Berke Bayrı, M.Sc.
Department of Scientific Computing
August 2026
Supervisor: Ömür Uğur (Institute of Applied Mathematics, Middle East Technical University, Ankara)
Abstract
This thesis investigates the use of Sparse Identification of Nonlinear Dynamics with Control (SINDYc) for Model Predictive Control (MPC) of a quad-rotor. A nonlinear quad-rotor model is identified using SINDYc and employed in both nonlinear and successive-linearization MPC formulations. The successive-linearization MPC is solved using an active-set algorithm and a CUDA-enabled OSQP implementation, while the nonlinear MPC is solved using Sequential Quadratic Programming. The identified model reproduces the dominant nonlinear dynamics with good agreement. All MPC implementations successfully track the prescribed trajectory. The CPUbased active-set solver provides the best computational performance and achieves faster-than-real-time execution, whereas GPU-enabled OSQP provides no computational advantage for the investigated problem size. The results demonstrate the applicability of SINDYc-based MPC and show that GPU acceleration may not be beneficial for relatively small optimization problems.
Keywords: SINDYc, GPU Parallelization, Real-Time Control, quad-rotor Dynamics, Nonlinear System Identification