An Improved Road Network Modeling and Map Matching for Precise Vehicle Localization

  • Authors:

  • Chenhao Wang;Zhencheng Hu;Naoko Hamada;Keiichi Uchimura

  • Affiliations:

  • Graduate School of Science and Technology, Kumamoto University, Kumamoto, Japan 860-8555;Graduate School of Science and Technology, Kumamoto University, Kumamoto, Japan 860-8555;Graduate School of Science and Technology, Kumamoto University, Kumamoto, Japan 860-8555;Graduate School of Science and Technology, Kumamoto University, Kumamoto, Japan 860-8555

  • Venue:
  • ICIRA '09 Proceedings of the 2nd International Conference on Intelligent Robotics and Applications
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper presents a precise and robust road model for Map Matching (MM). Map matching approach is generally used to move vehicle localization output to the nearest road segment in order to improve the accuracy of vehicle lateral localization. Most of the MM approaches adopted piece-wise line segment road network model, which will generate large bias at curve segment or turning point on the intersection. In this paper, a two-dimensional parabola road network model is employed in order to correctly represent vehicle's real state (position and orientation). Furthermore, an advanced longitudinal MM approach is also proposed here by comparing current road geometry model with visual lane detection results to improve longitudinal accuracy. Simulation and real road tests verified the proposed approaches.