My goal is to systematically develop next-generation decision-making framework that allows robots to unifies planning and abstractions learned from the environment, and further enable multi-robot collaboration, safe human-robot interaction. This vision was gradually formed during my undergraduate research, which involved four projects that I led or co-led at four labs. Roughly, they are structured into three components: Planning, Robot Learning and Multi-Robot Systems.

To be more specific, my ongoing projects respectively focus on learning generalizable policies across diverse robotic embodiments (CMU Safe AI Lab) and designing collaborative multi-robot Task and Motion Planning (TAMP) methods (CMU ARCS Lab). My previous projects focused on search-based multi-robot motion planning (SJTU RAP Lab & UCI IDM Lab), bridging the gap between planning and real-world execution with algorithms that account for practical constraints (CMU ARCS Lab). Some of my works on handling agents with heterogeneous speeds are published (AAAI 2025, SoCS 2025), while others on solving time-sensitive tasks with kinematic constraints are under review. I enjoy how planning explores the state space with human-prior intelligence, as well as the learning strategies for their efficiency in indescribable modeling. However, they respectively fail in certain scenarios, such as handling unknown worlds, long-term exploration and reasoning, and collaboration across embodiments.

I prefer slow science. I intend to pursue this line of research throughout my Ph.D. and into my future career, ultimately leading a laboratory devoted to developing principled planning frameworks for assistive robotics to help people with disabilities. To achieve this, I plan to draw upon prior work in planning, multi-agent systems, robot learning, cognitive science, and human–robot interaction. I’m not an expert on everything and I’m always excited to learn. Please drop me an email if you’re interested，we could collaborate on some exciting projects!