Investigation of autonomous docking strategies for robotic operation on intervention panels

This paper presents a localization strategy for an AUV which autonomously docks on intervention panels. A brief review of past research and working solutions of docking motivates the proposed choice of the strategy. It combines a ranging sonar localization technique featuring a modified particle filter at large distance and a visual model-based pose estimation using on-board camera at close distance to the docking panel. The particle filter solution is enhanced for effective exploration of the vehicle states without increasing the computational demand. It operates in an environment with a known map and has a real-time performance. The pose recognition algorithm derives from POSIT and is optimized for robustness. The visual docking utilizes a set of point-light markers which guarantees good accuracy at a large range of angles. Mentioned strategies are proven in a number of simulations as well as practical tests.