Molecular Communication (MC) channels inherently possess significant memory
and nonlinear dynamics due to diffusion and receptor kinetics, often posing
challenges for reliable data transmission. This work reconceptualizes these
intrinsic properties as computational resources by framing a canonical
point-to-point MC channel, consisting of ligand diffusion and reversible
ligand-receptor binding at a spherical receiver, as a physical reservoir
computer (PRC). We utilize the time-varying fraction of bound receptors as the
reservoir’s internal state, employing time-multiplexing to generate
high-dimensional virtual nodes without explicit recurrence. Only a linear
readout layer is trained via ridge regression. Through deterministic mean-field
modeling and particle-based spatial stochastic simulations, we demonstrate the
MC system’s capability for complex temporal processing by successfully
performing next-step prediction on standard chaotic time-series benchmarks
(Mackey-Glass and NARMA10). Performance, quantified by Normalized Root Mean
Square Error (NRMSE), exhibits a non-monotonic dependence on key system
parameters (receptor kinetic rates, diffusion coefficient, transmitter-receiver
distance), revealing optimal operational regimes. These findings validate the
potential of using MC channel as effective and low-complexity computational
substrate.

Cet article explore les excursions dans le temps et leurs implications.

Télécharger PDF: