Token Usage Reduced by 74.9%: Real2Gym Turns Simulated Experience into Reusable Real-World Skills
Token用量减少74.9%：Real2Gym把仿真经验变成真机可复用技能
Real2Gym is a research initiative proposed by institutions such as the Shanghai Artificial Intelligence Laboratory, aiming to construct interactive simulation environments through video, enabling robots to learn and transfer skills to the real world. This method combines geometric constraints, interaction keyframe correction, and physics engine validation, significantly improving task success rates and efficiency while reducing model invocation frequency.
Real2Gym presents a novel approach to bridge simulation and real-world robot operation by converting human or robot demonstration videos into interactive environments. The method emphasizes geometric accuracy, interaction correction, and physics-based validation, enabling agents to learn and reuse skills across different tasks and environments.
Source: QbitAI — Robotics on WeChat · Read original article ↗
English translation unavailable. Choose the original article to read it.
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- Environment
- SimulationOpen source S5Open source S12Open source S17
Reported numbers
Reported success rate
87.5%
What remains unknown
Not established in the collected evidence: Control, Data origin.
Reported performance applies to the described task. It does not establish general autonomy or deployment readiness.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
针对这一问题，来自 上海人工智能实验室、上海交通大学、浙江大学、中国科学技术大学、香港中文大学和香港大学 的研究团队提出 Real2Gym ：从人类或机器人视频构建可交互的仿真环境，让Agent在其中探索、积累技能，再将经验用于真实机器人。
Open source S5
执行轨迹随后导回Blender，与原视频对照，核查视觉对应和物理可行性。
Open source S12
在仿真里修正失败再回到真机 在Franka机械臂上，团队测试了方块入盘、挂杯、杯子放入微波炉并关门，以及盘子上架四项任务。
Open source S17
在这些环境中的Agent评估里，复用首轮经验后的第二次执行，平均成功率达到 87.5% ，GPT-6 Astra Direct Mode基线为70.8%；策略执行的 Token用量减少约74.9% 。
Open source S16
Token用量减少74.9%：Real2Gym把仿真经验变成真机可复用技能
Open source S1
Source：QbitAI — Robotics on WeChat · mp.weixin.qq.com