A robust adaptive course-keeping control algorithm is designed to deal with the course-keeping problem for under-actuated ships with rudder faults, gain uncertainty and marine disturbances.

By combining the robust neural damping technique and adaptive approach, numerous neural network (NN) weights can be compressed horizontally, and only two gain-related adaptive learning parameters need to be designed to compensate for both the gain uncertainty and unknown fault parameters. The proposed controller is proven to be semi-global uniform and ultimately bounded (SGUUB) through Lyapunov analysis. Finally, the Nomoto mathematical model is established using "Yukun", and the effectiveness and superiority of the course-keeping algorithm is illustrated by carrying out comparison experiments under marine interference conditions.

The results show that the average rudder angle of "Yukun" under rudder failure is reduced by 51%, significantly improving control performance.

The results of this study can provide references for tackling the course-keeping control problem of under-actuated ships.