Embedded Files
Our research group studies questions at the intersection of AI, engineering, and neuroscience. We are particularly focused on brain-computer interfaces, robot learning, and shared autonomy to build assistive devices. We also use AI to model and understand how populations of neurons carry out computations in the brain.
Projects
Non-invasive BCIs
We build non-invasive brain-computer interfaces (BCIs) to help people with paralysis control robots. We record and decode EEG in real-time with state-of-the-art AI to control robotic arms and computer cursors.
Video: A paralyzed participant controls a robotic arm with neural signals recorded through a headcap. Full video here.
Shared autonomy
We design algorithms for AI copilots to assist humans with tasks.
Video: Diffusion Sequence Copilots (DiSCo) enables humans to better control robots and cars. DiSCo performs shared autonomy with diffusion policy. More information here.
Offline reinforcement learning
We design algorithms for offline goal-conditioned RL to train better policies from offline data.
Video: Subgoal Advantage-Weighted policy bootstrapping (SAW) translates the benefits of hierarchical policies to flat policies. More information here.
EMG interfaces
We build EMG interfaces to enable participants to play games and control robots.
Video: Non-invasive EMG control of Pacman and Mario Kart.
Google Sites
Report abuse