cs.RONov 17, 2026

Task Convergent Flow Matching for Generative Robotics

Authors: Lorenzo Paiola, Maximilian Stölzle, Giorgio Grioli, Daniela Rus, Cosimo Della Santina

Organizations: SoftBots for Human Cooperation and Rehabilitation, Istituto Italiano di Tecnologia, Genova, Italy · Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology, USA · Research Center “E. Piaggio” and Dept. of Information Engineering, University of Pisa, Italy · Department of Cognitive Robotics, Delft University of Technology, The Netherlands

Abstract

Successful manipulation requires the manipulated objects to progress toward a desired task outcome. Generative policies can produce plausible robot motions without ensuring this progress, allowing successive action chunks to stall or undo earlier changes to the scene. We introduce a task-centric inference-time refinement framework for flow matching policies that evaluates candidate action chunks through parallel physics-based rollouts and searches for motions satisfying sufficient conditions for task convergence. Task-level performance funnels bound temporary regressions while requiring net progress over each chunk. We establish sufficient conditions under which repeated funnel satisfaction yields convergence throughout execution, and use interval-wise Lipschitz bounds to verify continuous-time funnel satisfaction from sampled rollouts. To accommodate contact-rich hybrid dynamics, we refine the flow's source seeds using derivative-free Covariance Matrix Adaptation Evolution Strategy (CMA-ES), constrained to radial quantile bands of the source distribution to help retain the learned motion prior. This formulation supports both Euclidean and pose-valued Riemannian Flow Matching (RMFM) policies. On PushT and three-cube stacking, refinement raises success from 31.3% to 88.5% and from 77.0% to 86.0%. On PushT, whose coverage score responds continuously to the motion of the block, a learned funnel lowers the average maximum loss of achieved progress by 19% relative to constraining the endpoint of each chunk alone, at a similar success rate. With the radial band, 11.3% of the executed motion windows are out of distribution, a rate within the range of held-out demonstrations, compared with 34.1% for an unconstrained search.

Figures & tables

Appendix figures & tables19 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

CardsList
  1. ConFlow: Constraints-Guided Learning with Flow Matching for Motion Generation

    Jul 15, 2026Nutan Chen, Jianxiang Feng, Marvin Alles +1Constrained Generative ModelingFlow Matching

  2. Manifold-Stable Flow Matching

    Sep 28, 2026Amirhossein Nazerian, Ali Pezeshki, Jianguo ZhaoConstrained Generative ModelingFlow Matching

  3. FlowDAgger: Human-in-the-Loop Adaptation of Generative Robot Policies in Latent Space

    Jul 9, 2026Michael Murray, Daphne Chen, Simran Bagaria +7Latent Action LearningRobot Policy Learning