Acrome Robotics | University of Maryland Case Study

The University of Maryland conducted a research study in which they utilized the Acrome Stewart Platform for the purpose of vision-based landing of a multirotor Unmanned Aerial System (UAS) on a dynamically moving platform emulating the motion of a ship deck. The UAS employed in this experiment was a quadrotor, meticulously designed and constructed in-house, and equipped with an integrated avionics system specifically tailored for autonomous vision-based navigation. To achieve the task of detecting and estimating the pose of the moving platform, a monocular camera was employed.

The simulated ship-deck motion was meticulously replicated on the platform, capable of exhibiting six degrees of freedom, including oscillatory movements of up to 10 cm in linear motion and angular amplitudes of up to 30 degrees, all at speeds reaching up to 50 cm/s. The researchers obtained various ship-deck motion scenarios with increasing levels of intensity from the Office of Naval Research (ONR) and faithfully simulated them using the aforementioned moving platform. The primary objective was to gain insights into the complexities associated with the estimation and tracking of large, stochastic motions.

The accuracy and reliability of the results obtained were cross-validated using an independent Vicon system. The study's most significant finding revealed that it is indeed feasible to precisely track the motion of a moving ship deck relying solely on vision-based methods. Remarkably, the vision-based ship-deck position estimation demonstrated an exceptional level of accuracy, with errors consistently falling within the narrow range of 1-2 cm. However, it was noted that ship-deck tracking exhibited a phase lag error that proportionally increased with the frequency of ship-deck motion. Importantly, the researchers identified a viable solution to mitigate this error: by estimating the ship-deck's velocity and incorporating it as feedback in the tracking controller, the phase lag error could be substantially reduced. This groundbreaking research showcases the potential of vision-based technologies in the field of ship-deck tracking and navigation.