Finite-Horizon URTSS-Based Position Estimation for Urban Vehicle Localization

Abstract

Vehicles in urban areas are facing the problem of lacking Global Positioning System (GPS) signals in the urban canyon environments. In this article, we present a finite-horizon Unscented Rauch-Tung-Striebel Smoother (URTSS)-based position estimation method for urban vehicle localization, which uses information from past, present, and near-future. To estimate vehicle pose, a nonlinear constant velocity state-space model is established. Sensors can only obtain immediate information, and we describe the near-future information obtained via predetermined paths using “pseudomeasurements.” The noise characteristics of pseudomeasurements are subsequently modeled and analyzed based on the fact that the uncertainty of pseudomeasurements increases over time. Using the URTSS, we design a finite-horizon estimator to achieve the fusion of past, present, and near-future information, which can perfectly yield a vehicle position estimation. Numerical simulation and road experiments are conducted to show that the proposed scheme is able to provide accurate position estimation in finite horizon when GPS signals are temporarily lost.