An algorithm for automatic dormant tree pruning

Abstract

Tree pruning is a labor and cost-intensive task. Still, it is a necessary activity that ensures a high yield of good quality products in horticulture and increases the overall health of trees in general. Extensive research has been done attempting to automate this labor-intensive procedure, lower the cost, and demand a skilled workforce. We introduce a new algorithm based on discrete differential evolution that simulates the pruning of virtual trees. Although pruning driven by differential evolution alone optimizes the overall tree light intake, it cannot maintain the distance between individual trees, nor can it shape trees into any of the growing forms. In the article, we show that adding additional steps into the pruning process, which is an initial trimming of the tree into a desired shape, can be improved significantly. We demonstrate our method by simulating the pruning of virtual trees and show that it provides results comparable to the results obtained by a human expert. By simulating the tree pruning over a few consecutive years, We show that our method is also capable of autonomous tree training toward the desired growing form.