Ship Steering Autopilot Based on ANFIS Framework and Conditional Tuning Scheme

The ever changing sea condition makes it difficult to use a single mathematical model to describe the nonlinear dynamic behavior of a vessel. Hence, the performance of model-based autopilots is generally inferior to that of modelfree designs such as fuzzy controllers or neural networks. This study combines the robustness property of fuzzy logic controllers and the learning capability of artificial neural networks to create an ANFIS (Adaptive Network-based Fuzzy Inference System) ship steering autopilot. In addition, a conditional tuning scheme is presented to increase the response speed of the proposed autopilot, while simultaneously reducing the overshoot. The performance of the proposed autopilot is evaluated by performing a series of course-changing and track-keeping simulations. The simulation results show that the proposed autopilot provides a more adaptive and robust control performance than a traditional PD fuzzy controller under typical sea conditions.