当前位置:主页 > 国内 >

金陵十三钗

WAIC观察:华为下场具身智能,能再成就一个“赛力斯”?_我的网站

wps

一 |     

    
Photos: Courtesy of Landspace
    Photos: Courtesy of Landspace
    

A rocket that once would have been discarded after completing its mission has now taken a different path - returning from the edge of space back to Earth. In the early hours on Wednesday, the Zhuque-3 Y2 reusable launch vehicle lifted off for a flight test, carrying not only a satellite payload but also a major test of China's emerging reusable rocket capabilities. After completing its orbital mission, the rocket's first stage attempted a controlled return and recovery, with its developers Beijing-based private firm LandSpace describing the mission as a key transition from technology verification toward engineering application for China's commercial aerospace industry.
About 137 seconds after liftoff, the rocket’s first and second stages separated. The second stage continued its flight and successfully placed the Honghu-03 satellite, independently developed by Hongqing Technology, into its designated orbit. And at around 7:43 am, the rocket’s first stage successfully made a soft landing as planned at the Zhuque-3 landing site in Minqin county, Northwest China’s Gansu Province, marking the successful completion of the flight test mission.
This mission marked China's first-ever recovery of a launch vehicle's first stage using landing legs, as well as the country's first land-based recovery of the first stage of an orbital-class launch vehicle, the developer highlighted.
The achievement came shortly after China's Long March-10B reusable rocket successfully completed its maiden flight and a sea-based recovery test on July 10, marking a wave of breakthroughs in reusable launch technology. The parallel progress of state-owned and private aerospace players through different technical approaches is creating a dual-track development landscape for China's future low-cost and high-frequency space transportation, some aerospace experts told the Global Times.
    
Photos: Courtesy of Landspace
    Photos: Courtesy of LandspaceDesigned to fly again

Unlike traditional expendable rockets, which are discarded after a single mission, reusable launch vehicles are designed around an entirely different concept: a rocket should not only complete a launch, but also return safely and be prepared for future missions.
According to LandSpace, for conventional expendable liquid launch vehicles, the rocket itself represents the majority of launch costs, while fuel expenses generally account for only 1-3 percent. The first stage typically makes up about 70 percent of the total rocket cost. As a result, repeated reuse of rockets - particularly the first stage - can significantly lower launch costs by amortizing the cost of the vehicle over multiple missions. For partially reusable rockets that reuse the first stage while discarding the second stage, the cost advantage becomes increasingly evident as the number of reuse cycles grows.
The Zhuque-3 is a reusable liquid oxygen-methane launch vehicle independently developed by Beijing-based private aerospace firm LandSpace for future large satellite constellation deployment missions, according to the company's statement.
The rocket adopts a two-stage configuration, with a diameter of 4.5 meters, a total length of 66.1 meters and a liftoff mass of about 570 tons. Its first stage is equipped with nine Tianque-12A liquid oxygen-methane engines, as well as recovery systems including grid fins, attitude control systems and landing legs, enabling autonomous return and soft landing after orbital missions, LandSpace said.  
The Y2 mission was conducted on the basis of experience accumulated from the previous flight test, with improvements focused on recovery reliability, flight control and engine performance.
LandSpace said it optimized the first-stage landing propulsion scheme by reducing the number of engines used during landing ignition, simplifying system design and improving reliability during the recovery phase. The company also added an autonomous safety control function capable of predicting landing points based on flight conditions and making corresponding safety adjustments.  
The mission also further verified key technologies including the use of stainless-steel rocket structures, liquid oxygen-methane propulsion and landing-leg recovery systems.
According to LandSpace, stainless steel provides advantages in manufacturing cost, processing efficiency, high- and low-temperature performance and post-recovery maintenance, making it suitable for reusable rocket development.  

Photos: Courtesy of Landspace
    Photos: Courtesy of LandspaceMore than a single rocket's return

The significance of Zhuque-3 Y2 extends beyond one flight test, as it reflects China's broader effort to develop reusable launch systems through multiple technological routes, according to Chinese experts.
Kang Guohua, a senior member of the Chinese Society of Astronautics and a professor of aerospace engineering at Nanjing University of Aeronautics and Astronautics, told the Global Times that China has become the second country after the US to develop orbital-class rocket recovery capabilities.
The US currently leads reusable rocket development, with SpaceX and Blue Origin advancing different recovery technologies. China, meanwhile, has seen both state-owned aerospace programs and private companies explore different technical pathways, creating a diversified development pattern, Kang said.  
According to Kang, the parallel development of different approaches demonstrates the advantages of combining national strategic capabilities with market-driven innovation.  
Pang Zhihao, a prominent Chinese space exploration expert and retired researcher from the China Academy of Spacecraft Technology, told the Global Times that the successful recovery of Zhuque-3 would carry special significance for China's private aerospace sector.
A landing-leg recovery approach would mean that private aerospace companies have completed the full process of developing a large-capacity reusable rocket system, validating key technologies including stainless-steel structures, multi-engine propulsion and recovery systems, Pang said.  
More to expect 
LandSpace said the first stage represents a significant share of the total cost of a liquid launch vehicle. Through repeated reuse, the cost of each launch can be gradually reduced, while production cycles can be shortened and launch frequency increased.  
Such capabilities are becoming increasingly important as countries and companies accelerate the deployment of large low Earth orbit satellite constellations, it added.
The Zhuque-3 has demonstrated the ability to support various commercial missions, including satellite constellation deployment, single satellite launches and rideshare missions, according to LandSpace.
The company said the rocket has achieved a maximum orbital payload capacity of 14.2 tons in expendable mode, while its reusable configuration is designed to support future constellation deployment needs.  
The Y2 mission also carried the Honghu-03 satellite developed by Hongqing Technology, combining reusable technology verification with a commercial launch application scenario. LandSpace said the mission verified the full chain of "design, manufacturing, testing, launch and recovery," laying the foundation for future commercial operations.  
The Zhuque-3 Y2 test comes as China's aerospace industry accelerates efforts to develop reusable launch vehicles.
According to Chinese experts, multiple companies are now exploring different recovery technologies, including vertical landing and other innovative approaches, reflecting the rapid growth of China's commercial aerospace ecosystem.  
Reusable liquid launch vehicle, Pallas-1 will begin its mission from Galactic Energy's independently built facility at the Dongfeng Commercial Aerospace Innovation Test Zone in northwestern China, where the company will soon complete testing, propellant filling and launch preparation operations, according to its developer Galactic Energy on Thursday.
Orienspace's Gravity-2 reusable large liquid launch vehicle is undergoing final ground testing ahead of its maiden flight, which is scheduled in the fourth quarter of 2026. The rocket is designed with a low Earth orbit payload capacity of 21.5 tons.
Space Pioneer is advancing verification and joint testing of its Yuanxingzhe-1 (YXZ-1) rocket, while i-Space's Hyperbola-3 reusable rocket has completed full-cycle ground verification of offshore recovery and landing technologies and is expected to conduct its maiden flight around the end of 2026.
Meanwhile, CAS Space's reusable Lihong-2 vehicle is being developed as a platform for verifying key recovery and reuse technologies, as well as conducting microgravity experiments. The vehicle is scheduled to complete its maiden flight and 100-kilometer-class recovery verification test in 2026.
Pang said reusable rocket technology could help lower launch costs, increase mission frequency and provide stronger support for emerging sectors such as satellite internet, space-based information networks and future large-scale space applications.  
For China's private space sector, the Zhuque-3 Y2 mission represents a transition from technology verification toward engineering application, according to LandSpace. The company said it aims to further build a "launch-recovery-inspection-reuse" cycle to support future high-frequency and low-cost commercial operations. 
Experts said reusable rocket technology could help reduce launch costs, increase launch frequency and provide stronger support for emerging sectors such as satellite internet and large-scale space applications.  
The future challenge for China's reusable rocket developers will be turning successful flight tests into repeatable commercial operations, as the country advances toward a new era of high-frequency space transportation, they noted.
Photos: Courtesy of Landspace
    Photos: Courtesy of Landspace

。    7月20日,在WAIC 2026世界人工智能大会上,华为首次完整展示全系列计算模组产品,包括网卡、SSD、DPU、RAID卡四大系列产品,如国内首发的800GE AI网卡,支持10万级节点超大规模组网,以及KVU方案等。         显然,华为在AI时代试图构建底层能力。其实在具身智能赛道,华为的打法同样如此,近期,华为云CloudRobo全流程具身智能开发平台正式开启公测,这也预示着具身智能产业的分工格局更加显化。

二 | 而华为再一次站在热门赛道的边界上,它其实不直接下场造机器人本体,不做终端品牌,把目标对准所有机器人的智能核心,试图整个行业的开发底座。                   熟悉华为的人可能已经意识到,此类策略打法与多年前华为切入智能汽车的路径如出一辙:不亲自下场造车,却靠技术赋能捧红问界,推动赛力斯完成逆袭。         其实不只是华为,纵观整个行业,今年以来,华为、腾讯、百度、阿里相继发布了各自的具身智能平台产品,而且无一例外都选择了“不做本体、做大脑”的路线,不亲自下场造机器人,而是为机器人厂商提供智能系统、开发工具和云服务底座。         只是在硬件形态更分散、场景更碎片化的机器人赛道,科技巨头们还能复制此前在其他赛道的神话吗?也许很难了。         01.          云厂商下场抢做机器人大脑供应商          绕过身体,直奔大脑,是近年来某些科技公司押注具身智能的玩法。         其中,华为在具身智能赛道的角色定位其实也很明确,它目前没有推出自造自产的机器人领域的终端品牌,它更像是打算做机器人大脑的核心供应商,用云端平台把全行业的开发效率抬升到新的层级。         再例如腾讯,腾讯Robotics X实验室首次以完整技术矩阵参展,发布了Tairos具身智能开放平台,开源了混元具身大模型HY-Embodied系列,还展示了机器人本体互连技术RoboFusion。         大厂们鲜有直接做硬件产品,这是为什么?          站在行业视角,过去几年人形机器人、工业机械臂、四足机器人等产品密集亮相,可多数团队的开发模式仍停留在手工作坊阶段。         不少具身智能企业都要从零搭建数据体系、训练专属模型、反复调试仿真、适配不同硬件本体,大量重复劳动消耗了核心研发资源,真正能投入到场景创新的精力被严重挤压。

三 | 行业走到从样机演示向规模化落地跨越的关键节点,本质上依然卡在底层开发效率的瓶颈上。

四 |                    此前人形机器人普遍定位几十万元级。价格下探的核心推手,是供应链成熟与规模化生产落地,关节模组、电机、减速器等核心硬件成本持续走低,可一旦硬件“身体”不再是行业稀缺资源,智能“大脑”的价值便愈发凸显。以仙工智能的招股书为例,其控制器业务毛利率高达79.8%,软件业务更是达到89.3%,而机器人整机的毛利率只有38.4%,配件业务仅15.7%。所以这也是一些资本追捧大脑公司的核心原因,涉及控制的大脑业务是个香饽饽。         按照这个逻辑来理解,华为云推出的CloudRobo,理论上也在瞄准这个方向,根据华为云官方在INSPIRE创想者大会上发布的信息,这是全球首个覆盖数据生产、模型开发、仿真评测、本体接入、部署运维全链路的具身智能开发平台,核心目标就是把具身智能开发从手工作坊带进工业化时代。

五 |                    优艾智合、华沿机器人等20多家企业已经宣布首批入驻CloudRobo平台,华为还同步启动了“百模千态,云聚共赢”的生态合作计划。         具身智能发展到今天,的确需要通用开发底座来降低行业门槛,让更多团队不用在底层技术上反复消耗,华为云试图成为机器人开发者的公共工具集,体现的是其一以贯之的战略逻辑。         02.          华为当年不造汽车,现在也没直接造机器人          坚持不碰硬件本体,不是华为对机器人赛道缺乏野心,恰恰是它把汽车赛道验证过的成功逻辑,平移到了全新的产业领域。         具身智能赛道看似火热,实际落地的障碍远比外界感知的更多。

六 | 通常情况下,产业链内外都认为数据稀缺是第一道关卡,真实物理世界的交互数据采集成本高、周期长,多数团队都在小体量数据中反复训练,直接导致模型泛化能力不足,更换场景就容易失灵。

七 |          仿真与现实的偏差是另外的难题。虚拟环境中运行顺畅的策略,放到真实环境里往往频繁出错,真机调试又要消耗大量时间与硬件成本。再加上本体形态千差万别、接入标准不统一、端侧算力跟不上模型需求,这些问题叠加在一起,让全栈自研的投入产出比变得极低。         面对上述几个行业症结,CloudRobo平台底层依托昇腾AI芯片与盘古具身大模型,搭建起PB级数据底座、云原生模型生产引擎、全国产Real-Sim仿真评测体系三大核心能力,开发者接入后无需从零搭建基础系统,可以直接调用百万级开箱即用的数据资产与二十余个昇腾亲和模型,把精力集中在场景应用与功能创新上。其效率目标是机器人小时级上云、模型分钟级部署。                   所以结论非常明显,CloudRobo平台可以借助自身价值为别人赋能。这种定位的价值不难理解。移动互联网时代,开发者不需要自主研发操作系统与芯片,基于成熟的系统生态就能快速开发应用。

八 |          对华为而言,下场制造本体不仅要直面这些行业共性难题,还会偏离自己深耕数十年的主赛道。华为的核心优势始终集中在ICT领域,从芯片算力到大模型,从云服务到操作系统,这些能力才是它的立身之本。

九 |          就像当年在汽车行业,华为没有自建工厂造车,而是把智能座舱、高阶智驾、电驱系统这些能力打包输出给车企,靠技术赋能拿到产业话语权。         类似模式的商业价值已经被充分验证。此前华为与赛力斯联手打造的问界系列,持续站稳高端新能源市场的头部位置,问界M9累计销量仍在50万级豪车市场榜首。赛力斯负责制造与品控,华为负责智能化定义与技术输出,双方各守能力边界,完成了单靠任何一方都难以实现的市场突破。放到机器人赛道,这套分工逻辑也许同样成立。

十 |          不过,机器人赛道终究和汽车不同,汽车可以靠一套方案打出统一品牌,机器人的生态版图,注定更分散,也更难用单一模式收拢。但即便如此,我们依然观察到,华为的具身智能朋友圈仍要扩大。         03.          华为的具身智能圈正在扩大          华为具身智能朋友圈最显然的可能是人。         根据外界一些媒体的不完全统计,目前已经有至少15位华为前高管离职加入或创办具身智能创企,这些企业中最高估值400亿至500亿港元(折合人民币约347亿至433亿元)、单笔融资最高达4.55亿美元(约合人民币约31亿元)。         这批人带着华为的工程化能力、大模型技术积累和量产落地经验,成为具身智能创业的主力军。他们的技术路径、研发思路甚至供应链资源,都带着鲜明的华为印记,相当于把华为的技术基因,扩散到了整个赛道之中。

十一 |          落地到技术布局来看,这批华为系创业团队几乎覆盖了具身智能全产业链,业务横跨机器人本体、运动小脑、决策大脑,同时布局机器人操作系统、专用大模型与数据采集系统。其中,智元机器人的量产机器人已突破15000台,再比如,它石智航、智澄AI等企业搭建起完整软硬件产品矩阵;还有的目前已对外推出实体产品。

十二 |          这就很有意思了,华为试图越来越像国内具身智能领域的人才黄埔军校,其华为昇腾、车BU、诺亚方舟实验室以及华为云具身智能团队中的核心人才,在批量外溢到机器人赛道,形成一波具身智能圈的创业潮。         可即便如此,对于华为自身而言,也很难在机器人赛道复制鸿蒙智行那样的品牌体系。汽车是标准化程度极高的产品,乘用车的形态和使用场景高度统一,华为可以用一套智能方案,联合不同车企打出统一的鸿蒙智行品牌。         机器人却完全不同,工业机械臂、家用服务机器人、商用配送机器人、人形机器人,形态天差地别,场景需求各有侧重,没有任何一套方案能覆盖所有品类。华为能做的是底层技术助力,给不同厂商提供工具和底座,却很难像主导汽车品牌那样,去定义每一款机器人的产品形态。         不构建统一品牌不意味没有合作机会,有意思的是,华为鸿蒙偏偏却已经深入行业场景中来了,早在2025年4月,华为已经发布M-Robots OS 1.0版本,所谓M-Robots OS是以开源鸿蒙为核心底座的分布式异构多机协同机器人统一操作系统,这也是全国首个基于开源鸿蒙打造的机器人操作系统。                   今年华为更新了M-Robots OS 2.0开源新版本,完成双足人形机器人、四足仿生机器人、机械臂等20余种机器人形态适配,还与创想未来、德壹医疗等多家企业实现了相应产品落地。         通过这些玩法,一方面证明了鸿蒙系统能给机器人带来生态联动的差异化优势,另外也给更多机器人厂商提供了参考,接入鸿蒙生态,就能直接复用十几亿设备的生态资源,不用从零搭建自身的智能体系。         回过头来看,从此前智能汽车到机器人,对于整个产业来说,有华为这样的玩家下场铺设底座,也许不是一件坏事。

Current article:http://www.lichuorenrenyitankanmeimu.cfd/news/20260826_525351.ppt

Published on:17:44:20


相关新闻

最后更新

热门新闻