Philip (Yizhou) Huang

I am a Ph.D. student in the Robotics Institute at Carnegie Mellon University, supervised by Prof. Jiaoyang Li . I completed my MSc in Computer Science at the University of Toronto, co-advised by Prof. Florian Shkurti and Prof. Tim Barfoot. I received a Bachelor of Applied Science in Engineering Science from the University of Toronto.

I have been fortunate to work under many fantastic researchers in robotics. Previously, I collabrated with Prof. Fabio Ramos from Nvidia on task and motion planning. In undergrad, I was involved in the student design team aUToronto, where we competed and won the SAE Autodrive Challenge. I also spent one amazing summer in Israel with Prof. Amir Degani and the summer before that with Prof. Angela Schoellig.

Email  /  CV  /  Google Scholar  /  LinkedIn  /  Twitter  /  Github

Research

My research focuses on developing planning algorithms for robot and robot teams to accomplish long-horizon tasks in dynamic and uncertain environment. I am particuarly interested in solving problems in real-world robotics and have worked on many different applications, such as multi-robot assembly, autonomous surface vessels (ASV) for environmental monitoring, and self-driving cars. I have also worked on continual learning for robot dynamics modeling and control.

If you have any questions / want to collaborate, feel free to reach out and send me an email! I am very excited to talk with more people and learn about your work!

Publications
Field Testing of a Stochastic Planning for ASV Navigation Using Satellite Images
Philip Huang, Tony Wang, Florian Shkurti, Timothy D. Barfoot,
IEEE Transactions on Field Robotics (T-FR), 2024
doi / arXiv / Video /

Field testing and system descriptions of our GPS-, vision-, and sonar-enabled autonomous vessel for water-quality monitoring in freshwater lakes.

STAMP: Differentiable Task and Motion Planning via Stein Variational Gradient Descent
Yewon Lee, Andrew Z. Li, Philip Huang, Eric Heiden, Krishna Murthy Jatavallabhula, Fabian Damken, Kevin Smith, Derek Nowrouzezahrai, Fabio Ramos, Florian Shkurti
Submitted to ICRA 2025
Website / Preprint / Video /

Leveraging parallel and differentiable simulation to efficiently search for multiple diverse plans with gradient-based variational inference.

Stochastic Planning for ASV Navigation Using Satellite Images
Yizhou Huang, Hamza Dugmag, Timothy D. Barfoot, Florian Shkurti
ICRA, 2023
Website / arXiv / Code / Video /

A robust route-planning algorithm uses satellite images as a corase map to plan water sampling routes for autonomous surface vessels (ASV) given environmental disturbances.

Continual Model-Based Reinforcement Learning with Hypernetworks
Yizhou Huang, Kevin Xie, Homanga Bharadhwaj, Florian Shkurti
ICRA, 2021 and Deep RL Workshop (NeurIPS 20)  
Website / arXiv / Video / Code

Task-conditioned hypernetworks can be used to continually adapt to varying environment dynamics in lifelong model-based reinforcement learning, with a fixed-size replay buffer.

Zeus: A system description of the two-time winner of the collegiate SAE autodrive competition.
Keenan Burnett, Jingxing Qian, Xintong Du, Linqiao Liu, David J. Yoon, Tianchang Shen, Susan Sun, Sepehr Samavi, Michael J. Sorocky, Mollie Bianchi, Kaicheng Zhang, Arkady Arkhangorodsky, Quinlan Sykora, Shichen Lu, Yizhou Huang, Angela Schoellig, Timothy D. Barfoot,
Journal of Field Robotics, 2021  
arXiv / Video

System design and development of the winning self-driving car in the AutoDrive Challenge, as well as lessons learned.

Theses
Improving Regularization-based Continual Learning with Hypernetworks
Yizhou Huang
Bachelor Thesis
Supervisor: Florian Shkurti
Thesis pdf/ Presentation

In continual learning, self-tuning network (STN) is a memory-efficient and easy-to-implement hypernetworks architecture with strong performance on many benchmarks.

Projects
Learning Heuristics for Minimum Latency Problem with RL and GNN
Course Project
University of Toronto, MIE 1666 Machine Learning for Mathematical Optimization
Final Report/ Presentation/ Code

RL can be applied to the Minimum Latency Problem by using a graph attention network to encode stochastic policy for constructively building partial paths, yielding solutions which are comparable to state-of-the-art, hand-engineered methods.

Linear-Time Dynamics using Lagrange Multipliers
Course Project
University of Toronto, CSC 2549 Physics-based Animation
Final Report/ Video/ Code

Re-implemented the SIGGRAPH 96 paper from David Baraff that introduced a linear-time sparse solver with Lagrangian multiplers in MATLAB. Verified that the time complexity of our re-implemented sparse solver is linear with serial chains and trees.

Evaluating Different Methods of Text Style Transfer
Course Project
University of Toronto, CSC 413 Neural Networks and Deep Learning
Final Report/ Code (Style Transformer)/ Code (ALE Model)

We measured the quantitative performance of recent language models for text style transfer, using three metrics and third-party models for fair comparison.

Hawkeye
Course Project
University of Toronto, MIE 324 Introduction to Machine Intellitenge
Final Report/ Code (>250 Stars)

Re-implementation of the paper PIXOR: Real-time 3D object Detection from Point Clouds using PyTorch.

Indoor Localization of a Crazyflie using a Ping-Pong Ball
Summer Research Project
CEAR Lab, Technion
Demo Video

Indoor localization and navigation of a Crazyflie nano-quadcopter with a RGB-D camera for pollinating sunflowers.

Controlling a Swarm of Crazyflies
Summer Research Project
Dynamic Systems Lab, University of Toronto
Demo Video/Technical Report

Designing software framework and simulations for flying a swarm of 9 Crazyflie nano-quadcopters indoors.

Misc
Reviewer, IROS 2022, 2023
Reviewer, ICRA 2023, 2024
Reviewer, MetaLearn Workshop NeurIPS 2020
Graduate Student Mentor, PRISM Workshop 2022
csc477 Teaching Assistant at University of Toronto
CSC477 Fall 2021,
CSC317 Fall 2022,
CSC384 Winter 2023

This template is from Jon Barron's website. Here is the source code