Max Robotics — Global Service Robot Directory + US Certification

About

Haoran Geng is a rising researcher at the intersection of embodied AI, robotic manipulation, and 3D object understanding. His work focuses on enabling autonomous robotic systems to generalize across diverse environments and tasks — a fundamental challenge in modern robotics. Geng's most influential contribution, **ManipLLM** (2024, 65 citations), advances object-centric robotic manipulation by harnessing multimodal large language models to predict contact points and end-effector directions, significantly improving generalizability beyond simulator-trained categories. This work positions him at the forefront of integrating large language models into physical robotic systems. His follow-up work, **Ag2Manip** (2024, 12 citations), tackles the persistent challenge of cross-embodiment transfer by introducing agent-agnostic visual and action representations, allowing robots to learn novel manipulation skills despite differences in hardware and reward sparsity. More recently, **PhysPart** (2025) extends his research into physically plausible 3D part completion for interactable objects, bridging generative modeling with practical robotics simulation and fabrication. Together, these contributions reflect Geng's commitment to building robust, generalizable robotic intelligence — making him a compelling voice in the next generation of embodied AI research.