Hi there 👋 I'm Feiyang Ma, a MSCS student at UC San Diego.

I'm interested in building autonomous system that can interact with real world effectively and safely. Previously, I have worked on combing flow matching model and RL algorithm to improve trajectory quality with a learned forward network. I also explored equipping diffusion policy with Recurrent State Space Model, aiming to mitigate distribution shift problem of robotic tasks in offline learning. I'm currently working on self supervised Multimodal alignment between 2D image tokens and 3D point tokens in autonomous driving scenario.