Successful technological innovations typically follow a path in which they start with a somewhat crude initial implementation that is refined over time. The computer, the automobile, the airplane — the same basic story plays out time and again. We are in the midst of that process playing out yet again in the world of 3D printing. Early 3D printers produced rough-hewn objects, and as technological advancements piled up on top of one another, those prints became much better and far more detailed. Now that print quality has reached such a high level, new frontiers are being explored. One of these is the development of programmable materials.
In contrast to traditional, static 3D prints, programmable materials can incorporate motion, perception, and structure all in a single object. Current approaches to create these materials rely on complex, error-prone processes and multimaterial prints. That complexity may be a thing of the past, however, thanks to the recent work of a team of researchers at MIT. They have developed structures with networks of sensors directly incorporated into them that can be produced by a single run of a 3D printer, and with just a single material.
The team developed a method to 3D print a lattice structure that contains networks of air-filled channels. As the lattice is deformed by being squeezed, bent, or stretched, the pressure of the air in these channels changes. By monitoring these pressure changes, it can be determined exactly how the structure is moving. These properties allow sensors to be added only to the outside of the structure, but allows those sensors to gain insights about what is happening throughout the material. Traditional approaches would require sensors to be embedded throughout the structure, which is technically much more difficult to engineer.
The channels were created using digital light processing 3D printing in which the structure is made from a pool of resin. The resin is hardened into a precise shape by light activation — a specific pattern of light is projected to define that shape. The process does require a bit of hand-holding at present, however. As the resin hardens, a residue remains trapped in the sensor channels that has to be manually flushed out and intricately cleaned.
The researchers tested out their technique using handed shearing auxetics (HSA) to create a soft robot. HSAs have the ability to be able to be twisted and stretched at the same time, which makes them useful as actuators in soft robotics, but these same properties make them very difficult to sensorize with traditional approaches. After 3D printing their robot and collecting data for 18 hours to train a neural network to recognize the robot’s motions, the team ran some tests. They were very excited by the results that they saw — in fact, they had difficulty even distinguishing the ground-truth from the neural network’s predictions.
With these successes in the rearview mirror, the researchers are now looking towards some next steps to push this technology further forward in the future. They are exploring the idea of creating human-machine interfaces or soft devices that have sensing capabilities built into their structure. The work outlined in this research has a lot of potential to make very complex robots, and other devices, with onboard sensing in a very simple manner.