Human and Ideal Spatial Navigation. Manuscript Submitted for Publication to Psychological Review. Lost in Virtual Space: Studies in Human and Ideal Spatial Navigation

We describe three human spatial navigation experiments that investigate the role of perception, memory, uncertainty and decision strategy on human spatial navigation performance. These studies were conducted in virtual reality indoor environments that were visually impoverished. To better understand the effect of these variables on human navigation performance, we developed an ideal navigator model for indoor navigation based on Partially Observable Markov Decision Process (Cassandra, Kaelbling, & Littman, 1994; Chung, 1960; Kaelbling, Littman, & Cassandra, 1998). The model is designed to navigate through visually sparse indoor environments where multiple states (positions and orientations) in the environment can produce the same visual percept. The model is designed to minimize the number of actions (translations and rotations) required to move from an unknown starting state to a specific goal state within the environment. We used the model to compute human navigation efficiency by computing the ratio of the number of actions required by the ideal navigator relative to the number of actions taken by the human subjects. Experiment 1 investigated the effect of increasing the layout size on spatial way-finding efficiency and found that subjects ’ efficiencies decreased as layout size increased. We investigated whether this reduction in navigation efficiency was due to visual perception (Experiment 2), memory, uncertainty or decision strategy (Experiment 3). The results revealed a significant reduction in efficiency as the layout size increased. By contrast, reducing visual information did not have