Abstract
This paper proposes an adaptive dynamic positioning (DP) control method based on a multi-observer fusion architecture with minimal sensor requirements. A sliding mode observer is designed based on a high- and low-frequency superposition model to filter high-frequency state variables, while a finite-time convergence disturbance observer estimates unknown time-varying low-frequency disturbances online. For efficient handling of model uncertainties, a single-parameter learning neural network is implemented that requires only one parameter to be estimated online. The control system employs auxiliary dynamic systems to handle input saturation constraints and considers thruster system dynamics. Theoretical analysis demonstrates the stability of the observer-fusion control strategy, while simulation results based on the SimuNPS platform validate its effectiveness in state estimation and disturbance rejection compared to traditional sensor-dependent methods.