Vehicle Localization for Autonomous Vehicles Using Environmental Magnetic Field Incorporating Artificial Land Markers

Vehicle localization is one of the key technical factors for autonomous vehicles (AV) on the road. It requires high accuracy, precision, and robustness to various conditions. Although methods using GNSS or Lidar are popular, organisms in nature such as birds have long used the earth’s geomagnetic field to navigate themselves. Here, we propose a method using an environmental magnetic field (EMF): the combined magnetic field of Earth’s geomagnetic field, and the magnetic field induced from nearby man-made objects such as manholes and underground structures. The proposed method includes generating an EMF map using Gaussian process (GP) regression, utilizing the Monte Carlo localization algorithm, and using magnetic markers in places where magnetic features are absent. The proposed method was validated in simulation. In the simulation, GP regression was able to generate an EMF map from actual magnetometer data, retaining the features of the original measurement. The simulation showed that accurate localization can be achieved by the MCL algorithm and adding magnetic markers enhances the accuracy. Combined with the existing methods, the proposed method can enhance the robustness of AV localization.