Previous Abstract Return to Session E9 Next Abstract

Session E9: PNT for Autonomous Systems

Austere Runway Navigation: Autonomous Landing of a small UAS using Two-Point Visual-Servoing
Liam McKenna, Clark Taylor, Jeremy Gray, ANT Center, AFIT, Center; Jason Bowman, AFRL/RQ
Location: Ballroom B
Date/Time: Wednesday, Jun. 3, 9:35 a.m.

Fixed-wing autonomous landing presents a high safety risk, demanding high performance to avoid incidents. In austere environments, autonomous landing is particularly challenging (conventional runways are typically well-marked and equipped with lighting systems, sometimes additionally ground-based localization). For a fixed-wing sUAS vehicle to land on an “Austere Runway”, the guidance and control system cannot rely on conventional sources of truth to perform a safe landing. To address this challenge, a methodology for sUAS guidance and control during landing on an Austere Runway is proposed, which uses egocentric information such as inertial measurements and visual cues, which will be known a priori.
In this work, a simple runway endpoint detection algorithm is demonstrated to provide control inputs for the landing of a fixed-wing sUAS on approach to an Austere Runway. It is assumed that the selected runway’s location and appearance are known a priori (such as that obtained from satellite imagery). The experimental setup includes a detailed simulation of a simple sUAS sensor suite composed of an Inertial Measurement Unit (IMU), Computer Vision System, Barometer, & Airspeed Sensor, which are integrated into a Software-In-The-Loop (SITL) simulation framework for aircraft dynamics and flight control. Visual feedback about the environment landmarks is produced using a simplified mapping from world position to inferred position in the camera frame. The sensor information is combined to produce inferential feedback about the aircraft’s approach to a closed-loop controller. The resulting system leverages computer vision and visual servoing to control the sUAS into the proper runway alignment and glide path in an austere setting.
A robustness analysis is performed to evaluate the performance of the visual-servoing controller. Success criteria for a landing is defined according to designated touchdown velocity and pose requirements. A set of initial conditions are defined for an in-flight sUAS approaching the austere runway and evaluated, including either alignment or rotation offsets in approach start. Final results from our analysis shows that the system is capable of robust landing control when the runway is in view, based on fulfillment of the landing criteria, even if the aircraft’s approach is entirely misaligned with the runway centerline.
The implications of this system extend to fully autonomous landing of platforms in austere or minimally-prepared and GPS-denied environments, which could provide immense capability for the distribution of autonomous platforms in environments which are unsuitable for conventional systems.

Previous Abstract Return to Session E9 Next Abstract