EARLY FIRE DETECTION AND INTELLIGENT EVACUATION IN HIGH-RISE BUILDINGS

Abstract

The construction of high-rise buildings in Uzbekistan, particularly in Tashkent, is proceeding at a rapid pace. However, fire detection and evacuation management in such buildings still rely on point-type sensors and static exit signs. Based on an analysis of existing scientific literature, this article examines four research areas aimed at making evacuation adaptive to the situation: multi-sensor and video-based early detection, computer modeling of smoke propagation, evacuation modeling and dynamic route selection, and system integration using digital twins. Key methods, published results, and their limitations are summarized for each area. Multi-sensor systems combined with machine learning reduce false alarms; however, a model trained in one environment loses accuracy when applied to another. While deep learning-based video detectors are evolving, literature reviews highlight issues such as dataset bias and false alarms in real-world conditions. Tools like the FDS software—which simulates fire and smoke propagation—and evacuation models that track individual movements allow for a comparison between the Available Safe Egress Time (ASET) and the Required Safe Egress Time (RSET). Meanwhile, pathfinding algorithms such as A* and Dijkstra can reroute people as conditions change. However, most research on dynamic routing is two-dimensional and fails to account for stairwells, where smoke spreads most rapidly due to the chimney effect. Full-scale tests have demonstrated that dynamic exit signs are more noticeable than static ones. This article proposes a four-layer system adapted to conditions in Uzbekistan—comprising sensing, AI analysis, a simulation-based digital twin, and human routing—and outlines a verification plan that does not rely on predetermined outcomes. Key research gaps identified include the logic of vertical evacuation, the lack of local training data, and the absence of testing in actual buildings in Uzbekistan.