​

Quadcopters are 4-motor unmanned aerial vehicles with rotors

positioned on the edges of a square. They have applications in

the military, package delivery, surveillance, etc. This project’s

goal was to model the dynamics and control of quadcopters.

These systems are difficult to control because there are six

dependent and only four independent variables. The research

question was how linear single-variable controllers such a

Proportional Integral Differential (PID) would compare with

advanced nonlinear multi-variable controllers such as Model

Predictive Control (MPC) under modeling errors, disturbances

and complex reference trajectories. The hypotheses were that MPC would perform better in the absence of model error due to its predictive capabilities, PID would work better with model error due to the mild nonlinearities and the fact that MPC relies heavily on the model, MPC would work better for smaller sampling times when disturbances and model error exist, and MPC would be computationally intensive but more intuitive to tune. Simulation results from scenarios involving load lifting, blade flapping, wind resistance and complex flight patterns confirmed the hypotheses. Overall, MPC reacted ahead of PID, anticipating changes in reference trajectories. This improved performance is justified for high flight accuracy in constrained spaces and when a large CPU is available. For low accuracy open air flights, PID would perform satisfactorily in most conditions. Finally, although MPC is not required for single quadcopter control, the nonlinear optimization framework is ideal for the precise control of a swarm of drones where positional constraints can be explicitly stated in the optimizer.