CMA-ES is known for its robustness for solving optimization problem but at the cost of computational time, yet it has been applied in real time by simplifying the problem that is to be solved. CMA-ES proved itself quite successful in searching for the optimal control that controls the ship to follow the reference trajectories in real time (0.8 s paced-simulation timestep in MATLAB Simulink). This can be considered as a novel attempt to investigate the use of an evolution strategy in the topic of automatic berthing or trajectory-tracking control. In addition, the course-changing experiment shows that CMA-ES gave quite a good performance in generating the optimal rudder angle that made the ship to reach the desired course and maintain it.
展开▼