Improving navigation using active landmark recognition

Navigation of underwater vehicles is a very demanding task, especially in confined environments. The risk of damage, expense and sometimes danger of recovering a vehicle makes improving the system's awareness of its environment a research priority. We propose here a novel approach for underwater navigation systems that combines data fusion from a variety of conventional on-board sensors (such as heading, depth, tilt, short-baseline sonar positioning), with an active vision system to perform attentive landmark recognition using optical and sonar imagery.