Pneumatic memristor enabled by elastic multistabilities

Abstract

• •
  Introduction of the first pneumatic memristor integrating memory and flow regulation
• •
  History-dependent, non-volatile mechanical memory enabled by elastic bistability
• •
  Programmable pneumatic logic gates enabling non-von Neumann mechanical computation

Mechanical intelligence refers to the ability of a mechanical system or device to process and respond to its environment autonomously through its physical structure and materials, without relying heavily on electronic computers or software control. Such systems are attracting growing attention due to their potential for enhanced robustness, faster response, and superior material adaptability. A critical capability of an intelligent mechanical system is the capacity to process and learn from past stimuli, which requires both memory and in-material computation. In this context, memristors, recognized as the missing circuit element in electronics, serve not only as memory units but also as computation elements, forming the foundation for non-von Neumann architectures that integrate data storage and processing within the same physical medium. Here, inspired by the electronic memristor, we present the first pneumatic memristor, featuring a bistable beam that alters its configuration in response to previous pressure inputs and can be fully reset. Experiments and modeling confirm that the device “remembers” previous pressure inputs and subsequently adjusts airway resistance autonomously. Moreover, by assembling pneumatic memristors into programmable logic gates, we demonstrate the feasibility of mechanical information processing directly within embodied soft hardware. This design advances the development of compact, intelligent, and electronically independent architectures for mechanically intelligent systems, paving the way toward fully integrated mechanical computation.

Mechanical intelligence enables processing and response through mechanical structures without electronic control. Such systems are gaining interest for their robustness, rapid response, and material adaptability. A critical requirement for mechanical intelligence is the ability to process and learn from past stimuli, which demands memory and in-material computation. Memristors provide this functionality by integrating data storage and processing within a non-von Neumann architecture. Here, inspired by electronic memristors, we present the first pneumatic memristor, which uses internal bistable beams to alter its configuration based on prior pressure inputs and can be fully reset. Experiments confirm its memory behavior, and assembling these units into programmable logic gates demonstrates mechanical information processing without electronics.

Mechanical intelligence enables processing and response through mechanical structures without electronic control. Such systems are gaining interest for their robustness, rapid response, and material adaptability. A critical requirement for mechanical intelligence is the ability to process and learn from past stimuli, which demands memory and in-material computation. Memristors provide this functionality by integrating data storage and processing within a non-von Neumann architecture. Here, inspired by electronic memristors, we present the first pneumatic memristor, which uses internal bistable beams to alter its configuration based on prior pressure inputs and can be fully reset. Experiments confirm its memory behavior, and assembling these units into programmable logic gates demonstrates mechanical information processing without electronics.

1. Download: Download high-res image (148KB)
2. Download: Download full-size image