| Summary: | In this thesis we analysed the behaviour of crowd flows in large gatherings, taking a case study of the corridors where the Sa'yee ritual are performed. Of the current simulation models few of them addressed disabled persons in their model calculations and they did not include real-life data in order to analyse the crowd behaviour. None of the crowd modelling and simulation studies of the Hajj crowd, focused mainly on crowd behaviours during the Sa'yee ritual. We have proposed and developed a methodology to extract crowd characteristics from real-life videos with unknown camera angle and position. Agent tracking for different conditions of weather and various walking styles have been studied in the scope of our study. We further tested the feasibility of the use of drone to track agent behaviour in our methodology. We propose, design, and develop, a realistic and flexible state-based model of the Sa'yee ritual that can be mapped onto different agent-based systems, and then implement it in the modelling platforms Netlogo and FLAME. Further a comparison is made in terms of processing speed so that the model could be escalated to larger crowd. There are two enhancements carried out in the Sa'yee model and simulations, that makes this research novel and different than other contemporary researches. The first is the addition of different types of people (men, women and people with disabilities), and the second is the use of real video recordings from CCTV to analyse and model the walking behaviours of pilgrims. We carried out experiments that define the safety limit for the number of people in a group, and the results of defining a dedicated corridor for the disabled population. An empirical study shows that some aspects of the model such as agent density, behaviour, and speed match the real Sa'yee crowd. Therefore, the model can be used to study the general crowd behaviour in terms of agent density and speed of the agents with only slight modifications. The model is mapped from Netlogo to FLAME which validates that our model can be extrapolated to simulate larger crowd and huge number of agents. In addition, safety guidelines for better crowd management have been proposed to enhance crowd flow and reduce risks.
|
|---|
