A3 has published ISO 21423:2026, a new international standard for communication and interoperability among industrial mobile robots. This standard establishes common communication requirements for mobile robots from different vendors operating in the same facility. It specifies how robots, fleet managers, and other equipment can exchange information such as location, status, and tasking. Adopting this standard is expected to reduce path interference, improve system efficiency, and foster market competition. Furthermore, the standard covers communication with external systems like ERP tools and building automation.
NVIDIA Cosmos provides world foundation models and tools to address sensor adaptation and reliable deployment challenges in physical AI. The Cosmos Framework offers modular tools for training, optimization, and inference across various models. Simulation recipes allow for generating synchronized video and LiDAR data for applications like autonomous vehicles and robotics. Cosmos NIM microservices accelerate inference through techniques like quantization and speculative decoding on GPUs. Developers can use the framework to train models, perform reinforcement learning, and deploy optimized models using specialized skills.
Shareholders of Churchill Capital Corp XI approved the $2.5 billion SPAC business combination with Agility Robotics. This move allows Agility to become a publicly-traded company, with the transaction expected to close on November 4th. The funding will support commercial expansion, customer deployments, and manufacturing capacity for the Digit humanoid robot business. Agility already has customer deployments with GXO Logistics and is expanding into manufacturing. The company is pursuing commercial markets in logistics and manufacturing with a product roadmap for Digit 5.
Electric motors were used to build a physical model simulating muscle coordination. The model demonstrated that muscle-like behavior emerges from the physical arrangement of motors and their shared mechanical structure, not from continuous programming. This setup showed that motors can coordinate by briefly engaging a shared backbone, allowing forces to be transmitted collectively. This principle of coordination without a central conductor could lead to simpler and more adaptable designs for robots. These findings suggest a path toward creating more organic and adaptive robotic systems.
The market for humanoid robot batteries is projected to reach US$850 million by 2037, according to IDTechEx. This forecast is based on a projected ten-year compound annual growth rate of 34.2%. The analysis considers spare batteries for swapping alongside demand from new robots. The report evaluates various battery technologies against performance metrics for industrial, commercial, and domestic humanoids. The demand is also influenced by solutions like battery swapping to extend operational time for robots.
Brett Adcock shared a video of newly manufactured Figure 03 humanoids leaving BotQ and walking across the campus. The 19-second clip shows the robots moving toward a parking-lot crossing. This footage offers a glimpse into Figure's growing robot fleet outside the factory. The video does not provide updated figures on total production or customer deliveries. The increased robot numbers support internal research and development efforts for Figure.
Intrinsic Core was released as an open-source, ROS-compatible real-time control framework for industrial robotics. It provides a hardware-agnostic runtime, SDK, and framework for real-time control, motion planning, and grasp planning. The release includes a pre-configured setup, including simulation tools and a reference solution called the Open Machine Tending Solution (OMTS). Community members are actively building variations on OMTS, demonstrating its practical application in tasks like picking up and inserting objects. Intrinsic Core aims to be practical and open by fostering community collaboration.
Tutor Intelligence raised $75 million in a Series B round, totaling $117 million, led by Andreessen Horowitz. This funding will support training their frontier robot foundation models and deploying their robot workers, Sonny and Cassie. The company bridges the gap between robot capabilities and human capabilities by using real-world deployment data for training. Tutor utilizes internet-connected robots operated by remote "tutors" to perform real work in production environments. This approach creates a self-reinforcing cycle that makes robots smarter, more affordable, and enables broader workforce automation.
Increased robot speed necessitates a clear process for experienced operators to challenge machine decisions to prevent errors from spreading. Safety procedures must remain paramount, and challenges should not supersede physical safeguarding measures like emergency stops. Effective commissioning should include rehearsals involving false positives and conflicting instructions to test human response and decision-making authority. Systems must retain exception records detailing operator challenges, evidence, and responses across shifts. Ultimately, a successful system evaluation should include metrics for human-AI teaming and operator understanding alongside traditional performance measures.
CMU has established a new Robotics Innovation Center in Pittsburgh's Hazelwood neighborhood. The center spans 150,000 square feet, reflecting the university's history in robotics and commitment to innovation. It is located along the Monongahela River. The announcement was first published on the Robotics Institute Carnegie Mellon University website.
Tesollo’s four-finger robotic hand scores 6/10 overall, with a Degree of Freedom rating of 4/5 and a Strength rating of 2/5. This scoring highlights a greater capability for dexterity compared to grip strength within the assessment framework. The assessment emphasizes that flexibility and strength must be evaluated separately based on specific handling requirements. The hand is listed as in production with an advertised price of USD 12,000 from Tesollo in South Korea. Buyers should evaluate specific dexterity and force requirements rather than relying solely on the overall score.
YMBot D is a foldable, wheeled humanoid robot developed by Jiangsu Yunmu Intelligent Manufacturing Technology Co., Ltd. It measures 165 cm tall and weighs 100 kg, featuring 22 degrees of freedom and two articulated arms. The robot is designed for tasks like obstacle avoidance, sorting goods, and product assembly using its mobile base and dexterous hands. It is listed as in production with a price of USD 35,000 and utilizes an Intel 12th generation CPU. The robot represents a direction in humanoid development combining mobility with object handling for industrial applications.
Nvidia's humanoid reference platforms utilize mechanical hands from Chinese startup Wuji and hands from Singapore-based Sharpa. Both suppliers have significant ties to Shanghai, centralizing Chinese hardware development for Nvidia's robotics software research. These systems demonstrate that robot training packages rely on specialized suppliers for body and manipulation hardware rather than a single manufacturer. Practitioners note trade-offs between the hands concerning task performance, thermal behavior, and cost. Hand selection is thus framed as balancing these practical constraints in humanoid development.
Jingchu Elf is a wheeled service humanoid developed by Hubei Jingchu Humanoid Robot Co., Ltd. It stands 165 cm tall and combines a humanoid upper body with a two-wheeled base. The robot features LiDAR and depth sensing for autonomous obstacle avoidance and can sort goods. It is powered by an NVIDIA Ampere GPU and has a skill score of 2 out of 10, indicating early-stage capability in navigation and manipulation. The platform is listed as a prototype with a price of USD 25,000 for research and exploratory service applications.
Max is Meta's first humanoid robot prototype, developed with Fourier for home assistance research. Standing 150 cm and weighing 48 kg, it features humanoid form, five-fingered hands, and a face screen. Documented capabilities include obstacle avoidance, stair walking, and operating a dishwasher, linking movement with household object manipulation. Its skill score is low (3/10), placing it in the developing category for general-purpose robots. The prototype serves as a platform for researching physical systems for home assistance.
The YMBOT E3 is a modular bipedal humanoid robot designed for robotics education and research. Standing 140 cm tall and weighing 35 kg, it features an Intel i9 CPU and ROS 2 compatibility for its control architecture. It possesses capabilities like obstacle avoidance and goods sorting to provide practical experimental tasks for users. The modular design and articulated structure facilitate research into motion control and manipulation. As a prototype, it is positioned as an experimental system rather than a finished general-purpose robot.
ATOM is a 160 cm, 50 kg humanoid robot from Vbot designed for indoor environments like homes and offices. It features 31 degrees of freedom, electric joint actuators, and a fabric-covered exterior emphasizing quiet operation. ATOM has established capabilities in obstacle avoidance and sorting goods, which are preliminary functions. The robot integrates LLM capabilities and is currently listed as a prototype with a price of $25,000. Deliveries for ATOM are expected to begin by the end of 2026.
YMBot H is a wheeled, humanoid service robot developed by Jiangsu Yunmu Intelligent Manufacturing Technology Co., Ltd. It is designed for visitor engagement in public spaces, featuring a humanlike appearance, dual arms, and a mobile base. The robot measures 170 cm tall and 75 kg, with a maximum speed of 5.4 km/h and six-hour runtime. Its confirmed capability is obstacle avoidance for navigation in indoor environments, but its skill scores for human guidance and manipulation are very low. Ultimately, YMBot H is positioned as a preliminary service platform emphasizing mobile humanoid form over advanced physical task performance.