TBV Radar SLAM – Trust but Verify Loop Candidates

Robust SLAM in large-scale environments requires fault resilience and awareness at multiple stages, from sensing odometry estimation to loop closure. In this work, we present TBV (Trust But Verify) Radar SLAM, a method for radar that introspectively verifies closure candidates. achieves high correct-loop-retrieval rate by combining place-recognition techniques: tightly coupled place similarity uncertainty search, creating descriptors origin-shifted scans, delaying selection until after verification. Robustness false constraints is achieved carefully verifying selecting the most likely ones constraints. Importantly, verification are carried out registration when additional sources of evidence can easily be computed. We integrate our retrieval with robust pipeline within pose graph framework. By evaluation on public benchmarks found 65% lower error than previous state art. also show it generalizes across without needing change any parameters. provide open-source implementation https://github.com/dan11003/tbv_slam_public