Kim Nam Gyun 是韩国科学工艺院(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,因而可以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。
自穿平台让两种情形都不再需要额外的人手或复杂的训练。
除了专业情形,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,整个平台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合适与人体共存,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,取出笔者不慎吞食的电池等异物,还预计用于消化平台的手术,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,这套逻辑不只合适搜救废墟,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,当下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,让氧气得以借助。
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%。
视频喉镜的浮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,再手动引导管子借助弯曲的解剖路径,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,外观了一套软体机器人插管平台 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。

测试在尸体上开展,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. 来推进这项工艺的商业化,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,机器人工艺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能力,控制平台因而可以简单一些。
本文来源:https://www.ifanr.com/1686368
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。
Kim Nam Gyun is a postdoctoral researcher at the Korea Advanced Institute of Science and Technology (KAIST). Once, while riding a bicycle, he was suddenly hit by heavy rain; he wanted to put on a raincoat, but he couldn’t get it on with one hand, and he had no free hand to pull the zipper.
He thought that if a raincoat could put itself on automatically, it would be much more convenient.
This sounded like a small idea, but his research team turned it into a serious paper.
The team embedded soft, air-pressure-driven “vines” into the garment. Once pressurized, the vines pull the fabric upward and inward against the wearer’s body—like ivy climbing a trellis—and the process can be completed even if the person keeps moving.
The whole process takes about 10 seconds, and a complete protective suit is put on.
Professor Ryu Jee-Hwan explained: this type of robot advances by growing at its tip, so it can move stably on curved surfaces. It can pass through narrow gaps and conform to the shape of its surroundings, continuing forward whether the surface is slippery, sticky, or inclined.
As for applications, the researchers mentioned cleanrooms in chip factories and emergency services.
Semiconductor cleanrooms have strict procedures for changing; workers must replace a full set of dust-proof garments every time they enter or exit. The steps are cumbersome, and improper wearing can pose a contamination risk.
Emergency responders need to put on protective gear quickly when handling chemical incidents or infectious-disease scenes, and putting on such equipment usually requires assistance from a partner.
The self-dressing platform eliminates the need for extra manpower or complex training in both scenarios.
Beyond professional settings, the team also mentioned older adults and people with limited mobility. Reduced hand strength and restricted shoulder movement are the daily reality for many elderly people when getting dressed.
In an interview, Ryu Jee-Hwan said: while everyone is talking about AI, mechanical engineering can also bring change to the world.
Vine robots
In 2017, researchers in Stanford University’s Department of Mechanical Engineering brought the logic of vine growth into the lab. They made a sealed-at-both-ends soft tube, like an inside-out sock, folded inside itself. When air pressure is pumped in from a fixed end, the tube’s tip begins to turn outward—like turning a sock inside out—but the inversion keeps going.
The prototype used thin plastic film, a low-cost material; the air source was ordinary compressed air; and the platform had no motors or sensor arrays.
This robot can briefly sprint at up to about 35 kilometers per hour, pass through gaps narrower than its own diameter, lift objects weighing 100 kilograms, and autonomously coil into an antenna shape in confined spaces.
When traditional robots move, every joint undergoes displacement and every segment bears resistance from the external environment; if clamped, it gets stuck. The “body” of a vine robot remains stationary, with only the tip extending forward as new material continuously feeds from the interior to the front end.
Irregular cavities, curved passages, and terrain with surface obstacles are almost unsolvable for traditional robots; for a vine robot, they simply require adjusting how air pressure is distributed.
Intubation shortens the time—and the distance of experience.
When hard robots contact the human body, force transmission is often rigid, concentrating pressure at the contact point and making injury more likely.
The inherent flexibility of soft robots makes them naturally suited to coexist with the human body. A magnetic slime robot developed by Harbin Institute of Technology and The Chinese University of Hong Kong can enter the body to remove foreign objects such as batteries accidentally swallowed, and is also expected to be used in surgery on the digestive platform or to deliver drugs to specific parts of the body.
The vine robot’s method of propulsion eliminates the action of “pushing,” replacing it with “growing.”
The researchers soon realized that this logic is not only suitable for searching and rescuing in rubble; it can also go into other places.
The human body contains numerous winding pathways, including the airway, digestive tract, and blood vessels; however, current medical devices rely heavily on force, rigidity, and the operator's tactile sensation when navigating these structures.
The human airway evolved to block food and foreign objects; emergency intubation involves rapidly and accurately inserting a breathing tube through this pathway to facilitate oxygen delivery.
According to statistics from a research team at UCSB, the first-attempt failure rate for tracheal intubation at out-of-hospital emergency scenes is nearly 35%.
The emergence of video laryngoscopes has improved this situation, yet it still demands a high level of operator proficiency; the procedure requires visualizing the airway opening and manually guiding the tube through curved anatomical paths—a process heavily reliant on experience and skill.
Elliot Hawkes's team developed the SRIS (soft robotic intubation system) based on the fundamental principles of vine-like robots; it no longer requires operators to visually identify the airway before advancing the tube.
During the process, only the tip advances as the main body remains stationary, generating no friction and avoiding scratching the airway walls; it mimics a vine finding a suitable crevice and quietly threading through it. This "frictionless" characteristic allows it to autonomously navigate winding anatomical structures rather than being forcefully pushed through.
Tests were conducted on cadavers with 8 emergency responders—4 emergency medical technicians and 4 paramedics—who received only 5 minutes of training; 87% of them successfully completed intubation on the first attempt with an average time of 21 seconds.
Under identical conditions using video laryngoscopes, the success rate was 63% with an average time of 44 seconds.
The study has been published in *Science Translational Medicine*; lead author David Haggerty, who completed his doctoral research in Hawkes's lab, founded Vine Medical Inc. in 2022 to advance the commercialization of this technology, with the next step being clinical trials to verify device safety.
Haggerty speaks of "democratizing airway management," which means transforming a procedure previously reserved for senior specialists into a skill that can be mastered by general emergency responders after brief training.
Today, progress in robotic engineering is not always defined by more sensors, larger data sets, or more complex models; sometimes, key improvements stem from a single vine. Since the mechanical structure assumes part of the capabilities of sensing, steering, and shock absorption, the control platform can consequently be simplified.