Our research carried out in the BrainLinks BrainTools Excellence Cluster at the University of Freiburg focused around the question of how to use Machine Learning to interact with the (human) brain.
We worked on two central questions, which could finally be answered positively:
Can we read brain signals and use them to control complex devices like a robot arm? - Yes (see below)
Can we learn to control a complex system like a living brain slice to show a desired activity pattern? - Yes (see publication)
Combining EEG and Reinforcement Learning for Robot Control. We use electroencephalography (EEG) to communicate higher-level intentions to an autonomous controller learned via Neural Fitted Q-Iteration.
BCI for High-Level Remote Reaching and Grasping. Showcase of the prototype system developed during the first stage of the NeuroBots project.
NetControl (artikel) J Wülfing, SS Kumar, J Boedecker, M Riedmiller, U Egert (2018) Adaptive Long-term Control of Biological Neural Networks with Deep Reinforcement Learning. Neurocomputing, Volume 342, pp. 66-74.