MSR student, Robotics Institute · Carnegie Mellon University
I work on making robots reliable when they physically interact with the world.
I am a Master of Science in Robotics student at the CMU Robotics Institute, co-advised by Guanya Shi and Sebastian Scherer across the LeCAR Lab and the AirLab. My research runs from multi-robot motion planning under connectivity and safety constraints to learning-based control for aerial manipulation.
Before CMU I was at Brown University, where I earned an Sc.B. in Computer Science (Honors, magna cum laude) and an A.B. in International and Public Affairs, and worked with Nora Ayanian on trajectory planning for large robot swarms. Earlier I built autonomous-driving perception data infrastructure at Horizon Robotics.
Yutong Wang, Dongjae Lee, Xiaofeng Guo, Yuanzhu Zhan, Yufei Jiang, Bavin Saravanan, Muqing Cao, Jia Xie, Chenyang Mao, Sebastian Scherer, Junyi Geng, Guanya Shi
Conference on Robot Learning (CoRL) 2026
12 aerial manipulation tasks across underactuated, fully actuated, and overactuated robots — so you can tell whether the embodiment, the controller, or the policy is what failed.
Jaewoo Lee, Jeongyeon Seo, Sihyun Cho, Gyeongrak Choe, Yutong Wang, Bavin Saravanan, Jia-Bin Huang, Furong Huang, Sebastian Scherer, Guanya Shi, H. Jin Kim, Seungjae Lee, Dongjae Lee
Conference on Robot Learning (CoRL), Late-Breaking Results 2026
After each attempt, a coding agent writes its own diagnostic programs: View flies to new vantage points and Probe runs targeted physical tests, so it can find out why an aerial manipulation task failed and fix the task program.
Yutong Wang, Narek Harutyunyan, Sebastian Scherer, Guanya Shi
CoRL Workshop: The UMI Arena 2026
A frozen UMI policy leaves integral action to execution: anchoring each action chunk at the measured pose repeatedly reintroduces steady tracking error, while anchoring at the commanded pose can cause windup.
Neelay Joglekar, Bavin Saravanan, Yutong Wang, Varun Kandiyappan, Junyi Geng, Sebastian Scherer
IROS 2026 Workshop on Sim2Real and Classical Control 2026
Best Methodology Paper Award
Instead of treating a multicopter's downwash as a disturbance, use it as a manipulation tool: an RL policy trained on a simplified downwash model dribbles a ball and transfers to the real world.
Yutong Wang, Lishuo Pan, Yichun Qu, Tengxiang Wang, Nora Ayanian
IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 2026
A real-time MPC–CLF–CBF planner that keeps a fleet radio-connected in cluttered space, and pulls it back together when connectivity breaks. Flown on 8 nano-quadrotors.
Lishuo Pan, Yutong Wang, Nora Ayanian
arXiv preprint 2025
Splitting the workspace and replanning robots in parallel scales deadlock-free trajectory generation to 142 simulated robots and 24 Crazyflies.