A Navigation Algorithm for Pedestrian Simulation in Dynamic Environments

In this paper we introduce a novel online path planning method that mimics the human sense of navigation for finding a path between the current location (the ‘origin’) and the destination point. Most pedestrian simulation systems pre-compute a minimum cost (e.g. shortest length) path. Such an approach implies that the simulated agents have perfect knowledge of the infrastructure. Moreover, pre-computing the path can be numerically infeasible in situations where some of the obstacles are of dynamic nature, e.g. doors that might be closed or open, escalators which might operate or be under repair or obscure regions of parks which pedestrians choose not to frequent at night time. For such dynamic environments, the shortest path needs to be pre-computed for every out of many possible combination. The algorithm proposed in this paper assumes that the modelled agent may or may not have any prior knowledge of the infrastructure but is able to locate the direction of the destination at all times. The proposed navigation algorithm is an iterative procedure to maximize the values of smoothed distance and directional function from the destination to the origin. Such an algorithm is guaranteed to find a path from the origin to the destination if such a path exists at all. The proposed method has been implemented in a pedestrian multi agent simulation and has been tested on scenarios with different complexities.