A hybrid model for rodent spatial learning and localization

Recent work [3, 21 has explored a Kalman filter model of animal spatial learning the presence uncertainty in sens o y as well as path integration estimates. This model was able to successfully account for several of the behavioral experiments reported in the animal navigation literature [lo, 61. This paper extends this model in some important directions. It accounts for the observed firing patterns of hippocampal neurons [19] in visually symmetric environments that offer polarizing senso y cues. It incorporates mechanisms that allow for diflerential contribution from proximal and distal landmarks during localization. It also supports learning of associations between rewards and places to guide goal-directed navigation.