We present a generative predictive control (GPC) framework that amortizes sampling-based Model Predictive Control (SPC) by bootstrapping it with conditional flow-matching models trained on SPC control sequences collected in simulation. Unlike prior work relying on iterative refinement or gradient-based solvers, we show that meaningful proposal distributions can be learned directly from noisy SPC data, enabling more efficient and informed sampling during online planning. We further demonstrate, for the first time, the application of this approach to real-world contact-rich loco-manipulation with a quadruped robot. Extensive experiments in simulation and on hardware show that our method improves sample efficiency, reduces planning horizon requirements, and generalizes robustly across task variations.
@article{arxiv.2510.14643,
title = {Generative Models From and For Sampling-Based MPC: A Bootstrapped Approach For Adaptive Contact-Rich Manipulation},
author = {Lara Brudermüller and Brandon Hung and Xinghao Zhu and Jiuguang Wang and Nick Hawes and Preston Culbertson and Simon Le Cleac'h},
journal= {arXiv preprint arXiv:2510.14643},
year = {2026}
}