Task Convergent Flow Matching for Generative Robotics
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
| \displaystyle\hbox{\pagecolor{red!15}\displaystyle\sup_{t\in[0,H]}!\left[V(T(x_{w}^{k}(t)))-\rho_{k}(t)\right]} | \displaystyle\text{\footnotesize{\triangleright \lx@sectionsign\ref{sec:task_refinement}, \ref{sec:generation_refinement}}} | (1) | ||||
| \displaystyle\hbox{\pagecolor{blue!15}\displaystyle w=\psi_{\theta}(1\mid z_{0},o_{k})} | \displaystyle\text{\footnotesize{\triangleright \lx@sectionsign\ref{sec:flow_prior}}} | |||||
| \displaystyle\hbox{\pagecolor{green!15}\displaystyle u(t)=\kappa(x_{w}^{k}(t),w,t)} | ||||||
| \displaystyle\hbox{\pagecolor{orange!15}\displaystyle\dot{x}{w}^{k}(t)=f(x{w}^{k}(t),u(t)),\ x_{w}^{k}(0)=x_{k}} | \displaystyle\text{\footnotesize{\triangleright \lx@sectionsign\ref{sec:world_task}}} | |||||
| Success | Final score | Max drawdown [pts] | Verified | |||||||
| Objective | [%] | Mean | [%] | |||||||
| PushT | Stack. | PushT | Stack. | PushT | Stack. | PushT | Stack. | PushT | Stack. | |
| Nominal | 31.3 | 77.0 | .000 | .000 | 40.0 | 8.5 | 86.9 | 46.1 | n/a | |
| Endpoint-only | 86.8 | 86.0 | .560 | .084 | 28.3 | 6.8 | 62.8 | 46.1 | n/a | |
| 60.5 | 85.5 | .218 | .103 | 23.4 | 5.8 | 54.2 | 41.8 | n/a | ||
| 80.0 | 82.0 | .499 | .000 | 22.6 | 6.4 | 54.4 | 43.4 | n/a | ||
| Search | Succ. | LDJ | OOD % | |
|---|---|---|---|---|
| L-BFGS † | 70.0 | .765 | 12.0 | |
| Best-of- | 82.5 | .768 | 4.9 | |
| CEM | 79.0 | .848 | 8.0 | |
| CMA-ES , no band | 96.0 | .941 | 34.1 | |
| Ours, | 86.5 | .877 | 11.3 |
Appendix figures & tables19 assets
Supplementary material from the paper’s appendix.
Appendix
| Symbol | Meaning | R on G | B on R | Rule |
| Geometry and physical tolerances | ||||
| half-heights | model geometry | |||
| error scales | ||||
| vertical weight | design | |||
| lowering band | design; | |||
| carry clearance | ||||
| Symbol | Meaning | Square | Round | Rule |
| Geometry and physical tolerances | ||||
| lateral tolerance | half-width of the native success box | |||
| seated / approach height | model geometry; peg top plus nut thickness | |||
| aligned threshold | ||||
| tilt weight | of the threading tilt limit | |||
| yaw weight | ; none for the round nut | |||
| Parameter | Setting |
|---|---|
| Population / parents | ( ) |
| Initial step size / generation cap | |
| Covariance eigendecomposition | Every generation |
| Seed region | Band with , one global group; mean projected after each update |
| Meaningful improvement | , (relative to achieved improvement) |
| Patience | / generations (best plan not verified / verified) |
| PushT | Three-cube stacking | Two-nut assembly | |
| Demonstrations | 206 | 1000 | 1000 |
| Training iterations | 30k | 100k | 100k |
| Batch size / warmup | 128 / 500 steps | 128 / 1000 steps | 128 / 1000 steps |
| Prediction / history | 24 waypoints / 2 steps | 32 waypoints / 5 steps | 32 waypoints / 2 steps |
| Execution horizon | 24 steps | 32 steps | 32 steps |
| State features history |
| Task | Initialization | Tracking |
|---|---|---|
| PushT | Hard reset mode: block in , agent in | Position PD |
| Cube Stacking | Each cube placed uniformly in m with uniform yaw, overlaps rejected (the robosuite placement sampler ( Zhu et al., 2020 ) ) | IK with joint-space PD |
| Two-Nut Assembly | MimicGen NutAssembly_D0 reset: both nuts placed on the table with uniform yaw, the square nut at , and the round nut at , , robot at the dataset’s initial joint configuration | IK with joint-space PD |
| Gain | Stacking | Two-nut assembly |
|---|---|---|
| Task (position / rotation) | / | / |
| Task (position / rotation) | / | / |
| Joint / | / | / |
| Posture / | / | / |
| DLS damping | ||
| Twist limit (linear / angular) | m/s / rad/s | m/s / rad/s |
| Success | Final score | Max drawdown [pts] | KL | Verified [%] | |||||||||||||
| Funnel | [%] | Mean | Mean | Executed | Nominal | ||||||||||||
| PushT | Stack. | PushT | Stack. | PushT | Stack. | PushT | Stack. | PushT | Stack. | PushT | Stack. | PushT | Stack. | PushT | Stack. | ||
| Nominal | 31.3 | 77.0 | .638 | .637 | .000 | .000 | 40.0 | 8.5 | 86.9 | 46.1 | .472 | .181 | n/a | n/a | |||
| Endpoint-only | 86.8 | 86.0 | .923 | .804 | .560 | .084 | 28.3 | 6.8 | 62.8 | 46.1 | .053 | .060 | n/a | n/a | |||
| No envelope | 60.5 | 85.5 | .855 | .744 | .218 | .103 | 23.4 | 5.8 | 54.2 | 41.8 | .153 | .065 | n/a | n/a | |||
| 80.0 | 82.0 | .910 | .715 | .499 | .000 | 22.6 | 6.4 | 54.4 | 43.4 | .052 | .084 | n/a | n/a | ||||
| Success | Native | V-score | Verified [%] | |||||||
| [%] | score | Final | Drawdown [pts] | KL | ||||||
| Funnel | mean | Mean | Mean | Executed | Nominal | |||||
| Nominal | 80.3 | .873 - | .875 - | .205 - | 5.4 - | 43.2 - | .192 - | n/a | ||
| Endpoint-only | 84.5 | .915 | .923 | .401 | 3.6 | 30.7 | .060 | n/a | ||
| No envelope | 86.8 | .927 | .928 | .394 | 3.4 | 30.6 | .080 | n/a | ||
| 86.3 | .920 | .923 | .358 | 2.8 | 24.3 | .057 | n/a | |||
| Success [%] | Max drawdown [pts], learned funnel | ||||
|---|---|---|---|---|---|
| Potential | Fixed funnel | Learned funnel | Mean | Median | |
| Nominal | |||||
| Terminal tolerance | |||||
| Phased, gate residual | |||||
| Phased, residual (main) | |||||
| Executed | Nominal | |||
| Task | Funnel | Sampled | Estimated | Estimated |
| PushT | Plateau, | |||
| Plateau, | ||||
| Plateau, | ||||
| Linear | ||||
| Calibrated, global | ||||
| Score | DD [pts] | Executed | Planned | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Search | Succ. | Max | Final | Mean | KL | LDJ | Jump | OOD % | KID | OOD % | KID | ||
| Nominal | 31.5 | .000 | .795 | .688 | 33.3 | 67.5 | .363 | .544 | 4.0 | .259 | 3.2 | .286 | |
| Adam † | 60.5 | .647 | .927 | .918 | 19.1 | 51.6 | .045 | .652 | 5.8 | .162 | 4.5 | .181 | |
| L-BFGS † | 70.0 | .765 | .942 | .938 | 19.5 | 47.4 | .012 | .662 | 12.0 | .196 | 10.7 | .154 | |
| Best-of- | 82.5 | .768 | .948 | .940 | 19.9 | 50.5 | .054 | .513 | 4.9 | .175 | 4.0 | .194 | |
| CEM | 79.0 | .848 | .954 | .950 | 18.7 | 47.3 | .014 | .600 | 8.0 | .143 | 6.2 | .141 | |
| Final score | Drawdown [pts] | |||||||
|---|---|---|---|---|---|---|---|---|
| Search | Succ. [pts] | Mean | Mean | KL | LDJ | Jump | ||
| Nominal | [ ] | [ ] | [ ] | [ ] | [ ] | [ ] | [ ] | [ ] |
| Adam † | [ ] | [ ] | [ ] | [ ] | [ ] | [ ] | [ ] | [ ] |
| L-BFGS † | [ ] | [ ] | [ ] | [ ] | [ ] | [ ] | [ ] | [ ] |
| Best-of- | [ ] | [ ] | [ ] | [ ] | [ ] | [ ] | [ ] | [ ] |
| CEM | [ ] | [ ] | [ ] | [ ] | [ ] | [ ] | [ ] | [ ] |
| Funnel | PushT | Stacking | Two-nut assembly | |
|---|---|---|---|---|
| Nominal | 0.01 | 0.01 | 0.01 | |
| Endpoint-only | 1.17 | 3.84 | 6.18 | |
| No envelope | 0.77 | 2.02 | 3.26 | |
| 1.41 | 4.97 | 8.34 | ||
| Plateau | 1.06 | 3.93 | 6.11 | |
| 1.12 | 3.76 | 6.20 |
| Search | Time [s] | Iterations | Evaluations |
| Nominal | 0.01 | – | 1 |
| Adam † | 1.33 | 20.3 | 21 |
| L-BFGS † | 1.35 | 14.5 | 36 |
| Best-of- | 0.20 | 1 | 200 |
| CEM | 1.52 | 40.4 | 808 |
| MPPI | 1.52 | 42.0 | 841 |
Robot Motion Generation
Robot motion generation creates feasible, task-directed trajectories for robots to move, manipulate objects, and interact with people and environments.
See all 215 papers