Our team is studying the foundations in kinematics and dynamics for humanoid robot locomotion and for kick, squat, dive and get-up motions. We apply reinforcement learning to improving the gait of the humanoid robots.
Our robots use cameras as primary sensors for localization and object detection. We are focusing on light-invariant feature extraction and fast object detection using a combination of traditional computer vision algorithms and deep learning.
Our behavior projects range from lower level single robot behavior (path planning and decision making) to team coordination (position and role calculation, coordination in a multi-agent system, work distribution).