Localization of Mobile Manipulator in Vineyards for Autonomous Task Execution

Abstract

Although robotic systems have found their place in agriculture, there are still many challenges, especially in the area of localization in semi-structured environments. A robotic system has been developed and tested to perform various tasks in the steep vineyards of the Mediterranean region. In this paper, we describe a method for vine trunk localization, based solely on the visual recognition of vine trunks by neural networks fed by an RGB camera. Assuming that the height of the first wire in the vineyard is known, the proposed method is used to determine the location of vines in the immediate vicinity of the all-terrain mobile manipulator—ATMM-VIV—needed for spraying and bud suckering. The experiment was conducted in a slightly inclined vineyard to evaluate the proposed localization method.