Neural network control of buoyancy-driven autonomous underwater glider

This chapter presents a mathematical model and motion control analysis of a buoyancy-driven underwater glider. The glider mathematical model, which includes the presence of disturbance from the water currents, has been designed by using the Newton-Euler method. In order to predict and control the glider motion, a neural network control has been used as a model predictive control (MPC) as well as a gain tuning algorithm. The motion has been controlled by six control inputs: two forces of a sliding mass, a ballast pumping rate, and three velocities of water currents. The simulation results show the analysis of the motion control system for both neural network control approaches, and a comparison with the Linear Quadratic Regulator (LQR) controller is also included. The results show that the model is stable, and the neural network controller of MPC produced better control performance than the neural network gain tuner and the LQR, where the accuracy value of the MPC is 94.5 %.