Multi-Sensor Based Adaptive Fusion Scheme for Position Estimation of Multirotor UAV Systems in GPS-Denied Environments

The use of Uncrewed Aerial Vehicles (UAVs) has seen significant growth across a wide range of applications over the past decade. UAVs are now commonly used for outdoor tasks such as surveillance, monitoring, inspection, delivery, and aerial photography, where Global Positioning System (GPS) is available and in most cases sufficient for localization, as large distances to obstacles make high-precision position estimation less critical. In contrast, indoor environments present a challenge for UAV localization, as GPS signals are unavailable, making reliable positioning a primary concern in such areas. This work focuses on the implementation of a novel Kalman Filter based estimation technique that delivers reliable position information to a coaxial octorotor Uncrewed Aerial Vehicle (UAV) in GPS-denied environments using inexpensive LIDAR technology. Accurate position estimation of a UAV is a primary challenge in navigation in a GPS-denied environment. Use of environmentally embedded sensors (such as motion tracking system) is a popular choice, however, their use is not feasible in many applications. Use of onboard sensors for localization is desirable in these applications and is an active area of research where a variety of onboard sensors, each with its own limitations and capabilities, such as LIDAR, cameras, optical flow, and other proximity sensors have been used. However, developing a robust method for localization that works in a variety of operating conditions is still challenging. The main contribution of this paper is to introduce an adaptive Kalman Filter framework for precise position estimate of a multirotor UAV system in GPS-denied environments solely utilizing LIDAR sensor and Inertial Measurement Unit (IMU) technology. The LIDAR sensors that were used within this research include a Unitree 4D LIDAR L1 in combination with a Garmin LIDAR-Lite V3. The system was developed on a Nvidia Jetson Orin Nano and a Cube Orange Flight Controller with PX4 Autopilot software. Point-LIO 3D SLAM was used to map the environment and obtain position information. Robot Operating System (ROS) performs the data communication between the sensors, the on-board computer, and the flight controller. Lastly, the system and its performance were validated by utilizing a high-precision OptiTrack Motion Capture system as a ground truth reference. The Root Mean Square Error (RMSE) analysis demonstrated that the fusion and predictive capabilities of the AKF algorithm significantly improved the position estimation accuracy.