RTA SAFE CONTROL FOR DRONE LONGITUDINAL DYNAMICS USING LQT AND CONTROL BARRIER

Unmanned Aerial Vehicles (UAV) are platforms controlled without human direct interaction. With decision-making left to the onboard computer, safety measures need to be taken into account. Run Time Assurance (RTA) and Control Barrier Functions (CBFs) are gaining high interest as of late, as methods used to implement safety through adjusting the primary control input, computed according to some control law, to guarantee safety by satisfying a set of boundary constraints. This thesis aims to use RTA and CBF methods to satisfy safety constraints on a quadrotor system in both one-degree-of-freedom (1-DOF) and two-degrees-of-freedom (2-DOF) configurations. The controller used is the Linear Quadratic Tracker (LQT) optimal controller, which tracks a certain reference signal known beforehand, using offline calculations. Following the reference signal may force the quadrotor to go beyond certain boundary constraints, and by using optimization-based RTA filter and CBF methods, the controller aims to choose an alternative path to satisfy the boundary constraints, with the least adjustments on the primary controller input. The continuous-time and discrete-time system representations are used and compared for the 1-DOF and 2-DOF motion of the quadrotor. CBF extensions such as discrete-time Control Barrier Function (DCBF) and High Order Control Barrier Function (HOCBF) are examined. For the 2-DOF motion, two different controller architectures (full order and cascaded) are used and compared, along with safety implemented for both controllers. The control strategies without and with RTA are validated in simulations and in experimental studies showing good results.