Buoyancy-driven underwater drones, commonly referred to as underwater gliders, require accurate and energy-efficient pitch control for long-duration ocean observation. Conventional PID-based pitch controllers may cause repeated movement of the internal shifting mass even after the commanded pitch angle has been reached, resulting in unnecessary actuator operation and cumulative energy consumption over repeated dive cycles.
In previous work, an Adaptive Neuro-Fuzzy Inference System (ANFIS)-based pitch controller was developed to reduce unnecessary internal-mass motion while maintaining pitch-tracking performance. Its performance was evaluated using a nonlinear six-degree-of-freedom dynamic simulator incorporating experimentally derived power-consumption models of the buoyancy engine and pitch actuator. The simulation results showed a 57.05% reduction in pitch-control energy consumption compared with conventional PID control.
The present study extends this work by validating the proposed controller under untethered sea-trial conditions using a 1,000 m class buoyancy-driven underwater drone. Sea trials were conducted off Pohang, Korea, without a safety tether. The PID and ANFIS-based controllers were compared in terms of the cumulative travel of the internal shifting mass and the horizontal travel distance of the vehicle.
The sea-trial results showed that the ANFIS-based controller reduced the cumulative travel of the internal shifting mass by 58.4% compared with PID control, while the horizontal travel distance increased by 12.6% under the tested conditions. These results demonstrate that the proposed learning-based pitch-control strategy can effectively suppress unnecessary actuator motion in a real ocean environment. The sea-trial validation further supports the applicability of ANFIS-based pitch control for improving the operational energy efficiency of buoyancy-driven underwater drones during long-duration ocean missions.