Our ability to point to unseen locations suggests that we build some kind of a map in our head. While the nature of such a map has been controversial, it is clear that it is not the kind of map produced in robots using the popular probabilistic SLAM-based approaches. The latter map is precise and complete while the former is imprecise and incomplete.

In this talk, I will present a new computational theory of perceptual mapping that shows how an inexact and imprecise map emerges from a process of remembering and forgetting local environments visited. Local environments are perceptually defined as snapshots of the environment taken at intermittent points during a journey through the environment. The emergent map is both dynamic and transient and provides an egocentric rather than an allocentric view of the environment visited. The theory has been successfully implemented on a mobile robot equipped with a laser sensor and an odometer. I will also compare the performance of our algorithm with those using SLAM-based approaches.