Much Ado About Noising: Dispelling the Myths of Generative Robotic Control
This work clarifies misconceptions in robotic control, potentially guiding more efficient policy designs, though it is incremental as it refines existing understanding rather than introducing a new paradigm.
The paper investigates the success of generative control policies in robotics, finding that their advantage comes from iterative computation with supervised intermediate steps and stochasticity, not from capturing multi-modality or complex mappings, and shows that a lightweight two-step policy matches or outperforms flow-based models.
Generative models, like flows and diffusions, have recently emerged as popular and efficacious policy parameterizations in robotics. There has been much speculation as to the factors underlying their successes, ranging from capturing multi-modal action distribution to expressing more complex behaviors. In this work, we perform a comprehensive evaluation of popular generative control policies (GCPs) on common behavior cloning (BC) benchmarks. We find that GCPs do not owe their success to their ability to capture multi-modality or to express more complex observation-to-action mappings. Instead, we find that their advantage stems from iterative computation, as long as intermediate steps are supervised during training and this supervision is paired with a suitable level of stochasticity. As a validation of our findings, we show that a minimum iterative policy (MIP), a lightweight two-step regression-based policy, essentially matches the performance of flow GCPs, and often outperforms distilled shortcut models. Our results suggest that the distribution-fitting component of GCPs is less salient than commonly believed, and point toward new design spaces focusing solely on control performance. Project page: https://simchowitzlabpublic.github.io/much-ado-about-noising-project/