Position and Heading Control of an Autonomous
Underwater Vehicle using Model Predictive Control

Abstract

Autonomous Underwater Vehicle (AUV) is currently being used for scientific research,
commercial and military underwater applications. AUV requires autonomous guidance and
control systems to perform underwater applications. This Thesis is concerned with position
and heading control of AUV using Model Predictive Control.
Position control is a typical motion control problem, which is concerned with the design of
control laws that force a vehicle to reach and maintain a fixed position. The position control
of body fixed x-axis to a fixed point using MPC toolbox of MATLAB is done here. System is
modelled Using INFANTE AUV hydrodynamic parameters. There is physical limitation on
thruster value.

Heading control is concerned with the design of control laws that force a vehicle to reach and
maintain a fixed direction. There are physical limitations on control input (Rudder deflection)
in heading control also a high yaw rate can produce sway and roll motion, which makes it
necessary to put constraint on yaw rate. The MPC have a clear advantage in case of control
and input constraints. To avoid constraint violation and feasibility issues of MPC for AUV
heading control Disturbance Compensating (DC) MPC scheme is used. The DC-MPC
scheme is used for ship motion control and gave better results so we are using the proposed
scheme to AUV heading control.
A 2 DOF AUV model is taken with yaw rate and rudder deflection constraints. Line of sight
(LOS) guidance scheme is utilised to generate the reference heading, which is to be followed.
Two types of disturbances are taken constant and sinusoidal. Then simulation has been done
for standard MPC, M-MPC and DC-MPC. A (DC) MPC algorithm is used to satisfy the state
constraints in presence of disturbance to get a better performance.
Standard MPC gives good result without disturbance. But in case of disturbance yaw
constraint is violated. At many time steps the standard MPC has no solution for given yaw
rate constraint at those time steps the constraints have been removed. The M-MPC satisfies
the constraints. The DC-MPC gives better result in comparison to standard MPC and
Modified MPC. The steady state oscillations are less in DC-MPC as compared to M-MPC for
sinusoidal disturbances.
The minimization of extra cost function in DC-MPC makes the result better than M-MPC. By
solving the extra cost function we try to make response close to that of without disturbance.
The only added complexity in DC-MPC is ni-dimensional optimization problem. Which is
very less compared to Np*ni, complexity of M-MPC. Where ni is the dimension of control
input and Np is value of prediction horizon. The feasibility of DC-MPC scheme largely
depends on the magnitude of disturbance. If disturbance is too large then this scheme is not
feasible.