Niharika Sabhlok

Tuesday, January 20, 2026, 11:00 - Tuesday, January 13, 2026, 12:00

SR 01-012

Abstract: This thesis investigates model predictive control (MPC) formulations for operating a leading-edge inflatable (LEI) pumping airborne wind energy (AWE) system efficiently and safely. The core idea is to first compute a high-quality periodic optimal trajectory for the kite and use this as an initial guess for the MPC formulations. This trajectory then serves as the backbone for designing and comparing different nonlinear model predictive control (NMPC) formulations in closed-loop simulations.
Flight trajectory optimisation is critical for the model predictive control design for AWE systems, since a good initial guess can strongly influence convergence and solution quality. In the first part of this thesis, periodic optimal control problems are solved using a dynamic model of the Kitepower LEI AWE system embedded in the AWEbox framework. A parameter study over the trajectory period, time discretisation and regularisation weights is conducted to identify numerically robust pumping cycle. The resulting periodic solution serves as a refined initial guess for the subsequent optimisation problem in which the controllers are designed.
In the second part, the system model and the periodic reference orbit are imported to an NMPC framework-TuneMPC and used to construct the prediction model and reference trajectory. Several MPC controllers are designed including Tracking MPC, Economic MPC and economically tuned tracking MPC. Their performance is evaluated in closed-loop simulations for different prediction horizons and disturbance scenarios.
The results show how strongly the closed-loop performance of the AWE system depends on the choice of MPC formulation and horizon length. Using a good periodic reference from trajectory optimisation significantly improve convergence, feasibility, and power output. The thesis thus demonstrates a practical workflow that links periodic trajectory optimisation and NMPC for LEI-based AWE systems