This paper presents a novel approach to simulate human wayfinding behaviour incorporating visual cognition into a software agent for a computer aided evaluation of wayfinding systems in large infrastructures. The proposed approach follows the Sense-Plan-Act paradigm comprised of a model for visual attention, navigation behaviour and pedestrian movement. Stochastic features of perception are incorporated to enhance generality and diversity of the developed wayfinding simulation to reflect a variety of behaviours. The validity of the proposed approach was evaluated based on empirical data collected through wayfinding experiments with 20 participants in an immersive virtual reality environment using a life-sized 3D replica of Vienna's new central railway station. The results show that the developed cognitive agent-based simulation provides a further contribution to the simulation of human wayfinding and subsequently a further step to an effective evaluation tool for the planning of wayfinding and signage.