TwinBridge: Digital Twin Transfer Strategy and Rewards
What happened
Shu'ang Sun and other authors released the TwinBridge framework, which uses digital twins to provide policy initialization and dense visual rewards for real-world online reinforcement learning. Its digital twin is built from 3D Gaussian splatting combined with physics simulation; simulation states give task events and geometric progress, used for simulation demonstration pretraining and reinforcement learning, then distilled into a visual reward model. In real interaction, the model's predictions are combined with execution history to compute dense rewards, supporting human-in-the-loop policy optimization. The authors claim that on five real laboratory tasks, using at most 25 minutes of real interaction per task, the final average autonomous success rate reached 98%.
AI 根据报道生成 · 6 days ago更新
Reporting timeline
Follow the reporting to explore different perspectives.
- arXiv cs.ROTwinBridge: Using digital twins to transfer policies and rewards for real-world reinforcement learning in laboratory robotics
TwinBridge proposes a digital twin-driven framework that provides policy initialization and dense visual rewards for real-world online reinforcement learning. The method builds digital twins by combining 3D Gaussian splatting with physics simulation; simulator states provide task events and geometric progress, which are used for pretraining with simulated demonstrations and for reinforcement learning, then distilled into a visual reward model. Combined with execution history, this computes dense rewards and supports human-in-the-loop policy optimization.
Activity for this event
- Current
- 3
- Peak
- 10Nov 5 14:00
- Past 24 hours
- -49%
The trend compares participants observed throughout the same period, so its range may be smaller than the current interest count. Move or tap to inspect each hour; use the arrow keys on a keyboard.