Abstract
Spatial navigation tasks often require us to memorize the relation between visual landmarks and the goal. Landmarks might only provide ambiguous or unreliable cues, making it necessary to make predictions about landmark reliability when learning a new location. These predictions might be enhanced with prior knowledge. How does anticipation about landmark properties affect learning of a spatial location? We studied navigation performance in two different virtual reality homing tasks. The participants walked on a treadmill surrounded by six large flat panel displays and navigated in a simulated desert environment or a parking lot. In both tasks they learned the position of a goal, determined by different landmarks. They were then relocated to a new position and had to return to the previously learned location. In the desert scene we tested how participants integrated ambiguous landmark cues and how conflicting cues change affect performance. In the parking lot scene we tested how participants choose reliable landmarks when locating a position in a cluttered environment that provides many possible landmark cues. The participants integrated the spatial information from each landmark near-optimally to reduce spatial variability. When the conflict becomes big, this integration breaks down and precision is sacrificed for accuracy. A probabilistic model based on the performance with only one ambiguous landmark can predict this integration performance. Participants use prior knowledge about object properties to select presumably reliable objects when they are freely choosing landmarks in the cluttered parking lot scene. Here we find that they use information from their day to day environment in order to avoid navigational errors by unreliable landmarks. How this knowledge is gained and to which extent is it used in increasingly complex scenarios is currently under investigation.
Meeting abstract presented at VSS 2017