Sensor fusion based vSLAM system for 3D environment grid map construction

Abstract

The objective of this paper is to develop a mobile robot system to construct stereo vision and sonar fusion grid maps of 3D environment structure, which is based on a visual simultaneously localization and mapping (vSLAM) system. We first solve the problem of large textureless area for stereo camera by projecting laser dots onto them, and design the system algorithm to avoid the visual features of vSLAM disturbed by the projected laser dots. Meanwhile, some functions are integrated to make the projection efficient and avoid laser dots projected toward humans. The 3D grid map construction methods are also well designed to ensure the confidence of each grid even when stereo and sonar range data are noisy. We successfully demonstrate the concept through experiments, and the grid map is sufficient for future navigation and path planning tasks.