Even without sensory input, an animal can estimate how far it has moved by integrating its velocity, a process called path integration. The entorhinal cortex (EC) and hippocampus seem to be involved in path integration, and in an animal's perceived location in space. However, path integration is highly susceptible to accumulating errors. A real animal avoids this problem by incorporating sensory input (e.g. vision) and updating its perceived position. The best path integration models do not yet incorporate this sensory-updating feature. In this paper, we extend one such model to enable sensory updating, and demonstrate its effectiveness in a series of computer simulations of spiking neural-network models.
展开▼