Navigation Strategies for a Vineyard Robot

摘要

<abstract> <b>Abstract.</b> The implementation of technology in European vineyards is occurring at a slower pace than in newer production zones such as Australia, Chile, South Africa, or the USA, where vast production fields favor the incorporation of automated systems. The VineRobot project emerges with the purpose of enhancing the management of vineyards through the combination of robotics, precision farming, and information technology. The project is sponsored by the European Commission, and aims at designing, developing, and deploying a novel use-case agricultural robot endowed with non-invasive biosensors to map vegetative growth in vines and red grape maturity. This paper explains the navigation strategies devised for the robot in its autonomous motion along the rows, and provides the first-year results on automatic steering using a stereoscopic vision camera as primary sensor for surrounding awareness and trajectory search. The vision model confines the robot universe into a set of situations occurring inside a look-ahead space 5-m wide and 8-m long at the height occupied by the vine canopy. Field experiments conducted in 2015 in a commercial vineyard showed stable behavior for low speed and revealed important sources of errors at higher speeds due to significant differences between vision-calculated angles and measured wheel angles. In addition to the effect of mechanical components in the navigation results, row perception was occasionally challenged by adjacent rows when canopy gaps appeared along the way. These findings will be the starting point for the project’s upcoming stage, which attempts to enhance the robustness of inside-row guidance and execute U-turns at the headlands.