A Real-Time Simulation Environment for Aircraft Vision-Based Autonomous Landing System Design

In the strive for always more safe and autonomous operations for civil aviation the landing phase remains by far the most challenging. Robustifying the already existing automatic landing system, classically based on exteroceptive sensors, with vision-based solutions have recently demonstrated promising perspectives. Nonetheless, the design and validation of such novel algorithms comes at enormous costs and logistic effort for installation, in-flight data collection and testing. In this paper a novel simulation platform, named SCHEMIN, for rapid conception, prototyping and testing of vision-based control solutions at very low cost and time-scale is presented. The software/hardware architecture is implemented to guarantee tight and high-fidelity coupling between flight dynamics and visual environment models, allowing for state observation, computer vision processing, virtual sensor data fusion and full aircraft control. The system is also able to easily and safely simulate system failures and off-nominal degraded conditions at any level. The platform has been validated in open-loop against real flight-test data recordings. The proposed architecture will also play an important role for conception and testing of learning-based algorithms for autonomous landing in a close future.