Viscoelastic mechano-bits for increased storage density and in-memory mechanical computation

In-memory computing, in which memory and computation are collocated, offers a potential approach to address some of the current challenges with mechanical computing systems. However, purely mechanical embodiments enabling simple encoding of physical signals, in-memory computation, better scaling of mechanical elements for computing units, and, crucially, low overhead information decoding remain rare to date. We design temporal coding-inspired viscoelastic mechano-bits capable of exhibiting temporal decay and, consequently, rank order coding, the output of which can be directly read out (decoded) by their mechanical stiffness. Leveraging the non-abelian behavior that arises from rank order coding, we design a mechanical memory system that can exhibit order-dependent behavior. This system can experimentally realize non-volatile, order-dependent memory, thereby expanding the available mechanical storage by factorial scaling compared to the exponential scaling associated with combinational logics. The mechanical memory is encoded using bistable mechano-bit state change and decoded using global stiffness change, therefore enabling simple decoding from force measurements. We derive the limits on the maximum available memory storage in an elastic block of bits and establish the fundamental boundaries on the memory expansion of viscoelastic blocks of bits. The introduced temporal coding and improved information processing scaling both promise to significantly expand the capability for realizing physical intelligence using the common bistable unit motif commonly used in mechanical metamaterials.