Advanced PANDA for high speed autonomous ambient noise data collection and boat tracking - system and results

Many underwater acoustic recording applications require a high speed data acquisition system that is not contaminated by noise from the support vessel. This calls for autonomous recording system that is self-contained. Many of such systems are either bulky in size such as moored data buoy, or conventional bottom mounted system with acoustic release system; small in data capacity such as miniature acoustic recorder or tags; or needs high maintenance costs such as AUV monitoring. We present an alternative autonomous system that is cost effective, small and provides directivity capability. The Advanced Pop-up Ambient Noise Data Acquisition (A-PANDA) is the next generation underwater acoustic recording system developed by the ARL. The system now has increased CPU power, enhanced sampling rate, increased numbers of channels, digital control, oven controlled real time clock, compass, and highly programmable through a home-built trusted-realtime scheduler. Each A-PANDA has enough data storage and battery capacity to allow deployment periods from tens of hours of continuous recording, to weeks of burst recording. System retrieval can be easily performed through an acoustic release or a pre-programmed time release. A-PANDA can be fully recovered and leaves no anchor or deadweight behind, introducing minimum or no environmental impact to the site of study. A typical deployment procedure involves a small vessel ferrying the A-PANDA to a desired location, deploying the system, leaving the vicinity and returns only for retrieval; hence the acoustic recording is free of self-contamination from the surface vessel. We present work on the use of the A-PANDA to form a random array for the purpose of tracking surface vessels. Since each A-PANDA has a triangular array with three hydrophones, our approach uses high resolution techniques such as the MUSIC algorithm to provide DOA estimations to targets from each A-PANDA. Localization can be achieved by combining the DOA estimations for all of the A-PANDAs. The developed algorithm makes use of cosine packet transforms to identify likely tonal content from surface vessels, prior to MUSIC beamforming, such that only chosen frequencies are beamformed. This reduces the computation, and reduces the clutter in the beamformer output. Finally a Myriad filter is used to track the output of the beamformer over time. When compared against a model based Kalman filter the Myriad filter requires no dynamic model of the surface vessel, and does not have a tracking lag. Simulation results and field trial results to test the DOA estimation performance of a single A-PANDA show a promising tracking performance with an average bearing error of 2 degrees. The maximum range was limited to 650 m due to physical constraints.