@ARTICLE{Botterill-etal-2016,
author = {Tom Botterill and Scott Paulin and Richard Green and Samuel Williams and Jessica Lin and Valerie Saxton and Steven Mills and XiaoQi Chen and Sam Corbett-Davies},
title = {A robot system for pruning grape vines},
journal = {to appear in the Journal of Field Robotics},
year = {2016},
volume = {PP},
issue={PP},
pages={1-31},
abstract={This paper describes a robot system for automatically pruning grape vines. A mobile platform straddles
the row of vines, and images them with trinocular stereo cameras as it moves. A computer vision system builds a
3D model of the vines, an AI system decides which canes to prune, and a six degree-of-freedom robot arm makes the
required cuts. The system is demonstrated cutting vines in the vineyard. This paper's main contributions are the
computer vision system that builds 3D vine models, and the test of the complete integrated system. The vine models
capture the structure of the plants so that the AI can decide where to prune, and are accurate enough that the
robot arm can reach the required cuts. Vine models are reconstructed by matching features between images,
triangulating feature matches to give a 3D model, then optimising the model and the robot's trajectory jointly
(incremental bundle adjustment). Trajectories are estimated online at 0.25m/s, and have errors below 1\% when
modelling a 96m row of 59 vines. Pruning each vine requires the robot arm to cut an average of 8.4 canes. A
collision-free trajectory for the arm is planned in 1.5s/vine with a Rapidly-exploring Random Tree motion planner.
Total time to prune one vine is two minutes in field trials, which is similar to human pruners, and could be
greatly reduced with a faster arm. Trials also show that the long chain of interdependent components limits
reliability. A commercially-feasible pruning robot should stop and prune each vine in turn. },
url={http://www.hilandtom.com/tombotterill/pruner-preprint.pdf}
video={http://www.hilandtom.com/vines.mov}
}