Max Robotics — Global Service Robot Directory + US Certification

About

YuXuan Liu is a leading researcher at the intersection of imitation learning, meta-reinforcement learning, and robotic perception. Their work addresses fundamental challenges in autonomous systems, particularly how agents can efficiently acquire skills from observation and explore complex environments. Liu’s most influential contribution is the concept of "imitation from observation," introduced in a 2018 paper (287 citations), which enables agents to learn behaviors directly from raw video without requiring access to an expert’s actions or internal states—a breakthrough for real-world applications. Their 2017 work on learning invariant feature spaces for skill transfer across different morphologies (117 citations) further advanced cross-embodiment learning. In meta-reinforcement learning, Liu’s 2018 paper (181 citations) pioneered structured exploration strategies that adapt to task structure, outperforming task-agnostic methods. More recently, Liu has contributed to robotic perception with self-supervised instance segmentation via grasping and distributional uncertainty modeling (Latent-MaskRCNN), both published in 2023. With over 630 total citations, Liu’s research bridges theory and practice, enabling robots to learn more efficiently from limited data and transfer skills across diverse scenarios—work that has significant implications for autonomous systems, human-robot interaction, and embodied AI.