Max Robotics — Global Service Robot Directory + US Certification

About

Xuanlin Li is an emerging robotics and embodied AI researcher whose work sits at the intersection of robot learning, 3D visual representations, and scalable simulation. His most significant contribution is ManiSkill3, a GPU-parallelized robot simulation and rendering framework designed to overcome critical limitations in existing platforms — namely, restricted scene diversity and poor sim-to-real transferability. Already accumulating 13 citations since its 2025 release, ManiSkill3 positions itself as an open-source foundation for scaling generalizable robotics research. Li has also made meaningful contributions to understanding how 3D point cloud representations can enhance robot manipulation learning. His "Frame Mining" paper demonstrated that coordinate frame selection — often overlooked — profoundly affects policy performance, offering a cost-free improvement for 3D-based manipulation systems. Complementing this, his systematic comparison of 3D versus 2D visual representations in reinforcement learning fills an important gap in the field's understanding of when and why 3D inputs provide advantages. More recently, Li has tackled the reproducibility crisis in real-world robot evaluation, proposing simulation-based assessment pipelines for generalist manipulation policies. Collectively, his work advances both the infrastructure and foundational understanding needed to build more capable, generalizable robots.