VITA: Vision-to-Action Flow Matching Policy

Researchers developed VITA, a new AI framework that streamlines robot policy learning by directly flowing from visual inputs to actions without requiring conditioning modules. The system achieves 1.5-2x faster inference speeds while maintaining or improving performance compared to existing methods across 14 simulation and real-world robotic tasks.

• →VITA eliminates the need for visual conditioning during action generation, reducing computational overhead significantly.
• →The framework uses an action autoencoder to map raw actions into structured latent space aligned with visual representations.
• →Flow latent decoding prevents latent action space collapse during training by anchoring the generation process.
• →Testing across 9 simulation and 5 real-world tasks shows 1.5-2x speed improvements over conventional methods.
• →The noise-free and conditioning-free approach represents a meaningful advance in robotic policy learning efficiency.