川崎KALEIDO 8.0人形机器人展示下一代协作技术

https://youtu.be/IlZ7msliF1s

标题 / Title

(3) 川崎KALEIDO 8.0人形机器人展示下一代协作技术（MANUS AI代理）- YouTube
(3) Kawasaki KALEIDO 8.0 Humanoid Robot Demos Next Gen Collab Tech (MANUS AI AGENT) - YouTube

---

字幕翻译 / Transcript Translation

(00:00)

中文: 日本川崎重工业公司刚刚展示了他们最新的Kido 8.0机器人演示，展示了几项新功能和AI能力。那么，它现在能做什么呢？首先，一个令人印象深刻的演示是人类和Kido合作搬运一张桌子，这项任务乍一看很简单，但实际上需要极高的协调水平。例如，当人类推动或拉动桌子时，Kido脚底嵌入的力传感器会立即测量重心的变化，然后机器人调整其
英文: Kawasaki Heavy Industries from Japan just unveiled their latest Kido 8.0 robot demonstration with several new features and AI abilities. So what can it do now? To start, one impressive demonstration involved a human and Kido teaming up to move a desk, which is a task that seems simple at a glance but actually requires an extreme level of coordination. For instance, when the human pushes or pulls on the desk, Kido’s force sensors embedded in the soles of its feet instantly measure shifts in the center of gravity, then the robot adjusts its

(00:36)

中文: 踝关节的扭矩以保持平衡，并与人类伙伴保持同步。这种流畅的协作突显了Kido在现实世界中补充人类努力的能力。在硬件方面，川崎公司在Kido的肘部集成了轻量级的紧凑型力传感器，类似于其踝关节的设计。这一 additions 使机器人能够检测作用于前臂的力和力矩，而不仅仅是手腕。这种精确的力检测增强了Kido处理精细或复杂任务的能力，从
英文: ankle torque to stay balanced and in sync with its human partner. This fluid collaboration highlights Kido’s ability to complement human efforts in the real world. And on the hardware front, Kawasaki has integrated lightweight, compact force sensors into Kido’s elbows, mirroring the design used in its ankles. This addition allows the robot to detect forces and moments acting on its forearm beyond just the wrist. Such precision in force detection enhances Kido’s ability to handle delicate or complex tasks, from

(01:08)

中文: 搬运物体到在体力需求较大的场景中与人类合作。这一升级使第八代Kido比其前代产品更具适应性，并为传感器数据收集设定了新标准。但也许第八代Kido最引人注目的特点是其重新设计的头部，现在配备了一个球形显示器，取代了早期的无机设计。新显示器现在使用内部投影仪来展示各种图像，包括视频、文本和类似人类的面部
英文: lifting objects to collaborating with humans in physically demanding scenarios. This upgrade makes the eighth generation of Kido much more adaptable than its predecessors and sets a new standard for sensor data collection. But perhaps the most visually striking feature of the eighth generation Kido is its redesigned head, now equipped with a spherical monitor replacing the earlier inorganic design. The new monitor now uses an internal projector to display a wide range of images, including videos, text, and humanlike facial

(01:41)

中文: 表情。这个显示器将Kido转变为一个社交机器人，支持自然互动，并能够以一种亲切的方式传达意图和情感。无论是微笑以安抚人类合作者，还是文本提示以澄清任务，球形显示器增加了一层互动性，以弥合技术与人类联系之间的差距。由于川崎公司在软件和硬件方面不断完善第八代Kido平台，另一个关键升级是机器人
英文: expressions. This display transforms Kido into a social robot that supports natural interactions and is capable of conveying intentions and emotions in a relatable way. Whether it’s a smile to reassure a human collaborator or a text prompt to clarify a task, the spherical monitor adds a layer of interactivity to bridge the gap between technology and human connection. And because the eighth generation Kido platform has been continuously refined by Kawasaki in terms of both software and hardware, another critical upgrade is the robot’s

(02:13)

中文: 用于自主任务执行的先进平衡控制系统。该系统通过实时修正机器人的脚部着陆位置来保持稳定性，此外，安装在机器人腰部的姿态传感器持续监测其倾斜和速度，并将数据输入强化学习算法，计算每一步的最佳着陆点。此外，这一技术升级确保Kido能够从几乎跌倒的情况中恢复过来，使其在动态环境中成为一个关键特征。至于机器人的
英文: advanced balance control system for autonomous task execution. This system employs real-time correction of the robot’s foot landing positions to maintain stability. Plus, a posture sensor mounted on the robot’s waist continuously monitors its inclination and speed, feeding data into a reinforcement learning algorithm that calculates the optimal landing spot for each step. Additionally, this tech upgrade ensures that Kido can recover from near falls, making it a critical feature in dynamic environments. As for the robot’s

(02:42)

中文: 未来功能，川崎公司还在开发一个远程操作系统，利用市售的动作捕捉设备。该系统允许人类远程控制Kido，执行危险环境中的远程任务等功能。此外，川崎公司还在开发一种混合操作模式，其中常规任务由机器人自动化完成，而非常规任务则由人类操作员管理。这种方法不仅提高了效率，还为将人工智能集成到
英文: future features, Kawasaki is also implementing a teleoperation system that leverages commercially available motion capture devices. This system allows humans to control Kido remotely for functions like remote task execution in dangerous environments. Plus, Kawasaki is also working on a hybrid operational model where routine tasks are automated while non-routine tasks are managed by human operators. This approach not only enhances efficiency but also serves as a stepping stone for integrating artificial intelligence into

(03:15)

中文: Kido的框架中奠定了基础。此外，机器人的远程操作系统意味着一个全新的数据收集水平，通过记录人类控制操作的真实世界工作数据，川崎公司旨在训练AI系统，最终使Kido能够完全自主地执行任务。至于机器人的价格，它仍然远高于15万美元，商业推广遥遥无期。但还有另一款名为Vega的新型通用机器人刚刚发布，它也被设计为执行各种
英文: Kido’s framework. Plus, the robot’s teleoperation system implies a whole new level of data collection. By recording real-world work data from human-controlled operations, Kawasaki aims to train AI systems to eventually enable Kido to perform tasks completely autonomously. As for the robot’s price, it’s still well above $150,000, with a commercial rollout nowhere on the horizon. But there’s another new general-purpose robot named Vega that just released, and it’s also designed to perform a wide range of

(03:45)

中文: 任务，但有一个特点：它配备了高负载手臂、灵巧的手部、可折叠的躯干和手臂结构、全向底座和高容量电池。Vega为各种操作需求提供了一个机器人平台，其设计优先考虑适应性和便携性，具有扩展的reach，可在不同环境中实际使用。事实上，机器人可以延伸至2.
英文: tasks, but with a twist. Built with high-payload arms, dexterous hands, a foldable torso and arm structure, an omnidirectional base, and a high-capacity battery, Vega offers a robot platform for various operational needs, with its design prioritizing adaptability and portability with an extended reach for practical use across different settings. In fact, the robot can extend itself up to 2.

(04:13)

中文: 2米以进入高处区域，并可折叠成紧凑的尺寸以便于运输。此外，其垂直工作空间范围也是2.2米，每个手臂支持15的标称负载，使其适用于涉及适度提升或定位的任务。至于电源，Vega的电池提供超过10小时的操作时间，当两个手臂都在使用而没有负载时，有可能超过30小时以进行长时间活动。Vega拥有超过36°的自由度，配备了需要协调运动和精确度的任务，其手臂具有13°的
英文: 2 meters to access elevated areas and folds into a compact size for easier transport. Plus, its vertical workspace range is also 2.2 meters, with each arm supporting a nominal payload of 15, making it suitable for tasks involving moderate lifting or positioning. As for power, Vega’s battery provides over 10 hours of operation, with the potential to exceed 30 hours when both arms are in use without a payload for prolonged activity. With more than 36 degrees of freedom, Vega is equipped for tasks requiring coordinated movement and precision, with its arms featuring 13 degrees of

(04:48)

中文: 自由度，而躯干和头部各有3°的自由度，以实现更大的机动性，同时空间感知有助于导航其周围环境。此外，全向底座允许在任何方向移动，增强了其在制造设施、仓库或研究空间等环境中的实用性。Vega的设计支持材料处理、组装或物流等应用，同时提供reach、负载能力和机动性的平衡。至于价格，根据其功能和自由度，这款机器人可能售价
英文: freedom, while its torso and head each have 3 degrees of freedom for even more maneuverability, while spatial perception aids in navigating its surroundings. On top of this, the omnidirectional base allows movement in any direction, enhancing its utility in environments such as manufacturing facilities, warehouses, or research spaces. Vega’s design supports applications like material handling, assembly, or logistics, while offering a balance of reach, payload capacity, and mobility. As for price, this robot will likely cost

(05:21)

中文: 在8万美元至13万美元之间。与此同时，数字生物学公司in silico刚刚推出了一家全自动实验室，现在集成了人形机器人以增强自动化和效率。事实上，这个实验室通过整合机器人来处理细胞培养、高通量筛选、下一代测序、细胞成像和基因组分析等任务，扩展了in silico的AI驱动的药物发现平台。这一举措简化了工作流程，同时提高了实验精度，并且该实验室
英文: between $80,000 to $130,000 based on its features and degrees of freedom. Meanwhile, digital biology company in silico just launched a fully automated laboratory, now integrating humanoid robots to enhance automation and efficiency. In fact, this lab expands in silico’s AI-driven drug discovery platform by incorporating robots to handle tasks such as cell culture, high-throughput screening, next-generation sequencing, cell imaging, and genomic analysis. With this move streamlining workflows while improving experimental precision, and the lab

(05:55)

中文: 建立在in silico的端到端系统之上，用人形机器人取代了自动引导系统，这些机器人最终将由公司的人工智能远程控制。这些机器人生成的数据直接输入panda omix平台，以完善目标假设和验证过程。总的来说，这种互联的设置将通过结合机器人自动化和AI驱动的洞察力来加速药物发现。最后，由中国一个小团队开发的通用AI代理在
英文: builds on in silico’s end-to-end system, replacing automated guided systems with humanoids that will eventually be remotely controlled by the company’s artificial intelligence. With these robots generating data that feeds directly into the panda omix platform to refine target hypotheses and validation processes. Altogether, this interconnected setup will accelerate drug discovery by combining robotic automation with AI-driven insights. And finally, a general-purpose AI agent developed by a small Chinese team is going viral after its

(06:28)

中文: 推出后迅速走红，但现在正面临对其能力和可用性的审查。由总部位于北京的小公司butterfly effect创建的Manus，能够处理复杂的任务。至于能力，Manus提供自主任务执行，如构建网站、规划旅行、创建仪表板、教育内容比较等。这是因为Manus集成了编码、网页浏览和数据分析等工具，以实现无缝执行和工作流程自动化，并通过移动设备进行实时进度跟踪以进一步
英文: launch but it’s now facing scrutiny over its capabilities and availability. Created by Butterfly Effect, a small company based in Beijing, Manus is capable of handling complex tasks. As for abilities, Manus offers autonomous task execution such as building websites, planning trips, creating dashboards, educational content comparisons, and more. This is because Manus integrates tools like coding, web browsing, and data analysis for seamless execution and workflow automation with real-time progress tracking via mobile to further

(07:02)

中文: 增强透明度。然而，其邀请制的推出和有限的服务器容量限制了访问，引发了对其声称的质疑，并邀请与中国deep-seek AI进行比较。针对访问限制，Manus AI的产品合伙人将其归因于服务器短缺，因为他们没有为公众的反应做好准备，当前版本仍处于早期阶段。launch的进一步复杂化是，该团队的X账户被暂停，批评者指出Manus依赖于
英文: enhance transparency. However, its invitation-only rollout and limited server capacity have restricted access, raising doubts about its claims and inviting comparisons to China’s deep-seek AI. Addressing the access constraints, a product partner at Manus AI attributed them to server shortages due to being unprepared for the public response, with the current version being in its early stages. And to add further complications to the launch, the team’s X account was suspended, and critics are noting Manus’ reliance on

(07:34)

中文: 未披露的大型语言模型，而不是像deep-seek这样的新基础技术，因为该系统目前被相当模糊地描述为一个使用多个模型的多代理框架，并计划开源其部分元素。随着团队努力扩展其基础设施并澄清其技术，Manus仍然是一个测试案例，展示了新兴AI参与者如何在日益竞争激烈的环境中平衡创新、可访问性和可信度。
英文: undisclosed large language models rather than new foundational technology like deep-seek, as the system is currently described quite vaguely as a multi-agent framework that uses multiple models with plans to open-source some of its elements. As the team works to scale its infrastructure and clarify its tech, Manus remains a test case for how emerging AI players navigate the balance between innovation, accessibility, and credibility in an increasingly competitive landscape.