Workflow
具身智能之心
icon
Search documents
8个实战,彻底讲清VLA的各类方案
具身智能之心· 2025-12-08 01:11
Core Viewpoint - The article discusses the challenges and advancements in the VLA (Variable Learning Algorithm) field, emphasizing the importance of real machine data and practical experience in achieving effective results in embodied intelligence applications. Group 1: Data Collection - Data collection methods for VLA primarily include imitation learning and reinforcement learning, with remote operation, VR, and full-body motion capture being key techniques [8][9] - The quality of data collected is crucial, and methods like real2sim2real are highlighted as important for effective data acquisition [8] Group 2: VLA Training - Before deploying models in real machines, simulation debugging is essential, especially when real machine data is insufficient [10] - Training techniques are critical, with challenges in fine-tuning models and achieving good results with small data sets [10] - Some algorithms, like ACT, are easier to train, while others, such as π0 and π0.5, require more intricate techniques and experience [10] Group 3: VLA Deployment - After training, models often need to be "slimmed down" due to their large parameter sizes, which poses challenges for deployment on edge chips [12] - Techniques like quantization and distillation are necessary to minimize parameter size while maintaining performance [12] Group 4: Educational Initiatives - The article introduces a practical course aimed at helping individuals learn VLA effectively, covering various aspects such as hardware, data collection, algorithms, and deployment [13][16] - The course is designed for a wide audience, including students and professionals looking to transition into the embodied intelligence field [27]
SpaceX估值8000亿美元超OpenAI,IPO就在明年
具身智能之心· 2025-12-08 01:11
Core Viewpoint - SpaceX is poised to become the highest-valued private company globally, with a potential valuation of $800 billion, surpassing OpenAI's $500 billion valuation [1][2][12]. Valuation Insights - SpaceX is negotiating a new round of internal stock sales that could elevate its valuation to $800 billion, nearly equivalent to Switzerland's GDP of $900 billion [1]. - The company's valuation has doubled from $400 billion in July 2023 to $800 billion in less than six months [3][10]. - There is speculation that the final valuation could be around $560 billion if shares are priced at approximately $300 each [11]. Business Overview - Founded in 2002, SpaceX is a private aerospace and space transportation company based in Texas, aiming to reduce space launch costs and facilitate human colonization of Mars [5][6]. - SpaceX has two core business segments: rocket launches and satellite services, with a dominant position in the rocket launch market [17][18]. Satellite Business - The satellite business is a significant driver of SpaceX's high valuation, with the Starlink division operating around 9,000 satellites, contributing to revenue [19][23]. - SpaceX's revenue for 2023 is projected to reach $15.5 billion, slightly exceeding OpenAI's expected revenue of $13 billion [25][26]. IPO Expectations - SpaceX is expected to go public in the second half of next year, potentially ahead of OpenAI, which is rumored to file for an IPO in 2026 [16][26]. - There is a possibility that SpaceX may bundle its rocket and satellite businesses for the IPO, rather than spinning off Starlink as a separate entity [24].
看到字节最新的GR-RL丝滑系鞋带,真的怕了......
具身智能之心· 2025-12-07 03:03
编辑丨具身智能之心 本文只做学术分享,如有侵权,联系删文 >> 点击进入→ 具身智能之心 技术交流群 更多干货,欢迎加入国内首个具身智能全栈学习社区 : 具身智能之心知识星球 (戳我) , 这里包含所有你想要的。 点击下方 卡片 ,关注" 具身智能 之心 "公众号 作者丨 Yunfei Li等 在机器人操作领域,视觉 - 语言 - 动作(VLA)模型虽已实现多任务泛化,却始终难以突破长时程精细操作的核心瓶颈——以系鞋带为例,该任务需同时满足 "毫 米级穿洞精度""柔性鞋带操控""多步误差规避" 三大要求,传统方法或通用 VLA 模型均因能力短板无法胜任。 字节跳动提出的 GR-RL 框架 ,以 "问题定位 - 方案设计 - 落地验证" 为逻辑主线,通过多阶段训练 pipeline 将通用 VLA 模型转化为长时程精细操作专家,成为首个 能自主完成多鞋孔系鞋带的学习型模型。 论文题目:GR-RL: Going Dexterous and Precise for Long-Horizon Robotic Manipulation 项目链接:https://seed.bytedance.com/gr_rl 作者单位 ...
今年大家最关注的具身方向原来是这些?
具身智能之心· 2025-12-07 03:03
最近正在准备为具身行业起草一份非常丰富的研报,预计明年的第一季度公布。因为涉及的内容和方向 非常多,包括具身公司的融资、产业、政策、算法、落地、出口等多个模块,所以也非常想了解下大家 都在关注哪些内容,侧重点应该在哪里。 国内具身产业与政策 国外具身产业情况 具身公司融资、业务情况 具身数采相关 具身算法优化部署相关 机器人边缘芯片相关 具身下游产业发展 具身产业人才结构与需求 具身公司上市辅导等 其它 微信扫码填写,只需10s 为了更好服务大家,我们也简单做个调研,涉及以下板块,支持多选哦~ ...
深扒PI π*0.6迭代式强化学习思路:VLA+在线RL,实现自我进化
具身智能之心· 2025-12-07 03:03
见证具身浪潮,书写智能新纪元 以下文章来源于具身纪元 ,作者具身纪元 具身纪元 . 更多干货,欢迎加入国内首个具身智能全栈学习社区: 具身智能之心知识星球(戳我) ,这里包含所有你想要的! 在Physical Intelligence 最新的成果π 0.6 论文里,他们介绍了 π 0 .6迭代式强化学习的思路来源: 其中有我们熟悉的Yuke Zhu的研究,也有他们自己(Chelsea Finn、Sergey Levine)的一些研究,我们之前对这些工作一直有跟踪和介绍。此外,还有来自国内具身智能团队的 工作,比如清华大学、星动纪元的研究。 随着π*0.6的发布,VLA+online RL成为了一个行业共识的非常有前景的研究方向 深扒了Π*0.6的论文,发现它不止于真实世界强化 学习 英伟达也来做VLA在真实世界自我改进的方法了 大语言模型从SFT到RL的发展方向也逐渐在具身研究中清晰明朗。 一、为什么VLA+RL很重要 编辑丨 具身纪元 点击下方 卡片 ,关注" 具身智能之心 "公众号 >> 点击进入→ 具身 智能之心 技术交流群 图注:VLA模型依赖研读微调 在具身智能(Embodied AI)领域,科学家 ...
具身智能的黄埔军校,都有哪些东西?
具身智能之心· 2025-12-07 03:03
最近在为大家收敛具身科研的几个重点模块:行业内容、本体形态、算法、还有部署的一些方案,已经汇总 在我们的社区内部。 现已梳理了行业正在从事具身大脑、本体研发的公司(突然发现本体也卷不太动了......),以及一些比较活跃的 具身实验室。除此之外,还提供了很多行业研报,供大家判断具身的发展与周期。 本体方面,推荐几款适合科研的产品:SO-100系列、openarm系列、XLerobot系列等; SO100及升级版本,能上一些VA和VLA的算法,常见功能可以实现了; Openarm是一款双臂任务框架,目前有几家公司开始生产相关本体,缺乏移动能力,一些叠衣服、pick and place也都能满足。但从数据采集来看,VR版本更舒服。 算法层面,目前我们收拢了关于vla(训练、无需训练方式、vla+RL、vla+世界模型、vla轻量化、部署等)、 vln(时间语言、目标导航、点导航等)、运控(强化、MPC、WBC)、仿真(通用、真实)、触觉感知等多 个方向路线; 部署层面,目前大多集中在云端推理。边缘侧基于索尔的方案已经逐渐落地(vla模型),除此之外,类似于 小鹏这类公司基于自研芯片已经完成vlm/vla的部署。1 ...
已经有7所高校,在悄悄地设立具身专业了......
具身智能之心· 2025-12-06 03:11
点击下方 卡片 ,关注 "红岸" 公众号 >> 点击进入→ 具身智能之心 技术交流群 更多干货,欢迎加入国内首个具身智能全栈学习社区 : 具身智能之心知识星球 (戳我) , 这里包含所有你想要 的。 前两天分享了清华具身研究院和上交的具身专业开设,除了这两所,还有另外6所双一流高校正在申请增 设"具身智能本科专业"。以下为教育部公示的名单。 | 学校名称 | 专业名称 | 学位授予门类 | 申报类型 | 申请表 | | --- | --- | --- | --- | --- | | 北京航空航天大学 | 具身智能 | 工学 | 尚未列入目录的新专业 | 下载 | | 北京理工大学 | 具身智能 | 工学 | 尚未列入目录的新专业 | 下载 | | 北京邮电大学 | 具身智能 | 工学 | 尚未列入目录的新专业 | 下载 | | 东北大学 | 具身智能 | 工学 | 尚未列入目录的新专业 | 下載 | | 上海交通大学 | 具身智能 | 工学 | 尚未列入目录的新专业 | 下载 | | 浙江大学 | 具身智能 | 工学 | 尚未列入目录的新专业 | 下载 | | 西安交通大学 | 具身智能 | 工学 | 尚未列 ...
字节前技术负责人联手清华姚班校友创业!
具身智能之心· 2025-12-05 16:02
编辑丨 机器之心 点击下方 卡片 ,关注" 具身智能之心 "公众号 自「造词大神」Andrej Karpathy 提出「Vibe Coding」这个概念后,它的热度就居高不下。 只需一句提示词描述「感觉」和意图,AI 就能直接生成可运行的代码,这种近乎魔法的编程体验让一众开发者叹为观止。 输入提示词: write a python code that visualizes how a traffic light works in a one way street with cars entering at random rate (编写一个 Python 代码,可视化单行道中交通信号灯的工作情况,车辆以随机速率 驶入), AI 就能在几秒钟内生成一个完整的动画模拟程序,包括交通灯的红黄绿切换逻辑、车辆的随机生成机制、停车和通行的判断规则,甚至还配上了流畅的可视化界面。 但惊喜过后,问题也随之而来。Vibe Coding 虽然擅长快速原型开发和单脚本编写,但在面对企业级复杂工程时仍显得力不从心。 受限于上下文窗口、推理深度 以及 Agentic 模式缺失, 它往往难以精准定位大型代码库中深埋的 Bug,也极易在 ...
对话多个行业大佬!VLA与RL方案在真机上的部署怎么样啦?
具身智能之心· 2025-12-05 16:02
Core Viewpoint - The article discusses the implementation challenges and advancements of VLA (Variable Latent Action) algorithms and Reinforcement Learning (RL) in robotics, focusing on their practical applications and future developments in the field of embodied intelligence [3][13]. Group 1: Guest Speakers - Wei Sui, Vice President of Diguo Robotics, has extensive experience in developing 2.5D and 3D vision algorithms for robotics and autonomous driving, leading a team that created a comprehensive 4D labeling system, with millions of chips shipped [5]. - Zhang Qiang, Chief Researcher and Academic Committee Director at Beijing Humanoid Robotics, specializes in humanoid robot motion control and multimodal perception, contributing to the development of core RL algorithms for humanoid robots [6][8]. - Wang Tiancai, Partner at Yuanli Lingji, has published over 30 papers in top international conferences and is a core author of notable algorithms in end-to-end autonomous driving [9][10]. - Yu Chao, Assistant Professor at Tsinghua Shenzhen Research Institute, focuses on decision intelligence driven by reinforcement learning, with over 50 published papers and significant academic recognition [11][12]. Group 2: Key Topics Discussed - The article addresses the pain points in the architecture and models of VLA, exploring how to enhance the overall motion control of robots [16]. - It discusses the integration of VLA with RL for better real-world application, including considerations for hardware selection and lightweight implementations [16].
最新分层VLA模型:使用失败的演示数据,也能优化VLA模型!
具身智能之心· 2025-12-05 16:02
作者丨 Jeonguen Park等 编辑丨具身智能之心 本文只做学术分享,如有侵权,联系删文 >> 点击进入→ 具身智能之心 技术交流群 更多干货,欢迎加入国内首个具身智能全栈学习社区 : 具身智能之心知识星球 (戳我) , 这里包含所有你想要的。 点击下方 卡片 ,关注" 具身智能 之心 "公众号 研究背景与核心问题 现有VLA模型的局限 视觉-语言-动作(VLA)模型是机器人操作任务的核心技术,传统模型依赖人类遥控收集的成功演示数据训练,但数据采集过程中自然产生的大量失败尝试(如抓 取不稳定、碰撞等)常被当作噪声丢弃。这些失败数据蕴含着政策脆弱点的关键信息——揭示了哪些动作序列不可行、哪些场景下容易出错,而单纯依赖成功数据 的模型难以应对复杂环境中的不确定性,在未见过的场景中鲁棒性大幅下降。 核心挑战与研究目标 核心挑战在于如何有效整合离线数据中的失败信号:模仿学习(IL)中直接惩罚易失败动作容易扭曲政策,而强化学习(RL)虽能通过奖励信号自然处理失败数 据,但需要合适的框架承载。研究目标是构建一个分层VLA模型,将失败经验转化为结构化学习信号,通过显式的规划机制实现"失败感知推理",在不改变机器人 核 ...