Bumblebee: A Path Towards Fully Autonomous Robotic Vine Pruning

摘要

Dormant season grapevine pruning requires skilled seasonal workers, but they are becoming less available. As workers hasten to prune more vines in less time due to the short-term seasonal hiring culture and low wages, vines are often pruned inconsistently, leading to imbalanced grapevines. In addition, grapevines cannot be pruned selectively using currently existing mechanical methods, thus manual follow-up operations are often required, further increasing production cost. In this paper, we present the design and field evaluation of a rugged and fully autonomous robot for end-to-end pruning of dormant season grapevines. The proposed design incorporates novel camera systems, a kinematically redundant manipulator, a ground robot, and novel algorithms in the perception system. The presented research prototype robot system was able to spur-prune a row of vines from both sides completely in 213 s/vine with a total pruning accuracy of 87%. Initial field tests of the autonomous system in a commercial vineyard have shown significant variability reduction in dormant season pruning when compared to mechanical prepruning trials. The design approach, system components, lessons learned, future enhancements, as well as a brief economic analysis are described in the manuscript.