AI encompasses various technologies, with Machine Learning (ML), Deep Learning (DL), and Generative AI (GenAI) being the driving forces (Figure 1). However, recent discussions around AI have predominantly focused on GenAI due to its ability to create new content such as text, images, and music by identifying patterns from extensive datasets available on the internet. Although the ability to generate realistic content is a significant milestone, GenAI is not intended for optimal process control and automation in physical environments. This article focuses on ML, a broader and more foundational category within AI, which has widespread applications across various domains including CEA. We will provide a primer on ML, explain why it is transformative to the CEA industry, and highlight current challenges to effective ML solutions for efficient greenhouse control.
Traditional greenhouse operations often require growers to manually program control rules into process control systems. Let's consider irrigation for greenhouse tomato as an example. Growers typically divide irrigation into various periods throughout the day, each associated with a different solar radiation-sum threshold. A common rule among growers is to maintain a stable drainage (leachate) ratio, for example 30% of the total irrigation.
volume. However, achieving a fixed drainage ratio is challenging because it is influenced by multiple environmental factors beyond solar radiation alone. As a result, growers frequently need to adjust parameters in response to changing weather patterns, relying heavily on their experience and intuition. They observe factors such as solar radiation and humidity throughout the day, along with historical drainage data, to make necessary adjustments.
Read more: e-gro.org