leveraging compliance

A soft robotic system is composed of soft, compliant materials. Soft robots manipulate fragile objects and tolerate large disturbances from their environments better than their traditional “rigid robot” counterparts.

They’re also a perfect example of the shortcomings of our traditional design practices. Soft structures are governed by non-intuitive mechanics, are challenging to simulate, are fabricated differently than traditional low-DOF mechanical systems, and are ill-represented by traditional component/subassy/assy hierarchies.

As a result, soft robot design requires the development of multi-domain expertise (e.g. advanced computer-aided design , nonlinear simulation, and advanced fabrication) - expensive! to fill this gap, I spent three years writing SoRoForge - an open-source design tool which uses implicit geometry modeling, automatic execution of nonlinear FEA simulations, and 3D printing to rethink our design practices. Read about design philosophy and technical underpinnings of the platform in the IEEE Transactions on Automation Science and Engineering.

geometry descriptions: sculpting with math

Instead of relying on familiar but ultimately fragile CAD representations for complex 3D designs, I invoke computational network representations of implicit geometry functions to describe robot shapes.

The core representations of designs in [SoRoForge](https://github.com/MacCurdyLab/SoRoForge are, in contrast to traditional CAD:

• volumetric and stable as opposed to surface-based and fragile
• accessible to human and artificial design intelligence
• intimately linked spatial discretizations used for simulation

high-performance simulations

The best models strike a balance between two competing, antagonistic goals:

• realism / accuracy
• computational cost

I advocate the use of shell finite elements in simulating pneumatic soft actuators - this method strikes a more favorable balance between the two objectives than traditional volumetric FEA and helps speed up design exploration. Conference publication for details!

forward compatability with automated design

These, accurate simulations enable computational design - the automatic production of mechanical design solutions that satisfy high level constraints. I built a computational design program around these scripted simulations of soft actuators, and asked it to produce strong, flexible soft actuator designs without any manual input.

The results were remarkable - after running overnight, the program identified bending soft actuator structures that look very similar to geometries which took years to develop manually! Read more about this research here.