A more realistic simulation of pedestrian based on cellular automata

The simulation of pedestrian has been studied for a long time from various points; however, most of them have not considered the psychological and terrain factors for pedestrians. Firstly, we develop the fundamental model of pedestrian simulation based on Cellular Automata. To consider the terrain factors, this paper embeds neural network into these agents to ensure intelligent and realistic. The data of training neural network are collected from real pedestrians, which make agents more realistic. This model can help us find the best path of pedestrians´ simulation in both single and many pedestrians. This method can be widely used in the business strategy and security control. These implementations are based on the open-sourced toolboxes in Scilab, including Neural Network Toolbox and Cellular Automata Toolbox.