Mesoscopic movement model of deliberative pedestrian agents

The agent based simulation is often utilised for modelling of complex systems composed of large number of autonomous entities. The computational complexity of such models depends on the chosen level of details and naturally increases with increasing number of modelled entities. The computationally intensive microscopic modelling of entities´ behaviour might often be substituted with mesoscopic modelling, even without sacrificing much of the validity of the model. The paper presents mesoscopic agent based movement model of pedestrians. Two layer agent based architecture ABAsim is utilized. Lower layer dynamic agents represent pedestrians; they are capable of autonomous reasoning (utilising BDI paradigm), but their movement is restricted due to the mesoscopic nature of the model. Managing agent of upper level is responsible for modelling of the movement. This agent collects requests from pedestrian agents, assists them with navigation, resolves conflicts between pedestrians and moves them. Proposed movement model utilizes cellular infrastructure model with cell of various sizes (in order to guarantee validity of the model, while maintaining computational efficiency). Interactions among pedestrians moving between cells are modelled with the help of specialised graphs that are also briefly discussed.