GPS/odometry/map fusion for vehicle positioning using potential function

Abstract

In this paper, we present a fusion approach to localize urban vehicles by integrating a visual odometry, a low-cost GPS, and a two-dimensional digital road map. Distinguished from conventional sensor fusion methods, two types of potential functions (i.e. potential wells and potential trenches) are proposed to represent measurements and constraints, respectively. By choosing different potential functions according to data properties, data from various sensors can be integrated with intuitive understanding, while no extra map matching is required. The minimum of fused potential, which is regarded as position estimation, is confined such that fast minimum searching can be achieved. Experiments under realistic conditions have been conducted to validate the satisfactory positioning accuracy and robustness compared to pure visual odometry and map matching methods.