|
Abstract : |
Abstract---We describe a computer model that reproduces many observed features of rat navigation behavior, including response properties of place cells and head direction cells. We discuss issues that arise when implementing models of this sort on a mobile robot. I. Rat Navigation As they navigate through their environment, rats appear to be employing several types of spatial representations. One type defines "places " based on the views they afford of distal landmarks [14]. Place cells in hippocampus, which fire when the rat is in a particular region of space, are known to be sensitive to visual cues (see [21] for a review). Rats' sense of place, as reflected in their navigation behavior, has also been shown to rotate in synchrony with the rotation of landmarks, but they fail to recognize the environment when landmarks are permuted [29]. This suggests that the animal's sense of place is not based on single landmarks but rather on landmark configurations. A number of computer models of visually-driven place cells have been described [3, 26, 27, 28, 33]. Any such model based on purely visual inputs, in which cells are dependent on more than one landmark, should, |
