Cyber-physical systems and AI-enabled decision architecture for smart agricultural production and logistics

Recent advances in Cyber-Physical Systems (CPS), Artificial Intelligence (AI), and optimisation-based decision models have improved the sensing, monitoring, and prediction capabilities of distributed agricultural systems. Many existing smart agriculture solutions, however, focus on data acquisition and analytics, with less attention to translating real-time information into operational decisions. To address this gap, this study develops the CPS-AI Decision Architecture for Agricultural Systems (CADAS), an integrated cyber-enabled decision architecture that combines CPS infrastructure, AI-based analytics, and optimisation-based decision models. CADAS is evaluated through two case studies. The first case study examines greenhouse crop-stress monitoring, in which local image-derived stress estimation is embedded as dynamic priority coefficients in an inspection-routing and task-allocation model. Simulation experiments show that CADAS achieves higher stress-detection rates, shorter delays, reduced travel distance, and improved resource utilisation compared with alternative approaches. The second case study considers the application of CADAS to a distribution network for perishable agricultural products. The results show that a collaborative real-time truck-sharing system reduces total travel distance and waiting time in a network of independent distributors. CADAS contributes to production research by demonstrating how cyber-enabled architectures can operationalise AI insights using structured decision models.