Fangyu Sun(孙方宇)

Research interests: reinforcement learning, sim-to-real system, and end-to-end robotic control.

Biography

I am Fangyu Sun (孙方宇-in Chinese), currently a direct Ph.D. student at the School of Automation and Sensing, Shanghai Jiao Tong University (上海交通大学自动化与感知学院). My research interests include reinforcement learning, differentiable simulation, and end-to-end robotic control. I am advised by Prof. Danping Zou, and I am a member of the ViSYS Lab at SJTU.

News!

Research

Curriculum Reinforcement Learning for Quadrotor Racing with Random Obstacles

Fangyu Sun, Fanxing Li, Yu Hu, Linzuo Zhang, Yueqian Liu, Wenxian Yu*, Danping Zou*,

ICRA 2026

Curriculum reinforcement learning for end-to-end agile quadrotor flight in complex obstacle courses, addressing visual perception and dynamics robustness challenges.

paper / code

E2E-Fly: A Comprehensive End-to-End Learning-based Framework for Quadrotor Bridging the Sim-to-Real Gap

Fangyu Sun, Fanxing Li, Linzuo Zhang, Yu Hu, Renbiao Jin, Shuyu Wu, Wenxian Yu*, Danping Zou*

In Submission

A systematic study of end-to-end quadrotor policy learning from simulation to real-world deployment.

paper / project

Ask to Land: Learning an End-to-end Open-vocabulary Landing Policy via Differentiable Physics

Fangyu Sun, Yang Deng, Jiahao Cui, Fanxing Li, Guanyan Wu, Yuxiang Huang, Wenxian Yu*, Danping Zou*

In Submission

Open-vocabulary autonomous quadrotor landing with a large-model front end and differentiable simulation.

paper / code

Vector Field Augmented Differentiable Policy Learning for Vision-Based Drone Racing

Yang Su, Feng Yu, Yu Hu, Xinze Niu, Linzuo Zhang, Fangyu Sun, Danping Zou*

IEEE RAL 2026

A novel approach to differentiable policy learning for vision-based drone racing, incorporating vector field information to enhance control performance.

paper

VisFly: An efficient and versatile simulator for training vision-based flight

Fanxing Li, Fangyu Sun, Tianbao Zhang, Danping Zou*

ICRA 2025

A novel simulator for training vision-based flight policies, offering efficient and versatile capabilities for real-world deployment.

paper / code

Cross-domain active learning for electronic nose drift compensation

Fangyu Sun, Ruihong Sun, Jia Yan*

Micromachines 2022

A machine learning approach for compensating drift in electronic nose sensors, utilizing cross-domain active learning techniques to enhance performance and reliability.

paper

Competitions

Awards

Service

Resume

CV / Resume