MDE-VIO: Enhancing Visual-Inertial Odometry Using Learned Depth Priors

Researchers have developed MDE-VIO, a new AI system that dramatically improves robot and drone navigation in challenging environments. By integrating learned depth priors into a visual-inertial odometry backend, it prevents system divergence where traditional methods fail. Crucially, it achieves this while staying within the strict computational limits of edge devices. The method reduces Absolute Trajectory Error by up to 28.3% on benchmark datasets like TartanGround and M3ED.