Pedestrian unrestrained behaviour, sudden movements and vulnerability are elements, which can highly affect road safety, especially when interacting with motorized vehicles. Therefore, it is important to have a deep insight in pedestrian behaviour. A way to tackle this issue is micro-simulation. Modern micro-simulation tools, indeed, allow, thanks to the implemented mathematical formulation of the problem, to model and repeat a real situation in a virtual environment. Nevertheless, they need to well-fit the real observed behaviour: the calibration step allows to make the model reliable, by adapting selected, influential model input parameters. By dealing with pedestrian issues, software Vissim/Viswalk has been selected for micro-simulation, which implements Helbing's Social Force model. This model is based on several parameters, like relaxation time, side preference, strength and range of pedestrian interactions, amount of anisotropy, parameters governing the forces among pedestrians, noise, number of reacting pedestrians, queue order and straightness, which need to be set by the user when creating the model, but they can be hardly measured. This paper presents a selection of the recalled input parameters, on which statistical tests are carried out to understand their influence on the behavioural output – crossing time - that is supposed to describe pedestrian crossing behaviour. This is the first step towards the development of a new calibration methodology, which will keep advantage of artificial intelligence tools to fine-tune micro-simulation input parameters.This paper presents a selection of the recalled input parameters, on which statistical tests are carried out to understand their influence on the behavioural output – crossing time - that is supposed to describe pedestrian crossing behaviour. This is the first step towards the development of a new calibration methodology, which will keep advantage of artificial intelligence tools to fine-tune micro-simulation input parameters.

pedestrian, micro-simulation, input parameters, statistics