Human Motion Likelihood Representation Map-Aided PDR Particle Filter

Abstract

Indoor localization systems are seeing increasing demand. Those for pedestrians are receiving a particular focus. Some of these systems leverage inertial measurement unit (IMU) data collected from a device worn by the pedestrian. The IMU data are used to predict and estimate the pedestrian’s location. This article proposes a system based on a pedestrian dead reckoning (PDR) and particle filter (PF) with a human motion likelihood grid and floor map filtering. We set an evaluation method by creating pedestrian ground-truth landmarks and by measuring statistical properties at these landmarks allowing the comparison to similar techniques. The algorithms, implementation, landmarks, and data used for the experiments of this article are available as free Open Source.