The Neurorobotics Lab develops machine learning techniques for robotics and neurotechology. We focus on data-efficient deep reinforcement learning methods that provide robust and adaptive solutions, and target applications on real-world devices. The lab is headed by Asst. Prof. Joschka Boedecker and is part of the Dept. of Computer Science, University of Freiburg, as well as the Cluster of Excellence BrainLinks-BrainTools. You find our main website and publication list here.
Reinforcement Learning
We develop state-of-the-art methods that provide robust, sample-efficient and adaptive solutions, focusing on the Q-Learning framework. Checkout our work on Constrained Q-learning, constrained imitation via Deep Inverse Q-Learning with Constraints, Composite Q-Learning and Model-assisted Bootstrapped DDPG.
Reinforcement Learning
We develop deep RL methods that learn decision policies for autonomous driving. Checkout our work on handling dynamic scenes using Deep Sets or Deep Scenes and constrained imitation by Deep Inverse Q-Learning with Constraints (see Foundations of RL).
Supervised Learning
We develop supervised deep learning methods that learn to detect and predict diseases from medical data automatically. Checkout our work on our early seizure detection device SeizureNet or the dynamic architecture AdaptiveNet to predict disease progression based on medical records.
