🔥36氪•Stalecollected in 1m
Virtual Point Joins OpenLET Embodied AI Committee
💡Access a massive, collaborative open-source dataset ecosystem for training embodied AI and humanoid robots.
⚡ 30-Second TL;DR
What Changed
Virtual Point joins 9 founding members including Peking University and Baidu Cloud.
Why It Matters
This collaboration strengthens the infrastructure for embodied AI by standardizing data collection. It provides researchers with better access to high-quality datasets for training humanoid robots.
What To Do Next
Explore the OpenLET dataset repository to identify relevant training data for your robotics simulation projects.
Who should care:Researchers & Academics
Key Points
- •Virtual Point joins 9 founding members including Peking University and Baidu Cloud.
- •OpenLET community aims to aggregate embodied AI data resources for open sharing.
- •The initiative focuses on accelerating the practical deployment of embodied AI technologies.
- •Over 70 organizations from academia and industry are involved in the ecosystem.
🧠 Deep Insight
Web-grounded analysis with 14 cited sources.
🔑 Enhanced Key Takeaways
- •The OpenLET Embodied AI Open Dataset Community, which Virtual Point has joined, is primarily led by Leju Robotics, a high-tech enterprise specializing in humanoid robots, rather than Robot Era as initially stated in the article summary.
- •The OpenLET dataset, introduced by Leju Robotics, is notable for being the first to combine tactile dexterous manipulation with full-body dynamic motion, integrating precise fingertip pressure matrix sensing and six-axis force feedback with control signals across 41 joints.
- •Robot Era, a key player in the embodied AI ecosystem, has developed its own end-to-end large robotic model called ERA-42, which integrates multimodal information (vision, language, touch, body posture) and can learn new complex tasks in under two hours with minimal data, demonstrating strong generalization capabilities without pre-programming.
- •China's embodied AI market is experiencing rapid growth, projected to expand at a Compound Annual Growth Rate (CAGR) of 51.10% by 2032, with the country having launched nearly 100 embodied AI robotic products since 2024 and capturing approximately 70% of the global market.
- •The OpenLET initiative aims to address the critical shortage of high-quality, real-world multimodal data for robot brain development by focusing on open-sourcing datasets and toolchains, establishing a collaborative governance framework, and accelerating the integration of research with industrial applications.
📊 Competitor Analysis▸ Show
| Feature/Platform | OpenLET (Leju Robotics) | Open X-Embodiment (Google DeepMind et al.) | AgiBot World (AGIBOT) |
|---|---|---|---|
| Leadership | Leju Robotics | Google DeepMind and 21+ global institutions | AGIBOT (Shanghai-based) |
| Launch Date | March 2026 | October 2023 | April 2026 |
| Scale | >1 million downloads, >60,000 minutes of data, 117 atomic skills | >1 million real robot trajectories, 60 existing datasets, 22 robot embodiments, 527 skills, 160,266 tasks | Million-scale real-world dataset, >1 million demonstration trajectories from 100 robots |
| Key Data Types | Tactile dexterous manipulation, full-body dynamic motion, fingertip pressure, 6-axis force feedback, 41 joints | Diverse robot types (single-arm, bi-manual, quadrupeds), various behaviors and household objects | Real-world scenarios (commercial, home, everyday), free-form data collection, 1:1 digital twin simulation data |
| Technical Focus | Addressing real-world data bottleneck, integrating data collection, open access, data trading | Standardized data formats, cross-robot generalization, pre-training for generalist policies | Comprehensive real-world scenarios, versatile hardware platforms, full-process quality control, imitation learning |
| Pricing | Open-source (free access) | Open-source (free access) | Open-source (free access) |
| Benchmarks | Null | RT-X model shows positive transfer and improved capabilities across multiple robots | Null |
🛠️ Technical Deep Dive
- OpenLET Dataset (by Leju Robotics):
- Integrates tactile dexterous manipulation with full-body dynamic motion.
- Features fingertip pressure matrix sensing and six-axis force feedback with ±0.5% precision.
- Includes control signals spanning 41 joints.
- Comprises over 60,000 minutes of open-source data covering 117 atomic skills across industrial, commercial, and household scenarios.
- Robot Era's ERA-42 Model:
- An end-to-end native robotic large model.
- Achieves unified generalization by integrating full-modal information: vision, language, touch, and body posture.
- Designed for real-time adaptability without manual feature design or pre-programming, processing all data through a streamlined neural network.
- Can learn new tasks within two hours using minimal data and continuously expands its skill repertoire.
- Incorporates world model technology, enabling it to understand the physical world and predict future behaviors, enhancing anti-interference capabilities.
- When combined with Robot Era's XHAND1 five-finger dexterous hand, it can perform over 100 complex operational tasks using various tools.
🔮 Future ImplicationsAI analysis grounded in cited sources
The collaboration within OpenLET, particularly with Leju Robotics' leadership, will significantly accelerate the standardization of embodied AI data formats in China.
Leju Robotics' initiative to launch China's first open-source embodied intelligence dataset community, coupled with its focus on integrating data collection, open access, and data trading, suggests a strong push towards unified data infrastructure, which is crucial for standardization.
Robot Era's ERA-42 model, with its rapid learning and generalization capabilities, will enable faster deployment of versatile humanoid robots in industrial and commercial sectors.
The model's ability to learn new tasks in under two hours with minimal data and its end-to-end design for real-time adaptability will reduce development cycles and increase efficiency for various applications.
The increasing investment and policy support in China for embodied AI will solidify its position as a global leader in humanoid robot development and deployment.
China's embodied AI market is projected for exceptional growth, with significant government backing and a high volume of new robotic products, indicating a strategic national focus on this technology.
⏳ Timeline
2016
Leju (Shenzhen) Robotics Co., Ltd. founded.
2023-08
Robot Era (Xingdong Jiyuan) founded by Chen Jianyu.
2024-01
Robot Era's full-size humanoid robot demonstrates autonomous snow walking. [cite: 6 (from previous search)]
2024-12
Robot Era officially launches its ERA-42 end-to-end large robotic model.
2025
Leju Robotics delivers its 100th full-size humanoid robot.
2026-03-18
Leju Robotics launches China's first OpenLET Embodied AI Open Dataset Community.
📎 Sources (14)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
📰
Weekly AI Recap
Read this week's curated digest of top AI events →
👉Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: 36氪 ↗