The population growth, and the consequent raise in food demand, are worldwide problems which are strictly related to the capacity of agricultural systems to withstand these increasing requests, despite climate changes and the lack of manpower. In the last decades, nations, companies and research groups are facing this problem, with the aim of introducing new techniques to optimize and preserve the available resources, as well as developing new tools aimed at increasing productivity and efficiency of human labor. Indeed, Precision Agriculture merges agronomy and farming techniques to data science and robotics in order to support and improve farmers' decision making and the whole production process. Mobile robotics can be the turnkey to bridge automation and agricultural worlds. In fact, farmers still carry out a large part of the operations manually, in particular within orchards and greenhouses. For orchard agricultural robots to become established, it is necessary to be able to navigate autonomously through harsh environments, perceiving the surrounding space and adapt to it, while carrying the required tools for performing tasks such as spraying, or thinning. Motivated by the previous framework, this thesis focuses on the design and assembling of an Unmanned Ground Vehicle for precision agriculture specifically tailored for orchards, greenhouses, and horticulture. In particular, the contributions of this thesis are multiple, from on hand mechatronic design of the vehicle is presented. Moreover, we developed an optimized model-based design tool for tracked vehicles, which leverages their dynamic models in order to drive the sizing procedure. In the end, we describe how autonomous navigation is achieved, proposing novel algorithms for GPS-free in-row localization and navigation, exploiting the semi-structured nature of the environment. We also report how we improved the path-following performance of tracked vehicles, designing an adaptive observer for slip estimation and compensation.
R Tazzari (Thu,) studied this question.