Defining and Automating Optimal Dormant Pruning

Pruning is an essential horticultural practice but is very labor intensive. Previous attempts at mechanical pruning have mostly involved non-discriminating hedging that leads to poor light distribution within the tree canopy and reduced fruit quality. In this study we developed and evaluated pruning heuistics for grapes and apples. These pruning “rules” led to horticultural outcomes, such as yield and fruit quality, broadly similar to those resulting from pruning by commercial human pruning crews. Vine and tree canopies were captured using cameras and sensors, then reconstructed accurately in 3D. Pruning heuistics were then applied to the reconstructions to determine optimal pruning points. This work has applications for the development of autonomous pruning, with a robotic grapevine pruner currently being tested and evaluated. Another potential application of this work is in the development of more effective training methods for human pruning crews. Grower attitudes towards the science of pruning appeared to depend on the size of their orchards.