Chen Guangda
Head of Planning and Control Algorithms for Fuxi Robot at NetEase
Chen Guangda, Head of Planning and Control Algorithms for NetEase Fuxi Robotics. Senior Engineer (Associate Professor Level), Ph.D. in Computer Science from the University of Science and Technology of China, Postdoctoral Fellow at Zhejiang University's School of Control Science and Engineering. Authored multiple papers in premier robotics conferences and journals including ICRA, IROS, and RAL. Specializes in intelligent construction machinery, leading the end-to-end development and deployment of NetEase Lingdong's excavator planning-control system and unmanned loader algorithm system. His technical achievements have resulted in over 50 patent applications (with 12 granted), covering key areas such as autonomous planning, environmental perception, and human-machine collaboration. The “Unmanned Loader System for Mixing Plants” he developed achieved China's first fully unmanned operation across the entire workflow, with a labor efficiency ratio exceeding 120%. It has been successfully deployed in multiple large-scale infrastructure projects nationwide, driving the intelligent transformation of construction machinery.
Topic
Data-Driven Algorithms: Scalable Implementation of End-to-End Operations for Autonomous Loaders
In the construction machinery industry, labor shortages, safety risks, and efficiency bottlenecks have made the unmanned transformation of wheel loaders a key breakthrough for industrial upgrading. This talk focuses on NetEase Lingdong’s large-scale deployment in real industrial scenarios such as mixing stations, providing an in-depth analysis of how to build a data-driven technological closed loop to systematically address the full operational workflow challenges of unmanned wheel loaders—from “mobile navigation” and “precise scooping” to “intelligent stacking.” Key points of the talk include: 1. Achieving centimeter-level precision in articulated vehicle trajectory control and dynamic path planning through a combination of world models and reinforcement learning. 2. Optimizing operational efficiency and energy management during the scooping process using embodiment-based perception and predictive modeling. 3. Innovatively applying diffusion models to overcome multi-joint coordination bottlenecks in stacking operations, increasing task efficiency by over 70%. These core algorithms, validated through tens of thousands of real operations, have enabled unmanned wheel loaders to perform 24/7 “lights-out” operations, improving human productivity by 130% and providing a replicable and scalable pathway for intelligent upgrades in construction machinery.
