Andreas Doerr

Learning to control - Accurate force control is a key requirement for the low level control of many robotic systems. Active impedance control necessary for safe interaction with the environment is based on fast and precise force tracking even in the context of massively changing loads (object handling) or walking/grasping. We explore concepts from self-learning control and adaptive control to avoid exhaustive manual tuning of controllers and to quickly adapt to variable inertia and contact forces.

I explore how to use Covariance Matrix Adaption optimization (CMA) for direct policy search to learn behavior and motions on humanoid robots. The reinforcement framework is leveraged to learn from demonstrations, given by a human teacher and to improve the learned behavior consecutively to achieve a higher level objective. In general, due to the derivative free optimization any objective function can be imposed to the desired motion which itself is represented as an optimal control problem.

My work for the Stratospheric Observatory for Infrared Astronomy (SOFIA) has a strong focus on system engineering. As part of my thesis the cooling system design and structural design for SOFIA's imager systems (fine- and wide field imager) has been accomplished. Further effort has been put into testing the optical system using a Shack Hartmann sensor for wave front measurements and into the electronics design of additional lab equipment.

The Akaflieg Stuttgart is a group of mainly aircraft engineering students at the University of Stuttgart working on the design and construction but also the operation of glider and general aviation airplanes. Currently, our group is working on a two seated motor glider usable as a powerful towing aircraft.

Entwicklung und Implementation von Algorithmen im Bereich bildbasierte Umgebungsanalyse, Pfadplanung und Reglung im Bereich unbemannter Landfahrzeuge.

Industrieller Grundpraktikum, Einführung in industrielle Fertigungsabläufe.

Maschinelles Lernen, Robotik, künstliche Intelligenz

Regelungstechnik, Systementwicklung