WRC回顾 | 五天三项新突破 星源智跑出具身智能“加速度”

This article summarizes three core breakthroughs Xingyuan Intelligence achieved in embodied intelligence at the 2026 World Robot Conference, as detailed below:

1. Breakthrough in interaction capability: Xingyuan Intelligence, in collaboration with a Nanjing University team, completed the world’s first real-world verification of heterogeneous multi-agent coordination. Two humanoid robots worked together to turn a jump rope while a quadruped robot dog jumped continuously, achieving real-time synchronized perception and decision-making across robots of different forms. This breakthrough breaks the limits of single-agent intelligence and advances embodied intelligence toward multi-agent autonomous collaboration.

2. Breakthrough in model technology: The company open-sourced its first JEPA-WAM world model with only 0.5B parameters, which eliminates the need for large-scale pre-training datasets and drastically reduces computing and deployment costs. Complementing the existing ω-EVA model, the new model enables a key leap for world models from merely predicting environments to powering robot decision-making.

3. Breakthrough in hardware products: Xingyuan launched the N5C liquid-cooled edge computing platform, which delivers 3 to 15 times the heat dissipation capacity of traditional air-cooled systems. The platform is adapted for complex industrial and outdoor scenarios, while also reducing size and maintenance costs to complete the company’s edge product portfolio.

Xingyuan Intelligence’s founder notes that successful embodied intelligence deployment requires accounting for three bottom lines: effectiveness, efficiency and cost, and edge-side processing chips ("edge brains") are a mandatory requirement for the deployment of high-level embodied intelligence.

Xingyuan Intelligence’s brand promotion and technology roadmap unveiled at the World Robot Conference offer valuable insights for brands in the embodied intelligence sector, as outlined below:

1. Brand marketing: Xingyuan created highly visual and viral demos including heterogeneous multi-robot rope jumping and human-robot table tennis matches. Leveraging the traffic of a top-tier industry exhibition, it quickly established brand recognition as a technology leader. This approach is a strong reference for peer technology brands.

2. Industry trends: Embodied intelligence has moved beyond the concept demonstration stage into real-world validation. The market and industry increasingly prioritize practical value, so brands must build competitive advantages by focusing on efficiency improvements and cost reduction for real-world use cases.

3. Product strategy: Xingyuan has built full-stack capabilities covering algorithms, models, hardware and use cases, completing the value chain from underlying technology to real-world deployment. This aligns with the current embodied intelligence development direction of "cross-ontology, cross-scenario, deployable and mass-producible," helping it gain greater trust from the industry.

The conference revealed the latest opportunities and development logic in embodied intelligence, offering operational guidance for sellers in related sectors. Key takeaways are as follows:

1. Market opportunities: There is strong industrial demand for intelligent robots that can replace human workers. Mature deployment solutions already exist for high-risk repetitive work scenarios such as high-altitude operation and cargo handling, where current loading and unloading efficiency matches that of human workers. These solutions help customers cut costs and boost productivity, opening up enormous market opportunities.

2. Product selection and operation logic: Embodied intelligence is no longer a competition of empty concepts. The industry ultimately measures success by effectiveness, efficiency and cost. Sellers should not rely solely on technical hype for product selection and marketing; instead, they must highlight the practical problems their products solve for customers to win market recognition.

3. Cooperation opportunities: Xingyuan has open-sourced its models and built a complete edge product portfolio covering diverse demands. Sellers can leverage these mature existing technology solutions to quickly enter different niche segments, reduce their own R&D investment, and launch deployable products faster.

The new embodied intelligence achievements released at the conference offer important insights for factories advancing intelligent upgrading and exploring business opportunities, as summarized below:

1. Product upgrading demands: Industrial operation scenarios are increasingly demanding higher robot stability, computing power and environmental adaptability. The newly launched N5C liquid-cooled edge computing platform solves the heat dissipation, size and stability challenges brought by high-level artificial intelligence, and is adapted to the complex high-temperature, dusty environments common in factories. Factories can introduce this new hardware to upgrade their existing robot fleets.

2. Commercial deployment opportunities: Embodied intelligence is already deployable for high-frequency, high-risk factory scenarios such as high-altitude operations and cargo handling, where it can fully replace traditional human high-altitude work and multi-device coordinated operation. The solution has already completed scenario validation, helping factories improve safety, boost efficiency and cut costs, with finished deployment cases at real enterprises available for reference.

3. Insights for digital upgrading: When advancing intelligent robot upgrading, factories should prioritize deployment in scenarios that deliver immediate practical value instead of blindly chasing unproven new technical concepts. Edge-side intelligence is a clear future direction, so factories should keep their focus on improving actual production efficiency.

Current industry dynamics in embodied intelligence point out clear development directions for technology service providers in this field, with key insights as follows:

1. Industry development trends: It is now clear that edge-side "brains" are mandatory for deploying high-level embodied intelligence, and technical development is focused on multi-agent cross-ontology coordination capabilities. The core value of world models now lies in powering robot decision-making and action. The industry has shifted from technology R&D to large-scale deployment, so service providers can build their business strategies around this direction.

2. Core customer pain points: Customers developing high-level embodied intelligence currently face common pain points including high data and computing costs, poor heat dissipation and stability for high-performance computing, and insufficient multi-robot coordination capabilities. There is strong demand for targeted solutions to these problems.

3. Reference solutions: Mature technical pathways are already available for reuse. The small-parameter JEPA-WAM model reduces data and computing costs; the liquid-cooled architecture solves heat dissipation challenges for high-performance computing; and the proprietary heterogeneous coordination framework enables cross-ontology cooperation between multiple robots. Service providers can build on this mature technology system to develop customized deployment solutions for clients across different industries and expand their business scope.

The development of the embodied intelligence industry points out new directions for the operation and growth of robotics industry platforms, as outlined below:

1. Industry demand for platforms: The embodied intelligence sector has already formed an ecosystem of open-source technology plus full-stack products. The industry needs platforms to provide open integration interfaces that connect technology developers with scenario partners to accelerate the translation of technical achievements into real applications. Platforms need to build dedicated open-source ecosystem integration sections to meet this industry demand.

2. Investment attraction and operation direction: The sector has already produced leading full-stack technology companies like Xingyuan Intelligence with multiple real-world deployment achievements. These technology-leading companies with proven deployed projects are high-quality investment targets. Attracting such companies can effectively boost a platform’s influence in the intelligent robotics space.

3. Risk mitigation: As the industry transitions from the concept stage to value-driven deployment, platforms should guide resident companies to anchor their development around the three core values of effectiveness, efficiency and cost. Platforms need to discourage excessive hype of unproven, non-deployable technical concepts to avoid industry bubble risks and support the healthy development of the sector.

The achievements and insights presented at the conference reveal the latest industry and technology trends in embodied intelligence, offering valuable reference for related research, as summarized below:

1. New technology and industry trends: On the technical side, multi-agent interaction has achieved real-world validation of heterogeneous cross-ontology coordination, advancing embodied intelligence from single-device pre-programmed skills to multi-agent autonomous collaborative decision-making. For world models, a complete dual-path complementary technology system has been formed, completing the core leap from predicting the world to powering robot decision-making. Edge computing hardware has also achieved technical breakthroughs, solving the heat dissipation and size challenges for high-performance computing deployment.

2. New core industry challenges: The most central question the industry currently faces is how to move embodied intelligence from demonstration to large-scale deployment. Industry consensus has emerged that deployment must be measured against the three core metrics of effectiveness, efficiency and cost, and it is now clear that edge-side brains are a mandatory requirement for high-level embodied intelligence deployment.

3. Industry development model: A development model of open-source technology ecosystem plus full-stack hardware products plus scenario-based deployment validation has taken shape. Advancing technology iteration through an open-source ecosystem and conducting scenario validation in partnership with industry players is a proven viable development pathway for the current embodied intelligence industry.