Stefan’s early research using tiny soccer-playing humanoid robots at Technische Universität Darmstadt in Germany prepared him well for software development on much larger humanoid robots that don’t play soccer at all. From the Darmstadt Dribblers RoboCup team to the DARPA Robotics Challenge Team ViGIR, Stefan has years of experience with robots that need to walk on two legs and do things while not falling over (much).
Almost all of the software that Team ViGIR used to control its ATLAS robot (named Florian) was ROS based. Stefan credits both the team’s prior experience with ROS, ROS’s existing software base, and its vibrant community for why Team ViGIR made the choice to go with ROS from the very beginning. Controlling the ATLAS robot is exceedingly complex, and Stefan takes us through the software infrastructure that Team ViGIR used during the DRC; from basic perception to motion planning to manipulation and user interfaces.
With lots of pictures and behind-the-scenes videos, Stefan describes how Team ViGIR planned to tackle the challenging DRC Finals course. The team used both high-level autonomy and low-level control, with an emphasis on dynamic, flexible collaboration between robot and operator. Stefan narrates footage of both of Florian’s runs at the DRC Finals; each was eventful, but we won’t spoil it for you.
To wrap up his talk, Stefan describes some of the lessons that Team ViGIR learned through their DRC experience: about ROS, about ridiculously complex humanoid robots, and about participating in a global robotics competition.
Next up: Gary Servin of Creativa77
Check out last week’s post: Mark Shuttleworth of Canonical