Velocity estimation using forward looking sonar

Abstract

The thesis investigates a method to estimate the forward velocity and heading rate of an autonomous underwater vehicle (AUV). Through relatively new technologies small AUVs are now able to mount a Forward Looking Sonar (FLS) on the vehicle's nose. This can be used for obstacle avoidance and feature based navigation. The sensor can also be used to estimate motion of the AUV, which can be useful for undersea navigation. The thesis focuses on a template matching technique used in computer vision. Two sequential sonar images are compared with the goal of finding the rotation and translation that best correlates the first to the second sonar image. The transformation which maximizes the correlation coefficient is then converted to forward velocity and heading rate through motion analysis. Experimentation shows that the method provides accurate estimates for both the forward velocity and heading rate of the AUV. Accuracy of the estimates for forward velocity was at the limitation of the resolution of the sonar. Using velocities estimated through image processing applied to FLS images entirely with software, the weight and energy resources currently required by standard measurement techniques could be used to increase the vehicles endurance or for additional payload capacity. Another benefit would be the reduction in acoustic and electrical interference with the FLS and side scan sonar, which would improve the vehicle's obstacle avoidance and mine-hunting capability. The vehicle could become more flexible in its capability to support additional roles vice specific missions. This method holds the promise for permitting smaller AUVs with a FLS to navigate undersea more accurately.