Diffusion policy: Visuomotor policy learning via action diffusion

Major category

Foundational work

Quadrant

Not applicable

Architecture

Not applicable

Prediction paradigm

Not applicable

Subcategories

Action policy foundationsDiffusion & flow-matching foundations

Source review status

Verified from primary sources

Category review. Diffusion Policy establishes observation-conditioned action diffusion and receding-horizon action-sequence control as a foundational visuomotor policy-learning method. The original RSS work appeared in 2023; the reviewed IJRR extension builds on that method. Its role is an action-policy foundation for later WAM action decoders, rather than a general pretrained backbone or tokenizer. Reading evidence