Apple's research team published the paper "Perceptive Humanoid Motion from Simulation" on arXiv, unveiling HALO, an autonomous control system for humanoid robots, and simultaneously open-sourcing the training and deployment code.
At CoRL 2026, NVIDIA launched the open-source world-action model series GR00T 2, comprising two versions, GR00T 2 and GR00T 2 Turbo. The former focuses on generalization and fine manipulation, while the latter has fewer parameters and can be deployed directly on the Unitree G1, with a single inference taking 124 milliseconds on the onboard Jetson Thor.
Why it matters: NVIDIA released the GR00T 2 series at CoRL 2026 and disclosed the scale of three types of pre-training data along with success rates for cross-embodiment transfer and few-shot adaptation, which can be used to assess the current capability boundaries of world-action models.
Gamma Intelligence, TeleAI and Unitree have jointly launched LocoManiLab, a unified evaluation framework for humanoid robot mobile manipulation. It covers five types of home spaces, 13 scenario instances and 234 tasks, with the Unitree G1 and Dex3 dexterous hands as the unified test platform.
On November 7, the General Embodied Brain Industry Innovation Consortium was launched at a side exchange meeting of the China International Import Expo in Shanghai. Initiated by MouShen Intelligence, it brings together 42 enterprises and research institutions, with members including Fudan University's School of Future Information Innovation, the National and Local Co-built Humanoid Robot Innovation Center, Unitree Robotics, and UBTECH.
The WISE2026 Business King conference will be held in Beijing Yizhuang on November 26-28. The first batch of speakers includes robotics entrepreneurs such as Unitree founder Wang Xingxing, Future Not Far Robotics founder Zhang Yi, and Ola Omni founder Zhou Shunbo. The conference has run for 14 years, and this year focuses on how AI and robotics can truly enter business. Early-bird tickets have only five days left.
Unitree Technology founder Wang Xingxing said in an interview at the World Internet Conference Wuzhen Summit that Unitree will focus on conquering robot brain technology in the future and seek cooperation globally, and is optimistic about physical AI technology. Unitree has previously defined three major strategic directions: improving large-scale robot mass-production capability to meet demand at the tens-of-millions and even hundreds-of-millions scale; advancing the deployment of physical AI small models suited to its own hardware through both in-house development and partnerships; and building out ultra-large-scale computing clusters.
On November 5, Agility Robotics listed on Nasdaq under the ticker AGLT, closing its first day at $11.51, up 2.2%, and falling to $8.75 as of November 9. Its prospectus shows Digit has accumulated more than 65,000 operating hours and handled over 100,000 totes, but revenue in the first half of 2026 was only $174,000 with a net loss of $362 million; the $300 million Digit 5 order comes from a single related party, and that model is not planned for delivery to early customers until the first half of 2027.
On November 10, Unitree Robotics rose more than 12% intraday, with its share price exceeding 800 yuan per share, overtaking Lianxun Instruments to become the third-highest-priced stock on the A-share market, with total market capitalization still above 300 billion yuan; based on the stage low of 404.03 yuan per share on October 15, the share price has rebounded nearly 100%.
On November 9, Unitree Robotics shares rose more than 15% intraday, with market capitalization briefly topping 300 billion yuan, before closing up 9.04% at 716 yuan. An Apple research team used a modified Unitree G1 to validate the HALO paper, completing 360 single-task tests across six real-world environments without prior mapping, with an average success rate of 91%. Unitree and Kunlun Manufacturing plan to establish a new joint venture to conduct industrial embodied intelligent robot solution business; as of the time of verification, the plan was still in its public notice period.
Unitree Technology's stock price rose over 10%, and its market cap returned above 300 billion yuan.
ForPC is a force-aware whole-body control framework that regulates a humanoid robot's cart-pushing speed through two-stage reinforcement learning. The method first infers force latent variables and state latent variables from onboard observations without external force sensors, then uses a residual policy to coordinate arm, waist, and leg movements; in simulation, cart-pushing speed RMSE is reduced by up to 68.2% compared with a locomotion-only control baseline, and lateral error on multi-curve routes is reduced by up to 74.2%.
Apple's research team published the paper "Perceptive Humanoid Motion from Simulation" on arXiv, unveiling HALO, an autonomous humanoid robot control system, and validating it on the Unitree G1.
Why it matters: Apple's HALO splits task orchestration, motion generation, and tracking into three layers, and reports comparative results of 91% in unfamiliar environments and 70% on composite tasks, offering a look at its layered design and data synthesis pipeline.
In late October, Unitree Robotics and Sequoia China announced the joint establishment of the "Shanyu Innovation Program," focused on discovering and investing in early-stage startup projects in embodied intelligence, robotics applications, and related upstream and downstream key technology fields. The article, citing Unitree's Q3 financial report, noted that its revenue was 665 million yuan, up 69.93% year on year; net profit attributable to shareholders was 113 million yuan, down 17.37% year on year; and R&D investment in the first three quarters was 263 million yuan, up 191.48% year on year.
On November 6, US Eastern Time, humanoid robotics company Agility (AGLT) closed down 14%, falling below the $10 issue reference price, after rising 2.22% against the trend the previous day. The SPAC listing had a redemption rate of 61.5% among existing shareholders, and the shortfall was filled by a $201 million PIPE led by Foxconn; selling pressure emerged after the backstop funds were absorbed. The article compares Figure AI's valuation of about $39 billion, Unitree Robotics' market value of 100 billion RMB, and Agility's listing valuation of about $2.5 billion, arguing that the secondary market is shifting toward pricing based on orders and delivery schedules.
PetroChina Group's Kunlun Manufacturing and Kunlun Capital signed an agreement with Unitree Robotics and Beijing Fanjie Intelligence to establish a joint venture focused on industrial embodied AI robot solutions. The PetroChina side will hold a combined 60% stake, while Unitree and Fanjie Intelligence will each hold 20%. The public notice period runs until November 13.
By 2026, daily rental for humanoid robots has fallen to the thousand-yuan range, down more than 70% from early 2025; daily rental for robot dogs has dropped below 500 yuan, and more than 90% of rental stalls in Huaqiangbei have withdrawn. The article argues that the root cause of falling rents is not only increased supply, but also that rental and sales share the same demand pool, dominated by entertainment performances and educational displays, with low repurchase rates. Manufacturer-controlled rental platforms absorb production capacity and inventory through subsidies, but the buffer layer is limited, and the industry is being forced to push robots into scenarios with sustained repurchases, such as industrial manufacturing and warehousing logistics.
Apple released the HALO humanoid whole-body motion control system, which procedurally generates 1,000 virtual residential scenes and synthesizes nearly 300 hours of paired data, enabling the Unitree G1 to master skills such as sitting on chairs, turning sideways to pass through a half-open door, crouching under a table, and pushing boxes.
Why it matters: Apple's HALO uses simulation-synthesized data to train humanoid whole-body skills, transfers them zero-shot to real scenes, and ablation experiments show the contribution magnitude of each module.
Unitree Technology founder Wang Xingxing said at the 2026 World Internet Conference Wuzhen Summit that next year will likely be a critical juncture for the robotics industry or robot models. He believes the core challenge in embodied intelligence passing the ChatGPT moment is aligning AI models' inputs and outputs with real robots, and that commercial service deployment will come faster than in the home and industrial sectors. Unitree will focus on conquering robot brain technology in the future and will seek cooperation globally.
A user has released free comparison content of the Unitree G1 EDU+ and the Basic G1, covering the unboxing, hardware differences, walking performance, and App features. The video is a side-by-side demonstration, and the creator is soliciting the community's views on the differences between the two versions on Reddit.
In 2026, financing in the primary market for embodied AI tightened and the IPO channel became congested, with valuation logic returning from grand narratives to business common sense. According to IT Juzi data, from the start of the year to October 20, domestic financing in this sector exceeded 600 deals with a total amount surpassing 160 billion yuan, and companies valued at over 10 billion yuan have reached 32, but in the past two months old-share transfer quotes were only 65% to 80% of the previous round's valuation. Industry consensus has shifted to delivery ledgers, cash collected, and the number of scenarios implemented, and whether 10,000-unit-level stable delivery can be achieved has become the new proof for a 10 billion valuation.
In Q3 2026, Unitree's revenue was 665 million yuan, up 69.93% year on year, while non-recurring net profit was 105 million yuan, down 17.42% year on year. R&D spending was 127 million yuan, up 249.58% year on year, showing revenue growth without profit growth.
On November 4, the Beijing Municipal Market Supervision Administration published a notice: CNPC Kunlun Manufacturing and Kunlun Capital plan to jointly establish a new joint venture with Unitree Robotics and Beijing Fanjie Intelligent Technology, focusing on industrial embodied intelligent robot solutions. The CNPC side will hold a combined 60% stake, while Unitree and Fanjie will each hold 20%.
The article compares Unitree and AgiBot's divergent approaches to embodied intelligence: Unitree integrates GPT-6 Astra into the G1 as an "external brain," with Wang Xingxing judging that general models will be able to handle planning after another one or two generations; AgiBot only lets Astra serve as System2, while System1 and the 100 to 1000Hz force-control closed loop are developed in-house, and GO-3 Preview tops three leaderboards: RoboDojo, RoboTwin, and RoboCosieum.
D-Robotics partnered with Unitree Robotics to launch an embodied AI education solution for primary and secondary schools, using the Sunrise S100 chip for integrated computing and control, paired with the Go2 quadruped and R1 humanoid bodies, and designing 15 themed courses totaling 180 lessons. High Torque Power released a full embodied AI education solution the same day, with courses already running regularly at schools including RDFZ and Dongzhimen Middle School. Qiyuan Robotics pushed its first OTA to the T1 and Q1, adding Hi Echo speaking practice and AI introductory courses.
After GPT-6 Astra's release, general large models can directly control robots to perform tasks; embodied data orders are rebounding, the value of body companies is being reassessed, and model companies building embodied brains now face competition on the same field as foundation model giants. In NVIDIA testing, Astra achieved a 66.3% success rate in zero-shot direct action output, higher than π0.5, but average single-inference time was 18.56 seconds, and one successful action consumes about 160,000 tokens, equivalent to $1.96. The industry currently prefers having general models handle planning and the semantic layer, while embodied models handle millisecond-level motion control and the execution layer.
The humanoid robotics industry is diverging between hype and real-world deployment capability: Unitree has already reported profitable results, Figure AI has secured a high valuation on the strength of its model narrative, Tesla is still exploring mass production, and some startups have run into delivery and funding difficulties. Figure published test results for Helix 2.5: in 30 unfamiliar households where no data had been collected, the robot completed three task types—tidying toys, folding towels, and making beds—with a 56% success rate under the full-task completion standard. 1X states on the NEO product page that early users receive "basic autonomous capability," and that complex tasks can be completed with scheduled remote supervision and assistance from experts.
On October 30, D-Robotics partnered with Unitree Robotics to launch an embodied intelligence education solution for primary and secondary schools. On the same day, High Torque Power released a comprehensive embodied intelligence education solution, and Qiyuan Robotics announced the first OTA for T1 and Q1, adding Hi Echo speaking practice and AI introductory courses. The D-Robotics solution is built around the Sunrise S100, which integrates a CPU, BPU, and MCU, paired with the Unitree Go2 quadruped and R1 humanoid bodies, and features 15 themed courses and 180 lessons covering elementary through high school. High Torque Power said the related courses have been regularly carried out at schools including Renmin University of China Affiliated High School and Dongzhimen High School.
The authors propose soccerbot, a hierarchical reinforcement learning framework that distills a structured latent motion prior from large-scale human soccer motion capture data, constraining policy learning within a dynamically feasible motion manifold. A single skill-conditioned policy thereby learns moving, dribbling, trapping, and kicking in a unified way, while maintaining human-like motion and fluid transitions during skill switching. After robustness-oriented simulation design, the framework transfers zero-shot to the Unitree G1 humanoid robot, completing the full soccer skill set in real-world environments.
RoboColosseum is a benchmark for testing policies on real robots, covering 150+ scenarios and hundreds of actual robots, with all rollouts (including failures) published on Hugging Face.
At the Wuzhen Summit, Unitree Technology founder Wang Xingxing proposed that 2027 is likely to become a tipping point for the robotics industry, defined as robots completing 80% of tasks via voice commands in 80% of unfamiliar scenarios. Unitree's 2025 revenue was 1.699 billion yuan, with 5,215 humanoid robots shipped, but in the first half of 2026 its non-recurring net profit fell 19.34% year on year; of the 4.2 billion yuan raised, 2.022 billion yuan is directed to intelligent robot model R&D.
Agility Robotics' SPAC merger closes today, and the company will list on Nasdaq tomorrow under the ticker AGLT, becoming the first pure-play humanoid robotics company listed in the US. Total fundraising is about $465 million, including a roughly $201 million PIPE led by Foxconn, about $164 million from the SPAC trust, and $100 million in equity financing received earlier.
At the 2026 World Internet Conference Wuzhen Summit, Unitree founder Wang Xingxing judged that next year is likely the tipping point for the robotics industry or robot models, based on the rapid capability gains of foundation models such as GPT-6. He said the biggest current bottleneck is aligning models with real machines in the "last few millimeters" of precision assembly and flexible manipulation. Unitree is advancing a physical AI robot self-evolution system: AI agents autonomously search papers and open-source solutions to generate control code, which is validated in simulation and then deployed for real-machine testing, followed by closed-loop iteration combining automated evaluation and human scoring.
The first CMG World Robot Mount Tai Climbing Competition concluded on the course from Hongmen to Zhongtianmen in Tai'an, Shandong. The Unitree A2 robot dog won the legged robot sprint with a time of 43 minutes 41 seconds (actual 41 minutes 41 seconds plus a 2-minute penalty), using fully autonomous mode throughout to perceive and make decisions in real time on irregular stone steps.
The first CMG World Robot Mount Tai Climbing Competition kicked off in Tai'an. A Unitree G1 Developer Edition robot, the same model featured in the Spring Festival Gala, withdrew due to an overheating failure during the climb and was carried down the mountain by staff on a stretcher. The course is 3.9 kilometers long with more than 3,100 steps, and other robots on site climbed unsteadily and needed towing assistance to finish the competition.
An analysis by Cninfo WAVE argues that robots have yet to enter factories and homes at scale, instead surging first into commercial performances, exhibitions, elderly companionship, and other scenarios, spawning a secondary-development and rental ecosystem. Daily rental for humanoid robots has fallen from a peak of 30,000 yuan a year ago to under 2,000 yuan for basic models, while QingTianZu has reached a valuation of 7 billion yuan; yet a robot 6S store in Guiyang still shows a 40% gap between monthly revenue and costs. The author judges this to be a transitional commercialization stage between demo and commercial deployment, with rental vacancy rates and store operating costs as the main risks.
HAMPER reproduces Unitree G1 basket carrying in Isaac Lab, using a PPO whole-body policy and a five-stage state machine to chain together approach, squatting, threading the handle, and lifting and carrying, with pretrained weights provided for real replay on a local machine.
Rockchip and SigmaStar reported a combined net profit of 1.174 billion yuan in Q3 2026, with Rockchip at 473 million yuan for the quarter, up 90.49% year on year, and SigmaStar at 701 million yuan, up 752.76% year on year. Rockchip's RK3588, RK3576 and other chips handle robot visual recognition, 3D mapping and path planning, and the company has launched a multi-sensor data acquisition solution; SigmaStar shipped more than 10 million robot vision chips in 2025 for a 23% global share, and its SSU9366 targets outdoor and companion robots.
On November 2, at the first CMG World Robot Mount Tai Climbing Competition, the Unitree G1 developer edition withdrew due to overheating and was carried down the mountain by staff on a stretcher. The author argues the problem is not just cooling components, but a zero-sum contest inside the chassis among compute, battery, and heat dissipation for limited space and power. The G1 has a 9000mAh battery and a runtime on the order of about two hours; during continuous ascent, joint motors sustain high-torque output, accelerating thermal degradation. The article also cites CAICT report data: in the first half of the year, global humanoid robot shipments exceeded 22,000 units, with more than 60% going to entertainment performances and science education/data collection, while intelligent manufacturing scenarios accounted for about 13%.
In the first CMG World Robot Mount Tai Climbing Competition, which began on November 2, a Unitree G1 Developer Edition robot of the same model seen on the Spring Festival Gala withdrew due to an overheating fault and was carried down the mountain by staff on a stretcher. The course is 3.9 kilometers long with more than 3,100 steps, and other robots on site climbed unsteadily and needed towing assistance to complete the race.
A research team at Seoul National University has proposed RISE (Responsibility-Induced Specialized Experts), a VLA framework that can collect demonstration data for locomotion and object manipulation separately and then learn jointly from them, enabling a robot to execute multi-stage tasks consecutively without full multi-stage demonstrations.