Objectives Multi-sensor fusion localization is widely used in indoor robot navigation. To overcome the brittleness of single-sensor localization in complex indoor environments with severe signal degradation and weak map features, this study designs a unified multi-source fusion localization framework.

Methods A robot localization system is constructed by integrating three complementary modalities: wheel/inertial measurement unit odometry, acoustic ranging, and single-line light detection and ranging (LiDAR) scan matching. A dynamic online calibration framework is proposed to decouple real-time pose output from quasi-static parameter estimation, which avoids the time-delay-induced projection errors and the divergence of coupled filtering that are common in classical approaches. A robust-loss-based adaptive weighted acoustic localization module is then embedded to suppress multipath and non-line-of-sight errors. On top of these modules, a particle-filter-based fusion algorithm with sliding-window optimization is built to fuse odometry, acoustic range and LiDAR map matching with adaptively determined observation confidence.

Results Experiments in long corridors, weak-feature regions, occluded zones and acoustic-degraded areas show that the proposed system significantly reduces long-term drift and abnormal jumps compared with odometry-only and acoustic-ranging-only localization methods.

Conclusions The proposed multi-source fusion scheme achieves stable, continuous and high-precision robot self-localization in complex indoor environments, and offers an effective solution for high-availability navigation on low-cost mobile platforms.