Neural Network Based Adaptive Control of Underwater Vehicle

A0012 Z. W. Xing Shenyang Institute of Automation, Chinese Academy of Sciences

A0013 X. S. Feng Shenyang Institute of Automation, Chinese Academy of Sciences

A0014 H. H. Chen Shenyang Institute of Automation, Chinese Academy of Sciences

Underwater Vehicle is a complex multivariable system with six degree-

of-freedom and the motions along each degree-of–freedom interact

each other. This brings many difficulties to the position and

attitude control of the underwater vehicle, especially in some

situations which require high precision control such as the underwater

vehicle docking with or tracking some underwater target precisely. So

we must consider the effects of the coupling among the variables and

introduce some necessary measures to eliminate or decrease the effect

of the coupling when we design the control system of the underwater

vehicle. Then we can design all the control-loop of the underwater

vehicle by means of single variable after decoupling. Decoupling

control is important for underwater vehicle to improve control

performance and extent their field of application.

Currently, there have two methods mainly used to deal with

multivariable decoupling control system. One is the method of state

space; the other is the method of modern frequency, that is the so-

called diagonal predominance, which resorts to Nyquist criterion to

design the decoupling control system. Both of the two methods are

based on the precision dynamic model of the system. As a typical

nonlinear, time-varying and coupled system together with

hydrodynamic uncertainties and external disturbances such as current,

the underwater vehicles precision dynamic model cant be obtained.

Generally, we can only design the contr ol system according to the

approximate model of the plant. In practical underwater environment,

the decoupling controller will be invalid due to the change of the

parameters of the vehicles dynamic model if we adopt the traditional

model-based decoupling control scheme. Base on the analysis discussed

above, the paper presents a neural network based online adaptive

decoupling control algorithm to cope with the uncertainty of the

parameters of the vehicles dynamic model. The parameters of the

adaptive decoupling controller will be regulated online continuously

according to the change of the parameters of the vehicle by means of

the method of system identification. The novel algorithm overcomes

the defects of the conventional decoupling control methods. A

simulation result is given to verify the valid of the algorithm

mentioned in the paper and some conclusions are stated at the end of

the paper.