Large-Scale Robotic 3-D Mapping of Urban Structures

Abstract

This article presents results for building accurate 3-D maps of urban environments with a mobile robot based on the Segway scooter. The goal of this project is to use robotic systems to rapidly acquire accurate 3-D maps which seamlessly integrate indoor and outdoor structures. Our approach uses an efficient implementation of the global scan alignment algorithm of Lu and Milios in order to integrate GPS, IMU, and laser data into globally consistent maps. The 3-D models acquired by the robot are analyzed for navigability using a multi-resolution evidence grid approach, and visualized using a meshing algorithm adapted from the computer graphics literature. Results are presented for a number of environments which combine indoor and outdoor terrain.