Guangyao Shi

Guangyao Shi

About me

I am currently a postdoc with Prof. Gaurav S. Sukhatme at USC. I was a Ph.D. candidate at UMD in ECE working with Prof. Pratap Tokekar. I received my M.S. and B.S. in EE at Harbin Institute of Technology. I mainly work on coordination and learning algorithms for multi-robot decision-making under uncertainty with an emphasis on information gathering applications.

Research Vision

My research goal is to identify the underlying computational principles and develope coordination algorithms that enable fleets of robots to sense (e.g., environmental monitoring, target search/tracking, exploration, infrastructure inspection) and reconfigure (e.g., cooperative transportation, construction, large objects assembly/disassembly) their environment in an autonomous manner that is both scalable and near-optimal. To achieve this goal, I have developed new algorithms and frameworks that unify tools and techniques from robotics, AI, operations research, and theoretical computer science communities. Moreover, my research focuses not only on creating coordination algorithms and studying their computational performance, but also on deploying these methods on mobile robots in physical settings. The core of my research focuses on coordinating the motion of multiple robots in large-scale environments—relative to robot size—to achieve collective goals, while accounting for individual robot capabilities (e.g., battery capacity, load capacity, field of view, communication range) and environmental uncertainties. The key contributions of my work include:

  • Multi-robot scientific information collection under uncertainty: planning paths for homogeneous/heterogeneous robots to gather information in dynamic environments under robot capability constraints and environmental uncertainty.

  • Human-in-the-loop multi-robot coordination: developing coordination frameworks using combinatorial optimization and LLMs to enable non-expert humans to interact with robotic teams through direct action feedback or language.

  • Towards a foundational model for multi-robot task planning: constructing datasets for multi-robot task allocation and routing optimization and leveraging it to build machine learning models that can translate natural language into optimization formulations and can accelerate finding the solution of these optimization problems.

Recent news

Honors and awards

  • The Ann G. Wylie Dissertation Fellowship (2023)
  • International Workshop on the Algorithmic Foundations of Robotics (WAFR) Travel Grant (2022)
  • ECE Outstanding TA Award (2019)
  • Clark School of Engineering Distinguished Graduate Fellowship (2017)

Teaching

  • ENEE 621 Estimation and Detection Theory
  • ENEE 307 Electronic Circuits Design