just because it lowered the cost of computing a single byte, but because the process of manufacture was robust enough to enable experimentation and optimization.

The process for creating a Glaucus is straightforward. The shape is modeled, including the interior channels, in SolidWorks. The exterior solids are used to form a cavity for two-part molds. The same is done with the interior channels. Those mold components are then subtracted from a slightly larger block to form what I have been calling a bathtub mold. I print the bathtub molds on a FDM printer and cast them in a molding silicone (selecting compatible silicones for the mold and final robot).

The interior channel molds get filled with wax. When the wax cools the components are cleaned and assembled into a support jig. The first exterior mold gets aligned and filled with silicone. When the silicone reaches its gel phase the wax becomes embedded and the support jig can be replaced with the second exterior mold. This is then filled with silicone. If you are careful with timing the two silicone pours chemically bond with one another and become a monolithic silicone slab.

Once the silicone cures the whole part is heated and the wax is removed. I use a hot oil bath these days, as the wax dissolves into the oil and the residue is easy to remove with dish soap. From there fasteners are attached to the air inlets and the robot is ready to take its first steps. You can find a more complete description of the process and the files to print and build your own here.

This whole process leaves nearly every crucial precision element to a robot (the 3d printers) or a passive process (like pouring silicone or wax). Changing designs requires a fairly minimal time investment if you have open access to 3d printers. The process can also be easily scaled up to producing multiple identical robots using the same process, or tooled for mass manufacture using HTV silicone.

What remains to optimize the process is design validation through simulation. There aren’t currently any simple tools for evaluating the behavior of a soft robot when pressure is applied. Still, this process offers fine-grain control over the interior structure and wall thickness, so optimization through iteration is not only possible – it’s how I got the Glaucus to walk.

You can find photo documentation of the project on Flickr, video of it in action on Youtube.