About

I am Tianyu Yang (杨天宇), a PhD student at Fudan University and the Shanghai AI Laboratory. I received my Master's degree from Shanghai Jiao Tong University and my Bachelor's degree from Xi'an Jiaotong University.

My research journey started in 3D vision and autonomous driving, where I worked on occupancy prediction and scene understanding for driving scenarios. Currently, my research focuses on embodied navigation — developing generalizable navigation policies, in particular post-training navigation diffusion policies that transfer across heterogeneous embodiments and challenging behaviors.

Education

Fudan University & Shanghai AI Laboratory Present
PhD Student · Embodied Navigation & Reinforcement Learning
Shanghai Jiao Tong University M.Sc.
Master · 3D Vision & Autonomous Driving
Xi'an Jiaotong University B.Sc.
Bachelor

Research Interests

Embodied AI Visual Navigation Reinforcement Learning 3D Vision

My research focuses on embodied AI and visual navigation, with interests in deep reinforcement learning, 3D vision, and generalizable robot policies.

Selected Publications

* indicates equal contribution. Links will be updated upon publication.

  1. X-NavDP: Generalizing Navigation Diffusion Policy to Novel Behavior and Embodiments with Group Q-score Reweighted Matching
    CoRL under review · Sep. 2025 - Jun. 2026
    Tianyu Yang, Yiming Zeng, Wenzhe Cai, Yuqiang Yang, Jiaqi Peng, Hui Cheng, Jiangmiao Pang, Tai Wang
    RL post-training for navigation diffusion policies that improves exploration and stability through self-bootstrapped perturbation and group Q-score reweighted matching.
    Project Paper Code
  2. ParkOcc: A Novel Dataset and Benchmark for Surround-view Fisheye 3D Semantic Occupancy Prediction in Parking Scenarios
    IEEE Transactions on Intelligent Transportation Systems · Jan. 2026
    Tianyu Yang, Yeqiang Qian, Chunxiang Wang, Ming Yang
    A surround-view fisheye 3D semantic occupancy dataset and benchmark for automated parking, covering calibration, garage-scale annotations, and fisheye-camera adaptation.
    Project Paper Code
  3. AdaptiveOcc: Adaptive Octree-based Network for Multi-Camera 3D Semantic Occupancy Prediction in Autonomous Driving
    IEEE Transactions on Circuits and Systems for Video Technology · Mar. 2025
    Tianyu Yang, Yeqiang Qian, Weihao Yan, Chunxiang Wang, Ming Yang
    An adaptive octree-based occupancy network that supports finer voxel granularity with faster inference and lower memory usage.
    Project Paper Code