Multi-Sensor Data Fusion for Autonomous Vehicle Navigation and Localization through Google Earth

摘要

Inexpensive implementation of localization and environment mapping are critical issues for safe urban autonomous driving. We propose a multi sensors fusion model based on particle filter to fuse GPS, laser and camera information. We estimate the path tracking errors caused by IMU, odometer and steering sensors and modify errors by detecting landmarks from Google maps through laser and vision fused system. We have checked the feasibility of proposed method by doing many experiments in our campus. It is confirmed that the positioning errors are within 10 cm for several paths with more than 1km distance including up and down slopes.