• agriculture
• localization
• mapping
• mobile robots
• RTK GNSS
• SLAM

This thesis work proposes a new 6D Simultaneous Localization And Mapping Algorithm
(SLAM) in an agricultural environment, which aims to localize a mobile robot, and, at
the same time, recreate a map of the environment. The environment considered is a
vineyard characterized by steep slopes and uneven ground.
Agriculture is one of the elds on which more research is being focusing in recent
years, with a particular interest in making one of the heaviest workplaces for the human
body more automated.
To enable a vehicle to move autonomously in an environment, it is essential that it be
able to locate itself and have conception of its surroundings. Agricultural environments,
compared to indoor or city environments, are highly unstructured. In particular, the
vehicle will have to deal with environments that are not
at, but rather with steep
slopes, and uneven terrain, which introduce vibrations that make classic dead reckoning
systems unusable.
To overcome this limitation, in this thesis work I developed a multi GNSS antennas
system to estimate position and orientation of the robot. Investigating how many antennas
to have to use, with what technology and with what geometry to arrange them on
the vehicle.
However, the GNSS system alone cannot be relied upon for localization, as it may not
always be available. Therefore, the multi-GNSS systems t into a larger SLAM algorithm
that aims to recognize the presence of vine trees and through them recreate a map of
the vineyard itself. Through the use of the latter, the vehicle will be able to locate itself
within the vineyard even in the absence of the GNSS signal.
The introduced system was designed to be used on the Robotnik Summit-XL Steel
and developed on ROS. For the testing of the system, I performed simulations on Gazebo,
evaluating the performance of the complete system, but also its behavior in case of GNSS
signal loss. Finally, I tested the algorithm on the real robot in a real outdoor environment.