Computational cost and implementation analysis of a wavelet-based edge computing method for energy-harvesting industrial IoT sensors

Abstract

The rapid advancement of Industrial Internet of Things (IIoT) has heightened the need for efficient data processing and transmission, particularly in energy-constrained environments. This study introduces a novel wavelet-based edge computing methodology designed specifically for low-power IIoT sensors using energy harvesting. Unlike existing implementations that rely on computationally complex instructions, this approach optimizes the wavelet transform (WT) for resource-limited microcontrollers (MCUs) without sacrificing data quality. By leveraging a lightweight assembly-level WT implementation, the proposed solution significantly reduces computational costs and energy consumption. A comprehensive analysis performed on ARM Cortex-M7 MCU on an industrial vibration dataset demonstrates energy savings of assembly language (ASM) up to 87% with discrete wavelet transforms (DWT) and 32.1% with fast wavelet transforms (FWT), compared to C-based implementations. This work is distinct in its ability to dynamically adjust data transmission levels based on available energy, ensuring robust operation in batteryless IIoT environments. Moreover, the method offers flexibility in signal reconstruction, supporting scalable compression ratios and facilitating long-term predictive maintenance applications, making it a pioneering step in sustainable industrial monitoring.