a seamless workflow for soft robot design and fabrication
What is SoRoForge?
SoRoForge is a suite of software tools written by the Matter Assembly Computation Lab (MACLab) to lower barriers to the design and fabrication of pneumatic soft actuators.
Diagram depicting the seamless flow of data between SoRoForge modules that handle three phases of soft actuator creation: Design, Simulation, and Fabrication. Soft actuator designs originate as interactive graph networks (user-created or loaded from a library), terminate as ready-to-fabricate print files, and may be iteratively simulated and modified along the way. Dotted boxes indicate distinct embodiments of a single design as represented in different SoRoForge modules.
Design
Design geometry in SoRoForge is represented by computational graph networks - very different from the feature trees you may be familiar with as an interactive CAD user!
These networks transform input coordinates in 3D space into scalar values that define whether that point is inside, outside, or on the boundary of a design geometry. SoRoForge features an interactive GUI that enables users to edit graph networks - adding, deleting, and inserting nodes, adding connections between existing nodes, changing bias and weight values, and more.
SoRoForge ships with a library of designs created in our lab, allowing new users to remix our work into their own designs! Check out this student user explaining the features of SoRoForge in this video, and take a tour of the GUI layout below.
SoRoForge graphical user interface for exploring multimaterial soft actuator design space. Exploration of a computational network is performed in the network editing pane A, while a rendering of the resulting actuator design is synchronously updated in the visualization pane B. Users can toggle seamlessly between design, simulation, and fabrication functions (C), while SoRoForge presents relevant information - for example the deformed state of a bending actuator in response to applied pressure, superimposed on the rest state currently visible. Users may also save out, and later load in, native SoRoForge files (D), as well as access a pre-defined library of designs including all those presented in this paper. The simulation in (B) is executed in 18s.
Simulation
Numerical simulation can help us predict the way a soft robot design behaves without needing to fabricate and physically test it. This saves precious time when many design iterations are required to arrive at a suitable design.
Typically, soft robots are simulated using computationally expensive 3D finite element (FE) simulations, requiring soft roboticists to manually convert their designs and set up a simulation model, then wait hours to see results. In SoRoForge, soft actuator designs are simulation-ready by default. With the push of a button, a large-displacement, contact-enabled FE simulation is dispatched to an open source numerical solver and results are retrieved and displayed in the GUI when the analysis terminates.
FE simulation of a soft actuator designed with SoRoForge which displays compound twisting and bending behavior in response to applied internal pressure. A and B give results of simulations executed using tetrahedral volumetric and triangular shell elements respectively. Contours in C and D show local L2 norm of difference in position predicted by shell simulation relative to reference volumetric simulation, using equivalent pressure load to reference simulation and load scaled by 1.6x, respectively. Applying a scaled load reduces normalized error in shell simulations, in this case bringing error to below 10% of actuator length.
SoRoForge leverages shell finite elements to accelerate the execution of simulations, reducing the typical solve time for a model by a factor of 5-10x! For a deeper look at shell finite elements, check out our recent conference paper on the topic, or watch below.
Fabrication
Soft robots have typically been fabricated using conventional manufacturing techniques such as casting, gluing, sewing, and overmolding. These methods are excellent for mass production of consumer products, but slow down design iterations and restrict the geometries soft robots can assume. Why not adopt an emerging fabrication method that releases fabrication constraints and is perfect for rapid prototyping - 3D Printing!
SoRoForge supports direct export of fabrication-ready STL files, the native format used by most 3D printers. SoRoForge provides ready-to-use printing profiles we’ve tuned over the years, allowing us to print complex soft actuator geometries capable of bending, grasping, and even walking!
Additionally, SoRoForge contains extensive test data used to characterize a wide array of inexpensive, commercially available soft 3D printer filaments suitable for fabricating soft actuators.
Characterization
The purpose of SoRoForge is to accelerate the design and fabrication of soft robots - so how do we measure the performance of a soft actuator when it’s finally ready for use?
We argue for the standardization of test metrics, procedures and equipment - and we have open-sourced the design of the Inflation Fixture we use in our own lab to test our own pneumatic soft robots! This general-use test fixture is capable of prescribing pressures, and simultaneously measuring flowrates, pressures, electrical resistances, displacements, and forces.
Our fixture is designed with modularity and hackability in mind - users can reconfigure it to suit their specific needs. Built around Festo pneumatic components and affordable, high performance Labjack data acquisition hardware, this fixture balances flexibility, cost effectiveness, and performance.
For example - check out the data we gathered on this antagonistic pneumatic bending actuator!
Simulated and empirical results for antagonistic bending actuator deformation under pressure loading. Simulation captures actuator behavior well when comparing tip displacement, though shell finite elements overestimate the stiffness of the structure. Computations executed in 42s, error bars indicate 2-sigma variation in measured angle, n=8 trials.
Our Mission
Soft robots offer key advantages over their traditional rigid counterparts. They are inexpensive to manufacture and require many fewer components, are robust to high forces and unexpected deformations, and can interact with their surroundings using a delicate touch. If you’d like a robot that can survive falling down the stairs, pick raspberries without bruising them, or handle delicate undersea organisms, a soft robot could be the answer!
However, soft robot design is challenging and cross-disciplinary. Human designers earn their design intuition through time-consuming design/build/test cycles, placing soft actuator design out of reach for many non-experts. Making matters worse, software currently used to create soft robots is inherited from the world of ’rigid design,’ creating an ill-fitting patchwork of design tools separated by bottlenecks.
We developed SoRoForge as a new paradigm for soft robot design, in order to lower barriers to entry and make soft robot design more accessible to curious experts in other fields. Even better, the SoRoForge workflow is forward-compatible with emerging computational design tools, setting the stage for researchers to blend human- and machine-powered design to create the next generation of soft machines.
