Brown University VENLab

with their environment. We have developed dynamic models of basic behaviors such as steering, obstacle avoidance, and target interception, which are combined to predict paths through complex environments. We also map out the visual information used to guide each behavior, such as optic flow. This work leads to models of "virtual humans" and crowd behavior, and has applications to robotics and computer animation. In studying navigation, our goal is to understand how humans develop and use spatial knowledge about the world to guide behavior. We study problems such as path integration, the influence of visual landmarks, the geometry of spatial knowledge, and the contribution of different sources of information (e.g. vision, proprioception, attention) to spatial learning. Our hypothesis is that successful navigation can be explained by a combination of abilities that are weaker than a Euclidean 'cognitive map.' This work has applications to architecture and the design of navigation tools like GPS.