Remote control of greenhouse cucumber production with artificial intelligence - results from the first international autonomous challenge

A. Elings*, I. Righini, H.F. de Zwart, S. Hemming, A. Petropoulou

*Corresponding author for this work

Research output: Contribution to journalArticleAcademicpeer-review


There is a need for remote greenhouse management. As farms become larger, the crop manager has difficulties monitoring all details of the various compartments. Also, finding skilled staff becomes more difficult and distant management of a crop production system requires new sensing technologies. At Wageningen UR Greenhouse Horticulture a competition on 'autonomous greenhouses' has been organized, in which Artificial Intelligence (largely) replaced human skills in greenhouse operation. The purpose was to test in this proof-of-principle the functionality of the approach, while improving production, product quality and resource use efficiency. Five multidisciplinary international teams participated. A reference compartment was operated by growers. Each team had available a 96 m2 greenhouse compartment to grow remotely a cucumber crop ('Hi-Power') from August to December 2018. Visiting the compartment was not permitted so all decisions were based on sensor output and observations on crop development and harvest. Compartments were equipped with standard actuators for climate control and fertigation, and some teams installed additional sensors. Teams decided on plant and stem density in advance. They remotely determined the continuously varying control setpoints, using AI algorithms, and provided instructions for leaf and fruit pruning on a weekly basis. Pest and disease management was WUR's responsibility and was no part of the challenge. All AI-algorithms classified light and CO2 as the most determinant production factors. The winning team, which had invested most in light, scored best on production at the cost of some resource use efficiencies. The winning AI-algorithm also out-performed the reference. Total fresh yield was closely associated with total number of fruits m-2, which is an aggregate of stem density and fruit pruning strategy.

Original languageEnglish
Pages (from-to)69-76
Number of pages8
JournalActa Horticulturae
Publication statusPublished - 17 Nov 2020


  • Artificial intelligence
  • Crop growth and development
  • Cucumber
  • Remote greenhouse management


Dive into the research topics of 'Remote control of greenhouse cucumber production with artificial intelligence - results from the first international autonomous challenge'. Together they form a unique fingerprint.

Cite this