Chapter 12 - An extreme learning-based adaptive control design for an autonomous underwater vehicle

Abstract

The chapter presents development of a new adaptive controller for an autonomous underwater vehicle (AUV) that is nonholonomic and underactuated in nature. An on-line sequential extreme learning machine (OS-ELM) model is used for designing adaptive proportional, integral, and derivative (PID) controller for steering the AUV to a desired depth profile. Conventional PID controller for controlling a complex six degree-of-freedom AUV is difficult and may lead to loss of controllability. We propose here a novel self-tuning PID controller that is designed using reduced order OS-ELM model of AUV for depth tracking. To resolve the uncertainties that arises due to ocean waves, ocean currents, model mismatch or parametric variations, OS-ELM is used to identify depth dynamics of AUV. Owing to the slow sampling rate of sensors such as Doppler velocity log sensor (sampling rate between 4 and 5Hz), the position control of AUV is inaccurate and thus the formulation of depth control law becomes difficult for AUV. Hence, the variable delay that arises in sensors measurement needs to be predicted on-line to improve the tracking performances of AUV. ELM structure is considered for time delay prediction at every sampling instant. The effect of variable time delay caused by sensors is thus handled appropriately in the proposed control design, thereby improving that the tracking performance of AUV is achieved in the face of variable delay and parameter variations in AUV dynamics. Simulations are performed using MATLAB and it is seen that the proposed PID controller exhibits efficient tracking performance during the depth control of AUV.