EXPO-AGRI: Smart automatic greenhouse control

Abstract

Predicting and controlling plant behavior in controlled environments is a growing requirement in precision agriculture. In this context sensor networks and artificial intelligence methods represent key aspects for optimizing the processes of data acquisition, mathematical modeling and decision making. In this paper we present a general architecture for automatic greenhouse control. In particular, we focus on a preliminary model for predicting the risk of new infections of downy mildew of basil (Peronospora belbahrii) on sweet basil. The architecture has three main elements of innovation: new kinds of sensors are used to extract information about the state of the plants, model predictors are generated from this information by non-trivial processing methods, and informative predictors are automatically selected using regularization techniques.