Model Predictive Controller (MPC)
controllers are commonly used in
control of process plants, but due
to the high computation load, they
have not been that popular for small
embedded systems. Due to big leaps
in the computational power of small
embedded computers, MPCs can be
implemented on a wide variety of
other embedded control problems. In
this thesis, we modify the software
and hardware of the original commercially available open-source platform
BlueROV2 to make it more suitable
and better optimized to run an MPC
and a Kalman filter. The code is
written in Python using object-based
programming principles. A comparison
between the original system and our
modifications is made. To implement
an MPC controller on this modified
system, the Remotely Operated Vehicle (ROV) is modelled in 6 Degrees
of Freedom (DoF), and the model
is validated. State feedback for the
controller is done using a Kalman filter
for sensor fusion of the Inertia Measurement Unit (IMU) and Underwater
Global Positioning System (UGPS).
Then, a comparison is made with
a Proportional–Integral–Derivative
(PID) and an LQR implemented on
the BlueROV2.