ML-based Model for Autonomous analysis of Spectral Data.

We propose to develop the foundation of a Machine Learning (ML) model for autonomous interpretation of spectroscopic data, and demonstrate in a cloud-based application for interpretation mineralogical data. Our goal is to demonstrate a tool that can process a range of analytical techniques with a high degree of automation and performance that rival that of expert users with conventional analytical software. The automation and ease of use will enable automated analysis of large quantities of data, allow non-experts to extract valuable high-level scientific products from raw data, and empower experts with a new approach to data analysis. The model demonstrated with this effort will provide the base on which methods for automated analysis spectroscopic data can be developed for implementation in autonomous rovers and spacecrafts.While our proposed approach can be ndash;and will bendash; extended to more analytical techniques, we are focusing our current development on two methods: X-ray diffraction (XRD), a well-established technique for identification and quantification of crystalline materials, currently deployed on Mars in Curiosity, and Raman spectroscopy, a more recent method that has shown increasing popularity over the last decade and that will be deployed in upcoming Mars missions including Mars 2020. XRD and Raman provide two different case studies on which we will ultimately develop a technique-agnostic analytical tool.