Modified Data Association for Underwater Localization and Map Building

The simultaneous localization and mapping (SLAM) algorithm could make up for the disadvantages of underwater navigation methods based on priori map, hence makes underwater vehicles to be truly autonomous. A modified data association method is proposed to lighten the systemic computational burden and improve the data association process. The method makes use of a two-step filtration to solve the ambiguities arisen by multiple observations falling into the validation gate of a single feature or an observation lying in the overlapping region of gates of multiple features. Experiments have been simulated to analyze the efficiency of the presented algorithm. Results show that the method could achieve a satisfactory association result with a computation complexity of O(mn) even when the dead-reckoning error is quite large, thus suitable to on-line data association for underwater SLAM implementations.