Managing Crowds in Hazards With Dynamic Grouping

Emergency navigation algorithms for evacuees in confined spaces typically treat all evacuees in a homogeneous manner, using a common metric to select the best exit paths. In this paper, we present a quality of service (QoS) driven routing algorithm to cater to the needs of different types of evacuees based on age, mobility, and level of resistance to fatigue and hazard. Spatial information regarding the location and the spread of hazards is also integrated into the routing metrics to avoid situations where evacuees may be directed toward hazardous zones. Furthermore, rather than persisting with a single decision algorithm during an entire evacuation process, we suggest that evacuees may adapt their course of action with regard to their ongoing physical condition and environment. A widely tested routing protocol known as the cognitive packet network with random neural networks and reinforcement learning are employed to collect information and provide advice to evacuees, and is beneficial in emergency navigation owing to its low computational complexity and its ability to handle multiple QoS metrics in its search for safe exit paths. The simulation results indicate that the proposed algorithm, which is sensitive to the needs of evacuees, produces better results than the use of a single metric. Simulations also show that the use of dynamic grouping to adjust the evacuees' category, and routing algorithms that have regard for their on-going health conditions and mobility, can achieve higher survival rates.