Path to autonomous soil sampling and analysis by ground-based robots

Good site characterization is essential for the selection of remediation alternatives for impacted soils. The value of site characterization is critically dependent on the quality and quantity of the data collected. Current methods for characterizing impacted soils rely on expensive manual sample collection and off-site analysis. However, recent advances in terrestrial robotics and artificial intelligence offer a potentially revolutionary set of tools and methods that will help to autonomously explore natural environments, select sample locations with the highest value of information, extract samples, and analyze the data in real-time without exposing humans to potentially hazardous conditions. A fundamental challenge to realizing this potential is determining how to design an autonomous system for a given investigation with many, and often conflicting design criteria. This work presents a novel design methodology to navigate these criteria. Specifically, this methodology breaks the system into four components - sensing, sampling, mobility, and autonomy - and connects design variables to the investigation objectives and constraints. These connections are established for each component through a survey of existing technology, discussion of key technical challenges, and highlighting conditions where generality can promote multi-application deployment. An illustrative example of this design process is presented for the development and deployment of a robotic platform characterizing salt-impacted oil & gas reserve pits. After calibration, the relationship between the in situ robot chloride measurements and laboratory-based chloride measurements had a good linear relationship (R²-value = 0.861) and statistical significance (p-value = 0.003).