Low-cost single-beam SONAR localization in AUVs: Systematic experiments with acoustic and optical perception

Abstract

This work presents a low-cost acoustic sensing framework for underwater robot (UWR) localization and perception, using single-beam sonars (SBS) and an external forward-looking sonar (FLS) for confined environments in place of costlier Doppler velocity logs (DVL) and onboard imaging suites. Instead of utilizing an SBS for depth estimation alone, an SBS unit each is also used in surge and sway directions for localization. A Kalman filter fuses sparse SBS measurements with inertial sensing via moving averaging and confidence-based gating, suppressing multi-path and orientation-driven anomalies. SBS-assisted fusion achieves sub-meter accuracy, reducing localization errors by more than 70% versus inertial sensing alone via acoustic corrections comprising just 27% of measurements. To further mitigate potential SBS tracking losses, an FLS is proposed to be deployed on an external parent platform, such that it can independently track multiple UWRs. Parent–child tracking achieves 0.063 m average error when combined with SBS estimates. Stabilized pose estimates anchor optical mosaicking and acoustic point-cloud reconstruction, cutting required image data to about 1% for mosaicking and 5%–25% for point clouds versus unassisted collection. Per-vehicle sensor cost falls to approximately $2000 for a ten-vehicle fleet, an order of magnitude below DVLs and ultra-short baseline alternatives.