Proactive Roaming Prediction using LSTM in Industrial Wi-Fi-Based AGV Networks

초록

With the rise of smart factories employing Automated Guided Vehicles (AGVs) and Autonomous Mobile Robots (AMRs), ensuring reliable wireless communication has become a critical requirement for safe and continuous operation. However, in industrial Wi-Fi networks, frequent roaming often cause unexpected disconnections, undermining communication reliability and operational stability. This study aims to proactively predict roaming events in industrial Wi-Fi networks using large-scale data collected from AGVs and access points (APs) in real-world automobile factory. An Long Short-Term Memory(LSTM)-based model is employed to forecast each AGV's next AP, and roaming likelihood is further analyzed using the entropy of predicted probabilities. Our model detects more than 80% of roaming events in advance, contributing to improved communication stability in industrial environments.

키워드