On October 13^th, 2022 at 11.30 am Solomon Pizzocaro, PHD Student in Information Technology, will hold a seminar on "Toward autonomous vineyard operations: development and in-field testing of a self-driving tractor" in DEIB Conference Room.

Agricultural robotics promises to bring high efficiency and accuracy to the management of crops and farms. Due to the long tradition and economic relevance of the wine industry in Europe, most of the interest is for applications in vineyards. The state of the art are vehicles capable of traversing the rows and/or performing headland turns autonomously.
In general, the controllers for these two tasks are not the same and don’t manage obstacles; in addition, the problem of reaching the first row is little addressed.

In this work, we propose a single controller that, given a path, drives the vehicle inside and outside the rows taking into account obstacles and crops vegetation. The proposed solution is framed on top of the ROS Navigation Stack, which makes it highly replicable for other applications. Path tracking is done on a local cost-map composed of two layers: a path cost-map, that pools the robot towards and along the given path; an obstacles cost-map, which is built from the LiDAR data. To remove leaves, branches, and other small objects from the obstacles cost-map, a row segmentation module is proposed with a custom implementation of the RANSAC algorithm, with a two-parallel plane model. We implemented the proposed modules for the navigation of a vineyard drone and tested it in a real vineyard.