Robotic precision thinning in apple production - using optimization and Bayesian modeling to assess potentials of automation in horticulture


Flower thinning is an important management step to maintain consistent yield and high fruit quality in apple production. As of today, thinning is mostly done using chemicals or tractor-mounted mechanical devices, which cannot do justice to the heterogeneity in both flower set and expected leaf area. In the near future, this task could theoretically be performed by autonomous robots. This could radically improve thinning precision. However, it remains unclear whether the benefits of such precision-thinning devices exceed their costs. This work aims to provide a detailed forecast of the economic potential of future precision thinning systems to support decisions in technology development and adoption. We developed a probabilistic model, which was fitted to experimental data on Elstar trees via Bayesian fitting. Using the decision analytical concept of Value of Control (VoC), we compared optimized to standard thinning in terms of revenue and yield. We found that the economic value of perfect flower thinning ranged from 898 to 3867 €/ha (90% credible interval). Yield is expected to increase by 24 % and revenue by 20%, while the average fruit quality decreased slightly due to a higher overall crop load.