ShanghaiJune 12, 2026 /PRNewswire/ — Recently, the Ministry of Industry and Information Technology and the State-owned Assets Supervision and Administration Commission launched the 2026 National Special Action for Humanoid Robots and Embodied Intelligence Real-Scene Practical Training, focusing on multiple key scenarios in industrial, service, and specialized fields. The initiative deploys six key tasks: building real-scene practical training spaces, forming innovative application consortia, tackling practical operational skills, strengthening real-scene application verification and routine deployment, reinforcing key element guarantees, and consolidating mature experience. This aims to accelerate the closed loop of “real-scene training—data accumulation—product iteration—large-scale deployment,” supporting the high-quality development of the humanoid robot and embodied intelligence industry.
This national special action emphasizes the development path of real-machine, real-scene practical training. Kepler Robot, as early as 2024, strategically positioned itself and precisely aligned with top-level policy guidance, establishing China’s only dedicated training ground for industrial real-scene practical training of humanoid robots. It is also one of the few standardized training carriers in the industry capable of supporting 7×24-hour continuous real-machine operations for humanoid robots, perfectly aligning with the policy direction of this special action. The policy dividends continuously provide strong policy support and industrial endorsement for the company’s full-stack technology implementation and commercial-scale expansion.
Large-Scale Industrial Real-Scene Real-Machine Training Drives Iterative Upgrades of Humanoid Robot Capabilities in Real Working Conditions
Currently, humanoid robot hardware is largely mature, but the industry generally suffers from critical shortcomings such as single-dimensional perception, lack of force-tactile data, and model incompatibility with tactile modalities. Addressing this industry pain point, Kepler leverages its self-developed native data collection system to build a high-fidelity real-scene data foundation. It has accumulated over 100,000 full-modal real-scene data points, replicating real task chains at a 1:1 ratio, covering diverse scenarios such as industrial, commercial, and home environments, and has developed over 20 industry-specific skills adapted to more than 40 real-world deployment scenarios. With its core advantages of high realism and multi-scenario coverage, Kepler has built a data barrier difficult for the industry to replicate, providing solid support for model iteration and real-scene practical training.
Unlike the industry’s common virtual simulation training model, Kepler, relying on its proprietary data foundation and VTLA native large model, is the first in the industry to implement a large-scale training system for real-scene full-modal data. All training is based on real scenarios, real tasks, and real data iteration, targeting and overcoming core industry challenges such as rigid precision assembly, flexible object manipulation, and multi-step complex operations. This comprehensively hones the robot’s practical capabilities, covering three core areas: industrial precision production, tool operation, and lifestyle services.
1. Ultra-High Precision Zero-Clearance Plug-in Assembly: Targeting the most difficult zero-clearance tasks in industrial precision production, overcoming traditional robot issues like force imbalance, insertion deviation, and rigid compression jamming. Leveraging force-tactile perception for dynamic posture and force control, it resolves problems such as force imbalance, insertion deviation, and workpiece wear, meeting the assembly requirements of high-end precision components.
2. Adaptive Large-Clearance Plug-in Assembly (approx. 2mm clearance): Addressing pain points in conventional industrial plug-in operations such as clearance deviation, alignment offset, loose insertion, and poor adaptability, the robot, trained on real-scene data, can identify clearance deviations, correct trajectories, and complete workpiece insertion with a 2mm standard clearance, balancing speed and precision, suitable for batch standardized production.
3. Multi-Step Gear Precision Assembly: Breaking through the limitations of single operations, it autonomously decomposes complex processes, completing tooth alignment, step-by-step engagement, and force self-check, capable of handling high-precision multi-process assembly tasks.
4. Flexible Cable Tie Bundling Operation: Solving the industry shortcoming where traditional robots cannot adapt to flexible deformable objects. Flexible cable ties are prone to deformation, slippage, and breakage under force, requiring extremely high precision in force control and operational trajectory flexibility from the robot.
5. Irregular Flexible Material Stacking Operation: Addressing the difficulties of flexible materials like clothing, which are prone to wrinkling, shifting, lack fixed shapes, and have no standard operational trajectories, the robot uses full-modal perception to dynamically adjust posture and force for neat folding, suitable for scenarios like home services and commercial logistics.
Furthermore, this model possesses efficient one-brain-multi-machine adaptability, compatible with all categories of robot hardware forms, breaking down single-device application barriers and enabling model capability reuse across devices and scenarios, laying a solid technical foundation for large-scale deployment. Leveraging multi-model target mapping and policy distillation technology, a single model set can adapt to humanoid robots with different degrees of freedom and structures, eliminating the need for repeated R&D and effectively reducing deployment costs.
Deep Digital-Model Synergy: Real-Scene Training Drives Leapfrog Improvement in Industrial Operational Efficiency
Closely aligning with the special action’s requirement to “strengthen real-scene application verification,” Kepler, through the deep coupling of its VTLA embodied large model and exclusive industrial real-scene training ground, has created a new technological paradigm for steady-state operations synergizing real-scene training and digital models. The equipment features an active force-tactile recognition system, allowing autonomous adaptation to flexible, fragile, and irregular special workpieces, completely solving the technical deficiencies of traditional pure-vision robots, which rely on passive perception and lack adaptive adjustment. Operational fault tolerance and environmental adaptability have achieved qualitative leaps.
For challenging industrial conditions such as assembly deviations, workpiece wobble, material deformation, and extremely small clearances, the robot can perform real-time force control correction, dynamic posture adjustment, and adaptive force control, completely solving problems like jamming, wear, misalignment, breakage, and poor adaptability found in traditional equipment. Measured data shows that compared to traditional pure-vision solutions, Kepler robots improve task success rates by 10%-20%, significantly reduce assembly failure rates, and achieve comprehensive superiority in precision, stability, and adaptability over traditional industry solutions.
Benchmark Industrial Scenario Deployment: Empowering High-End Intelligent Manufacturing, Creating a Dedicated Training Ground for Industrial Real-Scene Practical Training of Humanoid Robots
Kepler has taken the lead in achieving large-scale deployment across multiple industrial and commercial scenarios, validating its technological leadership with substantial real-world commercial results and leading the industrialization process of humanoid robots. It provides a “training ground” for robot applications in industrial settings.
Kepler has built a 1:1 full real-scene industrial training base, continuously honing core capabilities such as precision assembly, flexible grasping, and dynamic obstacle avoidance in robots. It has now achieved mature deployment in high-end industrial scenarios like automotive manufacturing, intelligent logistics, precision machining, and flexible production, comprehensively covering core working conditions in high-end manufacturing and adapting to complex workshop environments and 24-hour continuous operation needs.
Automotive Manufacturing Scenario: Achieving precise feeding of automotive welding parts, stably adapting to complex workshop conditions like high temperature and high reflectivity.
Laser Cutting Processing Scenario: Completing high-precision loading and unloading of laser cutting equipment, compatible with flexible grasping of multiple materials such as metal and composites.
Intelligent Production Line Logistics Scenario: Handling heavy-load intelligent material transport tasks in workshops, ensuring seamless coordinated operation of the entire production line.
Flexible Production Scenario: Adapting to the loading and unloading needs of multiple types of workpieces on flexible production lines, matching the modern production model of small batches and high variety.
Automated Wheel Hub Workpiece Feeding: Self-adapting to multiple specifications of wheel hub workpieces, completing precise workpiece feeding and fully automated pallet transfer in one go.
Expanding Beyond Industrial Boundaries: Developing Diversified Industry Commercial Solutions
Closely following the national strategic direction of “robots entering shopping malls and homes,” Kepler leverages its hardcore technological foundation deeply cultivated in the industrial sector. Using China’s only dedicated training ground for industrial real-scene humanoid robots as the core training carrier, it breaks down the barriers of a single industrial scenario, expands application boundaries, and creates diversified solutions for commercial services and vocational education. This achieves cross-domain reuse and value realization of industrial-grade embodied intelligence technology, precisely meeting the core requirements of the policy for multi-scenario real-scene training and large-scale deployment.
Leveraging its advantages in industrial-grade high-precision force control and full-modal perception, the Kepler K2 Bumblebee humanoid robot has completed comprehensive commercial real-scene adaptation training. It can adapt to various lifestyle and commercial complex scenarios, possessing standardized, large-scale commercial deployment capabilities.
Retail Scenario: The K2 Bumblebee can autonomously perform routine operations such as shelf inspection, product replenishment, goods sorting, and display organization. Adapted to the complex environment of high foot traffic and multiple interferences in retail stores, it can provide round-the-clock automated intelligent services, effectively reducing store operating costs.
Premium Service Store Scenario: Targeting the needs of small spaces and fine operations in coffee shops, the robot, with its industrial-grade high-precision force-tactile technology, can independently complete the entire process of fine, flexible services, including cup retrieval and beverage preparation, counter organization, and tool placement. This compensates for the shortcomings of traditional service robots, such as insufficient precision and rigid operations, helping humanoid robots enter commercial scenarios for people’s livelihoods.
Kepler Mechanical Qilin to Debut Soon: Fully Unleashing Industrialization Deployment Capabilities, Enriching the Real-Scene Training Collaborative Operation Matrix
Following the iteration of multiple category deployment solutions, Kepler has completed key technological upgrades for industrial scenario applications, proudly launching the world’s first dual-form heavy-load quadruped robot, the “Mechanical Qilin” series, including two standardized models: the riding model and the flatbed heavy-load model. Relying on its self-developed planetary roller screw linear actuator, the static standing maximum load capacity is ≥600kg; continuous endurance at full load exceeds 8 hours, overcoming the industry’s heavy-load endurance shortcoming. The Kepler Mechanical Qilin can achieve 45° slope climbing with IP-rated waterproofing, allowing traversal of complex alpine snowfields and muddy roads. It can operate in formation with the 75kg Kepler K2 Bumblebee humanoid robot, targeting border logistics, emergency rescue, and field material distribution, enabling round-the-clock unmanned heavy-load high-intensity transport operations. As this year’s major new product and an important member of the complete robot product matrix, the Kepler Mechanical Qilin will further assist Kepler Robot in building a humanoid robot training ground for continuous operation in industrial scenarios.
Anchoring National Training Policy Direction, Reshaping Industry Development Paradigm, Helping China’s Humanoid Robot Industry Lead Globally
From independently overcoming full-modal real-scene data barriers and developing native embodied large models, to being the first to implement industrial real-scene real-machine pre-training and digital-model collaborative commercial deployment, Kepler has completed a strategic upgrade from a single hardware manufacturer to a full-stack intelligent platform enterprise encompassing “hardware + real-scene training data + general large model.” This completely breaks the traditional industry development path reliant on virtual simulation and pure-vision solutions, preemptively aligning with all policy requirements of this national-level humanoid robot real-scene practical training special action.
Currently, Kepler is fully operating China’s only dedicated training ground for industrial real-scene practical training of humanoid robots, focusing on the real-machine training track for humanoid robots and embodied intelligence industrialization. By implementing the six major tasks of the national special action through a physical training carrier, it continuously promotes the high-quality, large-scale development of China’s humanoid robot industry.
In the future, Kepler will continue to deepen its “based in industry, expanding across domains” strategy, continuously broadening the application boundaries of robots, enabling embodied intelligence to truly empower all fields of intelligent manufacturing, commercial livelihood, and vocational education, helping China’s humanoid robot industry lead the global race.