Termsslug: sim-to-real-transfer

Sim-to-Real Transfer

仿真到现实迁移

The process of applying models trained in computer simulations directly to robots in the physical world.

Curated Intelligence

15 items · Page 1/2
IT之家 · 智能时代T2·Agents
84 pts
Apple Unveils HALO Humanoid Robot System With 91% Success Rate in Unmapped Environments

Apple's research team published an arXiv paper detailing HALO, a humanoid robot autonomy stack that layers a VLM task orchestrator, a DiT motion generator, and a reinforcement-learning tracker, deployed on a Unitree G1 with an onboard Mac mini for inference. Across 360 single-task trials in six previously unmapped real environments, HALO averaged a 91% success rate, though long-horizon composite tasks fell to 70% and the hardest combination to 50%. The team also open-sourced the data generation, training, and deployment code.

机器之心 (微信公众号)T2·Agents
38 pts
Tao Dacheng's Team Proposes Dual-Brain Embodied Framework for Reusable Robot Experience

Tao Dacheng, chief scientist at Daxiao Robotics, published a technical article outlining a "dual-brain embodied" framework in which SenseNova handles task understanding, planning and skill organization while the Kaiwu world model predicts physical states and action consequences. The goal is to distill real-world interaction experience, together with evidence and applicability conditions, into knowledge reusable across tasks, scenes and robot bodies. The piece is conceptual and provides no parameters, datasets or quantitative results.

InfoQ 中文T2·Agents
45 pts
LeXiang Tech Details Home Robot Deployment: Lightweight Arms & VLA

LeXiang Tech's CTO shared progress on home service robot productization at AICon, detailing lightweight heterogeneous dual-arm design, layered software architecture, and sim-to-real data loops. The talk explored synergies between rule-based and modular strategies, highlighted performance gaps in end-to-end methods under real-world constraints, and proposed the RSI self-evolution system.

量子位官网T2·Agents
78 pts
Tesla Starts Optimus Production; Musk Plans 'Robot School' for Training

Tesla has confirmed that the Fremont factory's Model S/X line has been converted to early production of Optimus humanoid robots, with the first batch meeting mass-production quality standards. Elon Musk plans to manufacture 10,000-20,000 units initially for training in a dedicated 'robot school,' targeting external deliveries by late 2027 and a long-term capacity of one million units annually.

IT之家 · 智能时代T2·Agents
68 pts
OpenAI Confirms Humanoid Robot Plans, Altman Prioritizes AI 'Brain'

OpenAI CEO Sam Altman confirmed the company will build its own humanoid robots, emphasizing that developing the intelligent 'brain' is more critical than hardware form factors. A demo prototype is expected in 2027, with short-term goals focused on infrastructure assistance and long-term personal task completion. OpenAI currently lists 27 robot-related job openings covering motion control, custom actuators, and ML infrastructure.

InfoQ 中文T2·Agents
38 pts
QCon Shanghai Embodied AI Track Preview: Focus on Real-World Data Loops and Sim-to-Real Alignment

The QCon Shanghai conference will feature a track on 'Data-Driven Embodied AI,' curated by Dr. Huang Yongtao from Ant Lingbo Technology. The session aims to explore scaling embodied intelligence through high-quality real-world data systems, optimized continuous learning loops, and bridging the sim-to-real gap. Currently, the content is limited to agenda previews and topic outlines, with no specific technical releases or open-source projects announced yet.

Rohan Paul (@rohanpaul_ai)T2·Agents
68 pts
WUJI Hand 2 Open-Sourced: Sim-to-Real Pen Spinning via MuJoCo and PPO on a 20-Joint Robot Hand

WUJI released and open-sourced the WUJI Hand 2, demonstrating high-dexterity pen spinning capabilities trained in simulation and deployed directly to hardware. The approach utilizes the mjlab toolkit within MuJoCo to run 4,096 parallel virtual copies, employing Proximal Policy Optimization (PPO) to learn from 40 reference motion trajectories for efficient sim-to-real transfer. Code and setup guides are now fully available for developers to reproduce this reinforcement learning-based dexterous manipulation paradigm.

Hugging Face BlogT1·Agents
68 pts
Strands Agents Integrates LeRobot to Bridge Hub and Robot Hardware

Amazon Strands Agents introduces deep integration with Hugging Face LeRobot, enabling a unified workflow from dataset download and model training to physical robot deployment. This solution eliminates tool fragmentation by using consistent data formats and policy interfaces, allowing seamless switching between simulation and real hardware. Developers can now drive physical robots directly via pre-trained policies on the Hub within a single agent loop.

OpenAI 官方博客T1·Agents
78 pts
OpenAI Robot Hand Solves Rubik's Cube via Sim-to-Real RL

OpenAI trained neural networks entirely in simulation using Automatic Domain Randomization (ADR) to enable a human-like robot hand to solve the Rubik's Cube. This breakthrough demonstrates that reinforcement learning can achieve unprecedented dexterity and zero-shot transfer from virtual to physical environments.