VLO: Vision-Laser Odometry for Autonomous Flight of Micro Aerial Vehicle

Abstract

This paper presents an onboard micro aerial vehicle (MAV) localization algorithm VLO using onboard multi-sensor system consisting of a camera, a laser scanner and an inertial measurement unit. On the basis of onboard processor, the VLO can operate in real time without any prior information and ground assistance. Besides, it shows a strong robustness since it can work in both small and large, indoor and outdoor environment. As the main sensing devices of this system, the camera and laser scanner generate different characteristic data. The VLO fuses these two kinds of data for a more sufficient information about environment. A filter and an optimizer are then designed to estimate the MAV poses with extra onboard sensors data. Finally, an incremental dense map is updated. Different with vision-based or laser-based odometry, this system has no requirements for environments such as strong texture or structured surroundings. The Gazebo-based simulated and real MAV systems are built together for algorithm validation. The simulated and real results show that our onboard odometry VLO performs strong robustness without any prior information and basic assumptions.