Autonomous grape vine pruning

"The project aims to be the first practical robotic pruning solution for cane-pruning (and spur pruning), most systems are hobbled by inaccurate scene understanding - if the plant structure is inaccurate or incomplete then attempting to make pruning descisions is almost impossible. Grape vines are complex plants with many overlapping vines, so it is important to not only recostruct the plant extremely accurately but also understand what's there. To achieve this we use a combiation of an autonomous many(12)-camera scanning vehicle and a secondary autonomous pruning vehicle. Our multi-camera scanner outputs highly accurate 3D models of plants. Following from this we have developed novel skeletonization algorithms which are more robust than current state-of-the art, allowing us to understand the branching structure extremely accurately, and make good pruning choices. Current development focusses on the physical pruning vehicle itself, which can make use of much of the infastructure and navigation systems of the scanning system, but poses some challenges such as re-localization with previously scanned plants, and the dexterity of the pruning arm itself. We are confident to produce prototypes in the next year and aim for robust systems following the year after."