Factored Diffusion Policies:Compositionally Generalized Robot Control with a Single Score Network

TL;DR

This paper introduces factored diffusion policies that use a single network with additive factor decomposition, enabling efficient and generalizable robot control across diverse task factors with theoretical guarantees.

Contribution

It proposes a novel factored diffusion policy framework with a shared network and a score decomposition method, reducing training complexity and providing formal control guarantees.

Findings

Achieves 90% success on held-out gates, matching an oracle.

Transfers zero-shot to unseen venues with +11.7pp success rate.

Reduces crash rate by 2.4 times in vision-based tasks.

Abstract

Robotic tasks are typically specified by a tuple of factors, such as the object to be grasped, the obstacles to be avoided, the color of the target, and so on. Collecting expert demonstrations for every combination of factor values grows combinatorially. We present factored diffusion policies: a single shared diffusion network trained with per-factor null-token dropout, whose score decomposes additively across factors at inference. Under approximate conditional independence between factors given the action-observation pair, this composition approximates the true joint score with a bounded uniform error, reducing the training-task budget from a product of factor cardinalities to a sum. A trajectory-tube certificate chains this score-level bound through the reverse-time sampling ODE and a contracting tracking controller into a closed-loop state-trajectory tube whose radius factors into an ODE-sensitivity constant and a per-factor score-error budget. Unlike compositional-diffusion methods for control that combine separately trained networks, we use one shared network. Drone racing experiments confirm both the generalization bound and the certificate. On state-based multi-gate racing, the factored policy passes 90% of held-out gates -- matching an oracle -- while a K-network composition baseline collapses to 3%; on vision-based single-gate traversal, it transfers zero-shot to an unseen venue with +11.7pp success-rate gain and 2.4X crash-rate reduction.