A probabilistic approach for the analysis of evacuation movement data

Abstract

This paper presents a probabilistic approach to analyse evacuation movement data. The approach relies on a detailed video analysis of people movement and pattern reconstruction. Conditional probabilities for travel path trajectories, walking speeds, and physical area occupied on stair landings are calculated for the evacuee population. The approach has been applied as a case study using data from an evacuation drill performed in a six-storey office building in the United States. The evacuation drill was filmed and occupant's behaviours on stairs were analysed using the new method. A comparison with the deterministic methods currently employed in engineering practice has been performed. The benefits of the probabilistic approach are discussed, including (1) a more accurate representation of people movement and (2) the use of probabilistic data for modelling purposes, i.e., model validation and model development.