DROID-SLAM in the Wild
This addresses the critical limitation of traditional SLAM methods that fail in dynamic scenes, which is important for robotics and AR/VR applications, representing a strong incremental improvement over existing dynamic SLAM approaches.
The paper tackles the problem of robust real-time SLAM in dynamic environments by introducing a differentiable uncertainty-aware bundle adjustment method that estimates per-pixel uncertainty from multi-view visual feature inconsistency. The system achieves state-of-the-art camera poses and scene geometry in cluttered dynamic scenarios while running at around 10 FPS.
We present a robust, real-time RGB SLAM system that handles dynamic environments by leveraging differentiable Uncertainty-aware Bundle Adjustment. Traditional SLAM methods typically assume static scenes, leading to tracking failures in the presence of motion. Recent dynamic SLAM approaches attempt to address this challenge using predefined dynamic priors or uncertainty-aware mapping, but they remain limited when confronted with unknown dynamic objects or highly cluttered scenes where geometric mapping becomes unreliable. In contrast, our method estimates per-pixel uncertainty by exploiting multi-view visual feature inconsistency, enabling robust tracking and reconstruction even in real-world environments. The proposed system achieves state-of-the-art camera poses and scene geometry in cluttered dynamic scenarios while running in real time at around 10 FPS. Code and datasets are available at https://github.com/MoyangLi00/DROID-W.git.