During the National People’s Congress and Chinese People’s Political Consultative Conference sessions, He Han, member of the National Committee of the Chinese People’s Political Consultative Conference and CEO of Tianyu Digital Technology, has submitted proposals centered on embodied intelligence, humanoid robots, computing power, and digital transformation initiatives for private enterprises.
He Han observed that previous discussions about humanoid robot technology primarily emphasized hardware components like motors and reducers—functioning as the joints and organs of these machines. Current advancements, however, now concentrate on breakthroughs in the “brain” and “cerebellum” of humanoid robots. The integration of embodied intelligence technologies, exemplified by models like VLA, with humanoid robots represents an evolutionary revolution from “mechanical shells” to “digital life.” This convergence substantially reduces entry barriers for embodied robots while rapidly multiplying application scenarios.
The core challenge facing embodied robot development remains the absence of universal platforms. He Han identified five critical deficiencies:
-
Absence of algorithm development platforms: Most companies must independently build foundational “brain” (cognitive decision-making) and “cerebellum” (motor control) systems from scratch, resulting in duplicated efforts, dispersed resources, and compromised efficiency for embodied robots.
-
Scarcity of standardized 3D data platforms: High-quality 3D datasets are rare, with data collection equipment being inaccessible and expensive. Non-standardized formats hinder deep training for embodied intelligence in robots.
-
Lack of unified certification systems: Incompatible hardware interfaces, communication protocols, and data formats across manufacturers prevent embodied intelligence from functioning across different robot platforms, restricting large-scale deployment of embodied robots.
-
Insufficient real-world testing environments: Without standardized scenario libraries or public testing facilities comparable to autonomous vehicle proving grounds, current embodied robots remain confined to limited applications like commercial guidance or academic research.
-
Deficiency in interdisciplinary talent development: While expertise in mechanics and automation exists, professionals skilled in both large AI models and robotics engineering are scarce, impeding innovation in embodied robots.
To address these challenges, He Han recommended:
-
Establishing a national development framework for embodied intelligence to coordinate progress across brain algorithms, cerebellum systems, and mechanical hardware for embodied robots. Encourage universal platform development to prevent redundant foundational work.
-
Accelerating standardization and ecosystem certification covering hardware interfaces, communication protocols, 3D datasets, performance metrics, and ethical safety for embodied robots. Implement “embodied intelligence compatibility certification” with subsidies for compliant products.
-
Opening application scenarios in government and state-owned enterprise domains including flexible manufacturing, healthcare rehabilitation, public safety, and emergency response to accelerate embodied robot technology iteration.
-
Creating cross-industry testing platforms through academic or corporate partnerships. Develop embodied intelligence robot testing centers with multi-scenario physical environments to evaluate perception, decision-making, and control capabilities.
-
Promoting open-source resources like code libraries, simulation environments, and datasets to foster collaborative innovation and reduce R&D barriers for embodied robot development.
He Han emphasized that integrated machines for embodied intelligence large models represent the strategic direction for universal platforms. These systems combine “algorithm + data + computing power” trinity through spatial intelligence MaaS platforms, 3D data repositories, and cloud-edge-device computing synergy. They create interoperable “brain” (perception-decision-control integration) and “cerebellum” (skill execution) development platforms enabling cross-platform compatibility—an “Android moment” for the embodied robot industry where manufacturers can achieve plug-and-play functionality.