SLAMReplay

SLAMReplay is a desktop + web app for robotics teams to capture, version, and deterministically replay SLAM runs across machines and code revisions. It ingests ROS/ROS2 bags, sensor calibrations, TF trees, and SLAM configs, then produces a “replay capsule” that can be rerun in a container to reproduce mapping/localization outcomes. The app automatically computes run-to-run diffs: trajectory drift, loop-closure events, map entropy/coverage, CPU/GPU usage, and failure signatures (e.g., lost tracking, bad time sync). It includes a small CI agent that runs nightly replays on a curated dataset so regressions are caught before field tests. This is not a new SLAM algorithm; it’s a reliability and debugging product that turns SLAM into something testable, comparable, and auditable—especially valuable when multiple engineers tweak parameters and sensor setups.