Fluidic memory and sensing for autonomous soft robots

Soft robots are incredibly promising for many applications, e.g. object handling, bio-medical devices, and exploration of unknown environments. However, an important drawback of fluid-driven soft robots is the need for electronics and hard, bulky components, such as valves and pumps, for their control. This limits their potential in real-life applications where tethers restrain their autonomy and soft-hard interfaces limit their resilience. To overcome these limitations, we remove electronics and hard components by embedding control elements directly in the fluidic circuits.

We use a soft hysteretic valve to cyclically actuate a soft robotic limb. We develop a bi-stable pneumatic capacitor to enable two discrete behaviors for the oscillator: a high-frequency state and a low-frequency state. We then implement a memory circuit that enables switching between the two states upon opening and closing of air channels, both in a long-term and a short-term fashion. To provide useful fluidic inputs to the circuit, we develop a touch sensor based on kinking tubes. Finally, we demonstrate the potential of our approach by embedding the memory and sensing elements on a crawling soft robot. The robot changes direction upon interaction with the environment, exploring unknown spaces autonomously.

Our work shines light on the opportunities given by the integration of non-linear mechanics into fluidic circuits, towards simple yet useful behaviors in soft robots, such as autonomy.