Neural Network Control for Fluid Drag Force Reduction in an Aqua-Robot Arm by Ejecting Air Bubbles.

Abstract

This paper deals with experimental examination of the feasibility of a neural network control scheme for reducing fluid drag force, which is developed to select the patterns of air-bubble injection from the surface of the 2 link-2 degree of freedom aqua-robot manipulator arm to result in reducing the fluid drag force acting on the manipulator arm. Two neural networks are implemented in the control system of the manipulator immersed in water ; one is for the inverse dynamics of the manipulator between the state variables and fluid drag force, and the other is for selection of an air-bubble injection pattern to minimize the fluid drag force in accordance with the manipulator motion. This neural network control scheme achieves a 30% reduction in fluid drag force for a high-speed motion during manipulator arm operation.