Integrated Emergency Self-Landing Method for Autonomous UAS in Urban Aerial Mobility
Bingze Xia, Iraj Mantegh, Wenfang Xie
Abstract
Applications of Unmanned Aircraft Systems (UAS) are growing fast in many areas, including Advanced Air Mobility (AAM) which requires the safe integration of aerial vehicles in an airspace that is shared by many other operators. Autonomous or automated UAS will have to a large extent rely on onboard or ground-based guidance and navigation, with no or minimum operator intervention, to perform their operations. Using GPS/GNSS data is a common way of navigation for existing UAS. Safe autonomous UAS operations require the capability for safe landing in case of abnormal situations, such as loss of GPS signal or weather effects. In this paper, a new automatic safe-landing method is proposed that can perform in GPS-denied or degraded environments. A multi-layer method is designed that applies the vehicle's Inertial Navigation System to navigate to a safe landing zone, and then with an Artificial Intelligence-based approach utilizes optical search and object detection to locate the landing area for landing. A 3D depth camera and fully convolutional neural network method are used to recognize the landing features and obstacles, integrated with Markov Decision Process to guide the aircraft safely without collisions towards the landing zone. A series of simulations are presented to test and validate the proposed system.