Aquatic modular self-reconfigurable robotic systems (MSRRs) have incredible potential for bringing practical, flexible, and adaptable robotic tools to challenging environments. They could build mobile ocean platforms or bridges for larger vehicles, act as ocean-going manipulators to perform repairs on infrastructure, or function as oceanographic research platforms, using reconfiguration to achieve precise spatial resolution when sensing or to improve energy efficiency when traveling over large distances. Development of aquatic MSRRs, however, is limited by the assumption that modules need to be capable of holonomic actuation, which makes them complex and expensive. This work challenges this limitation, presenting a novel underactuated aquatic robot called the Modboat that uses a single motor and passive flippers for propulsion and steering, and developing a capable aquatic MSRR that can dock, undock, reconfigure, and move as a collective using Modboats as its modules. Aquatic systems are further limited because conventional techniques assume that full flow models are needed to use ocean currents for navigation. Such flow models are rarely available, so practical deployments are limited to high thrust and energy-capacity systems. This dissertation challenges this assumption, demonstrating that limited knowledge of ocean gyres can be used for energy-efficient navigation even by low-thrust systems, and that this navigation can significantly expand the operational range of energy-limited robotic systems. Modboat modules are used to verify these results as an example underactuated and low-power robot.