Mutual information for the detection of crush conditions

Harding, Peter (2012) Mutual information for the detection of crush conditions. Doctoral thesis (PhD), Manchester Metropolitan University.

Preview

Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (5MB) | Preview

Abstract

Fatal crush conditions occur in crowds with tragic frequency. Event organizers and architects are often criticised for failing to consider the causes and implications of crush conditions, but the reality is that the prediction of such conditions o ers signi cant challenges. This thesis investigates the use of crush metrics in simulation environments, which can be used to help quantify the danger of crush conditions forming during real life evacuations. An investigation is carried out in the use of computer models for the purpose of simulating building evacuation. From this review we identify the most suitable methodologies for modelling crowd behaviour, and we detail the speci c areas of functionality which must be in place before modellers can incorporate crush analysis into an evacuation simulation. We nd that full treatment of physical force within crowd simulations is precise but computationally expensive; the more common method, human interpretation of simulation output, is computationally \cheap" but subjective and timeconsuming. A technique which admits a low computational cost alternative to the explicit modelling of physical force, yet still o ers a quantitative metric for the level of force present during an in silico evacuation is proposed. This technique and the precise manner in which we apply it to the problem of crush detection is shown and we present the results of initial experiments. To further test the ability of our technique to identify dangerous evacuation conditions, we recreate a well-known historical evacuation. Results of these experiments show that we do o er an e ective and e cient route towards the low cost automatic detection of crush, and an alternative approach to traditional methods.