Passive acoustic tracking for behavior mode classification between surface and underwater vehicles

Autonomous underwater vehicles (AUVs) pose significant communication challenges: vehicles are submerged for periods of time in which speed-of-light communication is impossible. This is a particular problem on low-cost AUV platforms, on which acoustic modems are not available to get vehicle state or provide re-deploy commands. We investigate one possible method of providing operators with a communication line to these vehicles by using noise underwater to both classify behavior of submerged vehicles and to command them. In this scheme, processing of data from hydrophone arrays provide operators with AUV mode estimates and AUVs with surface vehicle behavior updates. Simulation studies were used to characterize trajectories for simple transect versus loiter behaviors based on the bearing and time to intercept (TTI). A classifier based on K-nearest-neighbor with dynamic time warping as a distance metric was used to classify simulation data. The simulation-based classifier was then applied to classify bearing tracking data from passive tracking of a loitering AUV and bearing and TTI data from passive tracking of a transecting boat based on field array data. Experiment data was classified with 76 % accuracy using bearing-only data, 96% accuracy for TTI-only data and 99 % accuracy for combined classification. The techniques developed here could be used for AUV cuing by surface vessels and monitoring of AUV behavior.