Observability-Aware Mobility Parameter Estimation for Seamless Indoor-Outdoor Vehicle Positioning

Abstract

Seamless indoor-outdoor positioning remains challenging due to heterogeneous sensing, multipath and attenuation indoors, and time-varying GNSS quality outdoors. This paper presents a simulation-based framework for mobility parameter estimation of a 4-wheel vehicle transitioning from an indoor environment (Building A) through an outdoor segment and into a second indoor environment (Building B). The proposed system fuses GNSS (GPS/SBAS), BLE RSS, IMU, and odometry using a nonlinear state-space vehicle model. This study implement and compare the Extended Kalman Filter (EKF), Unscented Kalman Filter (UKF), Particle Filter (PF), and an Interactive Multiple-Model EKF (IMM-EKF) that combines rectilinear and curvilinear motion models. In addition, This work examine state observability under different sensor-availability conditions using an EKF-based local linearization approach. Simulation results show that multi-sensor fusion improves trajectory estimation under noisy observations; in the tested scenario, IMM-EKF achieves the most consistent overall accuracy (e.g.,  σ_x=2.51m, σ_y=3.11). Observability analysis indicates full observability when IMU is available (rank ), while GPS-only and odometer-only configurations are not fully observable. These findings support robust indoor-outdoor positioning by combining heterogeneous sensing with nonlinear filtering and observability-aware analysis.