Aim: The safety of people during meetings and public events requires an analysis of the conditions of evacuation and the proper organisation of escape in times of danger. Modelling the evacuation process is one of the options for analysing and planning the safe evacuation of the buildings, facilities and spaces during events. The aim of the article was to present the practical use of cellular automata for evacuation modelling and comparing the results with the results achieved using commercial software. Additionally, the objective of this work was to compare the cost of computational evacuation modelling of cellular automata with the “Social Force” model.
Project and methods: The authors used cellular automata on the grid with a fixed size of 0.5 m x 0.5m. The basic premise for the traffic model was the “Floor Field” static layer with Euclidean metric and the greedy algorithm. The determined value μ = 0.55 indicated the probability of the transition of a person to the neighbouring automat cell. The implementation of the model was made in “Python”, using the library for scientific computing “Numpy”, and the maths library “Math”. We compared the results of modelling the evacuation using the proposed model with the program “FDS + Evac” for room size 11.5 x 9 m with one and two emergency exits. Furthermore, the modelling was done and the results were juxtaposed with the results of the programs “FDS + Evac”, “Pathfinder” and “TraffGo”, using the example of a test of the International Maritime Organisation (IMO 9) covering the evacuation of 1,000 people.
Results: For both the evacuation modelling proposed by the authors and the IMO 9 test estimated evacuation times are consistent with the times estimated using programs applicable in fire-safety engineering. It was further found that the use of cellular automata for modelling the evacuation of 1,000 people allows modelling 20 times faster than in the “Social Force” model implemented in the “FDS + Evac” program. The average cost of modelling of 1 s of evacuation using cellular automata depends linearly on the number of persons subjected to evacuation, in contrast to the “Social Force” model, where the modelling time will increase exponentially with the number of people.
Conclusions: Evacuation models based on cellular automata, in contrast to the “Social Force” model, provide the ability to model the movement of large groups of people at a lower computing cost. The use of cellular automata allows the introduction of additional layers affecting the movement of people with a small increase in computational complexity. Introducing additional assumptions and layers to cellular automata models allow a more realistic representation of the evacuation with the small increase in the cost of the equipment used. Furthermore, a tool allowing fast estimates of the evacuation time in a manner accessible to engineers would help in the correct designing of buildings. Current commercial programs require expertise in the field of modelling. Using a simple interface with a fast algorithm estimating evacuation times can bring measurable benefits in terms of improving the safety of designed buildings and construction works.
Keywords: evacuation, modelling, estimating evacuation time, cellular automata, computational complexity
Type of article: original scientific article