Yingfeng Chen
Head of Agile Technology at NetEase and Head of Fuxi Robotics Algorithms at NetEase
Yingfeng Chen is the Head of the Fuxi Robotics Algorithm Division at NetEase and the technical lead for the construction machinery robotics business. His research focuses on robotics, reinforcement learning, and game AI. He received his Ph.D. from the University of Science and Technology of China (USTC), during which he participated in multiple RoboCup World Championships and won the world champion title in the service robot category in 2014. He was also involved in the development of the USTC “KeJia” robot series. Previously, he led reinforcement learning research at NetEase Fuxi, where he creatively applied cutting-edge reinforcement learning techniques to game AI and automated game testing, with successful deployment across several NetEase games. Currently, he is responsible for the R\&D and algorithm innovation of multiple construction machinery robots, including excavator robots and unmanned loaders. He has published over 60 papers in top AI and robotics conferences such as NeurIPS, IJCAI, AAAI, and ICRA, and received the Best Paper Award at ASE 2019. He has served as a long-term reviewer for conferences including NeurIPS, AAAI, and ICRA. He holds over 50 granted national invention patents and has been selected for the Hangzhou Young Science and Technology Talent Training Program and honored as a “Binhjiang District Outstanding Young Science and Technology Talent.”
Topic
Applications of Embodied Intelligence Technology in the Intelligentization of Construction Machinery
In recent years, embodied intelligence technology has attracted increasing attention from both academia and industry, offering promising potential for the widespread application of robotics. The construction machinery industry is a typical representative of advanced manufacturing, and the intelligentization of construction machinery is a crown jewel in this field. The integration of embodied robotics technology with intelligent construction machinery not only addresses urgent industry needs but also provides an ideal scenario for validating and advancing embodied intelligence. This talk will take human–machine collaborative excavation robots for automated loading as an example. It will first introduce the technical implementation and challenges of remotely operated excavators, then present the embodied intelligence data loop established based on these remotely operated excavators, and finally discuss current efforts and progress on end-to-end excavator automated loading models based on VLA. Outline: 1. Background: Integration of construction machinery intelligence and embodied intelligence technology 2. Introduction to human–machine collaborative remotely operated excavators 3. Data loop based on human–machine collaboration 4. Training and progress of end-to-end models based on teleoperation data
