Strengthening Generative Robot Policies through Predictive World Modeling

ArXi:2502.00622v3 Announce Type: replace-cross We present generative predictive control (GPC), a learning control framework that (i) clones a generative diffusion-based policy from expert nstrations, (ii) trains a predictive action-conditioned world model from both expert nstrations and random explorations, and (iii) synthesizes an online planner that ranks and optimizes the action proposals from (i) by looking ahead into the future using the world model from (ii