Gengshan Yang

Papers

Total Citations

346

H-Index

About

Gengshan Yang is pushing the boundaries of embodied AI, with key research spanning dense 3D perception and legged robot motion imitation. His most impactful contribution is **SplaTAM** (2024, 323 citations), a pioneering framework that introduces explicit 3D Gaussian representations for dense RGB-D SLAM. By replacing implicit neural fields with a volumetric, differentiable scene representation, SplaTAM achieves state-of-the-art performance in simultaneous localization and mapping—a critical capability for robotics and augmented reality. This work fundamentally addresses the limitations of prior non-volumetric methods, enabling faster, more accurate dense reconstruction. Yang also tackles the challenge of robot skill acquisition with **SLoMo**, a novel system that transfers skilled motions from casual, “in-the-wild” videos of humans and animals to legged robots. By synthesizing physically plausible key-point trajectories from monocular footage, SLoMo bridges the gap between unstructured visual data and robust robot control. His work demonstrates a rare ability to connect high-level vision with low-level dynamics, producing practical solutions for autonomous systems. With a growing citation footprint and innovations that directly advance both perception and locomotion, Gengshan Yang is a rising leader in the quest for robots that see and move with human-like fluency.