Evacuee Flow Optimisation Using G-Network with Multiple Classes of Positive Customers

Previous queueing theory based emergency navigation algorithms in built environments normally treat each significant location (such as doorways and staircases) as an "independent" queue and all the evacuees in a homogeneous manner. Hence, the interactions among linked queues caused by the re-routing instructions generated by the emergency navigation system, the panic behaviours such as evacuees not following the evacuation instructions, as well as the influence of diverse mobilities of evacuees are ignored. In this paper, we employ a Cognitive Packet Network based algorithm to customise distinct paths for diverse categories of evacuees. A G-network based model is used to analyse the latency on a path via capturing the dynamics of diverse categories of evacuees under the influence of panic and re-routing decisions from the navigation system. Moreover, by modelling the probabilistic choices of evacuees towards all the linked queues, the G-network model closely approximates the movement of the evacuees under the instructions of the Cognitive Packet Network based algorithm. The simulation results indicate that the use of the G-network model can improve the survival rates and ease the congestion during an evacuation process when there is a certain likelihood that evacuees do not follow evacuation instructions due to panic.