Evaluation of an acoustic detection algorithm for reactive collision avoidance in underwater applications

Abstract

This thesis sought to evaluate a vehicle detection algorithm based on a passive acoustic sensor, intended for autonomous collision avoidance in Unmanned Underwater Vehicles. By placing a hydrophone at a safe distance from a dock, it was possible to record the acoustic signature generated by a small motor boat as it navigated towards, and then away from the sensor. The time-varying sound intensity was estimated by Root Mean Square of the sound amplitude in discrete samples. The time-derivative of the sound intensity was then used to estimate the time to arrival, or collision, of the acoustic source. The algorithm was found to provide a good estimate of the time to collision, with a small standard deviation for the projected collision time, when the acoustic source was moving at approximately constant speed, providing validation of the model at the proof-of-concept level.