An architecture for passive joint localization and structure learning in reverberant environments

Area of Science:

• Acoustics
• Signal Processing
• Robotics

Background:

• Passive localization and tracking of mobile emitters in complex acoustic environments is challenging.
• Unknown structural features and occluders can disrupt line-of-sight, necessitating estimation of reflected raypaths.

Purpose of the Study:

• To develop a robust method for passive localization and tracking of mobile emitters.
• To jointly learn the reverberant three-dimensional (3D) acoustic environment, including reflective boundaries.
• To address the challenge of unknown structural features and potential occlusions.

Main Methods:

• A multistage global optimization and tracking architecture is proposed.
• Particle Swarm Optimization (PSO) is employed for joint estimation of the environment and emitter location.
• A 3D Hough transform-inspired algorithm is used for initial boundary localization.

Main Results:

• The developed architecture effectively estimates emitter position and acoustic environment features.
• The method demonstrates improved performance compared to approaches that do not leverage emitter motion.
• Reliability is validated in a reverberant watertank testbed simulating underwater acoustics.

Conclusions:

• The proposed multistage architecture provides a generalizable solution for passive emitter localization and environmental mapping.
• The joint estimation of emitter position and acoustic environment is feasible even with occlusions.
• The approach offers a significant advancement for acoustic sensing in challenging, reverberant settings.