Abstract. Indoor crowds can be regarded as dynamic obstacles and they could impede the movement of pedestrians. For instance, the scope of crowds could significantly impact accurate path choices. There are few discussions from an individual’s perspective to reveal the decision-making process of reasonable path choices. In this paper, we aim to address how an indoor pedestrian would walk in a crowded indoor environment. We propose a method to simulate a pedestrian’s path choices during wayfinding process. It is wrapped in a workflow that can capture near-real-time crowd uncertainty. First, the measurement errors of crowds are alleviated with a Bayesian filter. According to crowd locations at each moment, the method of Kernel Density Estimation (KDE) is applied to a grid model of the building to obtain the uncertainty map of the crowds. Based on a two-level spatial model, the logical path represented by a room sequence can be derived from a ’room-door’ network, while detailed path choices on the ’operational’ level are refined in each room. By considering both crowd uncertainty and time cost, path choices for a user are determined by using the A* algorithm at each time slot. By iterating the optimal path to the next door/the target in the current room, the user location of the next moment can be inferred. Our test initially validated the feasibility of the method on the path choices of a pedestrian. In the future, we plan to further conduct field tests with different users in crowded indoor environments.