AUV navigation around jacket structures I: Relative localization based on multi-sensor fusion

Underwater jacket structures or support legs of on-water platforms, such as ports and oil rigs, need to be periodically inspected for maintenance, environmental monitoring, and security reasons. Autonomous underwater vehicles (AUVs) can potentially make these tasks more inexpensive and reliable compared to conventional methods that involve the use of divers and remotely operated vehicles. This paper proposes a robust and practical selflocalization method for an underwater vehicle navigating around jacket structures, where the performance of conventional acoustic positioning suffers from multipath degradation. The key idea is to stochastically update the vehicle’s horizontal position and heading relative to the structures using two types of perceptional sensors, sonar and camera, assuming that the configuration of the structure is known. The performance of the method was verified with tank experiments using a jacket mock-up and the AUV Tri-Dog 1.