50 real demos

50 real demos

2000 sim demos

Success rate: 10/20

Success rate: 18/20

2x

2x

• Cotraining improves policy performance by up to 2-7x
• Scaling sim data improves performance and reduces sensitivity to \(\alpha\)

Real Only Training

Sim-and-Real Cotraining

Real-World Demo

Policy Rollout

(Cotrained)

Simulated Demo

• Human teleop
• Fix orientation, then translation
• Sticking & sliding contacts

• Similar to real-world demos

• Model-based planner
• Fix orientation and translation simultaneously
• Sticking contacts only

High-performing policies must learn to identify sim vs real

since the physics of each environment requires different actions

2x

2x

2x