UniSim-SLAM: Feed-Forward SLAM with Unified Sim(3) Optimization

2026-08-03

AI Summary

This work addresses geometric inconsistency and trajectory drift in long-sequence SLAM caused by input-view dependency in existing geometric foundation models, as well as the trade-off between low-latency two-view tracking and the richness of multi-view constraints during online deployment. The authors propose a tightly coupled front-end/back-end SLAM system: the front-end performs lightweight two-view keyframe tracking, while the back-end periodically optimizes multi-view submaps. A unified Sim(3) multi-level factor graph jointly refines global keyframe and submap poses. Novel depth-statistics-based scale anchoring and inter-submap similarity constraints are introduced to fuse predictions across heterogeneous local coordinate frames, enabling consistent cross-submap modeling. The method achieves state-of-the-art accuracy on TUM RGB-D and 7-Scenes under uncalibrated settings, reducing trajectory errors by 38.5% and 45.9%, respectively, compared to prior best approaches.

Abstract

Recent geometric foundation models enable feed-forward inference for SLAM, but their predictions are strongly dependent on the input view set, which leads to geometric inconsistencies and trajectory drift when results are chained over long sequences. Online deployment further exposes a trade-off between the low latency of two-view tracking and the constraint richness of multi-view inference. We introduce UniSim-SLAM, an integrated system that runs lightweight two-view keyframe tracking in the frontend and performs periodic multi-view submap refinement in the backend. To combine predictions defined in heterogeneous local coordinates with inconsistent scales, we formulate a unified multi-level factor graph on $Sim(3)$ that jointly optimizes global keyframe poses and submap poses. The graph integrates temporal view-to-view odometry edges, view-to-submap bridge edges with depth-statistics scale anchoring, and submap-to-submap tie and scale constraints to enforce consistent similarity relations across submaps. Experiments on TUM RGB-D and 7-Scenes show that UniSim-SLAM achieves state-of-the-art accuracy in the uncalibrated setting, reducing trajectory error by $38.5\% $ on TUM RGB-D and $45.9\%$ on 7-Scenes compared to prior best results. Project page: https://vision3d-lab.github.io/unisim-slam/

Problem

• SLAM
• geometric inconsistency
• trajectory drift
• feed-forward inference
• scale ambiguity

Innovation

• Sim(3) optimization
• feed-forward SLAM
• multi-view submap refinement
• scale-consistent pose estimation
• factor graph