Hybrid sensing, inventory accuracy, hidden demand detection, robust estimation, virtual sensor, perishable supply chains
This project introduces a hybrid sensing architecture aimed at enhancing the accuracy and reliability of inventory data within perishable supply chains, where measurement noise and unobserved demand events frequently impair decision‑making. The proposed system couples a low‑cost physical sensor with a virtual sensing layer that integrates a fast, robust state estimator and an outlier‑detection module. The estimator filters noisy measurements and quickly converges to the true inventory level, while the detection component identifies anomalies such as shrinkage and autonomously resets the estimator when necessary. This hybrid approach provides a cost‑effective alternative to high‑precision hardware, improving inventory visibility, operational resilience, and sustainability. The architecture is validated within an inventory management policy, demonstrating higher accuracy, timely anomaly detection, improved service levels, and stable stock exposure even under uncertain demand and unobserved disruptions.
