A Minimalistic Model of Visually Guided Obstacle Avoidance and Path Selection Behavior

In this study we present an empirical experiment investigating obstacle avoidance and path selection behavior in rats and a number of visually-guided models that could account for the empirical data. In the experiment, the animals were repeatedly released into an open arena containing several obstacles and a single feeder that was marked by a large visual landmark. We recorded and analyzed the animals' trajectories as they approached the feeder. We found that the animals adapted their paths according to the specific obstacle configurations not only to avoid the obstacles that were blocking the direct path, but also to select optimal or near-optimal trajectories. On basis of these results, we then develop and present a series of minimalistic models of obstacle avoidance and path selection behavior that are based purely on visual input. In contrast to standard approaches to obstacle avoidance and path planning, our models do not require a map-like representation of space.