Dynamic Positioning of Ships Using a Planned Neural Network Controller

Abstract

Dynamic positioning of vessels is a technique which uses ship-mounted propellers and lateral thrusters commanded by a control system to maintain a desired position in the horizontal plane as closely as required. The environmental disturbing forces are those due to wind, waves and current. Heretofore, control systems basing on the PID principle have been used. This paper continues the authors' earlier work on a new kind of neural network control system which has several advantages, among them a versatile objective function which is adaptive to the need of the operator (more precision or more savings in power) and a completely feed-forward control which is self-adaptive to changes in environmental forces, including nonlinear wave drift forces. Emphasis in this paper is placed on the quality of control, which is improved by including a velocity term besides the position term in the objective function. The performance is effectively enhanced. The station-keeping problem is generalized to include moving the ship by neural-net control to a new target position if required. Computer simulations were carried out with satisfying results.