Digital Systems for Smart Crop Management in Sustainable Agriculture

The digital and sustainable transition poses a significant challenge to agriculture. The overexploitation of natural resources such as soil and water is driven by climate change and the need to produce ever more food. Based on the concept of the living lab, this thesis presents the development of two frameworks to address two key agricultural issues: water scarcity and sex identification in dioecious plants. Various technological and agronomic solutions are combined synergistically to deal with these questions. The first framework integrates LoRaWAN-based Internet of Things architecture, artificial intelligence models, controllers and simulation models to optimize irrigation management and determine the most appropriate time and quantity of irrigation. This framework was tested and validated in open fields with tomato plants over three consecutive seasons, and in greenhouses with olive trees on an ongoing basis. Results obtained in the field and through probabilistic Monte Carlo simulations show that keeping soil volumetric water content within the optimal range between the wilting point and field capacity can lead to water savings of up to 30% and energy savings of up to 50% compared to the irrigation recommendations provided by Irriframe. The most promising solution is a fuzzy logic control system that integrates plant, soil and environmental data. The digital twin of an irrigation system was developed using fluid dynamic simulation models in Flownex®. The system's capabilities include assisting with the design and maintenance of an irrigation network, as well as applying advanced control logic to close the loop of data acquisition and processing, and to actuate the valve command. The second framework involves the early identification of the sex of dioecious species through the combination of computer vision and artificial intelligence models. Results achieved in a living lab focused on hops demonstrate 70% accuracy in classifying newly transplanted plants. This makes the solution as effective as the genetic techniques currently in use but reduces processing time and costs.