Controlled environment agriculture allows to maximize crop yields by providing plants with ideal growth conditions for 365 days a year, in a sustainable way and without the use of pesticides. The maximum exploitation of the productive potential of the plant, however, depends on the knowledge of the response of the latter to a wide spectrum of environmental stimuli, which can be acquired through a long and expensive experimental process. In this context, biophysical models of plant growth, which are based on physiological knowledge and minimize the need for calibration data, represent an opportunity to speed up the optimization process on new genotypes that are continuously introduced into the market. These mathematical models have been developed in open field conditions, or in greenhouses, and are not currently available for all species of interest in indoor farming. The aim of this thesis project is to extend the existing models to new species and/or cultivation techniques used in controlled environments. The calibration and validation of the models, possibly integrated with Machine Learning algorithms, will be carried out experimentally in a vertical farm.