Yale researchers have found a way to make flowing fluids remember previous events by using tiny flexible fibers that change shape under pressure. The discovery could help engineers develop soft robots that process information without traditional electronic controls.
The idea turns a well-known technical problem on its head. Most fluid systems respond to the conditions they are exposed to, but the Yale design can retain traces of previous flows. Its physical components effectively store information, so past events influence how the system behaves next.
Fibers that remember
In a microfluidics chip, researchers placed flexible fibers in a network of narrow channels. Each channel is about half the width of a human hair in depth.
As fluid moves through the channels, pressure pushes individual fibers between two stable configurations. Each position changes the path available to the fluid, creating a physical record of the system’s previous state.
Engineers could represent the two configurations as binary digits or zeros and ones. Unlike traditional computer memories, the arrangement does not require any electronic components to switch between states.
Abhineet Rajput, the study’s lead author and a graduate student in Professor Amir Pahlavan’s lab, explained the difference. “While traditional microfluidics relies on active elements, this system uses microfibers to passively store information,” Rajput said.
The fibers can also influence each other. Through their interactions, individual changes can influence the behavior of the larger network, creating a more complex form of physical memory.
Brain made of liquid
The research could open up another avenue for controlling soft robots. Such machines often have to bend, squeeze through tight spaces, or operate in environments that challenge traditional hardware.
A fluid controller could use pressure and flow rates to transmit signals. Channels would connect different parts of the system, while the flexible fibers would regulate fluid movement and store information.
Pahlavan sees particular value in environments where electronics reach their limits. “We could design a fully fluid controller that could be integrated into a soft robot and effectively act as its brain,” said Pahlavan, an assistant professor of mechanical engineering at Yale.
Electromagnetic interference can cause malfunctions in electronic systems in certain operating environments. A controller based on fluid motion could offer an alternative, although researchers have yet to show how reliably it can control a robot through practical tasks.
Can the system learn?
Memory is only part of the researchers’ ambitions. The next question is whether engineers can train fluid networks to elicit specific responses.
Training could change how a network responds to incoming data streams and give it the opportunity to adapt its behavior. Achieving this would extend the system beyond the storage of prior information to include physical learning.
The same principle could be of interest to biologists studying organisms that navigate their environments without brains. For example, slime molds can find efficient paths through mazes. Researchers also want to understand how fungi and plant roots respond to environmental changes.
Fluidic networks could provide a physical model for studying how interactions and accumulated experiences influence behavior. Whether the technology can ultimately learn useful tasks remains an open question.
The results appear in the journal Physical Examination X.
