Avoiding marine vehicles with passive acoustics

Passive acoustics represent a low-cost approach to detecting and avoiding other marine vehicles using limited computing and power resources. We present a robust safety behavior for long-term environmental sensing of rivers and estuaries demonstrated on a smartphone-driven, diving robot. The detection method is based on energy detection from the hydrophone output. Field testing demonstrates detection and avoidance of motorized vessels of various sizes with 94.1% accuracy and a 12.1% false positive rate. Dives are triggered with a median approach distance of approximately 700 m, and they are completed within 500 m of the detected vehicle.