Memorial University of Newfoundland

Purpose

Vertical take-off and landing (VTOL) aircraft are often used for transporting goods to and from inaccessible areas, surveillance, ground support, etc. which can be enhanced and de-risked by increasing the autonomous capabilities of the vehicles. One of the main challenges in flight autonomy is the detection of landing zones (LZ) and obstacles to facilitate the safe, reliable, and quick landing of VTOL vehicles. The current research for LZ detection makes use of simple geometric rules to identify the safe LZs (e.g., can the helicopter fit in the available obstacle-free area). These approaches may not provide reliable results since the terrain characteristics such as slope, type of terrain, type of surrounding obstacles that affect landing safety are not taken into account. In this work, a sensing and dedicated computing module for real-time LZ identification and obstacle detection using deep learning will be developed. A novel dual-sensor solution, combining Laser Detection and Ranging (LiDAR) and optical sensors, will be used, which allows identifying both geometric features and visual information of a scene with a high level of confidence. The performance of the proposed system will be evaluated using flight data captured using a VTOL drone.