A particle swarm optimization based multi-agent stochastic evacuation simulation model
MetadataShow full item record
How do we evaluate the evacuation efficiency of a building floor plan or an aircraft cabin? The most direct approach is to arrange evacuation drills for the evaluation purpose. However, several disadvantages have been associated with these drills. Firstly, these drills are usually considered dangerous, especially when a large number of participants are involved. Secondly, these drills usually require relatively long time and a large amount of money to prepare. Furthermore, the cost increases dramatically when the number of participants increases or when the drill fails and a new one needs to be redesigned. Lastly but not least, it is typically the case that only limited number of trials can be performed in each evacuation drill due to the time-consuming planning and enormous cost. Therefore, the trial results could easily be biased if the number of trials is not sufficient. The other approach to guarantee the evacuation efficiency is to keep floor plan designs in compliance with relevant prescriptive Building Code or Aircraft Safety Code. However, these codes are typically conservative in nature therefore hinder the innovation of floor plan designs. Performance-Based Code, on the other hand, allows engineers to design fire protection individually for each new building or aircraft using alternative methods other than prescriptive ones, with a maximum design freedom. To evaluate the evacuation efficiency conveniently and efficiently in performance-based design, computer simulation models, besides being cost efficient and eliminating the need of involving real participants, can perform repeated tests fairly easily with a built-in stochastic feature that enables a reasonable representation of appropriate behaviors across a spectrum of situations. This dissertation research introduces a Particle Swarm Optimization (PSO) based multiagent stochastic evacuation simulation model incorporating fire hazards and critical human behaviors. The model has two sub-models: Vacate and VacateAir . The former one is for building evacuation simulation and the latter one is for aircraft evacuation simulation. The model utilizes a modified PSO Algorithm as a path finding algorithm that directs evacuees to the final exit as well as dynamically adjusts evacuees' direction according to fire hazards and crowd movements. The fire data are pre-calculated in Fire Dynamics Simulator (FDS) and imported in the model for the use of conducting Human Tenability Analysis (HTA) . Critical human behaviors that identified in building and aircraft evacuation are simulated and their impact on the evacuation efficiency is evaluated. There are numerous advantages in applying modified PSO as the path finding algorithm. Application of this strategy overcomes several limitations of existing evacuation models, e.g. eliminating the need to divide the entire floor plan layout into grids and nodes therefore saving substantial computational expense and enabling a simulation of the continuous movement of evacuees, which outperforms the jagged and often unrealistic movement generated by traditional grid-based path finding algorithms. With these improved features and validations against published evacuation experimental data, Vacate (for building evacuation) and VacateAir (for aircraft evacuation) can help designers and fire protection engineers conduct the performance-based design of buildings and aircraft more conveniently. The parametric study of the effects of physical factors such as exit width, aisle width, seat pitch and evacuation motivation (competitive or cooperative), on evacuation efficiency not only provides valuable information to building and aircraft designers, but also opens a potentially new avenue in the future research work on the System-of-Systems (SoS) design approach by coupling the evacuation system with aerodynamic system, weight system, and airline resource allocation system.