A robotic arm framework for autonomous grapevine pruning: computer vision–based skeleton extraction and collision-free path planning

Grapevine pruning represents one of the most critical and complex agronomic operations, as it directly determines harvest quality and the overall longevity of the vineyard. Despite its central importance, this practice is currently characterized by high economic costs and a growing difficulty in sourcing skilled labor, factors that limit its scalability. In this scenario, the present thesis work is part of the research activities at Field Robotics, proposing the development of an autonomous robotic solution designed to operate in real agricultural environments, which are characterized by extreme biological variability. The core of the project lies in the design of a robotic pruning system capable of managing the entire operational pipeline: from environmental perception to the physical execution of the cut. Unlike traditional mechanization, the proposed solution aims to elevate the qualitative standard of the intervention through constant repeatability, reducing the subjectivity associated with individual operator experience. The integration of artificial intelligence and field robotics outlined in thiswork does not only address the need for cost efficiency but stands as a fundamental pillar for advanced precision agriculture. The achieved results lay the groundwork for resilient and data-driven vineyard management, capable of tackling the future challenges of the global wine sector in terms of both economic and environmental sustainability.