Daily AI and Robotics Wrap: Humanoid Capabilities Accelerate in Dexterity and Durability
Figure AI Unveils Figure 03 Humanoid Robot, Engineered for Mass Manufacturing and Enhanced Dexterity
Figure AI introduced its third-generation humanoid robot, the Figure 03, which represents a significant hardware and software redesign aimed at achieving a truly general-purpose machine capable of learning directly from humans and performing human-like tasks at scale. The company’s goal with the 03 model is to create a robot ready for high-volume production and deployment in both home and commercial environments.
A central feature of the new robot is its integration with Figure’s proprietary vision-language-action AI model, Helix. To enable this advanced AI, the Figure 03 features a completely redesigned sensory suite, including a next-generation vision system. This new camera architecture boasts twice the frame rate, one-quarter the latency, and a 60% wider field of view per camera compared to its predecessor. These enhancements are crucial for providing the Helix AI with a denser, more stable perceptual stream, which is essential for intelligent navigation and precise manipulation in complex and cluttered settings, such as residential homes.
The manipulators, or hands, of the Figure 03 have also received major upgrades. They now include embedded palm cameras with a wide field of view, allowing the robot to maintain visual awareness even when its main head-based cameras are blocked, such as when reaching into a cabinet. Additionally, the hands are equipped with new pressure sensors sensitive enough to register the weight of a paperclip, which enables better grappling of fine-textured materials and detection of slippage. The fingertips are now softer and more adaptive, allowing for more stable grasps of delicate items like eggs or tomatoes.
Beyond its enhanced capabilities, the Figure 03 was engineered from the ground up for mass manufacturing, a departure from traditional prototypes that are often time-consuming and expensive to produce. Figure has established a new supply chain and process, including a dedicated manufacturing facility called BotQ, to facilitate this scaling. The BotQ facility’s initial manufacturing line is projected to be capable of producing up to 12,000 humanoid robots per year, with an ambitious goal of producing a total of 100,000 robots over the next four years.
- **AI Integration:** Built around the proprietary Helix vision-language-action AI model.
- **Vision System:** Features a next-generation camera architecture with double the frame rate and significantly lower latency for improved perception.
- **Dexterous Hands:** Includes embedded palm cameras and highly sensitive pressure sensors for refined object manipulation.
- **Manufacturing Focus:** Engineered for high-volume production, with a dedicated facility, BotQ, aiming for 100,000 units in four years.
DEEP Robotics Launches DR02: The World’s First All-Weather Industrial Humanoid Robot
DEEP Robotics, a company focused on embodied intelligence for industrial applications, has officially unveiled its new generation of industrial-grade humanoid robot, the DR02. The robot is being touted as the world’s first full-size, all-weather industrial-grade humanoid, addressing a critical industry pain point where mainstream robots often struggle in real-world, complex outdoor and factory environments.
The DR02 is designed for durability and environmental resilience, featuring an IP66 protection rating, which signifies its ability to adapt to all-weather outdoor operations. It is capable of continuous operation across a wide temperature range, from -20℃ to 55℃, allowing it to seamlessly transition between dry factory floors, outdoor rainy conditions, low-temperature cold storage, and high-temperature workshops. This robust design is intended to overcome environmental constraints that have previously hindered the practical implementation of industrial-grade humanoids in complex scenarios like security patrols and certain factory operations.
In terms of intelligence, the DR02 is equipped with a high-computing-power unit that boasts a computing power of 275 TOPS (Tera Operations Per Second). This powerful unit provides the necessary momentum for real-time data processing and complex task decision-making, enabling the robot to respond quickly to environmental changes. Its multi-sensor system includes a LiDAR, a depth camera, and a wide-angle camera, creating a robust perceptual computing platform for accurate obstacle identification and optimal path planning. The modular design and high-performance hardware configuration allow the DR02 to handle diverse and complex operations while also being designed to unlock more advanced intelligence through future software iterations.
Google DeepMind Elevates Robot Intelligence with Gemini Robotics 1.5 Models
Google DeepMind has introduced two new artificial intelligence models, Gemini Robotics 1.5 and Gemini Robotics-ER 1.5, significantly boosting the intelligence and reasoning capabilities of robots. These models are specialized for robotics and are based on the Gemini large language model (LLM), building upon the first iteration of Gemini Robotics released earlier in the year.
The key advancement in the 1.5 iteration is the ability for robots to handle “long-horizon” and complex, multi-step tasks that were previously out of reach. While the original model could manage simple instructions, like placing a single banana in a basket, the new models enable a robot to perform more intricate tasks, such as sorting a selection of fruit into separate containers based on color. This capability represents a significant step toward robots assisting people in real-world use cases by allowing them to reason and execute more general, complicated commands.
The new models also possess the ability to use external tools, such as Google Search, to inform their decision-making and task execution. This feature further expands the robot’s ability to handle ambiguous instructions and deal with novel, never-before-seen experiences by accessing and utilizing prior knowledge, common sense, and physics-based reasoning. The ultimate goal is to move beyond simple, pre-programmed actions to a state of true general-purpose reasoning in the physical world.
Global Robotics Competition Intensifies as China Prioritizes Hardware and Deployment
The field of humanoid robotics is rapidly emerging as the next major battleground in technology, following the initial race in foundational artificial intelligence. Countries worldwide are now vying to dominate the market for embodied intelligence—AI-powered robots that can navigate and operate in the real world.
China is leveraging its manufacturing strengths and strong policy support to take a leading position in robot hardware, deployment, and overall ecosystem development. The country has established a comprehensive supply chain for key components, from actuators and sensors to batteries, enabling startups to rapidly fine-tune prototypes into affordable and scalable products. This tripartite alliance between the government, enterprises, and research institutions has been instrumental in accelerating what is being called “China Speed.”
The focus on rapid deployment is evident in the numbers; last year, China installed nearly 300,000 robots in its factories, surpassing the rest of the world combined, with over half of these being domestically made. This deployment on factory floors is crucial, as it generates the real-world data necessary to train the robots’ AI brains. Furthermore, entry prices for Chinese humanoid robots are dropping drastically, with some models being released for under \$6,000, making them highly accessible for deployment.
In contrast, the United States is focusing its efforts on the “brain” of the robots, maintaining distinct advantages in software, advanced AI chips, and foundational research. Companies like Google DeepMind and those utilizing NVIDIA’s Isaac GR00T foundation models are concentrating on developing the sophisticated AI needed for human-like reasoning and complex task execution. South Korea is also participating, focusing on specialized humanoid robots that leverage its existing strengths in semiconductors and electronics manufacturing.
According to a Morgan Stanley “Humanoid 100” report, Chinese companies accounted for 61% of the 66 humanoid robot models unveiled between 2022 and the end of last year, while U.S. and Canadian companies contributed 24%. This split highlights the current global strategy: the U.S. leads in the AI intelligence layer, while China is rapidly dominating the production and deployment of the physical hardware.