FutureVision
FutureVision is Rhoda AI's robot intelligence, which the company describes as a new approach based on video-predictive control. Its Direct Video-Action model predicts future video frames, turns them into robot actions and repeats the loop every few hundred milliseconds. It is pre-trained on internet-scale video and post-trained on smaller amounts of robot data, for industrial work in manufacturing and logistics.
- Repeats its predict-and-act process every few hundred milliseconds in a closed loop (press release)
- Long-horizon tasks learned with 10-20 hours of robot data (DVA research page)