Empowering the Development of Embodied Intelligence: ZTE’s Innovations

On February 24, 2026, ZTE Corporation hosted the “2026 Embodied Intelligence Robotics Academic Seminar” in Shanghai, in collaboration with the Shanghai Computer Society and the School of Computer Science at Shanghai Jiao Tong University. The seminar, themed “Human-Machine Symbiosis, Shaping the Future,” brought together over thirty experts and scholars from leading universities such as Tsinghua University, Shanghai Jiao Tong University, and Fudan University, as well as several cutting-edge technology companies. The discussions focused on breakthroughs in embodied intelligence robot technology, addressing industry bottlenecks, and pathways for large-scale implementation.

During the event, ZTE shared its innovative achievements and future plans in the field of embodied intelligence. Leveraging its extensive technological foundation in information communication, chips, operating systems, and AI, as well as rich practical experience in smart manufacturing, ZTE positioned itself as both a “core technology innovator and industry ecosystem promoter.” This dual positioning aims to accelerate the development of technologies related to embodied intelligence and support China in achieving independent innovation and high-quality growth in this field.

Driving Technological Innovation Through Practical Applications

ZTE adheres to the principle of “practical testing of technology, application empowering industry” and has taken the lead in validating and incubating technologies related to industrial robots and embodied intelligence at its Nanjing Binjiang Smart Manufacturing Base. This flagship facility for ZTE’s global smart manufacturing serves not only as a production site but also as an open “innovation laboratory” and “capability incubator.” Here, technologies for industrial robots and embodied intelligence have undergone comprehensive validation and iteration:

• Global Intelligent Scheduling: Utilizing the intelligent manufacturing brain for global scheduling, the base has achieved efficient coordination of numerous 5G cloud-based AGVs, robots, and machine vision systems, establishing a highly flexible automated production line.
• Overcoming Flexible Assembly Challenges: ZTE’s self-developed composite robot with dexterous hands and grippers successfully automates the precise insertion and removal of RJ45 connectors and optical modules through a visual-force control integration algorithm. This breakthrough addresses critical challenges in the flexible assembly industry, achieving an automation rate of over 85% on the production line.

In 2025, the Binjiang Smart Manufacturing Base was recognized as an “Excellent Intelligent Factory” by the Ministry of Industry and Information Technology, marking ZTE’s significant accumulation of technology and practical experience in the “AI+” smart manufacturing domain, thus providing solid scenario validation for the large-scale implementation of embodied intelligence robots.

Empowering Industry Development with “Connection + Computing Power”

ZTE believes that embodied intelligence is a crucial evolutionary form of artificial intelligence, extending from virtual cognitive decision-making to physical world interaction and execution. It is the key development direction for achieving an integrated loop of perception, cognition, decision-making, and execution in intelligent agents, serving as an essential carrier for the deep integration of the digital economy and the real economy. In line with this vision, ZTE has defined its core positioning and, supported by its strategic path of “connection + computing power” and robust technical foundations, aims to leverage its comprehensive advantages in chip and operating system technologies, communication networks, computing infrastructure, and full-stack AI technologies. By integrating these with its practical applications, ZTE seeks to create a collaborative embodied intelligence technology architecture that spans cloud, network, edge, and terminal, forming a differentiated competitive edge.

Building a Full-Stack Foundation: “Chip, Soul, Body, Brain, Network, Cloud”

Chip and Soul: ZTE has launched the automotive-grade heterogeneous chip “LAN Yue A1,” the first of its kind in China, designed to meet the computational needs of robots’ small brains, providing strong computing power along with high safety and reliability. The self-developed operating system for embodied robots, based on a deeply customized Linux kernel, integrates real-time enhancement, middleware for robots, AI foundational environments, and security enhancements, acting as the “central nervous system” for robots and ensuring the efficiency and determinism of the sensing-decision-execution loop.

Strong Body: Targeting advanced industrial manufacturing scenarios, ZTE has made breakthroughs in precision execution and bionic body technologies. Its self-developed dexterous hand, featuring a combination of rigid and flexible components, has 22 degrees of freedom and incorporates over 40 high-density sensors, enabling full sensory feedback for touch, angle, and position. This human-like design allows for precise execution of complex tasks such as grasping, pinching, twisting, and precision assembly, filling a significant gap in the domestic market for high-degree-of-freedom dexterous hands. The innovative lightweight rotary-moving hybrid joint configuration further enhances flexibility and performance while maintaining a balance among performance, cost, and energy consumption.

Intelligent Brain: ZTE has established a complete intelligent loop encompassing perception, planning, decision-making, and execution. The Embodied Brain, developed for general artificial intelligence (AGI), overcomes the limitations of traditional VLA models and achieves breakthroughs in intelligent agent alignment data structures, long-cycle task optimization, and generative reward models, providing a full-stack autonomous capability for robots. The integrated “Real-Simulation-Internet” three-in-one “Embodied Data Collection” platform and the RealMirror full-stack simulation environment utilize a combination of imitation learning and reinforcement learning to efficiently overcome data bottlenecks and accelerate algorithm optimization.

Robust Cloud and Network Collaboration: ZTE is capable of constructing a full-stack, domestically produced intelligent computing cluster and has already implemented several large-scale domestic intelligent computing center projects, providing robust, independent, and secure computing power for training and inference of large models in embodied intelligence. Additionally, leveraging the ultra-low latency, high reliability, and large bandwidth features of 5G-A/6G technology, ZTE has developed a cloud-network-edge-terminal collaborative framework, ensuring deterministic connections for multi-robot cooperative operations and remote real-time control, thereby solidifying the network foundation for large-scale industrial applications.

Open Source Collaboration: Striving Towards an Inclusive Future

The prosperous development of the embodied intelligence industry relies on an open and collaborative ecosystem. ZTE deeply understands industry development trends and actively assumes the roles of connector and enabler. Through standard formulation, technology open-sourcing, and ecological collaboration, ZTE promotes the co-construction and sharing of the embodied intelligence industry ecosystem, accelerating the process of technology democratization.

In terms of standard formulation, ZTE has actively participated in the National Robot Standard Technical Committee (TC591) and the Ministry of Industry and Information Technology’s Humanoid Robot and Embodied Intelligence Standard Technical Committee (MIIT-TC8), contributing to the drafting and formulation of several national and industry standards to help regulate the development order of the industry and clarify the direction of technological advancement.

Regarding technology open-sourcing, ZTE has fully open-sourced core tools such as the EmbodiedBrain planning model, RealMirror simulation platform, and Co-Sight intelligent agent framework. These tools have performed exceptionally well in international rankings and open-source communities, significantly lowering the R&D threshold in the industry. The EmbodiedBrain planning model has set milestone records in embodied intelligence task planning, achieving over a 15% improvement in task completion rates compared to previous models. All training data, weights, and codes for this model, as well as the innovative VLM-PlanSim-99 simulation environment, have been made publicly available.

The RealMirror simulation platform addresses key challenges in humanoid robots, such as high data collection costs and missing standardized benchmarks. It integrates comprehensive VLA benchmark testing for humanoid robots and utilizes generative models with 3D Gaussian splashing technology to create realistic environments for zero-sample Sim2Real transfer. The Co-Sight intelligent agent framework has achieved the highest comprehensive score in the authoritative GAIA benchmark and surpassed similar frameworks from OpenAI and Google DeepMind in the HLE benchmark test.

The specific open-source addresses for these tools are as follows:

• Co-Sight Intelligent Agent Framework: GitHub Link
• RealMirror Simulation Platform: GitHub Link
• EmbodiedBrain Planning Model: GitHub Link

In terms of ecological collaboration, ZTE closely collaborates with robotics companies, application partners, and developers by providing open hardware modules, software platforms, and development tools to deliver customized solutions for industrial manufacturing, warehousing logistics, and commercial services, thus accelerating the transformation of technological achievements.

During the 2025 MWC Shanghai event, ZTE, in collaboration with industry partners, launched the world’s first multi-intelligent heterogeneous collaborative innovation scenario based on 5G-A technology. This scenario utilizes a unified intelligent scheduling platform to achieve collaborative control of various manufacturers’ humanoid robots, showcasing ZTE’s highly open and system-level collaborative capabilities in embodied intelligence technology. Currently, this solution operates regularly, allowing robots to autonomously complete navigation, explanation, and interaction tasks without human intervention, supporting dynamic interactions based on environmental understanding. This marks a critical step in the integration of 5G-A and embodied intelligence, laying a solid foundation for new paradigms of human-robot collaboration in intelligent manufacturing and smart services.

Looking ahead, ZTE will continue to uphold its commitment to independent innovation, deepen the integration of “connection + computing power” full-stack technological innovations, and adhere to the principles of open cooperation. By collaborating with partners across industry, academia, and research, ZTE aims to address the challenges of large-scale applications of embodied intelligence, strengthen the technological foundation for independent innovation in the industry, and drive the advancement of embodied intelligence technologies from the laboratory into various sectors, contributing significantly to the upgrade of China’s high-end manufacturing and the high-quality development of the digital economy.