Extensive representations and algorithms for nonlinear filtering and estimation

Abstract

Most estimation problems in robotics are difficult because of (a) the nonlinearity in observation models; and (b) the lack of suitable probabilistic models for the process and observation noise. In this paper we develop a set-valued approach to estimation that overcomes both these limitations and illustrates the application to localization of multiple, mobile sensor platforms with range sensors. © 2008 Springer-Verlag Berlin Heidelberg.