Speeding up System Identification Algorithms on a Parallel RISC-V MCU for Fast Near-Sensor Vibration Diagnostic

Structural Health Monitoring (SHM) systems currently utilize a combination of low-cost, low-energy sensors and processing units to monitor the conditions of target facilities. However, utilizing a dense deployment of sensors generates a significant volume of data that must be transmitted to the cloud, requiring high bandwidth and consuming substantial power, particularly when using wireless protocols. To optimize the energy budget of the monitoring system, it is crucial to reduce the size of the raw data near the sensors at the edge. However, existing compression techniques at the edge suffer from a trade-off between compression and accuracy and long latency resulting in high energy consumption. This work addresses these limitations by introducing a parallelized version of an unconventional data reduction method suited for vibration analysis based on System Identification models. Our approach leverages the unique capabilities of GAP9, a multi-core RISC-V MCU based on the parallel ultra-low power (PULP) architecture. Compared to the sequential implementation, we achieve a maximum execution time reduction of ≈ 60 × and power consumption of just 48.3mW while preserving the spectral accuracy of the models.