Particle filter divergence monitoring with application to terrain navigation
Particle filters are an efficient tool for bayesian estimation in non-linear models. However, under certain circumstances, they are subject to divergence. Increasing the number of particles is not always possible so it’s interesting for many applications to evaluate the reliability of the estimate provided by the filter. In terrain navigation, trusting an erroneous estimate can be problematic for obvious reasons. This paper introduces a framework for detecting filter divergence in the case of scalar measurements. The detector is based on a sequential change detection algorithm and we illustrate its performance on several terrain navigation scenarios.