Hardware- and Vision-in-the-Loop Validation of Deep Monocular Pose Estimation for Autonomous Maritime UAV Flight

A new hardware-validated framework allows unmanned aerial vehicles to fly fully autonomously indoors while simulating the visual conditions of a ship at sea, according to research posted to the arXiv preprint server on June 17, 2026 [1][2]. The system, described in a paper by Maneesha Wickramasuriya, addresses a persistent barrier in maritime drone development: validating autonomous flight on actual vessels is expensive, constrained by weather, and carries operational risk [1][2]. The proposed vision-in-the-loop setup renders photorealistic maritime views that are processed onboard the drone by a deep transformer-based monocular pose estimator [2]. Delayed vision measurements are then fused with high-rate inertial measurement unit data through a delayed Kalman filter, producing consistent state estimates for geometric control [2]. The authors report that the framework captures embedded effects such as perception latency, asynchronous updates, and computational limits — factors absent from software-only simulations [2]. Experiments covering autonomous takeoff, trajectory tracking, and landing demonstrated stable closed-loop flight [2]. The work is positioned as a safe, hardware-realistic intermediate stage for developing maritime UAV autonomy before moving to shipboard trials [2]. The paper was submitted to arXiv’s robotics section at 15:18:11 UTC and is available as a 1,890 KB PDF [1]. arXiv, which began in 1991, is an open-access repository of electronic preprints that are moderated but not peer-reviewed; it surpassed two million articles by the end of 2021 and currently receives about 24,000 submissions per month [6]. The platform also hosts arXivLabs, a framework for community-built tools that appear on abstract pages, though new project proposals are temporarily paused while the development team modernizes arXiv’s infrastructure and migrates systems to the cloud [3][4].