Full Professor, University of Alberta
1 paper at NeurIPS 2025
We propose an action-guided diffusion policy that leverages a variational latent SDE and Vector-Jacobian Product-based update to model the dynamic interplay between perception and action for real-time robot manipulation.