Optimization-based Iterative Learning Control Applied to a Quadrotor System

Master Thesis Defense

Ephraim Eckl

Thursday, April 08, 2021, 16:00 - 17:00


In this work we develop an Optimization-based Iterative Learning Control (OptILC) approach for reference trajectory tracking problems with the aim to improve the tracking performance of a quadrotor in flights affected by the so-called ground effect. The OptILC is designed to improve the tracking performance through iterative learning from repeatedly executed tasks. In each iteration the OptILC generates a feasible reference and corresponding optimal open-loop controls with respect to a, possibly nonlinear and constrained, nominal model of the plant. The controls and the feasible reference are obtained by solving an Optimal Control Problem (OCP) with the objective to track a desired reference. Due to a model-plant mismatch the tracking
of the real system might be inaccurate, which is why the nominal model is iteratively corrected using the data of the previous iterations. As the OptILC is an open-loop control approach, we combine it with a Linear Quadratic Regulator (LQR) acting in closed-loop to make the overall setup robust against stochastic disturbances. We analyze the convergence properties of the approach with a simplified two-dimensional quadrotor and an actual quadrotor system in simulation. Finally, we apply the OptILC to a tethered quadrotor in an experiment, in which we show that the tracking performance of the quadrotor, affected by the ground effect, can be improved.

Online via Zoom
Meeting-ID: 627 9173 7415
Password: syscop2021