11.6.2 Sampling-Based Approaches

Since probabilistic I-space computations over continuous spaces involve the evaluation of complicated, possibly high-dimensional integrals, there is strong motivation for using sampling-based approaches. If a problem is nonlinear and/or non-Gaussian, such approaches may provide the only practical way to compute probabilistic I-states. Two approaches are considered here: grid-based sampling and particle filtering. One of the most common applications of the techniques described here is mobile robot localization, which is covered in Section 12.2.