Localisation of underwater autonomous vehicles in a swarm formation based on computer vision

A swarm of underwater autonomous vehicles can perform demanding tasks faster and more accurately than a single one because individual vehicles can cooperate, sharing complex tasks among the swarm’s participants. To facilitate such cooperation, the vehicles should demonstrate abilities to create various shape formations in which each participant localizes and follows its preceding neighbor. This simple task in land applications becomes very challenging in an underwater environment since the most popular solutions, such as GPS or radio systems, cannot be used. Among the available localization techniques, primarily based on high-priced acoustic methods, computer vision has been proven to be the most suitable due to the low cost of implementation and flexibility. Therefore, this work focuses on short-distance vehicle localization in a swarm formation. For this purpose, the underwater vision system was deployed to detect and localize the swarm participants equipped with lightning markers. The first part of the research was devoted to creating an image dataset using the proposed acquisition system. Then, an image processing algorithm utilizing image segmentation techniques was developed and tested. Finally, based on the segmentation results, the localization method employing a regression model to calculate distance was proposed. The results suggest that the developed solution allows for accurate localization up to 7 meters in the Baltic Sea. This distance is sufficient for the established swarm’s functionality.