Design and development of vineyard row following algorithms for agricultural robotic vehicles

The whole thesis project was realised in collaboration with STIIMA-CNR (Sistemi e Tecnologie Industriali Intelligenti per il Manifatturiero Avanzato - Consiglio Nazionale delle Ricerche) located in Bari. The need to improve the relationship between phenotyping and automation is increasing due to the world’s current sub-optimal and worsening situation regarding food production. It is therefore important to research and implement new methods in order to increase sustainability and food security worldwide. The first step we used to automate phenotyping is to enable the robot to move autonomously within the rows of vines. This is made possible by data acquisition through several Intel Real Sense D345s from which PointClouds are exploited. The points are used to construct a suitable plan that best fits the row. By extrapolating the data from the plane normal, the robot can recognise and adjust its angle to the row and also the distance so that it is always parallel. In addition, with the ultimate aim of improving the torque distribution in the four drive wheels, a system was developed to calculate the odometry of the robot, obtaining the x and y distances from the starting position and the rotation angle. The entire system was tested and verified through several indoor and outdoor tests, which yielded good results, thus validating the methods used. The collected data was further analysed and, through an offline study, a Kalman Filter was designed and tested to smooth the online data collected and thus avoid decision inaccuracies of the robot.