The aim of this dissertation is the multi-parametric analysis of the road safety and traffic behaviour of autonomous vehicles during their interaction with pedestrians for different, well-defined, levels of automation. For this purpose, an experimental procedure will be developed both in naturalistic driving conditions, for different vehicle types (i.e., passenger vehicle, minibus, bus), and in a driving simulator environment at automation levels 2 and 3 (according to SAE J3016). After an optimal analysis of the driving data, obtained from the above experiments, will be fed into traffic simulation software on an existing calibrated network in order to further study the interaction of autonomous vehicles with pedestrians at a network level. By executing the simulations, an extension to other levels of automation (SAE level 4 & 5), e.g., Connected Autonomous Vehicles, will be achieved and the results can be generalised. The impact of the autonomous vehicle-pedestrian interaction on road safety will be investigated using surrogate safety measurements and traffic behaviour, after appropriate analysis at both incident-level and network-level.