由Google Cloud赞助
选择您的第一个基因AI 使用案例
要从基因AI开始，首先关注能够改善人类信息体验的领域。
该系统说明在AI能够可靠地自行处理复杂的物理任务之前，远程人控如何能帮助机器人开发者部署机器。
4 读取
本田镇
国际汽车公司已经演示出一个新的机器人系统，使用户能够远程地执行复杂的任务，同时使用可调速的机器人手和AI辅助控制。
本月早些时候，该公司在日本的CEATEC 2026会议上展示了本田Avatar机器人，其潜在应用包括设备维护和医疗支持以及救灾和空间探索等。
机器人被设计成能紧密地模仿人类的德克力，其两只多指手相间各有16度的自由度。 本田的手技术还具有16个活性关节，最高为每秒180度。
在之前的演示中，系统执行复杂的任务，包括用针线接线，收紧小螺丝和操纵螺栓。
该设计旨在解决机器人领域持续存在的操纵问题。 虽然机器越来越聪明，更有能力导航复杂的环境，但执行复杂任务或处理脆弱物体的能力仍然是部署的挑战。
远程操作使人类能够远程控制或指挥机器人，随着自主能力的发展，它提供了弥补这一差距的途径。
George Chowdhury是ABI Research的一位分析家，称远程操作对于人造物的商业可行性是“关键”的。
“由于目前机器人的AI模型的脆弱性，当今的物理AI市场严重依赖远程操作。” “当自主系统失效时，决策者需要保证生产线不会长时间下降。”
根据Gartner的分析师Abdil Tunca的说法，本田的方法也可以在完全自治费用太高或技术上不切实际的情况下开放申请。 他指出，本田的做法并不是全新的。
“这是一种不寻常的做法，但不是第一个。 然而，本田的解决方案具有更多的商业潜力，"他称。
潜在的早期应用包括航空航天制造，电子装配和制药实验室，其中复杂的操纵任务可能需要人类的判断，但不一定需要运营商实际在场。
然而，Tunca 称并不期望Avatar机器人在大量制造或仓库业务中取代自主系统，因为重复的、可预测的任务更适合自动化。
“我的立场是，机器人的价值应该用商业结果来衡量，而不只是自主水平。”
本田并非独自依靠远程人类输入来使机器人更加实用。
远程操作已经用于仓库自动化和人造机器人，可以帮助系统从出乎意料的情况中复苏，并支持用于提高自主性能的数据收集。
乔杜里指出"加一机器人","一X"和"敏捷机器人"为例。
He 称 1X计划今年与机器人一起在家中生活，他承认自主的AI系统还没有准备好适应这些环境
称先生说，“大多数人造人销售商利用远程操作获取机器人基础模型的培训数据，并确保溶液正常运行”。 与此同时，机动机器人使用乔杜里称为“骑马和骑手”的系统，在这种系统中，当机器人遇到例外时，人类就指挥机器人，而不一定直接控制它的手臂和手。
对通卡来说，这种向人类监督的转变，而不是持续的人工控制，对于使电信业务在商业上可行至关重要。
“如果本田能优先考虑AI处理日常运动的共享控制架构，而人类只有在例外情况下才能干预，“这一突破将从一个控制一个机器人的操作员转变为一个控制整个车队的操作员。”
然而，将工作外包给远程运营商的概念，融资对工人流离失所和安全以及遥控机器人造成伤害时应如何分配责任感到关切，Chowdhury 称。
他称先生说，“工资仲裁的概念与电信业务并存”。 “说我们无法让人工智能工作，不得不依靠远程操作，但延迟程度很低，人形功能正常。 例如，在墨西哥的工人有可能在美国操作机器人 — — 公司可以利用较低的劳动力成本。”
远程运行还受到推出技术限制。 命令和传感器数据在操作员和机器人之间行走所需时间会影响其准确的反应，特别是在操纵微妙物体或在错误可能造成破坏的环境中工作时。
尽管存在这些挑战，电信业务可以帮助弥合实际AI系统的进展和商业部署之间的差距。
Tunca预测，到2028年，只有不到20个公司将人造机器人规模化生产，用于制造和供应链应用，这再次突出了从示范转向大规模收养的挑战。
称先生说，“这更加需要区分技术进步与商业准备状态。” “我认为，电信业务在弥合这一差距方面可能起关键作用，尤其是作为通过人类示威培训体能AI模型的机制。”
对本田来说，较长期的机会可能超越了机器人本身。 正如通卡所言，公司可以从操控数据，AI能力和人对机器人技能的转移中创造出比从Avatar机器人硬件本身更多的长期价值。
本文由 AI宝库自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
Sponsored by Google Cloud
Choosing Your First Generative AI Use Cases
To get started with generative AI, first focus on areas that can improve human experiences with information.
The system illustrates how remote human control could help robotics developers deploy machines before AI can reliably handle complex physical tasks on its own.
4 Min Read
Honda
The international automobile company has demonstrated a new robotic system that enables users to perform complex tasks remotely, using a combination of dexterous robotic hands and AI-assisted controls.
The company showed off its Honda Avatar Robot earlier this month at the CEATEC 2026 conference in Japan, with potential applications ranging from equipment maintenance and medical support to disaster response and space exploration.
The robot is designed to closely imitate human dexterity, with 16 degrees of freedom in each of its two multi-fingered hands. Honda’s hand technology also features 16 actuated joints, up to 180 degrees per second.
In previous demonstrations, the system performed intricate tasks including threading a needle, tightening a small screw and manipulating a bolt.
The design aims to address a persistent manipulation problem in robotics. While machines are becoming smarter and more capable of navigating complex environments, the ability to perform intricate tasks or handle fragile objects remains a challenge for deployment.
Teleoperation, which enables a human to control or direct a robot remotely, offers a way to bridge that gap as autonomous capabilities develop.
George Chowdhury, an analyst at ABI Research, said teleoperation is “critical” for the commercial viability of humanoids.
“The physical AI market today depends heavily on teleoperation due to the fragility of current AI models for robotics,” he said. “When autonomous systems fail, decision makers need the reassurance that a production line will not be down for an extended period.”
And Honda’s approach could also open up applications where full autonomy is either too expensive or technically impractical, according to Abdil Tunca, an analyst at Gartner. Honda’s approach is not entirely new, he noted.
“This is an unusual approach but not the first. However, Honda’s solution has more commercial potential,” he said.
Potential early applications include aerospace manufacturing, electronics assembly and pharmaceutical laboratories, where complex manipulation tasks may require human judgment but do not necessarily require an operator to be physically present.
However, Tunca said he doesn’t expect avatar robots to displace autonomous systems in high-volume manufacturing or warehouse operations, where repetitive, predictable tasks are more suited to automation.
“My position is that the value of robotics should be measured by business outcomes, not autonomy levels alone,” he said.
Honda is not alone in relying on remote human input to make robots more practical.
Teleoperation is already used across warehouse automation and humanoid robotics, where it can help systems recover from unexpected situations and support the collection of data used to improve autonomous performance.
Chowdhury pointed to Plus One Robotics, 1X and Agility Robotics as examples.
He said 1X, which plans to go live in the home setting with a robot this year, acknowledges that autonomous AI systems aren’t yet ready for these environments and intends to heavily teleoperate, capturing data as it goes, which the company will use to train foundation models.
“Most humanoid vendors use teleoperation to capture training data for robotics foundation models and to ensure solution uptime,” he said. Agility Robotics, meanwhile, uses a system Chowdhury described as a “horse and rider” approach, in which a human directs the robot when it encounters an exception without necessarily taking direct control of its arms and hands.
For Tunca, this shift toward human supervision rather than continuous manual control could prove crucial to making teleoperation commercially viable.
“If Honda can prioritize shared control architectures where AI handles routine movements, and humans intervene only during exceptions, economics can change significantly,” he said. “The breakthrough would be moving from one operator controlling one robot to an operator controlling an entire fleet.”
The concept of outsourcing work to remote operators, however, raises concerns about worker displacement and safety, as well as about how responsibility should be allocated when a remotely operated robot causes harm, Chowdhury said.
“The concept of wage arbitration exists with teleoperation,” he said. “Say we can't get the AI to work and have to rely on teleoperation, but the latency is minimal and the humanoid is functional. It's possible that, for example, workers in Mexico could teleoperate robots in the U.S. – companies could take advantage of the lower labor costs.”
Remote operation also introduces technical constraints. The time it takes for commands and sensor data to travel between an operator and a robot can affect how precisely it responds, particularly when manipulating delicate objects or working in environments where errors could cause damage.
Despite these challenges, teleoperation could help bridge the gap between advances in physical AI and commercial deployment.
Tunca predicted that fewer than 20 companies will scale humanoid robots into production for manufacturing and supply chain applications by 2028, again underscoring the challenges of moving beyond demonstrations to large-scale adoption.
“This reinforces the need to distinguish technological progress from commercial readiness,” he said. “I believe teleoperation may play a critical role in bridging that gap, especially as a mechanism for training physical AI models through human demonstrations.”
For Honda, the longer-term opportunity may extend beyond the robot itself. As Tunca put it, the company could create more long-term value from the manipulation data, AI capabilities and human-to-robot skill transfer than from the avatar robot hardware itself.
Auto-fetched and translated by AI Gem Box. All rights belong to the original author.