GreenhouseDT: Progettazione di un'architettura per gemelli digitali autoadattivi

This thesis presents the design and development of GreenhouseDT, an exemplar framework aimed at exploring and defining an appropriate architecture for self- adaptive digital twins. Digital twins, dynamic virtual replicas of physical systems, are increasingly used in various industries to enhance real-time monitoring, simula- tion, and optimization. However, creating a flexible and scalable architecture that supports self-adaptation remains a critical challenge. In this project, GreenhouseDT was developed as a low-cost, scalable research hardware setup which aims at being easily replicated in various domains. The system integrates a combination of sensors, Raspberry Pi devices, and a software stack to monitor environmental conditions such as soil moisture, temperature, and humidity. Leveraging the SMOL experimen- tal language, the digital twin performs decision-making and actuation operations. Another software layer handles self-adaptation by continuous synchronisation of physical and digital counterparts. This architecture supports both behavioural and architectural self-adaptation, allowing the system to adjust to environmental changes and structural modifications within the greenhouse. The study assesses the performance of the GreenhouseDT system, demonstrating its ability to provide insights into optimal architecture design for self-adaptive digital twins. Through practical experimentation, the framework’s strengths and limitations are evaluated, contributing to future research on digital twins in domains beyond agriculture. Potential enhancements, such as integrating ML driven operation on the greenhouse are also discussed, showing how this architecture can be further improved in the agriculture domain.