这机器人从嘴里钻进人体，居然是来救命的
Kim Nam Gyun 是韩国科学技术院（KAIST）的博士后研究员，一次骑车遭遇突降大雨，他想穿雨衣，可单手穿不上，腾不出手来拉拉链。
他想，如果一件雨衣可以自动穿好，那就方便多了。
这听起来像个小念头，但他的研究团队把它变成了一篇严肃的论文。
研究团队把由气压驱动的柔软「藤蔓」嵌进服装里，加压之后，藤蔓会带着织物贴近穿着者的身体向上爬，就像常春藤攀附在支架上，即使人不保持静止也能完成。
整个过程大约 10 秒，一套完整的防护服就穿好了。
Ryu Jee-Hwan 教授解释道：这类机器人靠在尖端生长前进，因此能够在弯曲的表面上稳定移动。它可以穿过狭窄的缝隙，顺着周围环境的形状生长，不论表面是滑的、黏的还是倾斜的，都能继续前进。
至于用在哪里，研究者提到了芯片工厂的洁净室和应急服务。
半导体洁净室对更衣有严格的程序要求，工人每次进出需要更换一整套防尘服，步骤繁琐，而且穿戴不当就会带来污染风险。
急救人员在处理化学事故或传染病现场时需要迅速穿上防护装备，而穿戴这类装备通常需要同伴协助。
自穿系统让两种场景都不再需要额外的人手或复杂的训练。
除了专业场景，研究团队也提到了老年人和行动不便者。双手的力量减退、肩关节活动受限，是许多老年人日常穿衣的真实困境。
Ryu Jee-Hwan 在接受采访时说：当所有人都在讨论 AI 时，机械工程同样可以给世界带来变化。
藤蔓机器人
斯坦福大学机械工程系的研究人员在 2017 年就把藤蔓生长的逻辑搬进了实验室。他们做了一根两端密封的软管，像一只翻过来的袜子，折叠在自身内部。当气压从固定的一端泵入，管子的顶端开始向外翻转，就像把袜子从内往外翻，但这个翻转过程不停。
原型机用的是薄塑料膜，材料便宜，气源是普通的压缩空气，整个系统没有电机和传感器阵列。
这台机器人能以最高约 35 公里每小时的速度短暂冲刺，能在比自身直径更窄的缝隙里穿过，能举起 100 公斤重的物体，还能在狭小空间里自主盘绕成天线形状。
传统机器人移动时，每一节关节都在位移，每一段身体都在承受外部环境的阻力，被夹住就卡死了。藤蔓机器的「身体」保持静止，只有尖端在向前扩展，新材料从内部不断补充到前端。
不规则的空腔、弯曲的通道、表面布满障碍的地形，对传统机器人来说几乎无解，对藤蔓机器人来说不过是改变一下气压分配而已。
插管，缩短的是时间，也是经验的距离
硬质机器人与人体接触时，力量的传递方式往往是刚性的，接触点上的压力集中，容易造成损伤。
软体机器人本身的柔性让它天然适合与人体共存，哈工大和港中大研究出的磁性粘液机器人可以进入人体内，取出我们不慎吞食的电池等异物，还预计用于消化系统的手术，或者将药物输送到人体的特定部位。
藤蔓机器人的推进方式消除了「推」这个动作，取而代之的是「长」。
研究人员很快意识到，这套逻辑不只适合搜救废墟，它还可以走进别的地方。
气道、消化道、血管，人体内部有大量弯曲的管道，目前的医疗器械在穿越这些结构时大多依赖推力，依赖刚性，依赖操作者的手感。
人体的气道在进化上是为了阻挡食物和异物的。急救插管就是要在这段路里快速、准确地送入一根呼吸管，让氧气得以通过。
据 UCSB 的研究团队统计，在院外的急救现场，气管插管的首次失败率接近 35%。
视频喉镜的出现已经改善了这个处境，但它对操作者的要求仍然不低。操作者需要先目视气道入口，再手动引导管子通过弯曲的解剖路径，这是一套强烈依赖经验和技巧的操作。
Elliot Hawkes 的团队从藤蔓机器人的基础原理出发，设计了一套软体机器人插管系统 SRIS（soft robotic intubation system）。它不再要求操作者先看清气道，再推送管子。
过程中，只有尖端在推进，身体不移动，不产生摩擦，不刮伤气道壁，就像一棵小藤蔓找到了合适的缝隙，然后安静地钻过去。正是这种「无摩擦」的特性，让它能在曲折的解剖结构里自主导航，而不被推着硬走。
测试在尸体上进行，参与者是 8 名急救人员，4 名急救医护技师，4 名辅助急救员，接受的培训时间是 5 分钟。结果，87% 的人在第一次尝试时成功完成了插管，平均耗时 21 秒。
同样条件下使用视频喉镜，成功率是 63%，平均耗时 44 秒。
这项研究已发表在《Science Translational Medicine》上，主要作者 David Haggerty 曾在 Hawkes 实验室完成博士研究，他在 2022 年创办了 Vine Medical Inc. 来推进这项技术的商业化，下一步是启动临床试验，验证设备的安全性。
Haggerty 说要「让气道管理民主化」，也就是让这项本来只属于高级专科医生的操作，变成经过短暂培训的普通急救人员也能完成的技能。
今天，机器人技术的进步未必总表现为更多传感器、更大数据和更复杂模型。有时，关键改进来自一株藤蔓。机械结构承担了感知、导向和缓冲的一部分功能，控制系统因此可以简单一些。
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
This robot is trying to save your life.
Kim Nam Gyun, a post-doctoral researcher at the Korea Institute of Science and Technology (KAIST), was hit by a heavy rain on a bike. He wanted to wear a raincoat, but he couldn't wear it with one hand, so he couldn't pull it.
He thought it would be easier if a raincoat could be worn automatically.
It sounds like a little idea, but his research team turned it into a serious paper.
The research team embeds the soft, air-driven vines into the garments, which, when pressurized, would climb up with fabrics close to the body of the dresser, just like Ivy climbing on the stairwell, even if people were not static.
The whole process was about 10 seconds, and a full suit was ready.
Professor Ryu Jee-Hwan explained that these robots are growing at the tip of the curve and are thus able to move steadily on the curved surface. It can go through narrow cracks and grow along the shape of the surrounding environment, whether the surface is slippery, sticky or tilted.
As to where to use it, researchers referred to the cleaning room and emergency services at the chip factory.
The semiconductor clean room has strict procedural requirements for changing clothes, and workers are required to replace a full package of dustproof clothing each time they enter or leave, the steps are cumbersome and inappropriately dressed, posing pollution risks.
First aid personnel need to quickly wear protective equipment when dealing with chemical accidents or scenes of infectious diseases, which usually require peer assistance.
The self-trading system no longer requires additional manpower or complex training for both scenarios.
In addition to the professional scene, the research team also referred to older persons and persons with reduced mobility. The erosion of the power of both hands and the restriction of shoulder joint activities are the real dilemma of many older persons in their daily clothing.
Ryu Jee-Hwan, in an interview, said that mechanical engineering could also change the world when everyone was discussing AI.
A vine robot.
In 2017, researchers from the Department of Mechanical Engineering at Stanford University brought the logic of growing vines into the lab. They made a two-end-sealed hose, like a flipping sock, folded inside themselves. When the pressure is pumped from a fixed end, the top of the pipe starts to flip out, as if the socks were flipping out from the inside, but the flipping continues.
The prototype is based on thin plastic membrane, low-cost materials, a general compressed air source, and no electrical or sensor arrays throughout the system.
The robot can sprint for a short time at a maximum speed of about 35 kilometres per hour, can pass through a narrower gap than its own diameter, can lift up 100 kg of objects and can orbit the shape of the antenna in a narrow space.
When traditional robots move, each joint moves, and each body withstands resistance from the outside environment and is stuck to death. The "body" of the vine machine remained static, with only the tipping edge extending forward and new material being added from the inside to the front end.
The irregular cavity, the curved passage, the surface of a barrier-filled terrain are almost incomprehensible for traditional robots, and simply alter the distribution of air pressure for vines.
Intubation, shorter time, distance from experience.
When rigid robots come into contact with the human body, power is often transmitted in a rigid manner, with concentrated pressure at the point of contact and prone to damage.
The softness of the flexible robot itself makes it natural to coexist with the human body, and the magnetic mucous robots that have been studied in Hajia and Hong Kong can enter the human body, extract foreign objects such as batteries that we accidentally swallow, and are expected to be used in the digestive system or to deliver drugs to specific parts of the human body.
The way in which the vine robots move was to eliminate the "push" and replace it with "long".
Researchers quickly realized that this logic is not only appropriate to search for and rescue the ruins, but that it can also go elsewhere.
Aerobics, digestive tracts, blood vessels, and a large number of curved pipes inside the human body, most of the current medical devices travel through these structures by force, rigidity and the hand of the operator.
The airways of the human body evolve to stop food and foreign matter. The first aid intubation is intended to be delivered quickly and accurately into a respiratory tube in this road to allow oxygen to pass.
According to UCSB ' s research team, the rate of initial failure of the tube intubation was close to 35 per cent at first aid sites outside the hospital.
The appearance of video larynx has improved this situation, but its requirements for operators remain high. The operator needs to look at the airway entrance and then manually direct the pipe through a curved anatomy path, an operation that relies heavily on experience and skill.
Elliot Hawkes' team designed a software robot intubation system SRIS (soft robotic institution system) based on the basic principles of the vine robot. It no longer requires operators to see the airways before pushing the pipes.
In the process, only the tipping edges are advancing, the body is not moving, there is no friction, there is no scratching of the airway walls, like a little vine that finds the right crack and then drills in silence. It's this "no friction" characteristic that allows it to navigate in a twisted anatomy structure without being pushed.
The tests were carried out on the body and were attended by eight first aid workers, four emergency medical technicians, and four auxiliary first aid workers, who were trained for five minutes. As a result, 87% of the people successfully completed the intubation at the first attempt, average latency 21 seconds.
Using video larynx under the same conditions, success rate is 63%, average latency 44 seconds.
The study was published in Science Transport Media, where David Haggerty, the lead author, completed a doctoral study at Hawkes Laboratories, founded Vine Medical Inc. in 2022 to advance the commercialization of the technology, and next step initiated clinical trials to verify the safety of the equipment.
Haggerty says that "democratization of air-channel management" means making it a skill that would have been the sole function of a senior specialist and that could have been performed by a general emergency worker with short training.
Today, advances in robotic technology do not always take the form of more sensors, more data and more complex models. Sometimes, the key improvement comes from a vine. The mechanical structure provides a partial function of perception, direction and buffer, so the control system can be simpler.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.