Stitching a sustainable future: 3D knitted robots weave their way to success
18 May 2023
Knitted robots created through additive manufacturing could revolutionise soft robotics, reducing waste and increasing performance.
Image: Harvard John A. Paulson School of Engineering and Applied Sciences
Soft robotics is gaining momentum, offering key advantages over their rigid counterparts, including enhanced safety and versatility. The use of soft materials, with motions powered by inflating and deflating air chambers, enables their safe operation in fragile environments and in close proximity to humans.
Textiles have emerged as a preferred material for constructing soft robots, particularly wearables, but the traditional 'cut and sew' methods of manufacturing have left much to be desired.
A no to 'cut and sew'
Now, researchers at the Harvard John A. Paulson School of Engineering and Applied Sciences (SEAS) have pioneered a new additive manufacturing method for soft robotics which uses 3D knitting. Departing from traditional 'cut and sew' techniques, this innovative approach enables the entire robot to be ‘printed’ through the knitting process.
“The soft robotics community is still in the phase of seeking alternative materials approaches that will enable us to go beyond more classical rigid robot shapes and functions,” says Robert Wood, senior corresponding author on the paper, who is the Harry Lewis and Marlyn McGrath Professor of Engineering and Applied Sciences at SEAS.
“Textiles are appealing since we can radically tune their structural properties by choice of their constituent fibres and how those fibres interact with each other,” Wood says.
“Using ‘cut and sew’ methods, you need to manufacture large sheets of textile material that you then cut into patterns that are assembled by stitching or bonding – and this typically involves a high level of human labour,” says Vanessa Sanchez, first author on the paper and a former PhD student in Wood’s lab.
“Every seam adds costs, and potential points of failure. For manufacturing complex robotic devices, this can be a big challenge.”
The new solution
Sanchez was intrigued by the concept of 3D knitting, which can produce seamless articles of clothing with little material waste. She wondered if the method could be adapted to create textile-based soft robots.
The team acquired a vintage punch card knitting machine and Sanchez connected with knitting experts from the Rhode Island School of Design and Parsons School of Design and Fashion Institute of Technology.
To automate the knitting process, Sanchez and the team also needed to develop software that could direct the knitting equipment – machines often several decades old – to make complex structures out of various types of yarns.
“In one instance, I had to trick the machinery – using a software program – into thinking that my computer was a floppy disk,” Sanchez says. After initial experiments were promising the team moved to a more modern, automated machine.
James McCann, an Assistant Professor at the Carnegie Mellon Robotics Institute, collaborated on the software. “The team wanted to develop and characterise a wide range of soft actuators – they weren't just building one pattern, they were building a whole set of parametric patterns,” McCann says.
“This is hard to do with traditional knitting design software, which is generally focused on developing single outputs by hand instead of easily adjustable parametric families of outputs.”
The team devised a workaround by describing 3D patterns using a "knitout" file format – a knitting description written in general-purpose programming languages. They then developed code to translate those knitout descriptions to run on their desired knitting machine.
“The cool thing about developing parametric patterns in a generic knitting format like knitout is that other groups with different types of knitting machines can use and build on the same patterns, without extensive translation effort,” McCann says.
Through conducting a series of experiments, Sanchez and her collaborators developed, for the first time, an extensive library of knowledge centring on how various knitting parameters impact the mechanical properties of the resulting material.
Testing 20 different combinations of yarn, structure, and more, the team characterised the effects of varied knit architectures on folding, unfolding, structural geometry, and tensile properties.
Using combinations of these structures, they demonstrated many different knit robot prototypes, including various types of gripper devices with bending and grasping appendages, a multi-chamber claw, an inchworm-like robot, and a snake-like actuator capable of picking up objects much heavier than the device itself.
“We wanted to create a library for engineers to draw from, to develop a variety of soft robots, so we characterised the mechanical properties of many different knits,” Sanchez says.
“3D knitting is a new way of thinking about additive manufacturing, about how to make things that could be reconfigured or redeployed. There are already industrial machines to support this type of manufacturing – with this initial step, we think our approach can scale and translate out of the lab.”
“I envision that programmable textiles will have a similar impact on how soft robots are made as fibre-reinforced composites have had on the construction of high-performance aircraft and automobiles,” Wood says.