SLAM in Low-Light Environments: Project Report

This project report benchmarks six RGB-based SLAM systems across five LaMARia sequences with varying difficulty and illumination: ORB-SLAM3, DSO, Kimera-VIO, OpenVINS, DPVO, and DPV-SLAM. Kimera-VIO is the only system to complete all five sequences and achieves the lowest relative pose error, but its absolute error grows because it lacks loop closure. DPVO and DPV-SLAM maintain tracking yet reach roughly 100 m absolute error in low light. ORB-SLAM3, DSO, and OpenVINS fail or diverge on most difficult sequences.