Simultaneous Localisation and Planning

Abstract

This paper proposes an algorithm for the solution of the Simultaneous Localisation and Planning problem. The solution is based on statistical inference in a Hidden Markov Model which proceeds in separate phases of localisation and planning. Each requires access to the same contextual model operationalised via the `prior dynamics', and is implemented using forward (localisation) and forward and backward (planning) message passing. I propose that this formalism provides a useful computational-level description of aspects of Hippocampal function.