Mitigating unbalanced GDoP effects in range-based vehicular Cooperative Localization

Abstract

In the context of Vehicular Ad Hoc NETworks (VANETs), we investigate the problem of range-based Cooperative Localization (CLoc), which combines absolute position information from on-board Global Navigation Satellite System (GNSS) with relative distance measurements to fellow mobile vehicles, for instance based on Impulse Radio - Ultra WideBand (IR-UWB). CLoc accuracy usually suffers from poor Geometric Dilution of Precision (GDoP) along the dimension orthogonal to the road, due to highly constrained VANET mobility and topology. This paper considers benchmarking and even combining two different approaches to resolve this problem. The first solution accounts for the vehicle's heading, making use of an Inertial Measurement Unit (IMU). The second one assumes prior lane detection, along with associated road information (e.g., lane width stored in a digital map). The data fusion scheme is based on a modified Extended Kalman Filter (EKF) capable of filtering out the unbounded noise process integrated at the gyroscope and/or using lane boundaries to numerically refine the density of posterior state estimates. The proposed strategies are shown to boost the probability to reach an accuracy of 0.5 m from 60% up to 85– 90% in comparison with conventional range-based approaches.