19 KiB
Robotics Weekly Review 2026-04-23
Week In Review
The same weekend that Tesla stationed Optimus at the Boston Marathon finish line to greet runners and pose for photos, a Chinese humanoid built by Honor ran the Beijing E-Town Half Marathon faster than any human ever has—finishing in 50 minutes and 26 seconds, seven minutes ahead of the human world record. The contrast captures the state of the humanoid race in miniature: American companies are building brand recognition while Chinese firms are pushing performance boundaries. A CNBC analysis published two days later confirmed that Chinese startups now dominate global humanoid shipments, even as U.S. rivals command far higher valuations.
Beyond humanoids, the week brought infrastructure advances that will shape how all robots learn and deploy. NVIDIA released GR00T N1.7 and Newton 1.0 during National Robotics Week, providing the open foundation models and physics simulation that developers need to move from lab prototypes to production systems. On the industrial side, ABB launched its PoWa cobot family to bridge the gap between collaborative and industrial robots, while Flex and Teradyne deepened their manufacturing partnership to scale intelligent automation globally.
The week also underscored how robotics is expanding into new domains and geographies. SoftBank and Matternet partnered to scale drone delivery across U.S. healthcare and logistics networks, Pudu Robotics opened a U.S. headquarters in Dallas to serve the Americas market, and USC received ONR funding for multimodal robot dexterity research that could transform how robots handle tools. At the biological frontier, Northwestern engineers printed artificial neurons that communicate with living brain cells—work with long-term implications for neuromorphic robot control systems that approach the brain's energy efficiency.
Items
Honor Humanoid Beats Human Half-Marathon World Record in Beijing
A humanoid robot built by Chinese smartphone maker Honor completed the Beijing E-Town Half Marathon on April 19 in 50 minutes and 26 seconds, beating the human world record of approximately 57 minutes set by Uganda's Jacob Kiplimo in March at the Lisbon road race. Honor's robots swept the podium—gold, silver, and bronze—all navigating the 21-kilometer course autonomously.
The winning robot's design was modeled on elite human athletes, featuring legs approximately 95 centimeters long and a powerful liquid-cooling system developed largely in-house to sustain performance over the full distance. The result marked a dramatic improvement from last year's inaugural event, when the top robot finished in 2 hours, 40 minutes, and 42 seconds—a fivefold improvement in pace within a single year.
The achievement sparked debate about what the record actually proves. Scientific American noted that the robots ran on flat, controlled roads rather than the varied terrain that challenges human runners, and that the engineering optimization for straight-line bipedal speed is a different problem from the general-purpose locomotion that would make humanoids useful in homes and factories. Still, the demonstration showcases China's rapid progress in legged locomotion and the engineering depth that comes from intense domestic competition among humanoid startups.
Source: Scientific American
Tesla Optimus Greets Runners at Boston Marathon
Tesla stationed its Optimus humanoid robot at the company's 888 Boylston Street showroom on April 19 and 20, directly along the Boston Marathon finish stretch, where it greeted runners from roughly 130 countries and posed for photos with hundreds of thousands of spectators. The appearance represented Optimus's most high-visibility public interaction to date, turning a global sporting event into an earned-media showcase at zero advertising cost.
The deployment served a dual purpose beyond marketing: it stress-tested human-robot interaction in a real, unpredictable public environment—a data point that controlled lab demonstrations cannot replicate. The contrast with the same weekend's Beijing half-marathon, where Honor's humanoid outran humans, highlighted the divergent strategies of the two leading humanoid ecosystems: China pushing raw performance benchmarks while the U.S. focuses on brand building and controlled commercial deployment.
Tesla continues to target late 2026 for the beginning of Optimus mass production, with portions of the Fremont factory being converted for Gen 3 manufacturing. At scale, Elon Musk has stated the robot could cost between $20,000 and $30,000—roughly in line with a typical car—though the company has not yet disclosed commercial orders or production volumes.
Source: Interesting Engineering
China Leads U.S. in Humanoid Robot Shipments as Valuations Diverge
Chinese humanoid startups took the top six spots in Omdia's rankings of global robot shipments in 2025 and are already delivering robots to factories and malls, according to a CNBC analysis published April 21. Yet a striking valuation gap persists: U.S. startup Figure commands at least $39 billion in valuation, and Apptronik achieved $5 billion in February, while China's AI2 Robotics has reached only 20 billion yuan ($2.93 billion) despite comparable or greater shipment volumes.
The gap reflects differing investor philosophies. Roughly 90 percent of U.S. venture capital flows into software, leaving a financing gap in hard tech that sovereign wealth funds—particularly from the Middle East—are beginning to fill for Chinese firms. Limx Dynamics, backed by China-based Future Capital, received its first foreign investor this year from Dubai-based Stone Venture, illustrating how the capital structure of the humanoid industry is becoming geographically complex.
The analysis raises a question that will define the next phase of the industry: whether the U.S. model of high-valuation, pre-revenue development or the Chinese model of lower-valuation, high-volume shipping will prove more durable. With Agibot having shipped its 10,000th cumulative humanoid in late March, China's lead in deployment volume is widening even as U.S. companies continue to attract disproportionate capital.
Source: CNBC
NVIDIA Releases GR00T N1.7 and Newton 1.0 for National Robotics Week
NVIDIA announced a suite of open tools for physical AI during National Robotics Week, which began April 20. The Isaac GR00T N1.7 model is now available in early access with commercial licensing, bringing generalized robot skills including advanced dexterous control to production-ready deployments. GR00T N1.7 enables robots to understand natural language instructions and perform complex, multi-step tasks using vision-language-action reasoning.
Alongside the model, NVIDIA released Newton 1.0, an open-source physics engine providing fast and reliable simulation for dexterous manipulation with accurate collision detection, realistic object contact, and stable simulation of complex systems with both rigid and flexible parts. The release also includes Isaac Sim 6.0 and Isaac Lab 3.0, expanded simulation capabilities that allow developers to model real-world scenarios and validate robotic systems before physical deployment.
NVIDIA frames these releases as the "three computers" needed to bring robots from research to production: GR00T serves as the robot's brain, Newton simulates its body, and Omniverse provides the training ground. By open-sourcing both the models and the physics engine, NVIDIA is positioning itself as the default development platform for robotics—extending its GPU ecosystem advantage from data centers into the physical world, just as it did with deep learning a decade ago.
Source: NVIDIA Newsroom
ABB Launches High-Speed PoWa Cobot Family
ABB Robotics launched its new PoWa collaborative robot family on April 22, designed to bridge the long-standing gap between traditional cobots that lack industrial speed and payload, and conventional industrial robots that require dedicated safety infrastructure. The PoWa family offers payloads from 7 to 30 kilograms, a best-in-class top speed of 5.8 meters per second, and what ABB describes as the longest reach and highest arm load on the collaborative robot market.
The six-model family is powered by ABB's OmniCore controller platform and integrates with the company's expanding suite of AI-powered software, including RobotStudio for simulation and Wizard Easy Programming for code-free deployment. ABB says a PoWa cobot can be unboxed and operational within an hour, with plug-and-play compatibility across a wide range of end effectors.
ABB estimates the collaborative robot market will grow by 20 percent annually through 2028, and PoWa targets the applications where that growth is expected to concentrate: high-speed machine tending, palletizing, screwdriving, and arc welding. The launch reflects a broader industry trend toward cobots that can operate at industrial speeds when humans are absent and automatically slow to collaborative-safe speeds when workers enter the shared workspace.
Source: ABB
SoftBank and Matternet Partner to Scale Drone Delivery Across U.S.
SoftBank Robotics America and Matternet announced a strategic partnership on April 17 to accelerate autonomous drone delivery deployments across healthcare, retail, and enterprise logistics in the United States. Matternet brings FAA Type Certification and Production Certification—the first drone delivery company to hold both—along with tens of thousands of completed flights across urban and suburban areas.
SoftBank Robotics America will support manufacturing, installation, and maintenance of Matternet's ground infrastructure systems and provide operational support for enterprise deployments. The partnership model addresses what both companies identified as the primary bottleneck: not the drone technology itself, but the ground-level logistics of deploying and maintaining delivery networks consistently across complex operational environments.
Matternet CEO Andreas Raptopoulos described the partnership as part of a broader shift toward "physical AI"—systems that move goods in the real world rather than just processing information. The collaboration positions autonomous drone delivery as a complement to ground-based autonomous mobile robots, collectively building out the last-mile logistics layer that e-commerce and healthcare networks increasingly demand.
Source: SoftBank Robotics America
Flex and Teradyne Robotics Expand Partnership to Scale Intelligent Automation
Flex and Teradyne Robotics announced on April 22 an expansion of their collaboration to accelerate intelligent automation across global manufacturing. The partnership builds on a 20-year relationship in which Flex manufactured Teradyne's semiconductor test equipment, now extending into a dual-role model where Flex both deploys Teradyne Robotics solutions in its own facilities and manufactures key robotics components for Teradyne's customers worldwide.
Teradyne's Universal Robots (UR) cobots and Mobile Industrial Robots (MiR) autonomous mobile robots play central roles in the partnership. Flex manufactures key UR components while simultaneously deploying UR cobots and MiR AMRs in its own production environments—creating a continuous feedback loop where real-world operational data informs manufacturing improvements.
The partnership integrates physical AI technologies designed to enable more adaptive, flexible automation within increasingly complex production environments. The arrangement illustrates how the robotics industry is maturing beyond standalone robot sales toward integrated automation ecosystems where the same company that builds the robots also uses them—generating the operational insights needed to improve reliability and reduce deployment friction for end customers.
Source: Flex
Pudu Robotics Opens U.S. Headquarters in Dallas
Pudu Robotics, a global leader in service robotics, officially opened its new U.S. headquarters in Richardson, Texas, on April 23 as part of a strategic expansion across the Americas. The facility in Sherman Tech Center combines office space, a product showroom, and on-site warehousing, serving as the central hub for nationwide operations, customer support, and regional coordination.
Raymond Pan, General Manager of the Americas at Pudu Robotics, stated the company's ambition is to serve one million people across the U.S. over the next five years. Pudu, which first entered the U.S. market in 2018, has built its business on service robots for hospitality, healthcare, and commercial cleaning—segments where autonomous mobile robots have achieved faster commercial adoption than humanoids due to simpler manipulation requirements and clearer ROI calculations.
The Dallas location was chosen for its logistics infrastructure, business-friendly environment, and access to enterprise customers. The opening reflects a broader pattern of Chinese robotics companies establishing physical U.S. presence to localize sales, support, and eventually manufacturing—a strategy that hedges against potential trade disruptions while bringing robots closer to the customers who use them.
Source: DC Velocity
USC Wins $750K Naval Research Award for Multimodal Robot Dexterity
Dr. Erdem Bıyık, an assistant professor at USC's Viterbi School of Engineering, was selected for the 2026 Office of Naval Research Young Investigator Program Award for a three-year, $750,000 project titled "Learning Dexterous Robot Hand Behaviors from Multimodal Human Feedback." The research introduces a new framework for teaching robots fine-grained manipulation through a combination of visual demonstration and natural language guidance.
Current approaches to robot dexterity training face a fundamental feedback problem: it is difficult to provide robots with the precise corrections needed to fix minute physical errors during manipulation tasks. Bıyık's framework addresses this by combining two feedback channels—visual demonstration, where robots learn by observing human actions, and natural language guidance, such as real-time verbal corrections like "grip tighter" or "rotate slightly left"—that together provide richer training signal than either modality alone.
While funded for naval applications such as shipboard maintenance and autonomous tool use, the multimodal feedback approach generalizes to any environment requiring high-precision dexterous manipulation, including warehousing, manufacturing, and domestic service. The project represents the growing recognition that robot dexterity—the ability to handle varied objects with human-like skill—remains the critical unsolved problem separating current robots from truly general-purpose machines.
Source: USC Viterbi School of Engineering
Northwestern Prints Artificial Neurons That Communicate with Living Brain Cells
Engineers at Northwestern University developed flexible, printed artificial neurons that successfully triggered responses from living brain cells in mouse brain tissue, according to research published in Nature Nanotechnology. The devices use electronic inks containing molybdenum disulfide and graphene, deposited onto flexible polymer substrates via aerosol jet printing—a fabrication method that is both low-cost and scalable.
Unlike previous artificial neurons that produce simple one-off electrical pulses, the Northwestern devices generate complex signaling patterns—single spikes, continuous firing, and bursting—that closely resemble biological neural communication. This richer signaling vocabulary enabled them to activate real neurons in tissue tests, demonstrating a new threshold of biocompatibility between electronic and biological systems.
For robotics, the implications run along two paths. First, the printed neurons advance the prospect of bio-electronic interfaces that could enable robots to communicate more naturally with biological systems—relevant for surgical robots, prosthetics, and any application where machines must interact directly with neural tissue. Second, and more speculatively, the work points toward neuromorphic control systems that process information using the brain's own signaling architecture. The brain operates five orders of magnitude more efficiently than conventional digital processors, and printed neurons that replicate its signaling patterns could eventually enable robot controllers that match biological efficiency—a critical requirement for untethered robots operating on battery power.
Source: Northwestern University