Core Insights - The development of embodied intelligence is currently hindered by a shortage of data, with the available data for this industry being only a fraction of that used by large language models [1][2] - The establishment of data collection factories across various regions in China, such as the Super EID Factory in Tianjin and others in Shanghai and Beijing, marks the beginning of efforts to address this data scarcity [1][2] Group 1: Data Collection Facilities - The Super EID Factory in Tianjin spans 12,000 square meters and focuses on producing data rather than physical products, with a daily capacity to collect up to 550,000 data points and an annual output of 200 million high-quality data points [2][4] - The Zhiyuan data collection center in Shanghai, covering 3,000 square meters, has already collected over one million high-quality data points across various real-world scenarios [3][5] Group 2: Data Collection Methods - Different companies employ various methods for data collection; for instance,帕西尼 uses human data collection where operators wear devices to capture actions, while Zhiyuan employs robotic teleoperation to gather data through repetitive tasks [4][5] - The data collection process involves capturing both simulated and real data, with real data being the most valuable yet challenging to obtain [2][3] Group 3: Industry Challenges and Standards - The lack of standardized data collection practices poses a significant challenge for the industry, making it difficult to reuse and transfer training data across different applications [6] -帕西尼 and Zhiyuan are working towards building an open ecosystem to enhance data availability and quality, with帕西尼 actively participating in establishing data collection standards [6]

据中国证券报消息，智元数据采集中心在上海浦东运营，通过"数据+AI"提升机器人智能化水平。自 2024年9月启动以来，该中心已采集超百万条高质量数据，覆盖多种真实场景。智元机器人开源了百万 真机数据集AgiBot World，并发布了通用具身基座模型GO-1，提升机器人学习效率。今年4月推出的 Genie Studio平台，为开发者提供一站式解决方案。预计2025年，智元机器人将进入量产化元年，推动 产品商用，出货量达数千台。公司已完成新一轮融资，支持其智能化进程。(中国证券报) ...

Group 1 - The core viewpoint of the articles highlights the challenges and advancements in the field of humanoid robots, particularly focusing on the need for training data to enhance their capabilities [1][2][3] - Humanoid robots are gradually demonstrating autonomy in complex scenarios, but they still face limitations in precision, speed, and generalization due to insufficient training data [1][3] - Major companies like Tesla and Google are actively working on creating training datasets, but they encounter high costs and long timelines in the process [2][3] Group 2 - The scarcity of training data for embodied intelligence models is a significant bottleneck, with estimates suggesting a million-fold difference compared to text data [2][3] - The largest datasets currently available for humanoid robots are only in the millions, which is inadequate compared to the billions of data points generated in the automotive sector [3] - The lack of sufficient data hampers the training of effective models, leading to a slow iteration cycle and limited real-world application [3] Group 3 - Synthetic data is emerging as a viable solution to the data scarcity issue, utilizing generative AI techniques to create data that mimics real-world scenarios [4][5] - Companies like Galaxy General Robotics are demonstrating the potential of synthetic data with models trained on datasets exceeding one billion entries, which are already being deployed in operational settings like 24-hour unmanned pharmacies [4][5] - Despite its advantages, synthetic data has limitations, particularly in generating multi-modal data such as tactile and auditory information, and concerns exist regarding the effectiveness of synthetic data in real-world applications [5][6] Group 4 - The "simulation to reality" transfer process is crucial for training embodied intelligence models, requiring a reduction in the gap between simulated and physical environments [6][7] - The National and Local Joint Innovation Center for Humanoid Robots is exploring ways to enhance data interoperability across different robot architectures to avoid redundant training efforts [7] - The center has developed a platform that collects data from over 100 robot configurations, aiming to facilitate better data sharing and training efficiency within the industry [7]

证券研究报告 具身智能数据:AI时代的石油 证券分析师 ：王紫敬 执业证书编号：S0600521080005 联系邮箱：wangzj@dwzq.com.cn 二零二五年六月五日 请务必阅读正文之后的免责声明部分 投资要点 1 ➢ 数据是推动具身智能技术快速突破和落地应用的关键。借鉴自动驾驶汽车发展的路径，数据对于具身智能同样至关重要。优质的数据集能够驱 动智能体感知与理解环境，能够加速具身智能模型的训练与部署，帮助机器人有效完成复杂任务。不同于大语言模型可以使用互联网海量信息 作为训练数据，机器人所用的具身智能模型没有现成数据可以使用，需要投入大量时间和资源进行机器人操作实践或仿真模拟，以收集视觉、 触觉、力觉、运动轨迹以及机器人本体状态等多源异构数据。符合通用标准、得到验证的数据集成为具身智能行业的刚需。当前具身智能本体 形态多种多样，应用场景千差万别，对于具身智能训练数据的需求也更为多元。目前业内仍有部分数据集主要聚焦在特定机器人、特定场景和 特定技能等方面，在整体通用性上有待提升。因此，构建高质量、多样化的感知数据集是不可或缺的基础工作，这些数据集不仅为算法训练提 供了丰富的素材，也成为了评估具身性能的 ...

腾讯正式入股"天才少年"稚晖君创业的机器人公司。 3月24日，证券时报·券商中国记者从天眼查获悉，近日，智元机器人关联公司上海智元新创技术有限公司发生工商变更，原股东临港新片区道禾前沿碳中禾 （上海）私募投资基金合伙企业等退出，新增广西腾讯创业投资有限公司、卧龙电驱等为股东，其中，广西腾讯创业投资有限公司为腾讯旗下投资公司。 与此同时，上海智元新创技术有限公司的原法定代表人舒远春卸任，由邓泰华接任；同时注册资本由约7637万元人民币增至约8046万元人民币。 据悉，该公司成立于2023年2月，经营范围包括智能机器人的研发、服务消费机器人销售、人工智能理论与算法软件开发等，现由桑蓬（上海）科技合伙企业 （有限合伙）、管鲍（上海）科技合伙企业（有限合伙）等及上述新增股东共同持股。 就上海智元新创技术有限公司的股东而言，不可谓阵容不豪华。除了此次新入股的腾讯，此前融资中已经吸纳了包括比亚迪、高瓴资本、软通动力、蓝驰创 投等机构，其中，新能源汽车巨头比亚迪持股上海智元新创技术有限公司2.38%股权。 曾经的华为"天才少年" 一家初创公司，获得业界多家知名企业或投资机构的青睐，背后与一位离职的华为"天才少年"有着密切关 ...

上周五，华为出身"天才少年"——彭志辉（智元机器人联合创始人兼CTO稚晖君）的官宣消息，刷屏朋 友圈，在微博也赚足热度。 3月1 0日早间，消息落地，"稚晖君"创业项目智元机器人（Ag iBo t）发布了 首个通用具身基 座模型——智元启元大模型(Geni eOpe r a t o r- 1 )。 据介绍，智元启元大模型开创性地提出了Visi o n -La n g u a g e -La t e n t-Ac ti o n (ViLLA)架构，该 架构由VLM(多模态大模型)+MoE(混合专家)组成， 实现了可以利用人类视频学习 ，完成小 样本快速泛化，降低了具身智能门槛，并成功部署到智元多款机器人本体。 智元机器人发布的视频展示，GO-1通过学习人类视频，助力机器人本体完成 递送水杯、制作餐点、迎接 外宾 等一系列任务。从性能来看， GO-1相比已有模型成功率大幅领先，平均成功率提高了32%。 经实 验发现，GO-1在"倒水""清理桌面""补充饮料"等任务中表现相对突出。 图：智元机器人公众号 据《科创板日报》报道，智元具身研究中心常务主任任广辉介绍，截至目前， 智元机器人已 经量产下线1 0 0 0台机 ...