Conceptual design of an autonomous rover with ground penetrating radar: application in characterizing soils using deep learning

In the pursuit to make agricultural production efficient, the earliest farmers used data in the form of notes of observations. In the current age of data, it has become easier to collect data over a wide spectrum of parameters. There are numerous sensing technologies for measuring processes on and over the surface of a field, typically mounted on satellites, aerial vehicles (drones), ground vehicle and static platforms. Recently, soil has been gaining increasing attention and recognition for its significance in not only increasing productivity but also stabilizing the environment. However, characterizing soil in a field is not trivial, especially when required to access the deeper layers and quantifying the essential contents – water, nutrients and organic matter. This paper presents a short review of applications of ground penetrating radars (GPR) in measuring soil content and structure. The focus is on deep learning constructs that have been used for interpreting and establishing correlations in the characterization of soils. The review serves to inform design considerations for a planned autonomous rover that will be used for surveying field soils in the Satakunta region of Finland. After the review, a conceptual design of an autonomous GPR rover is presented.