Spatial Learning for Robot Locialization

Although evolutionary algorithms have been employed to automatically synthesize control and behavior programs for robots and even design the physical structures of the robots, it is impossible for evolution to anticipate the detailed structure of speci c environments that the robot might have to deal with. Robots must thus possess mechanisms to learn and adapt to the environments they encounter. One such mechanism that is of importance to mobile robots is that of spatial learning, i.e., the ability to learn the spatial locations of objects and places in the environment, which would allow them to successfully explore and navigate in a-priori unknown environments. This paper proposes a computational model for the acquisition and use of spatial information that is inspired by the role of the hippocampal formation in animal spatial learning and navigation.