Unverified50% confidenceFactExact time
MicroDuck's walking, fall-recovery, and object-pickup actions are based on reinforcement learning, trained in a simulation environment and transferred to the real robot via domain randomization, with all underlying software written in Rust and supporting automatic firmware rollback.
1
Sources
50%
Confidence
Long-term
Relevance
9/10/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedMicroduck is designed specifically for sim-to-real reinforcement learning.76% similarUnverifiedMicroduck ships with 7 pre-trained behaviors out of the box.73% similarUnverified四足机器人的自我恢复能力来自强化学习的sim-to-real训练范式，通过在仿真环境中摔倒数百万次学习平衡策略64% similarUnverified域随机化（Domain Randomization）和系统辨识等技术在2016-2020年出现，大幅缩小了仿真与现实的差距，DeepMind、OpenAI等机构成功将仿真训练策略迁移到真实硬件63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/890131API
curl https://kongchang.com/api/v1/knowledge/claims/890131MCP
get_claim(id=890131)