Evaluation of two depth-holding algorithms for a buoyancy-controlled coastal float

For coastal floats, buoyancy control algorithms must provide high-accuracy depth control in shallow waters with fast currents and strong density gradients. Here, we compare two depth-holding algorithms through tank and field testing with the $\mu$Float, a coastal float designed for distributed array surveys of tidal currents and other water properties. A linear-quadratic-regulator (LQR) controller was designed and implemented on the $\mu$Float. The LQR controller was evaluated against a two-stage proportional-differential (2S-PD) previously used on the $\mu$Float. Relative to the original 2S-PD controller, the LQR controller reduced overshoot, motor actuation, and power consumption. Implementing a wide (±0.5 m) dead-band on the LQR controller proved particularly successful, reducing motor actuation by 80% and reducing power consumption by 40-60%. The original 2S-PD controlled remained superior in minimizing time to target depth and depth-holding accuracy. Future improvements on the LQR controller are discussed.