Sichang Su Destiny

PhD student

The University of Texas at Austin

Biography

I am a first-year Ph.D. student in Aerospace Engineering at The University of Texas at Austin, co-advised by Dr. Thinh Doan and Dr. Ufuk Topcu. I obtained my Master of Science degree in Mechanical Engineering from the National University of Singapore, under the supervision of Dr. Guillaume Adrien Sartoretti. Additionally, I am a visiting scholar at the University of Iowa, working under Professor Shaoping Xiao. I earned my Bachelor’s Degree in Engineering Mechanics from Zhejiang University.

My research interests lie in Reinforcement Learning and Robotics, with a recent focus on developing robotic generalist policies. My long-term goal is to enable agents and robots to efficiently solve complex, diverse tasks in real-world applications.

Please contact me at sichang_su@utexas.edu if you are interested in discussing relevant research topics or potential collaborations.

Interests

Reinforcement Learning
Robotics
VLA/VLM

Education

PhD in Aerospace Engineering, 2025-present
The University of Texas at Austin
MSc in Mechanical Engineering, 2023-2025
National University of Singapore
BEng in Engineering Mechanics, 2019-2023
Zhejiang University

Recent News

(Sept 2025) One co-authored paper, Reinflow (https://reinflow.github.io), has been accepted to NeurIPS, and one co-first-authored paper, HyGen (https://arxiv.org/abs/2408.13567), has been accepted to the NeurIPS 2025 Workshop@ARLET.

Sep 23, 2025 0 min read

(Oct 2024) I was invited to be a reviewer for ICLR 2025 and NeurIPS 2024 Workshop SafeGenAi.

Oct 5, 2024 0 min read

Projects

Enhancing Multi-task Generalization in Multi-agent Reinforcement Learning

In multi-agent reinforcement learning (MARL), achieving multi-task generalization to diverse agents and objectives presents significant challenges. Existing online MARL algorithms primarily focus on single-task performance, but their lack of multi-task generalization capabilities typically results in substantial computational waste and limited real-life applicability. Meanwhile, existing offline multi-task MARL approaches are heavily dependent on data quality, often resulting in poor performance on unseen tasks. Therefore, we want to develop effective Multi-task MARL methods that can achieve generalization across different tasks with little training costs. One related paper is HyGen.

Efficient Reinforcement Learning for Solving Multiple Tasks

Reinforcement Learning (RL) has gained significant attention for its potential to solve complex problems in areas such as robotics, sensor networks, and gaming. However, most recent RL approaches focus primarily on learning policies for a single task in simulated environments. This focus can create substantial challenges when transitioning to real-world applications, as the high costs and risks associated with interactions under partially learned policies can be prohibitive. Consequently, enabling agents to solve multiple tasks within a single training process is crucial for the practical deployment of RL in real-world scenarios.