Physical Reservoir Computing Using an Electret-Based MEMS Vibrational Energy Harvester

This study explores the potential of an electret-based MEMS vibrational energy harvester as a physical reservoir for physical reservoir computing (PRC). An equivalent circuit model representing the electromechanical behavior of the device was implemented using an electrical circuit simulator, and PRC operations were evaluated by applying acceleration inputs. Simulation results demonstrated that the proposed device inherently exhibits nonlinear dynamics and memory effects, enabling it to perform basic computational tasks and indicating its suitability as a physical reservoir. Furthermore, experimental evaluations were conducted using an actual MEMS vibrational energy harvester. Without employing any external feedback circuits, the device successfully solved parity tasks, confirming its practical computational capability. These results validate that the electret-based MEMS vibrational energy harvester intrinsically possesses the essential properties required for PRC, including nonlinearity, fading memory, and high-dimensional state responses. By integrating energy harvesting and computation within a single physical system, the proposed approach provides a promising pathway toward low-power, event-driven edge computing devices, particularly for autonomous sensing applications where both energy efficiency and on-site processing are critical.