Vehicle Positioning in Road Networks without GPS

Abstract

Estimating vehicle position on road maps is important for many ITS applications. Advanced Driver Assistance System (ADAS) may expect robust positioning invariant to day time, weather or simplifications induced by the topological representations of road maps. This paper describes a particle filter approach used to achieve vehicle positioning on freely available road maps. Anti-lock Braking System (ABS) and Electronic Stability Program (ESP) sensor data are used in the motion update model. Measurement updates solely rely on vehicle heading. As our results indicate, our approach is able to cope with the odometry error accumulation. We also found that our methodology is able to successfully localize and track a vehicle with a median error of 3.6m in a road network made of 380km of drivable roads with a performance comparable to a high-end INS unit.