Max Robotics — Global Service Robot Directory + US Certification

About

Yunhai Feng is a robotics researcher advancing the frontier of sample-efficient robot learning through model-based reinforcement learning (MBRL) and physical interaction. His work centers on three interconnected challenges: enabling robots to learn from raw sensory inputs with minimal real-world data, developing novel gripper designs for dexterous manipulation, and bridging the simulation-to-reality gap. Feng’s most impactful contribution is the SAM-RL framework (Sensing-Aware Model-Based RL), which integrates differentiable physics simulation and rendering to automatically build accurate world models from images—a critical step toward practical, data-efficient robot learning. This work, accumulating over 15 citations across multiple versions, addresses the long-standing bottleneck of sample efficiency in complex environments. His complementary research on visual pre-training for robot manipulation (7 citations) systematically investigates how large-scale real-world datasets can bootstrap pixel-based policies. On the hardware side, Feng’s Roller Grasper V3 (6 citations) introduces a non-anthropomorphic design with steerable rollers on each fingertip, enabling sophisticated in-hand manipulation—a skill essential for tasks requiring object reorientation. His work on finetuning offline world models in the real world further tackles the practical challenge of deploying pretrained models on physical robots. Through this synthesis of algorithmic innovation and mechanical design, Feng is helping to make robots that learn faster, manipulate more dexterously, and transfer seamlessly from simulation to reality.