MapDrift

MapDrift is a desktop + web app that automatically evaluates SLAM quality across repeated runs and environments, then flags where your map and pose estimates are silently degrading. You import rosbag/MCAP logs (or upload from CI), select a baseline run, and MapDrift produces a “drift report”: loop-closure health, map consistency heatmaps, trajectory divergence over time, and regression alerts tied to code changes. It’s built for robotics teams who don’t have ground truth and are tired of arguing over subjective RViz screenshots. The MVP focuses on repeatability testing: same route, different day/sensor settings/firmware, and it quantifies stability with clear pass/fail thresholds you define. It also stores run metadata (sensor serials, calibration hashes, parameter sets) so you can actually reproduce failures. This is not a SLAM algorithm; it’s the missing QA layer around SLAM.