Overview

Vision-Language-Action policies map camera observations, language instructions, and robot state into continuous action chunks. They share some runtime machinery with diffusion pipelines, but the user-facing workload is robot action inference rather than image or video generation.

This section keeps VLA policies separate from the diffusion model cookbook so robot deployments can document their own input schemas, policy endpoints, cache behavior, and control-loop performance targets.

OpenPI

<CardGroup cols={2}> <Card title="Pi0.5" mode="card" href="/cookbook/vla/OpenPI/Pi0.5" /> </CardGroup>