Key Insights from the Meituan Robotics Institute Academic Annual Meeting on Embodied Intelligence and AI‑Driven Robotics

On December 19, 2023, the Meituan Robotics Institute held its academic annual meeting at Tsinghua University Shenzhen International Graduate School. The event reviewed a year of progress, presented research achievements, and invited experts to discuss future trends in robotics.

The round‑table was chaired by Zhang Peng (Founder & President of GeekPark) and featured five distinguished scholars: Wang Tianmiao (Honorary Director of the Institute of Robotics, Beihang University), Zhang Hong (Chair Professor, Southern University of Science and Technology), Zhang Jianwei (Foreign Academician of the Chinese Academy of Engineering and German Academy of Engineering), Wang Yu (Professor, Department of Electronic Engineering, Tsinghua University) and Xu Huazhe (Assistant Professor, Tsinghua Institute of Cross‑Media Information, leader of the Embodied Intelligence Lab). Their discussion focused on three themes: technology development , application scenarios , and industrial prospects of robotics.

Core viewpoints:

Embodied intelligence was described as a “child” that learns and grows through human‑machine interaction, breaking tasks into a chain of sub‑tasks and using automatic programming to achieve autonomy.

Large‑scale models and big‑data were highlighted as the most effective entry points for advancing robotics, enabling higher‑level perception, decision‑making and control.

Industry experts emphasized that logistics, manufacturing, elderly‑care and service robots are the most promising near‑term markets, while scientific discovery (e.g., material and drug research) could benefit from continuous, 24‑hour robot operation.

Several speakers stressed the importance of industry‑academia collaboration . Zhang Hong noted that university research provides new theories, structures and algorithms, while companies like Meituan bring massive data and real‑world scenarios. Wang Yu pointed out the need for a unified “brain” to control heterogeneous sensors, suggesting that a common robotics operating system could lower costs and accelerate deployment.

Other notable points included:

The rapid progress of 3D point‑cloud cameras, lidar, and motor reducers, which are shrinking in size and becoming applicable to drones and small‑scale robots.

The concept of “top‑down” (research breakthrough) versus “bottom‑up” (industrial stability) as a guiding principle for successful technology transfer.

Future research directions such as generative simulation, continuous online learning, and multimodal sensing (vision, touch, hearing) for more robust embodied agents.

The meeting concluded with a call for deeper cooperation between academia and industry to address labor shortages, accelerate robot adoption in logistics and manufacturing, and explore new applications such as precision weed removal in agriculture.