A real-time vision-based stationkeeping system for underwater robotics applications

We describe a vision-based system for automatic ROV/AUV stationkeeping, based on an algorithm for computing the motion of a camera (and the vehicle) from the images of the ocean bottom. The vision module estimates the vehicle position (displacement from station) and feeds this information to the control system in order to generate the appropriate thruster signals. The current testing platform is a one-degree-of-freedom thruster system, submerged in a water tank and moving on a linear track. A PID controller has been used without any attempt to optimize transient response performance by choosing the best gains. The performance of the optical stationkeeping system is demonstrated in three selected experiments by comparing results with the data gathered from an optical encoder position sensor