Analysis of algorithms for autonomous driving of a telescopic handler

Autonomous vehicles have started revolutionizing the automotive industry about a decade ago, and are now starting to revolutionize the construction and agricultural fields. Construction and agriculture presents different challenges with respect to the automotive counterpart, some of them given by the environment, others given by the nature of the vehicle at study. The aim of this thesis is to analyze the sensors and the sequence of algorithms that are needed for AV applications, and apply them to the vehicle at study, the Merlo telescopic handler, in order to understand and highlight the challenges posed by this kind of vehicle in an agricultural environment. The analysis has been carried out on four different fields: path planning, Lidar data clustering, obstacle representation, navigation. Particular attention has been given to the clustering algorithms, in which both well established solutions and a very recent algorithm have been analyzed and implemented. The clustering and navigation algorithms have been tested using two different datasets, one belonging to the construction field and the other to the urban environment: the lack of freely available agricultural Lidar datasets is still a difficult point in the development of AV for agriculture. The challenges posed by the arm of the telescopic handler have been analyzed and visualized using the same datasets as above.