Pedestrian inertial navigation with building floor plans for indoor environments via non-recursive Bayesian filtering

In this paper, we present a pedestrian navigation method for indoor environments that uses a novel non-recursive Bayesian filter. Developed method takes building floor plans into account to compensate for error-prone heading measurements provided by a foot mounted micro electro-mechanical system (MEMS) inertial measurement unit (IMU). Zero velocity updates (ZUPT) are implemented intuitively to estimate sensor bias/drifts and subsequently dead reckoning (DR) equations calculate the pedestrian location from corrected velocity. Despite high accuracy in estimating traveled distance, navigation solution is unacceptable due to incorrect heading estimates. Methods to fix the heading estimation problem typically exploit particle filter (PF) framework. Limited computational power of IMU board, however, cannot afford a PF implementation therefore a novel approach alternative to recursive Bayesian filtering (RBF) is developed. Proposed filter uses the static floor plan density distribution and an extended motion model to eliminate heading errors in pedestrian location estimation.