Dr. Xubo Lyu

I like to build AI-driven applications


About Me

I hold a PhD degree in Computing Science from Simon Fraser University, supervised by Dr. Mo Chen. Currently, I am working as a system engineer at Ma Robot Responsible AI Inc. Previously, I have worked as a machine learning researcher at Huawei Technologies Canada Co. Ltd and as a software engineer at Horizon Robotics Inc.


My expertise lies in machine learning and reinforcement learning, with strong experience in developing and deploying ML systems at scale. I've worked extensively with deep learning frameworks and ML tools including Python, PyTorch, TensorFlow, Openai-baselines, stable-baselines, scikit-learn, numpy and jax.


Additionally, I bring technical depth in robotics and systems engineering, having worked on optimal control theory, ROS, Gazebo, C/C++, and embedded systems development (STM32, ESP32, Pi). I also enjoy doing hardware design like PCB prototype (EasyEDA), CAD (SolidWorks).


Scroll down to see more details or download my CV.

Eduation

Sep. 2018 - Jan. 2024

Msc in Control Science and Engineering at Beihang University.

Sep. 2015 - Apr. 2018

Bsc in Control Science and Engineering at Northeastern University.

Sep. 2011 - Jun. 2015

Work Experience

Machine Learning Researcher, Huawei Technologies Canada Co. Ltd.

Researched advanced multi-robot cooperative system with reinforcement learning. Proposed a novel algorithm for asynchronous multi-agent PPO over temporal extended actions.

May. 2021 - Mar. 2022

Machine learning Engineer, Horizon Robotics Inc.

Built machine learning models for lane line recognition. Developed user-friendly, interactive software for efficient image labelling.

Jul. 2016 - Apr. 2017

Publications

Koopman paper

Task-Oriented Koopman-Based Control with Contrastive Encoder.

Xubo Lyu, Hanyang Hu, Seth Siriya, Ye Pu, Mo Chen.

7th Conference on Robot Learning (CoRL), Atlanta, USA, 2023. Oral Spotlight (6.6% acceptance rate)

PDF | BibTeX | Website
async_mahrl paper

Asynchronous, Option-Based Multi-Agent Policy Gradient: A Conditional Reasoning Approach.

Xubo Lyu, Amin Banitalebi-Dehkordi, Mo Chen, Yong Zhang

IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Detroit, USA, 2023.

PDF | BibTeX | Website
TTR_reward paper

TTR-Based Reward for Reinforcement Learning with Implicit Model Priors.

Xubo Lyu, Mo Chen.

IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Las Vegas, USA, 2020.

PDF | BibTeX | Website
mbb paper

Optimal Control-Based Baseline for Guided Exploration in Policy Gradient Methods.

Xubo Lyu, Site Li, Seth Siriya, Ye Pu, Mo Chen

arXiv preprint arXiv:2011.02073.

PDF | BibTeX

Pet Projects

a linux board for robotic AI vision

aivisionH6 is one of my ongoing project that aims to provide a powerful but portable linux-driven camera board to apply deep learning method for computer vision tasks.

ddbot

ddbot is an AI-driven robot I built mainly for educational purpose to anyone who'd like to play with AI features on a robot. It is designed for running machine/deep learning algorithms for visual, audio and control tasks. I programme it using CircuitPython and design the structure using Solidworks.

fingerprint recognition software

fingertap is a simple fingerprint recognition software I built during my undergraduate. It implements several main machine vision algorithms using Python.

An library for classical RL methods

introRL implements a set of classic RL algorithms based on the book "an-introduction-to-reinforcement-learning" by Rich Sutton. I did it as at that time of 2017, it's not easy to find a easy-to-understand, complete set of algorithms on RL which is of vital help for freshers who want to get into this field. So I implemented this alongside reading that book.

vehicle crossroad navigation

crossNavDRL implements a DQN and DDPG algorithms to solve multi-agent navigation scenarios (mainly crossroads) in simulated environment based on SUMO.

Scholarships and Awards

Academic Service

I served as a Teaching Assistant for serveral undergraduate Computer Science courses, creating Q&A for assignments and exams, grading, holding office hours, leading project group and giving extra talks to provide help for students and professors.


Additionally, I actively volunteer as a paper reviewer for various conferences, including prestigious ones like IROS, ICRA, CoRL, RA-L, and L4DC.

Other Interests