VSLAM and Vision-based Approach and Landing for Advanced Air Mobility

Advanced Air Mobility (AAM) aircraft have many challenges in landing accurately and
safely in urban, suburban, and rural environments. Localization in large and open rural
environments could utilize GPS, but AAM aircraft in urban environments will encounter GPS
degradation. Another challenge involves flight operation time, i.e., flying during the day or at
night. There are different guidelines, landmarks, and landing light configurations at runways,
heliports, and vertiports for daytime and nighttime applications. Tailoring feature detection
methods for AAM approach and landing during the day and night pose different issues and
challenges. It is easier to detect edges, lines, and other runway markers during the day than at
night. Conversely, it is easier to see landing light configurations and patterns at nighttime than
daytime. Consequently, utilizing the same feature detector for daytime and nighttime operations
may not be feasible. This paper focuses on a vision-based precision approach and landing (PAL)
by comparing ORB SLAM 2, a Vision Simultaneous Localization and Mapping (VSLAM)
algorithm, and a novel EKF that combines onboard IMU measurements with coplanar pose
from orthography and scaling with iterations (COPOSIT). Conducting unmanned aerial system
(UAS) flight tests at NASA Armstrong Flight Research Center (AFRC) with landmarks and
fiducials distributed around the landing zone provides a simulated AAM approach and landing
data to test vision-based PAL methods to provide Alternative Position, Navigation, and Timing
(APNT) solutions for AAM PAL applications. The novel vision-based PAL EKF with IMU and
COPOSIT provides accurate state estimation when distributed landmarks and fiducials are in
the field of view.