Masters Thesis

Problem

• Quadrotor dynamics are nonlinear and highly coupled, making attitude stabilisation fragile under disturbances and modelling errors

• Many control strategies demonstrate excellent nominal performance, but their robustness claims often collapse once disturbances, uncertainties or model nonlinearities are introduced

Why it matters

• Real-world autonomous flight rarely operates under nominal conditions. Wind, actuator imperfections, and parameter uncertainty are the norm, not the exception

• Controllers that fail gracefully in simulation but degrade sharply under disturbance pose safety and reliability risks in practical deployment

Approach

• Developed a full nonlinear attitude model and implemented multiple state-of-the-art nonlinear controllers spanning model-dependent and robustness-driven design philosophies

• Designed a controlled, like-for-like evaluation framework to isolate robustness effects by testing all controllers under identical nominal and disturbed conditions

Key Insight

• Nominal performance is easy; robustness is not

• Because Variable Structure Control enforces the desired dynamics through a discontinuous control law that dominates modelling uncertainty, disturbances are driven into the sliding manifold and rendered dynamically irrelevant to the closed-loop behaviour once sliding is established

• Controllers with explicit disturbance handling (ISMC, SMC) significantly outperform model-dependent methods when disturbances are present.

Result

• ISMC achieved the best disturbance rejection and zero steady-state error; feedback linearisation and backstepping degraded under disturbances.