Integration of GNSS-receivers with dual foot-mounted INS in urban and indoor environments

Abstract

In safety-critical applications, including firefighter and law enforcement operations, infrastructure-free localization systems are typically required. These systems must provide accurate localization in all scenarios. Seamless indoor and outdoor localization and navigation, including in dense urban environments, are needed. Multi-sensor fusion algorithms constitute an integral part in all state-of-the-art indoor positioning systems. GNSS-receivers typically provide poor estimates of their own position uncertainty in dense urban and indoor environments, where significant position errors can be expected, which makes the design of a robust sensor fusion algorithm a challenge. Sensor fusion strategies for integration of a GNSS-receiver with footmounted inertial navigation systems (INS) are described and evaluated in this work. For a loosely coupled integration strategy, we suggest to use a cut-off criteria that governs when to discard the GNSS-positions and demonstrate that it can improve the position and heading accuracy in outdoor/indoor transition regions. Similarly, for a tightly coupled integration strategy, we suggest an approach with heavy-tailed measurement noise and demonstrate its capability to suppress inconsistent data and improve performance in the same regions.

Marginalized Bayesian filtering with Gaussian priors and posteriors

“…A onedimensional variant, in terms of mean and covariance, can be found in [14]. In [15], the presented techniques are used for robust tightly-coupled GPS-aided dead reckoning.…”