RealMan Robotics

RealMan Robotics is a Chinese robotics company founded in Beijing in 2018. Its legal operating name is RealMan Intelligent Technology (Beijing) Co., Ltd., rendered in Chinese as 睿尔曼智能科技（北京）有限公司. The company develops lightweight robotic arms, integrated joint modules, wheeled mobile manipulators, teleoperation systems, and data services for robot learning.^[1][2] Its RealBOT line is a family of wheeled humanoid development platforms intended for embodied-AI research, real-world data collection, remote operation, and service or industrial tasks.^[3][4]

RealMan Robotics is unrelated to Realbotix, the North American maker of lifelike social and companion robots. The similar English names refer to different companies, products, histories, and markets.

Scroll sideways for more →

Company history and positioning

RealMan states that it was established in 2018, developed four core robot components in 2019, expanded its lightweight humanoid-style arms in 2020, and introduced lifting, single-arm, and dual-arm mobile robots in 2021. Its Jiangsu production base began operating in 2022. The company later moved beyond selling components and arms toward what it describes as a three-part infrastructure stack: hardware, real-world robot data, and a distributed remote-operations network.^[1] These milestones and the infrastructure framing come from RealMan itself.

The company says it established a humanoid-robot data training center in Beijing in 2025 and reports more than 8,000 customers, 120 intellectual-property rights, and five global offices.^[1] It also reports that its robotic arms have a certified mean time between failures of 50,000 hours. The certification claim appears on RealMan's corporate materials, but the underlying test report was not independently reproduced for this article.^[1]

In March 2026, RealMan announced a financing round of approximately US$70 million. The company said the money would support product development, expansion of an AUTRON manufacturing facility, and international growth.^[6] Funding databases agree that a March 2026 round occurred, but differ on whether to label it Series B or Series C and do not independently substantiate all of RealMan's operating metrics.^[7][8] The round amount and intended uses should therefore be read as company-reported information.

Robotic arms and joint modules

RealMan's RM-series arms are designed around a controller integrated into the base, reducing the need for a separate control cabinet. The RM75 is a seven-degree-of-freedom arm with a 5 kg rated payload, roughly 610 to 627 mm reach depending on variant, and a stated positioning repeatability of plus or minus 0.05 mm. Standard and six-axis force-sensing versions are available.^[9] The company's catalog also includes six-degree-of-freedom RM65 arms, longer-reach RML models, economical ECO models, and RX-series arms used in its larger mobile platforms.^[1]

Published robotics research provides independent evidence that RealMan hardware is used as a laboratory platform, though it does not validate every manufacturer specification. An ICLR 2026 paper describing a dual-arm research system used two RM75-6S arms with third-party grippers.^[10] Other experimental work has used an RM75-6F for garment manipulation and an RM75-6F to demonstrate a bio-functional soft robot.^[11][12] These studies treat the RealMan arms as experimental equipment rather than evaluating the company or the arms against competing products.

RealBOT platform

RealBOT is RealMan's name for a set of wheeled humanoid and mobile-manipulation platforms, not a single fixed configuration. The company's current catalog distinguishes RealBot-01, RealBot-S2, and RealBot-L2, while earlier event material also used names such as RealBOT 1.0.^[3][13]

RealBot-01 combines a detachable humanoid upper body with a wheeled base. It has 21 active degrees of freedom: 14 across two arms, three at the waist, two in the lower body, and two at the neck. RealMan specifies two RX75 arms with a 5 kg payload per arm, stereo depth sensing, lidar, and a 40 Ah battery. The end effectors are configurable, and the upper body can be removed for use with another chassis or integrated with another RealMan dual-arm platform.^[3][4]

RealBot-S2 is a larger foldable wheeled humanoid with 20 active degrees of freedom, dual 5 kg-payload arms, four-wheel steering and drive, a working-height range up to 2.1 m, and a claimed operating time of about six hours. RealMan markets it for data collection, training, home services, retail, and industrial production.^[13] These specifications and use cases are manufacturer claims, not results from an independent field trial.

RealMan introduced RealBOT publicly as an open development and remote-work platform at CES 2026. Its demonstration connected operators in Beijing with a robot at the Las Vegas show floor. The company presented that link as an example of transoceanic teleoperation for tasks such as object delivery and transfer.^[14] Event posts document the demonstration, but they do not establish network latency, task-success rates, or autonomous performance under controlled testing.

Data and teleoperation systems

RealMan's data-services stack includes a Darwin teleoperation client and a Darwin device-management service. The teleoperation client combines arm controllers, multiple video channels, and localization maps so a human operator can control a remote robot. The management service covers device onboarding, task dispatch, compliance controls, and multi-tenant fleet operation.^[15] RealMan also markets data-collection systems and datasets for manipulation tasks in homes, retail spaces, and industrial settings.

This distinction matters when interpreting demonstrations. A RealBOT task may be autonomous, teleoperated, or used to record demonstrations for later model training. Hardware capability, remote human operation, and learned autonomy are different layers and should not be inferred from one another without task-specific evidence.

LingBot-VLA 2.0 collaboration

In the supplied September 1, 2026 X post, RealMan showed what it described as RealBOT training in action on material-handling tasks designed around industrial workflows.^[16] The post documents a company-selected demonstration. It does not establish whether the actions were autonomous or teleoperated, and it provides no task-success rate, comparison, or independent performance measurement.

In a separate LinkedIn article, RealMan said it was a core hardware partner for Robbyant's open-source LingBot-VLA 2.0 release and that its RS-02 data-collection robot contributed to the model's pretraining pipeline.^[17] That is a partnership claim from RealMan; the accompanying LingBot paper does not identify RealMan or RS-02 in its abstract.

The LingBot-VLA 2.0 paper reports a training corpus of about 60,000 hours, comprising 50,000 hours of robot trajectories across 20 robot configurations and 10,000 hours of egocentric human video. It describes a unified action representation for arms, end effectors, hands, waists, heads, and mobile bases, plus a future-prediction training objective.^[18] The public repository provides code and a 6-billion-parameter checkpoint under the Apache License 2.0.^[19] Those corpus and architecture figures describe the full Robbyant project. They should not be read as the amount of data supplied by RealMan, because neither RealMan nor Robbyant has publicly quantified RealMan's share.

Evidence boundaries

Most detailed RealBOT specifications, customer totals, production figures, reliability figures, and deployment descriptions originate with RealMan or its distributors. Academic papers confirm use of RealMan arms in research rigs, but do not independently validate the company's broader market, reliability, or autonomy claims. No public peer-reviewed evaluation located for this article compared a RealBOT platform with other wheeled humanoids under a common protocol. The supplied X post supports only RealMan's narrow description of a material-handling demonstration. The separate LingBot partnership claim is supported by RealMan's LinkedIn article and bounded against Robbyant's paper and repository.^{[16][17][18][19]}

See also

• Embodied AI
• Humanoid robot
• Robot learning
• Vision-language-action model
• Realbotix

References

1. ^RealMan Robotics. "About Us." Accessed September 2, 2026. realman-robotics.com/...about-us
2. ^RealMan Intelligent Technology (Beijing) Co., Ltd. Company profile. LinkedIn. Accessed September 2, 2026. cn.linkedin.com/...C%89%E9%99%90%E5%85%AC%E5%8F%B8
3. ^RealMan Robotics. "RealBot-01." Accessed September 2, 2026. realman-robotics.cn/...realbot-humanoid
4. ^RealMan Robotics developer documentation. "RealBot-01 Overview." Accessed September 2, 2026. develop.realman-robotics.com/...RealBot01
5. ^RealMan Robotics. Founder announcement for Eric Zheng. LinkedIn, November 2025. linkedin.com/...-activity-7397559856955641856-LYTW
6. ^RealMan Robotics. "RealMan Closes Nearly $70M Funding Round." March 4, 2026. realman-robotics.com/...about-us
7. ^Caplight. "RealMan Robotics." Accessed September 2, 2026. caplight.com/...realman-robot
8. ^CB Insights. "Realman Funding and Financials." Accessed September 2, 2026. cbinsights.com/...financials
9. ^RealMan Robotics. "RM75 Series." Accessed September 2, 2026. realman-robotics.com/...rm75
10. ^"RoboCOIN: An Open-Sourced Bimanual Robotic Platform." ICLR 2026 conference paper, hardware appendix. openreview.net/pdf
11. ^"Real-World Experiment Scene," paper appendix describing RealMan RM75-6F garment-manipulation hardware. 2025. openreview.net/...9bfff8e9377066d02214dbf61327.pdf
12. ^"A Bio-Functional Mimetic Robot for Versatile Tasks From Cross-Scale Manipulation to Limb-Tool Integration." 2026. pmc.ncbi.nlm.nih.gov/...PMC13469783
13. ^RealMan Robotics. "RealBot-S2." Accessed September 2, 2026. realman-robotics.com/...realbot-s2
14. ^RealMan Robotics. CES 2026 RealBOT teleoperation announcement. January 2026. linkedin.com/...-activity-7413948189021523970-zhm7
15. ^RealMan Robotics. "Data Services." Accessed September 2, 2026. realman-robotics.com/...data-services
16. ^RealMan Robotics (@RealmanRobotics). "RealBOT training in action" material-handling demonstration. X, September 1, 2026. x.com/...2094802463992021221
17. ^RealMan Robotics. "RealMan Joins Robbyant's LingBot-VLA 2.0 Open-Source Push." LinkedIn, September 1, 2026. linkedin.com/...-activity-7481723983902466048-ly9X
18. ^Wu, Wei, et al. "From Foundation to Application: Improving VLA Models in Practice." arXiv:2607.06403, July 2026. arxiv.org/...2607.06403
19. ^Robbyant. "lingbot-vla-v2: From Foundation to Application." GitHub. Accessed September 2, 2026. github.com/...lingbot-vla-v2