Predictive modelling for AGF traffic optimization in warehouses

Abstract

Automated Guided Vehicles (AGVs), a rapidly evolving system in response to the increased demand for warehouse efficiency and the growing complexity of supply chain operations, have become integral to modern warehouse systems along with Automated Guided Forklifts (AGFs). While these systems offer numerous benefits in enhancing productivity and reducing labour costs, AGF jams persist as a significant operational challenge, impacting overall equipment effectiveness (OEE) and raising operational costs. This Final Year Project focuses on developing a real-time predictive model for AGF jam occurrence using machine learning techniques, which aims to identify contributing factors and enable proactive measures for jam prevention. The project involved data extraction, pre-processing, and sorting from the system database for model training purposes. Multiple supervised machine learning models were evaluated and compared, with the best-performing model selected to predict AGF jams based on factors such as Mission source and destination position, AGF live coordinates, and time data. The project's outcome provides a basis for future work, including integrating the prediction model into an optimizer system, developing a task scheduler, and implementing a continues system review with the user, ultimately improving warehouse efficiency and reducing operational costs.