UAV Path Planning and Multi-Modal Localization for Mapping in a Subterranean Environment

The use of Unmanned Aerial Vehicles (UAVs), especially quadcopters, have become popular in academia and industry due to their small size and maneuverability. These UAVs can be programmed to autonomously execute missions that are usually difficult and risky for humans, such as subterranean exploration, infrastructure surveying, and even disaster response. However, inaccessible and remote environments pose a challenge in terms of navigation as they often lack access to Global Navigation Satellite System (GNSS) connections and lack features. To address these challenges, UAVs are equipped with multiple sensors to acquire different types of data. These include range, acceleration, and even images, which are fused to estimate a localization solution. The Autonomous Robotic Early Warning System for Underground Stone Mining Safety project, sponsored by Alpha Foundation, is conducted within the Statler College at West Virginia University (WVU). The project aims to map walls and pillars within a mine to analyze its structural integrity and safety. This thesis investigates the implementation of a path planning strategy to optimize the coverage of a wall and also the use of an error state Extended Kalman Filter (EKF) for sensor fusion to perform Simultaneous Localization and Mapping (SLAM). The experiments were carried out both in simulated and real world environments. In these experiments, the UAV was equipped with an Inertial Measurement Unit (IMU), laser altimeter, Ultra-Wideband (UWB) module (for ranging data), LiDAR for mapping, and an RGB-D camera to provide a Visual Odometry (VO) solution. In the simulation, the 3D reconstruction and odometry was compared to the ground truth, whereas the real experiment provided further insight into the strategy’s feasibility