Deployment of a passive acoustic collision alarm for autonomous underwater vehicles

Abstract

Autonomous underwater vehicles (AUVs) operate in increasingly busy environments, and avoiding collisions with ships is an important aspect of the vehicle's decision framework. Advances in low-cost AUV technology has further raised interest in solutions that minimize the need for specialized equipment. A proposed solution is to passively detect and track boats based on their noise. We have developed and demonstrated an algorithm that uses changes in acoustic power to estimate time to collision. A cylindrical propagation model is used for shallow-water environments, to relate measured power with range from a source. The time-derivative of this result is computed to estimate the time to closest approach, for a ship with unknown but constant acoustic source level. Experiments were performed with vessels of various characteristic frequencies and source levels. Acoustic measurements were recorded through off-the-shelf hydrophones. Estimates from the acoustic method were first compared with GPS-based measurements, and later with the AUV's on-board navigation data. Test missions were deployed to alter AUV behavior in response to approaching acoustic sources. Successful estimates and AUV responses were recorded, although improved signal processing and noise filtering is recommended for future implementations. [Work supported by Lincoln Laboratory and DARPA.]