Axis V2

Axis V2 is a robot-learning platform for improving trained policies through interaction with physical or simulated environments. A policy performs tasks, the system records its trajectories and human operators intervene when failures or uncertain states occur. Those corrections are converted into additional training data for subsequent policy updates.

The workflow supports human-gated DAgger, active rollouts and collection of verified demonstrations across multiple robot embodiments. Axis also provides deployment and data infrastructure so teams can manage the cycle from policy release through correction and retraining. The platform is intended for manipulation systems that require continuing improvement after the initial offline training stage.