An FPGA-Based Solution for Computing a Local Stationarity Measure From Satellite Data

The analysis of data variability from in-situ observations is essential for scientists and space mission controllers. Given the limited resources available on-board a spacecraft as well as the presence of the Field-Programmable Gate Arrays (FPGA) devices in modern spacecraft architectures, an efficient real-time monitoring solution should be deployed on these devices to use minimal computational and energy resources, and to reduce the main on-board computer utilization, thus making it available for other tasks. This paper describes the implementation of an algorithm for computing a local stationarity measure (LSM) on FPGA devices. The algorithm tests weak stationarity from the convergence of the partial means of the signal computed on subsets of increasing length, compared to the overall mean of the signal over a fixed-length running window; the window spans the entire signal. The algorithm is designed for an on-board implementation which monitors and detects changes of variables measured in-situ by scientific instruments (e.g., magnetometers). The design was tested with synthetic and real-time signals and provides results in very good agreement with a dedicated data analysis library specifically designed for the analysis of satellite data.

Automated summary

This paper describes the implementation of an algorithm for computing a local stationarity measure on FPGA devices, which efficiently monitors and detects changes in in-situ observation data, providing results in good agreement with a dedicated data analysis library.