Adaptive Neuro Fuzzy PID Controller for A Compact Autonomous Underwater Vehicle

An adaptive Neuro Fuzzy Proportional-Integral-Derivative (PID) controller for navigation of an compact autonomous underwater vehicle is presented here. The Autonomous Underwater Vehicle (AUV) considered for the study is an under-actuated system with three thrusters and a neutrally buoyant, modular, and closed-frame body. Mathematical model of the AUV is presented with system parameters estimated from detailed CAD model and Computational Fluid Dynamic (CFD) study. Line-Of-Sight (LOS) technique is used in the guidance system for path planning. The AUV model is a 3 Degree of Freedom (DOF) coupled non-linear system. A partitioning law is used to develop a model based PID controller for trajectory tracking operation. PID controllers are popular for its simplicity and ease of implementation, but for highly nonlinear system like AUV, the controller gains have to be tuned for different trajectories. Furthermore, the uncertainties in the system model and dynamic environment will affect model based controller. Neuro- Fuzzy controller is developed to handle dynamic environmental forces and unknown system behavior. Here a neural network model of the system is fitted with the experimental data and the fitted model is used with the PID system to adapt to different working environments. The controller is successfully simulated for 3D trajectories and results are discussed. Comparative simulation in presence of external disturbance forces showed better performance by Neuro-Fuzzy-PID than the PID controller.