Intelligent control systems for unmanned aerial vehicle

Unmanned Aerial Vehicles (UAVs) have played an essential role in military and civilian domains. The research in this thesis contributes to the field of Intelligent Control Systems (ICSs) and especially achieving reliable and convenient autonomous control for Rotary Wing UAVs (RUAVs). In particular, the challenge of adapting to unmodelled dynamics and disturbances such as changing payloads in mid-air is tackled. UAVs can carry extra weights such as sensors, cargo and even underslung loads which are known as the payload. Many strategies have been developed to stabilize the drone with changing payload but they all assume the payload to be rigid and the Centre of Gravity (CoG) to be static and known. Variations in the payload mass and its type during flight, can dramatically affect the dynamics of the drone, requiring a controller to adapt to maintain satisfactory closed loop performance. A scenario where a fleet of delivery drones could be launched from a larger aircraft (like a weather balloon) in mid-air with random pose attitude is also not yet explored. Finally, uncertainties such as unmodelled dynamics and wind gusts pose challenges to flight operations, so ICSs are essential to deal with these uncertainties but not enough attention is given to the design and development of non-model-based ICSs. Motivated by these research gaps, this thesis tackles the control problem of handling payload with changing CoG and pose independent launch in mid-air. To address these problems and to achieve the desired trajectory tracking control, a novel non-model based ICS called the Bidirectional Fuzzy Brain Emotional Learning (BFBEL) control system is presented. The proposed control system merges fuzzy inference, neural networks and a novel Bidirectional Brain Emotional learning (BBEL) algorithm based on reinforcement learning. The proposed BFBEL controller is capable of adapting rapidly from scratch and it is introduced to control all the Six Degrees of Freedom (6DOF) of the RUAVs. To expand the applicability of the proposed controller, both Single-Input-Single-Output (SISO) and Multi-Input-Multi-Output (MIMO) architecture are developed. The two RUAV models considered for this research are the Quadcopter UAV (QUAV) and the Helicopter UAV (HUAV). The SISO version of BFBEL control system is applied to QUAV to address the problem of handling external payload with varying CoG and weight. The MIMO version of the BFBEL control system is applied to a HUAV to address the problem of pose independent launch in mid-air. Both systems are evaluated with simulations and the problem of handling external payload with uncertain CoG is verified with experiments. Finally, the flight capabilities and control performance are compared with a conventional Proportional Integral Derivative (PID) controller scheme under the same control scenarios.