Estimation and control of a manipulating unmanned aerial vehicle

A Manipulating Unmanned Aerial Vehicle (MUAV) is an aerial platform equipped with a mechanism to physically interact with its environment. The interaction is realized by means of robotic arm(s), vehicle body or suspension cable(s) and enables a wide range of novel applications including perching, grasping, pick–and–place, load transportation, etc. However, this is a challenging task as MUAVs are inherently unstable platforms with highly nonlinear and coupled dynamics, often associated with moving parts that result in complex modelling, estimation and control problems. This thesis deals with the problem of estimation and control of MUAVs. First, a comprehensive literature survey covering various aspects of MUAVs such as those related to modelling, estimation and control is presented. In the first approach for MUAV state estimation and control, effects of robotic manipulator on Unmanned Aerial Vehicle (UAV) dynamic equations of motion is treated by adding process noise with unknown noise statistics to conventional UAV dynamic model. With that in mind, state estimating and control of a UAV by means of conventional Kalman filters and their adaptive counterparts are formulated. Having designed Linear Quadratic Regulator (LQR) laws, it is shown that adaptive Kalman filters provide accurate satisfactory estimation and overall control of a UAV, even with simultaneous uncertain process and measurement noise statistics. Next, in order to improve the estimation and overall control performance of the previous approach, full nonlinear and coupled dynamic modelling of a MUAV based on Euler–Lagrange formulation is presented. Then, a General Unscented Kalman Filter (GUKF) is proposed to accomplish full state estimation of a MUAV, along with LQR control laws. Finally, in order to improve the execution time of GUKF, a computationally–efficient UKF known as Scaled Spherical UKF (SSUKF) with estimation and overall control performance comparable to GUKF is formulated. It is shown that both UKF–based algorithms result in satisfactory estimation and setpoint/trajectory tracking of quadcopter UAV and its robotic manipulator, even in scenarios with increased noise level and a period of total outage of sensory data.