A Multi-Modal SLAM Cooperative System for UVs operating in GNSS-denied Environments

Rapid, infrastructure-free situational awareness indoors is critical for disaster response but is hampered by GNSS and fixed infrastructure; key research gaps remain: (i) infrastructure-free cooperative localization techniques are under-explored, (ii) common simulator studies rely on idealized, stationary Ultra-Wideband (UWB) noise models that overestimate field performance, and (iii) there is limited integration and repeatable evaluation of SLAM-enabled leader–follower pipelines under realistic conditions. We present a simulation-driven leader–follower framework where a SLAM-equipped leader supports low-cost followers using an empirically derived, distance-dependent LOS/NLOS UWB measurement model and a dual-Monte-Carlo Localization with measurement-driven proposals. Across three SLAM methods, Hector SLAM yielded the lowest mapping RMSE; the proposed follower algorithm achieved the best LOS localization trade-off (RMSE = 0.206 m; CPU \((\approx)\) 58.7%; RAM \((\approx)\) 401 MB) and degraded to RMSE = 0.630 m under NLOS, comparing favorably with VINS-Mono. Simulator tests show CoppeliaSim demands far less CPU than Gazebo, and the results indicate that measurement-aware, infrastructure-free cooperative localization combined with lightweight SLAM offers a practical, resource-aware path toward deployable UV teams for search-and-rescue.