Site Li

Papers

Total Citations

H-Index

About

Site Li’s research bridges the critical gap between perception and control in autonomous systems, with a primary focus on semantic segmentation for self-driving vehicles and model-based reinforcement learning (RL) for robotics. His most impactful work, “Importance-Aware Semantic Segmentation in Self-Driving with Discrete Wasserstein Training” (2020, 54 citations), addresses a fundamental limitation of standard cross-entropy loss in semantic segmentation. By introducing a discrete Wasserstein training framework that prioritizes importance-aware pixel classification, Li’s method significantly improves mean Intersection-over-Union (mIoU) performance—a key metric for safe autonomous navigation. This contribution directly enhances how self-driving cars perceive and react to their environment, making his work highly cited in the computer vision and robotics communities. In parallel, Li’s paper “MBB: Model-Based Baseline for Efficient Reinforcement Learning” (2020, 3 citations) tackles the data inefficiency of model-free RL by proposing a model-based approach that leverages system dynamics for more sample-efficient policy learning. Though newer, this work signals his commitment to developing practical, data-efficient algorithms for complex robotic tasks. Li’s research is notable for its dual emphasis on theoretical rigor and real-world applicability, positioning him as a rising figure in autonomous systems and machine learning.