Interacting Multiple Model Particle Filters for GPS/INS Integration

Abstract

This paper presents the interacting multiple model (IMM) particle filters with application to navigation sensor fusion. Performance evaluation for various single model nonlinear filters as well as nonlinear filters with IMM framework is carried out. A high gain (high bandwidth) filter is needed to response fast enough to the platform maneuvers while a low gain filter is necessary to reduce the estimation errors during the uniform motion periods. The IMM estimator obtains its estimate as a weighted sum of the individual estimates from a number of parallel filters matched to different motion modes of the platform. Based on a soft-switching framework, the IMM algorithm allows the possibility of using highly dynamic models just when required. The use of an IMM allows exploiting the benefits of high dynamic models in the problem of vehicle navigation. Some results presented in this paper confirm the improvements.