Abstract
M.Ing.
Linear control system theory is well developed and has lead to a number of control system
types with well-defined design methods that can be applied to any linear system. Unfortunately,
no system in nature is truly linear. As a result, such non-linear systems must be
represented by a linear model that is accurate over some region of the operating states of
the system. The success of linear control theory in commercial applications is testament
to the fact that some types of systems can be adequately represented by a linear model.
However, systems with time-varying dynamics or non-linearities such as input or operating
state saturation cannot always be adequately controlled by linear control systems.
For that reason, non-linear control techniques must be investigated. This project aims
to investigate Non-linear Model Predictive Control theory and practical implementation
in the context of developing an autopilot for an Unmanned Aerial Vehicle based on a
miniature helicopter.
A non-linear model of the dynamics of an X-Cell Spectra G radio-controlled helicopter
was developed based on the existing literature. A number of experiments were performed
to determine the parameters of this model. Significant future work exists in designing
additional ground experiments since certain parameters are difficult to measure safely in
the laboratory. Additional work to improve the accuracy of the model at high airspeeds,
as well as incorporating a more accurate yaw dynamics model, is also required.
Following this, a Non-linear Model Predictive Control autopilot was simulated using
MATLAB®. The simulation tested the effects of control system parameters such as
control horizon and sampling period, as well as the sensor noise susceptibility and its
ability to handle wind as a random disturbance.
The results determined adequate control system parameters for level flight as well as
landing the helicopter under ideal conditions. Simulations in which sensor noise and wind
were added showed that the control system is significantly affected by sensor noise and
that it cannot hover in the presence of wind. A real-time implementation was not achieved
during this work; however, several directions for future research have been discussed.