Because real-time wireless communications are impossible underwater, underwater remotely operatedvehicles (ROVs) are connected to a surface station by a physical link. Recent works investigate limiting themechanical effects exerted by this cable by using intermediary ROVs to control its shape. This configuration iscalled an underwater robot chain and is the focus of the current thesis, which investigates its self-localizationusing embedded sensors.While ROV localization with respect to its environment can be estimated from its embedded sensors, the cablecan be taken as an advantage to localize the ROV at its end point given a knowledge of the cable’s 3D state.Two complementary approaches are studied: the proprioceptive localization of the chain based on anestimation of the 3D state of its cable parts using inertial measurements, and exteroceptive, multi-agentlocalization with respect to the environment using visual simultaneous localization and mapping techniques.