Max Robotics — Global Service Robot Directory + US Certification

About

Xuemeng Yang is a researcher at the forefront of 3D scene understanding for autonomous driving and robotics. Her work centers on solving the critical challenge of interpreting sparse, real-world LiDAR point clouds—a task essential for machines to navigate safely. Yang’s major contributions lie in developing novel deep learning architectures that fuse multi-scale context for semantic scene completion, a process that jointly estimates 3D occupancy and object labels. Her most cited paper, “Semantic Segmentation-assisted Scene Completion for LiDAR Point Clouds” (2021, 43 citations), pioneered the use of semantic features to guide the completion of occluded outdoor scenes, directly addressing the sparsity problem inherent to LiDAR data. She further advanced this field with the “Up-to-Down Network” (2021, 20 citations), which efficiently integrates multi-scale information for robust 3D perception. Demonstrating versatility, Yang also explored decentralized multi-agent coordination with her hierarchical learning approach for formation movement (2021, 8 citations), and developed LessNet (2022), a lightweight model balancing efficiency and accuracy for large-scale point cloud segmentation. Through these contributions, Yang is shaping the future of intelligent perception systems, making autonomous agents more aware of their environment.