MUSE-SLAM: Multi-Sensor Underwater State Estimation SLAM

• Larissa e S. Gomes FURG
• Paulo L. J. Drews FURG

Resumo

This paper presents MUSE-SLAM, a robust multi-sensor fusion framework designed for state estimation of Autonomous Underwater Vehicles (AUVs) operating in GPS-denied and visually degraded environments, such as underwater. The system is based on an Extended Kalman Filter (EKF) and four complementary sensors to overcome the drift commonly associated with traditional dead-reckoning methods. The sensors are: an Inertial Measurement Unit (IMU) for high-frequency motion tracking, a Doppler Velocity Log (DVL) for velocity relative to the seafloor, a pressure sensor for absolute depth measurement, and a mechanically scanned imaging sonar (MSIS) used to detect environmental landmarks. An important aspect of MUSE-SLAM is its asynchronous sensor integration strategy, which processes timestamped measurements in chronological order, allowing for varying sensor update rates and accommodating communication delays. The approach is validated using a real-world dataset. Experimental results highlight the clear benefits of the multi-sensor setup: when comparing different sensor combinations, the full configuration (using all four sensors) showed significantly improved performance.

Palavras-chave: Simultaneous localization and mapping, Tracking, Sonar measurements, Sea floor, Sonar, Time measurement, Sensor systems, Sensors, State estimation, Sonar detection, SLAM, underwater, robotics, sonar, multi-sensor