Machine Learning Based Optimal Evacuation Route Guidance AR Navigation System in Indoor Fire Situations

초록

In this paper, we propose an optimal evacuation route guidance system for indoor fire emergency environments using machine learning and AR technologies. For user indoor positioning, we have implemented BLE beacon devices, and then collect received signal strength indication (RSSI) data of beacon signals at the user's smartphone according to the locations. Based on the data collected, a deep learning model is trained to determine user locations. Q-learning is utilized to derive optimal evacuation paths for individuals, and comprehensively reflects information such as user location, fire locations, and congestion situations in evacuation routes. Derived evacuation paths are intuitively provided through the Android AR application we implemented. Through performance evaluations based on the RSSI data collected in practice, user location estimation showed high accuracy and the Q-learning based evacuation path can derive the shortest time route without passing risky areas. Through the system proposed in this paper, we can help individuals evacuate quickly and reduce human casualties. © 2022, Korean Institute of Communications and Information Sciences. All rights reserved.

키워드