| Helicopter Unmanned Aerial Vehicles (UAVs) present a challenging control problem since their dynamics are nonlinear, underactuated and non-minimum phase. Although it is inherently an applied research field, due to the difficulty of building and maintaining an experimental platform relatively few experimental results exist in the literature. The approach followed in this thesis is to combine rigorous analysis with thorough experimental testing. This testing ensures validity of the designs. We present our experimental platform which is designed to be flexible so that it can accommodate nonlinear control research. Existing accounts of helicopter testbeds focus on hardware details. Since autopilot software design and development also requires a significant investment, we describe our implementation which has been released as open source for the benefit of the community. Due to the intractability of existing helicopter models, many control designs use non-physical inputs. We propose simple, invertible expressions relating the non-physical inputs to the physical inputs. In particular, modelling of the main rotor typically results in complicated expressions with extensive state dependence. While it is unlikely that angular velocity has a significant influence on the thrust, we show using experimental results that previous attempts to simplify this expression using a hover assumption are invalid during vertical flight. The platform is validated using a model-based PID control law. This control is derived using passivity to ignore nonlinear terms which do not affect stability. Among the class of vehicles with similar flight capabilities, helicopters possess a coupling between the rotational inputs and translational dynamics which is unique. This coupling is sometimes referred to as the Small Body Force (SBF) and is ignored in the literature for controller synthesis. We derive an experimentally-validated control design which accounts for the effect of the tail rotor in the SBF. In addition, we show why the contribution of the main rotor flapping in the SBF cannot be compensated using the same approach, and give a robustness analysis of their effect on the closed-loop. Finally, based on recent results we propose a control design which accounts for state constraints by enforcing bounds on translational velocity and roll-pitch travel. |
