Neural Dynamics Augmented Diffusion Policy
Ruihai Wu*, Mingtong Zhang*, Haozhe Chen*, Haoran Lu, Yitong Li, Yunzhu Li
Published in Under Review
To reduce the number of required demonstrations for skill learning, we propose dynamics-guided diffusion policy. This method leverages learned dynamics models, which can explicitly model the interactions in a much wider space than the regions just covered by expert demonstrations.