Micro-meta-structures for computational sensors with built-in memory

Introduction

Sensory input in integrated systems is expected to increase with the entrance of AI and Internet-of-Things, requiring systems to become efficient and autonomous. The proposed research aims to introduce and study a new type of smart structure, dubbed meta-structures (MS), composed of repeating a unit cell to create a structure with new abilities such as multistability, non-volatility, and configurability.

Applications of Meta-Structures

Such structures can be used to design autonomous sensors with built-in memory and computational abilities, allowing the formation of a new class of smart micro-electromechanical systems (MEMS) with edge computation and in-memory programming (IMP).

In the aggregate, such smart sensors can lessen the dependency on a CPU and increase the autonomy of an overall system, while enabling distributed and parallel computations.

Limitations of Current MEMS

Current MEMS-based structures are mono- or bistable, and as such are limited to registering one or two values in a sensor/mechanical memory/logical gate. However, recent studies have shown that an MS can break free from a two-bit structure.

Breakthrough in Multi-Valued Structures

Indeed, in a recent breakthrough, we have shown that in the presence of electrostatic actuation, a micro-MS becomes multi-valued, with three stable equilibria. This discovery opens a gateway to a paradigm shift that goes beyond the study of new structures, to the formation of a new class of MEMS.

Future of MEMS with Mechanical-Based Computation

This new class of MEMS will be able to incorporate mechanical-based computation with IMP capabilities. Such an unconventional approach has the potential to augment traditional capabilities, introducing new abilities such as:

• Reduced leakage and power consumption
• Reconfigurability
• Decreased footprint
• Compatibility with harsh environments (i.e., high temperatures or electromagnetic radiation)
• Reversible computing