At the 2026 World Robot Conference Forum, held in Beijing from August 19-22, Xie Shaofeng, Chairman of the OpenAtom Open Source Foundation and Chairman of the Ministry of Industry and Information Technology's Humanoid Robot and Embodied Intelligence Standardization Technical Committee, stated that while the humanoid robot embodied intelligence industry appears to be thriving, it actually faces three major challenges: divergent technical routes, a lack of standards, and difficulties in sustainable industrial development.

The first challenge is the fragmentation and disconnection of technical routes. This divergence, which can gradually evolve into a structural problem affecting resource allocation, results transformation, and long-term innovation, means the industry urgently needs to establish common industrial rules. He noted that the industry has yet to form a convergent technical paradigm. From hardware configurations to software architectures, companies are almost working in isolation, choosing vastly different technical paths. This divergence not only leads to a significant waste of research and development resources but also makes it difficult to pool the industry's collective strength to tackle common challenges such as the coordination of cerebellar motor control and cerebral cognitive planning.

The second challenge is the absence of product standards. Xie pointed out that a lack of standards leads to difficulties in product finalization and mass production, challenges in quality assessment, safety assurance, and cost reduction, as well as hindered market promotion. This series of systemic constraints weighs down the entire industry's development and weakens its foundation for sustainable growth.

The third challenge is the dilemma of sustainable industrial development. In his view, many companies remain at the prototype development and functional demonstration stage, with unclear paths for commercial application. Without unified common technical support, startups and new entrants must bear extremely high technical thresholds and trial-and-error costs. Disorderly competition and misallocation of resources can easily give rise to industry bubbles.

He also proposed a solution: using open source to build a unified technical architecture. This architecture primarily comprises four components: operating systems, control systems, communication protocols, and a data foundation. He emphasized that the core capability of embodied intelligence lies in learning from real-world data. However, the current proliferation of various data formats has created numerous data silos. He stressed the need to establish a common standard covering data collection, annotation, format, and quality control, to build large-scale, multimodal standardized datasets, and to continuously supply high-quality, shareable data resources to the ecosystem.