Education & Research Industrial Software & Algorithm Provider Collaborative
Yao Maoqing Discusses the Evolution of Physical AI Through Model and Data Integration
Original from leaderobot.com: Yao Maoqing on the Key to Evolving Physical AI: Model and Data Flywheel

Yao Maoqing Discusses the Evolution of Physical AI Through Model and Data Integration

On July 19, during the 2026 World Artificial Intelligence Conference, Yao Maoqing, Senior Vice President and President of the Embodied Business Division at Zhiyuan, shared insights on the technological pathways for scaling physical AI. Zhiyuan has developed a three-phase training architecture of 'pre-training, post-training, and continuous learning' to advance its VLA and WAM technology routes towards the unified World Reasoning Action Model (WRAM).

The integration of data is facilitated by Mifeng Technology, which utilizes the MEgo series of collection terminals and the MEgo Engine governance platform to create a comprehensive physical AI data infrastructure. This infrastructure supports data collection, governance, training, and deployment feedback, ensuring that real-world data continuously enhances model evolution. The collaborative model and data iteration system has already been validated in real industrial scenarios.

Yao emphasized that 'models determine the starting point, while data defines the outcome.' He expressed the ambition of Zhiyuan and Mifeng to collaborate with the global academic community, industry, and developer ecosystem to accelerate the evolution of physical AI in real-world applications. No further timeline was disclosed at the time of publication.

Editor's Note

The advancements in physical AI highlighted by Yao Maoqing reflect a significant trend in the industry towards integrating model training with real-world data. This approach not only enhances the capabilities of AI systems but also aligns with the growing demand for intelligent solutions in various sectors. As organizations seek to leverage AI for practical applications, the collaboration between technology providers and academic institutions will be crucial for driving innovation and adoption.

RobotToday Initiative

Robotics needs a service framework.

RSF defines a common language for robot service capability, lifecycle operations, certification pathways, and service-provider networks.

Share

Related Articles/News

Humanoid Hack Tokyo 2: Disaster Rescue Robot Wins in Physical AI Competition

Humanoid Hack Tokyo 2: Disaster Rescue Robot Wins in Physical AI Competition

The 'Humanoid Hack Tokyo 2' event took place from July 11 to 12 at the GMO Humanoid Lab in Shibuya, Tokyo, showcasing humanoid robots in practical applications. Participants developed new robot applications using the Unitree G1 humanoid and technologies like Vision-Language-Action and Physical AI. This event is significant as it highlights the rapid growt...

Nubia NaviX Ultra: ByteDance Doubao Phone Gen 2 Launches with Advanced AI Features

Nubia NaviX Ultra: ByteDance Doubao Phone Gen 2 Launches with Advanced AI Features

The second-generation ByteDance Doubao phone, the Nubia NaviX Ultra, was unveiled at the WAIC. This device features multi-app task queuing, a proactive memory system, and the ability to operate autonomously across more than ten platforms. This launch is significant as it showcases advancements in mobile technology, particularly in AI integration. The Nubi...

Gadgets
Honor Robot Phone Launches with 4DoF Gimbal and YOYO AI at WAIC

Honor Robot Phone Launches with 4DoF Gimbal and YOYO AI at WAIC

The Honor Robot Phone has made its global debut at the WAIC, showcasing a 4DoF titanium mechanical gimbal system. This innovative device features YOYO proactive AI, a 200MP camera, and advanced stabilization technology with CIPA 5.5, enhancing its functionality and user experience. This launch is significant as it represents a breakthrough in mobile techn...

Gadgets
Qingche Intelligent Showcases Comprehensive Data Systems and Robotics at WAIC 2026

Qingche Intelligent Showcases Comprehensive Data Systems and Robotics at WAIC 2026

From July 17 to 20, 2026, Qingche Intelligent participated in the World Artificial Intelligence Conference (WAIC) in Shanghai, focusing on bridging models and real-world intelligence. The company showcased a complete technology system from real-world data collection to model training and robotic applications, emphasizing the importance of real-world data ...

Embodied Intelligence Data Management Robotics
WAIC 2026: Insight AI Unveils First Embodied Semantic Intelligence System, insightOS Semantic

WAIC 2026: Insight AI Unveils First Embodied Semantic Intelligence System, insightOS Semantic

At the 2026 World Artificial Intelligence Conference in Shanghai, Insight AI launched the world's first embodied semantic intelligence system, insightOS Semantic. This system represents a pivotal shift in the embodied intelligence industry, moving from mere demonstration capabilities to practical applications in real-world scenarios. The significance of t...

Embodied Intelligence Robotic Systems AI Technology
WAIC 2026: Embodied Intelligence Enters Industrial Deployment Era

WAIC 2026: Embodied Intelligence Enters Industrial Deployment Era

WAIC 2026 in Shanghai (July 17–20) saw 200+ embodied intelligence companies, 300+ live robots, and first confirmed humanoid deployments in logistics and pharmacy. RobotToday engineering analysis.

Market and Business News
ROBOTTODAY Weekly July 13 – 17, 2026

ROBOTTODAY Weekly July 13 – 17, 2026

This week in robotics: Hyundai takes full ownership of Boston Dynamics, Xpeng targets 1,000 Iron humanoids a month, LimX Dynamics raises $200M pre-IPO, and Agility opens a Silicon Valley physical-AI hub. July 13–17, 2026.

RobotToday Weekly
Inside Guident: How Harald Braun Is Building the Safety Layer Every Autonomous Vehicle Needs

Inside Guident: How Harald Braun Is Building the Safety Layer Every Autonomous Vehicle Needs

Guident CEO Harald Braun on the RMCC architecture, IRLS risk identification system, and why California’s 2026 AV mandates validate six years of platform development.

Market and Business News
Gravity 4D Launches First Module to Enhance World Models for Robotics

Gravity 4D Launches First Module to Enhance World Models for Robotics

On July 17, Extreme Intelligence unveiled Gravity 4D WAM, the first module of its embodied intelligence framework, at the 2026 World Artificial Intelligence Conference (WAIC). This launch addresses a critical issue in current visual language models: the discrepancy between visually plausible predictions and physical reality. Gravity 4D aims to transition ...

Robotics AI Machine Learning
Stardust AI Launches Lumo-2: Innovative Robot Action Model for Home Automation

Stardust AI Launches Lumo-2: Innovative Robot Action Model for Home Automation

On July 15, Stardust AI introduced its second-generation embodied base model, Lumo-2, which is the industry's first household latent world-action model. This launch includes the physical AI symbiotic agent, Agent Philia, enhancing their full-stack architecture of AI models, embodied operating systems, and rope-driven entities. The company will showcase it...

Household Robotics Physical AI AI Models
PHANES AI Launches TouchWorld Tactile Model for Enhanced Robot Dexterity

PHANES AI Launches TouchWorld Tactile Model for Enhanced Robot Dexterity

PHANES AI, established by 28-year-old Yang Shuo from HIT, has introduced TouchWorld, a tactile foundation model designed to enhance robotic manipulation capabilities. This model enables robots to perform precise physical tasks by incorporating a sense of touch, marking a significant advancement in robotic dexterity and interaction with their environment. ...

Technology
ROBOTTODAY Weekly July 6 – 10, 2026

ROBOTTODAY Weekly July 6 – 10, 2026

This week in robotics: AI² Robotics raises $735M for wheeled humanoids, Unitree G1 performs first live surgery, Dexmal launches DexOS platform, Agility Robotics announces SPAC, and Ondas acquires DZYNE for $875.8M. July 6–10, 2026.

RobotToday Weekly

Related Suppliers

Agibot 智元创新

AGIBOT is a leading developer of general-purpose embodied intelligent robots, offering a one-stop development platform for Embodied AI alongside integrated data solutions. Key products include the open-source AgiBot X1 and the AgiBot G1 Universal Embodied Intelligent Robot.

Humanoid

Agribot (PO)

Agribot (Poland) manufactures compact hydraulically-powered autonomous agricultural robots for vineyard and orchard row work, open-field crop transport, grass mowing, and winter road maintenance. Compatible with mowers, tillers, sprayers, plows, and spreaders.

Soar Robotics

The product focuses on autonomous reasoning agents that integrate cognitive AI with physical capabilities. Utilizing advanced algorithms, these agents can perceive environments, process information, and execute actions in real-world scenarios. Key technologies include machine learning, computer vision, and robotics, enabling seamless interaction between digital intelligence and physical systems. This approach aims to enhance operational efficiency across various industries.

RLWRLD

RLWRLD (rlwrld.ai), a South Korea-based Physical AI startup founded in 2024, develops robot foundation models (RFMs) enabling human-like vision, cognition, and dexterous manipulation for real-world autonomy. Focus: scalable embodied AI for high-difficulty tasks, precise industrial deployment, reinforcement learning-enhanced reasoning (e.g., Robot-R1), raised $14.8M seed funding.

Software & Algorithm Provider

Deplace AI

The product focuses on deploying Physical AI solutions in industrial environments, specifically factories and warehouses. It evaluates various AI models based on performance metrics and real-world applications. The platform facilitates task automation by analyzing specific use cases, providing data-driven insights, and ensuring alignment with key performance indicators (KPIs) to optimize return on investment (ROI) in robotic deployment scenarios.

MAUM.AI Inc.

MAUM.AI (South Korea) develops Physical AI foundation modules for embodied systems. Key tech: MAIED (edge robotic brain, multi-chip VLA support), BODA (vision), SUDA (language), WoRV (motion), MAAL (action/LLM). Products include JINDO BOT (autonomous quadruped security robot), Mate Box (retrofit brain for Unitree quadrupeds/humanoids), on-premise LLMs. Focus: autonomous decision-making, task automation in security, industrial, and service applications.

CynLr

CynLr develops vision-guided robotics with object intelligence. CyRo dual-arm system uses CLX vision stack (liquid lens optics, hierarchical depth mapping, force-correlated perception) for real-time grasping of unknown/reflective/transparent objects without pre-training. Features force-aware control and human-like dexterity for adaptive assembly in unstructured manufacturing environments.

Manipulation & Grasping Mobile Manipulator

Abaka AI

Abaka AI offers a robust data platform that supports the entire AI data lifecycle, including data collection, annotation, and processing. It provides pre-curated multimodal datasets for various applications such as LLM training, speech recognition, and image generation. The platform ensures compliance with GDPR/CCPA and features automated annotation capabilities, enhancing efficiency and accuracy in data handling for AI development.

Aweta

Post-harvest sorting/packing systems for fruit/veg (e.g., apples, tomatoes, avocados); sensor-based grading (external/internal quality via cameras, NIR, weight/size); roller/cup singulators, high-capacity lines; no field robotics.

Agriculture Software & Algorithm Provider Logistics & Supply Chain

Ellips

Post-harvest AI grading/sorting systems; TrueAI/TrueSort software with multispectral cameras for external/internal quality (defects, color, size, shape), 360° inspection, handles 19+ fruits/veg (e.g., apples, potatoes, citrus); high-speed lines, no field robotics.

Wayve (Wayve Technologies Ltd.)

Wayve develops advanced autonomous driving software utilizing deep reinforcement learning and computer vision techniques. Their platform leverages end-to-end neural networks to process real-time sensor data, enabling dynamic decision-making in complex urban environments. The software architecture is designed for scalability and adaptability, allowing for continuous learning from diverse driving scenarios and integration with existing vehicle systems for improved safety and efficiency.

Devanthro

Devanthro's Robodies integrate advanced robotics, artificial intelligence (AI), and teleoperation technologies to deliver continuous support in domestic environments. These humanoid robots utilize machine learning algorithms for task optimization and employ virtual reality (VR) for operator interaction. The system operates in unstructured environments, enhancing user experience while generating valuable real-world data for iterative improvements in functionality and autonomy.