An Agent-based Approach to Modelling Pedestrian Behaviour

The modelling of pedestrian behaviour in a real-world environment is a complex problem, mainly due to the unpredictable nature of human decision making. Agent-oriented simulation moves away from traditional all-knowing and "controlling" simulations and towards reality, where pedestrians exhibit different behaviours depending on their knowledge of the environment and other personal characteristics. We investigate the behaviours that pedestrians may exhibit, the different techniques used for pedestrian modelling, and the appropriateness of each technique for particular domains. The classification framework developed will play a role in the decision making process for planning and design of pedestrian areas. We then explore the agent-based approach in more detail, in particular the belief-desireintention (BDI) architecture, by presenting the development of a sample model using Prometheus, an agent-oriented design methodology, and JACK, an agent-oriented programming language. Although the BDI architecture is useful for high-level decision making, further work is required in representing and updating the environment.