Isovists, Occlusions and the Exosomatic Visual Architecture

Recently, simulation agents (or animats) using an exosomatic visual architecture (EVA) have been shown to correlate well with observed pedestrian movement in both building and urban environments. An EVA uses a grid overlaid on a two-dimensional plan of a system to record the locations visible from the current grid square. The agents are allowed to roam freely in the environment, and lookup visual information from the EVA in order to guide them through the plan. This allows many agents to navigate concurrently using visibility relationships. However, while good correlation between observed physical and virtual systems has been shown, experiments to date have been based on agents which move stochastically to visible locations. This leads to them congregating in large open vistas, where there are more visible locations. In contrast, when people are observed, they tend to follow the edges of spaces to move or take direct routes across open spaces to the far side. Here we hypothesize that rather than using open space to guide them, people instead use the visual clue of an occluding edge to indicate where further movement potential may lie. We supplement the information in the EVA with details of the isovist at each location, to supply the locations of occluding edges from each grid square. We show that these new agents follow paths much more similar to observed pedestrians using an open space. We speculate that the invariance of the occlusion points within a plan may thus lead to an economic skeletal mapping of the environment, and possible basis for a cognitive map.