Junfeng Long「龙俊峰」

Hello there! I am Junfeng Long, a researcher focusing on Control, Reinforcement Learning and Robotics. I am currently visiting Hybrid Robotics at UC Berkeley advised by Prof. Koushil Sreenath. Previously, I was a research intern at Shanghai Artificial Intelligence Innovation Center(a.k.a. Shanghai AI Laboratory), where I was working on Reinforcement Learning and Robotics Control advised by Dr. Jiangmiao Pang. I received my Bachelor's degree in Computer Science and Technology from ShanghaiTech University. During my undergraduate, I worked with Prof. Youlong Wu as an undergraduate research assistant on Information Theory and its Application in Distributed Systems.

My research interest mainly falls on Reinforcement Learning, Optimization, Control and Robotics. I am particularly interested in the intersection of machine learning and control theory and apply them to real robotic systems to achieve agile and robust robotic locomotion, manipulation and interaction. My goal is to bring together the strengths of machine learning (generalisability and inference friendliness) and control theory (robustness, theoretical guarantees) to push forward the frontier of robotic systems. Besides, I am also interested in the theory of optimization and learning. I am always open to new ideas and collaborations. If you are interested in my research or have any questions, please drop me an email.

Google Scholar  |  Github  |  Twitter  |  Email 

We proposed GRUtopia, the first simulated interactive 3D society designed for various robots. It features (a)GRScenes, a dataset with 100k interactive and finely annotated scenes. (b) GRResidents, a LLM driven NPC system. (c) GRBench, a benchmark posing moderately challenging tasks.

We present the H-infinity Locomotion Control, an adversarial framework improving the control policy's ability to resist external disturbances with H-infinity performance guarantee.

We propose TOP-Nav, a novel legged navigation framework that integrates a comprehensive path planner with Terrain awareness, Obstacle avoidance and close-loop Proprioception.

We present the Hybrid Internal Model, a method enabling the control policy to estimate environmental disturbances by only explicitly estimating velocity and implicitly simulating the system's response.

We propose a coded scheme to reduce the communication latency by exploiting computation and communication capabilities of all nodes and creating coded multicast opportunities.More importantly, we prove that the proposed scheme is always optimal, i.e., achieving the minimum communication latency, for arbitrary computing and storage abilities at the master.

Updated at November 2024.
Template from Jon Barron.