Fusion of Spatially Referring Natural Language Statements with Random Set Theoretic Likelihoods
Adrian N. Bishop, Branko Ristić · IEEE Transactions on Aerospace and Electronic Systems · 2013
Localisation via the fusion of spatially referring natural language statements is considered here. The contribution lies in the underlying problem formulation and a robust modelling framework. Random-set-based estimation is the underlying mathematical formalism. Each statement generates a generalised likelihood function. A Bayesian filter is outlined that takes a sequence of likelihoods generated by multiple statements. The idea is to recursively build a map over the state space that can be used to infer the state.