Navigation and Control of Flexible Wing UAV Using Vision System

This thesis presents an advanced guidance and control system for unmanned aerial vehicles (UAV) with flexible wing. The control system is based on a stereo vision system and advanced fuzzy logic algorithms that can detect wing deflections and shapes. The thesis proposes a novel Deflection-Detection-Vision-System (DDVS) to control a flexible wing of unmanned aerial vehicle (UAV). The technique measures the deflection of the flexible wing with a stereo camera and determines the three-dimensional (3D) coordinates to identify the wing shape. In addition, the fuzzy logic algorithm classifies the shapes and determines the flight parameters, such as the speed, angle of attack and roll angle. The Deflection-Detection-Vision-System (DDVS) consists of a stereo camera positioned at the back end of the wing structure that reads the deflection of chosen landmark locations on the flexible wing for each image instantaneously. The DDVS characteristics and dynamic parameters were tested in wind tunnel.Controlling an autonomous UAV with flexible wing can be difficult using classical methods. An autopilot controller based on an intelligent controller known as the Adaptive Neuro-Fuzzy Inference System (ANFIS) algorithm was developed, and it applies the neural networks and fuzzy logic features in hybrid control architecture. To achieve optimal performance, three ANFIS modules were designed to control the altitude, heading angle and speed of the flexible wing UAV. The longitudinal motion controller and the inner loop (pitch rate feedback) of the longitudinal system are designed first, then a pitch tracker with an ANFIS controller is developed. The design of the altitude and speed controllers is related to the guidance and control system (outer loop controller) using the ANFIS controller design. The ANFIS controller performance is compared and evaluated with the Proportional-Integral-Derivative (PID) controller. The lateral motion control is performed by an inner loop controller, that includes roll rate feedback, and the roll tracker is done with ANFIS controller. The proposed ANFIS controller was chosen because it has better performance than the classic controller. The fusion algorithm based on Adaptive Unscented Kalman Filter (AUKF) was integrating measurements from an accelerometer sensor and DDVS.