A Software Test Suite for Data Acquisition and GPU-based Computations on IoT Edge Devices for High Frequency Power Grid Monitoring

This paper discusses the design and testing results of a highly configurable software suite for data acquisition and GPU-based computation suitable for Internet-of-Things devices supporting both the I2C protocol and equipped with GPUs that support CUDA. The IoT device used is the NVIDIA Jetson TX2, which ingests analog sensor data via the I2C protocol. Thus, to ingest data, a development board was developed, and to test the data acquisition and processing pipeline, a suite of software was developed so that automated tests can be performed repeatedly using the same testing parameters. This suite of software coordinates several devices for testing, communicating over TCP/IP. To take advantage of GPU capabilities in the NVIDIA Jetson TX2 platform used in the experiments in this paper, FFT-based analysis was performed using two different algorithms, one exploiting the on-board GPU capabilities. The results show the great potential of these capabilities for edge computations by comparing several FFT algorithms w.r.t. to their processing speed. The analysis from experiments conducted with both CPU and GPU-based algorithms show how IoT devices with GPU capabilities can meet the real time computational requirements posed by high frequency power grid monitoring applications.