Fengjun Yang

Robotics PhD Student | UPenn GRASP Lab 26’

Philadelphia, Pennsylvania, United States

About

I am a Ph.D. Student in the GRASP Laboratory at the University of Pennsylvania experienced in machine learning, control theory, and robotics. I am graduating in May 2026 and seeking full-time industry positions.

Experience

  • University of Pennsylvania (5 yrs 10 mos)
    • Graduate Research Assistant
      Sep 2020 - Present · 5 yrs 10 mos

      Develop learning-based algorithms for robot planning and control. Projects include: • Residual Dynamics Learning for Quadrotor Proximity Flight: Learned structured residual dynamics from data to improve controller robustness against unmodeled effects. • Learning Controller-Aware Regularizer for Trajectory Planning: Learned regularizers to adapt trajectory planners to a given low-level controller. • Vision Language Model (VLM)-Guided Trajectory Optimization: Led a team to implement a pipeline to synthesize acrobatic maneuvers for quadrupeds by using VLMs to generate trajectory optimization problems iteratively. • Distributed Learning-Based Control using Graph Neural Networks (GNN): Developed and implemented a GNN-based algorithm in PyTorch for co-learning distributed controllers with their communication network; Conducted rigoroustheoretical analysis of the approach on a class of systems with graph symmetry.

    • Teaching Assistant
      Sep 2021 - May 2022 · 9 mos

      ESE 6050 - Convex Optimization ESE 2040 - Decision and Models

  • Research Scientist Intern at Honda Research Institute USA, Inc.
    May 2024 - Aug 2024 · 4 mos

    Designed and trained a Graph Neural Network (GNN) policy via reinforcement learning to solve joint multi-robot task allocation and path planning under communication constraints.

  • Robotics Intern at The Boston Dynamics AI Institute
    May 2023 - Aug 2023 · 4 mos

    - Designed and implemented a ROS2-based software stack integrating large language models with a pre-built map to translate natural language input to navigation commands. - Ran hardware experiments on a Spot quadruped robot.

  • Research Assistant at Stanford University School of Engineering
    Feb 2020 - Aug 2020 · 7 mos

    Algorithms for multi-robot task allocation using game-theoretic methods. Mentor: Prof. Negar Mehr, Prof. Mac Schwager

  • Research Assistant at Stanford University School of Engineering
    Feb 2019 - Jul 2019 · 6 mos

    Algorithms to coordinate robot taxi fleets by combining online model predictive controller with offline reinforcement-learned heuristics. Developed a pruning algorithm to sparsify road networks based on travel demands using network flow optimization. Implemented the algorithm in Python using Gurobi and networkx. Mentor: Dr. Matt Tsao, Dr. Ramon Iglesias, Prof. Marco Pavone