Towards Learning Manipulation Skills Under Wind Influence

Teodor Tomic

Abstract: The talk presents two topics that offer new perspectives in aerial manipulation. The first is learning of robotic manipulation skills, based on reducing the parameter search space using a skill formalism framework, adaptive impedance control and finally using evolutionary optimization to find skill hyperparameters.
Experimental results show fast learning rates, good generalization and superhuman performance for complex mainpulation tasks using industrial manipulators. The second topic is discrimination between physical interaction and aerodynamic wrenches acting on flying robots, which may improve robustness of aerial manipulation in harsh environments.
The approach is based on a fault detection and identification scheme using the external wrench, an aerodynamics model, and a wind speed estimate to achieve this task without the use of additional onboard sensors

Bio:

Teodor Tomić is a postdoctoral researcher at the Technical University of Munich (TUM) in Munich, Germany, and a planning and controls engineer at Skydio in Redwood City, CA, USA, working on fully autonomous flying robots. He received the Dr.-Ing. degree in EE/CS at Leibniz University of Hannover in 2018 under the supervision of Prof. Sami Haddadin, for which he received the Georges Giralt PhD Award for best European PhD in robotics in 2019. He received his BSc and MSc degrees in mechanical engineering in Zagreb, Croatia, in 2008 and 2010, respectively.
From 2008 to 2010 he worked as a GNC System Engineer an omnidirectional airhip at Hipersfera in Zagreb, Croatia, and from 2011 to 2017 he was a Research Scientist with the German Aerospace Center (DLR), Institute of Robotics and Mechatronics in Wessling, Germany.

%d bloggers like this: