Aerial Robots On the Way to Underground: An Experimental Evaluation of VINS-Mono on Visual-Inertial Odometry Camera

In recent years, applications of unmanned aerial vehicles (UAV) have been well received by a wide range of users due to its versatility and cost-effectiveness. When the UAVs operate outside without any surroundings, it can be possible to integrate global positioning system (GPS) into the UAV. However, GPS is ineffective to determine the position of UAV when the signal is weak or absent, such as operating in an indoor environment and tunnel network. To accurately determine the position of a UAV, another approach is to endow different sensors such as LIDAR, inertial measurement unit (IMU) and camera. In this paper, we present an experimental evaluation of a visual-inertial odometry application using VINS-Mono state estimator coupled with visual-inertial sensors. The objective of the experiment is to determine the effectiveness of the setup in an indoor environment, where the GPS signal is fully absent. In the experiment, a UAV capable of self-localization has been constructed with the integration of onboard computer, camera, and IMU. First, camera and IMU calibrations, as well as their hardware synchronization, have been set. Second, experimental evaluations have conducted to compare the UAV localization performance with ground truth data at motion analysis laboratory equipped with OptiTrack tracking system. Finally, a comparison study is given to determine the robustness and reliability of VINS-Mono state estimator and the UAV system using various flight velocities and environment features settings. Based on extensive evaluations, it is capable to localize a UAV in a GPS-denied environment, with an average root mean square error kept under 25 cm over an average flight duration within 2 minutes.