Unverified50% confidenceFactExact time
Microduck 专为 sim-to-real（仿真到现实）强化学习设计
1
Sources
50%
Confidence
Long-term
Relevance
8/31/2026
First Seen
Sources
Related Entities
Related Claims
Unverified强化学习方案通常基于MuJoCo、Isaac Gym等物理仿真引擎构建，再通过Sim-to-Real迁移技术部署到真实硬件71% similarUnverifiedMicroduck 出厂即搭载 7 种预训练行为65% similarUnverified超级马里奥强化学习环境通常基于gym-super-mario-bros库，输入处理常用灰度化、堆叠连续4帧、下采样到84×84像素等技巧62% similarVerified增强学习领域用'sim-to-real gap'描述在模拟环境训练的机器人策略因现实物理差异而在真实世界失效的现象62% similarUnverified预训练+微调范式源自NLP和计算机视觉领域，其核心是大规模数据预训练学到通用底层特征，下游任务用少量标注数据适配60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/828439API
curl https://kongchang.com/api/v1/knowledge/claims/828439MCP
get_claim(id=828439)