Fengze Xie

Ph.D. Candidate

Computing & Mathematical Sciences, California Institute of Technology

My research bridges model-based and data-driven methods for safe, data-efficient robot control and planning in the real world. I develop algorithms that combine physics-based structure with learned components such as meta-learning, adaptive control, and equivariant neural networks. These methods enable robots to operate reliably across quadrotors, ground vehicles, fixed-wing UAVs, and legged platforms with minimal data and strong safety assurances.

I am co-advised by Yisong Yue and Soon-Jo Chung, working across the Autonomous Robotics and Control Lab and Yue Lab. I received my M.S. in Electrical Engineering from Caltech in 2021, and my B.S. in Computer Engineering and Engineering Physics from the University of Illinois Urbana-Champaign in 2020.

Scholar LinkedIn X CV
Fengze Xie

News

Research

Journal Publications

MAGIC robot platforms and terrain
T-RO Best Paper Nominee, FMNS @ ICRA 2025 Best Poster Runner-up, ROAR @ RSS 2025

MAGICVFM — Meta-learning Adaptation for Ground Interaction Control with Visual Foundation Models

Elena Sorina Lupu*, Fengze Xie*, James A. Preiss, Matthew Anderson, Jedidiah Alindogan, Soon-Jo Chung

An offline meta-learning algorithm to build a residual dynamics and disturbance model using both Visual Foundation Models (VFM) and vehicle states. Integrated with composite adaptive control to adapt to terrain and vehicle dynamics changes in real time, with mathematical guarantees of stability and robustness.

PDF Video
MLA-MPC block diagram
Lemniscate MPC with disturbance Lemniscate MPC NF+FF Lemniscate MPC NF
L-CSS

Meta-Learning Augmented MPC for Disturbance-Aware Motion Planning and Control of Quadrotors

Dženan Lapandić, Fengze Xie, Christos K Verginis, Soon-Jo Chung, Dimos V Dimarogonas, Bo Wahlberg

A disturbance-aware motion planning and control framework for autonomous aerial flights. Combines a predictive control scheme with an online-adapted learned disturbance model and a tracking controller based on contraction methods, providing safety bounds near obstacles.

PDF

Conference Publications

Narwhal algorithm flowchart
ICRA 2026

A Narwhal-Inspired Sensing-to-Control Framework for Small Fixed-Wing Aircraft

Fengze Xie*, Xiaozhou Fan*, Jacob Schuster, Yisong Yue, Morteza Gharib

An end-to-end sensing-to-control pipeline for fixed-wing UAVs combining bio-inspired hardware instrumentation, physics-informed dynamics learning, and convex control allocation. Inspired by the narwhal whale's protruding tusk, we use in-house developed multi-hole probes to measure airflow upstream, reducing force-estimation error by 25-30%.

PDF Website Video
HMAC flowchart
ICRA 2024

Hierarchical Meta-learning-based Adaptive Controller

Fengze Xie, Guanya Shi, Michael O'Connell, Yisong Yue, Soon-Jo Chung

HMAC learns from multiple types of disturbances using Hierarchical Iterative Learning and Smoothed Streaming Meta-Learning. It achieves an average improvement of 26% in tracking performance over Neural-fly on Crazyflie quadrotors with multi-source disturbances.

PDF Website Video
MS-HGNN Architecture
A1 robot simulation Solo robot simulation
L4DC 2025

Morphological-Symmetry-Equivariant Heterogeneous Graph Neural Network for Robotic Dynamics Learning

Fengze Xie*, Sizhe Wei*, Yue Song, Yisong Yue, Lu Gan

MS-HGNN integrates robotic kinematic structures and morphological symmetries into a unified graph network. By embedding structural priors as inductive biases, it ensures high generalizability, sample and model efficiency across various multi-body dynamic systems.

PDF Website Code
Online policy optimization framework
COLT 2024

Online Policy Optimization in Unknown Nonlinear Systems

Yiheng Lin, James A. Preiss, Fengze Xie, Emile Anand, Soon-Jo Chung, Yisong Yue, Adam Wierman

A meta-framework that combines a general online policy optimization algorithm with a general online estimator of the dynamical system's model parameters. We develop Memoryless GAPS (M-GAPS), yielding the first local regret bound for online policy optimization in nonlinear time-varying systems with unknown dynamics.

PDF

Preprints

MS-PPO framework
Preprint

MS-PPO: Morphological-Symmetry-Equivariant Policy for Legged Robot Locomotion

Sizhe Wei, Xulin Chen, Fengze Xie, Garrett Ethan Katz, Zhenyu Gan, Lu Gan

A morphology-informed GNN architecture that is provably equivariant with respect to the robot's morphological symmetry group actions. Achieves superior training stability, symmetry generalization, and sample efficiency on Unitree Go2 and Xiaomi CyberDog2 robots.

PDF Website
Bad initialization recovery
Episodic vs non-episodic comparison M-GAPS on Traxxas car
Preprint

Fast Non-Episodic Adaptive Tuning of Robot Controllers with Model-Based Online Policy Optimization

James A. Preiss, Fengze Xie, Yiheng Lin, Adam Wierman, Yisong Yue

Online algorithms to tune robot controller parameters on a single trajectory without episodes or state resets. M-GAPS finds near-optimal parameters more quickly than episodic baselines and rapidly adapts to heavy unmodeled wind and payload disturbances on both quadrotors and a 1:6-scale Ackermann-steered car.

PDF
Learned RRT overview: data-driven heuristics biasing tree search
Preprint

Joint-Space Multi-Robot Motion Planning with Learned Decentralized Heuristics

Fengze Xie, Marcus Dominguez-Kuhne, Benjamin Riviere, Jialin Song, Wolfgang Hönig, Soon-Jo Chung, Yisong Yue

A method for multi-robot motion planning that biases centralized, sampling-based tree search with decentralized, data-driven steer and distance heuristics. While plain RRT fails for 4+ robots, our method plans for up to 16 robots in a 65-dimensional space, illustrating effective decomposition of high-dimensional joint-space problems.

PDF

Education

Caltech logo
Sep 2021 – Present

California Institute of Technology

Ph.D. in Computing and Mathematical Sciences

Advisors: Yisong Yue, Soon-Jo Chung

Caltech logo
Sep 2020 – Jun 2021

California Institute of Technology

M.S. in Electrical Engineering

Advisor: Yisong Yue

UIUC logo
Aug 2016 – May 2020

University of Illinois, Urbana-Champaign

B.S. in Computer Engineering; B.S. in Engineering Physics

GPA: 3.98 / 4.00

Honors & Awards

Simoudis Discovery Prize, Computing & Mathematical Science, Caltech 2023 – 2024
Dean's List, The Grainger College of Engineering, UIUC 2016 – 2020
Bronze Tablet (Grainger, Top 3%), UIUC 2020
Graduation with Highest Honors, Computer Engineering, UIUC 2020
Graduation with Highest Honors, Engineering Physics, UIUC 2020
Chancellor's Scholar, UIUC 2017 – 2018
Edmund J. James Scholar, The Grainger College of Engineering, UIUC 2017 – 2018

Teaching

Winter 2026

CMS 155 — Machine Learning & Data Mining

Teaching Assistant, Caltech

Lectures: Latent Factor Models [video], Embeddings [video], Hidden Markov Models [video 1] [video 2], Reinforcement Learning [video] [slides]
Spring 2024

CDS 212 — Optimal Control and Estimation

Teaching Assistant, Caltech

Lectures: CVX/CVXPY [slides], Reinforcement Learning [notes]

Service & Mentorship

Service

  • Summer 2024, Caltech, Co-organizer for SURF Program at ARCL lab.
  • Conference Reviewer: IEEE Conference on Decision and Control (CDC), IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Conference on Robot Learning (CoRL).
  • Journal Reviewer: IEEE Robotics and Automation Letters (RA-L), IEEE Control Systems Letters (L-CSS), IEEE Transactions on Automatic Control (TAC).

Mentorship

  • Adam Johansson (Summer 2024, co-mentored with John Lathrop) — “Identifying Constraints From Safe Demonstration Trajectories Using Gaussian Processes.”
  • Leonid Pototskiy (Summer 2024, co-mentored with John Lathrop) — “CrazyPong.”