Close

Tao Chen

Robotics Software Engineer

Download Resume

About Me

I am a software engineer specialized in robotics. I enjoy programming and building robots. I like hands-on projects, enjoy working in teams and collaborating with others. I have extensive experiences in research & software development, with machine learning, control, and estimation.

Currently I am working at an autonomous driving startup. My main job is to develop and evaluate sensor fusion algorithms for multi-object tracking on our production vehicles. I get to see self-driving cars in action every day \^_^/ cool～～ I was also the main developer on an internal real-time visualization tool that replaced Rviz within our company.

Previously I worked at the Robotics Embedded System Lab at the University of Southern California. As a student researcher, my focus was on quadrotor controls combined with machine learning. We were approaching quadrotor control with reinforcement learning to learn a neural network that was capable of low-level, safe, and robust control of quadrotors.

Recent News

Southern California Robotics Symposium 2019

April 27, 2019

I was selected to present a spotlight talk for my recent project:)

University of Southern California

March, 2019

I won the Master's Best Research Award 2019 \o/

Experience

Xsense.ai

Software Engineer

I joined Xsense.ai in December of 2019. Xsense.ai is a startup that develops autonomous driving technologies. It's part of an EV startup in China, named Xpeng Motor, that already has two vehicles on sale in China. I joined as a software engineer to develop a real-time visualization tool to replace Rviz within the company. Later I was transferred to the sensor fusion team to work on algorithm development, and I have been enjoying it.

Robotics Embeded System Lab

Student Researcher

I joined the Robotics Embedded System Lab in May 2018, where I worked with Artem Molchanov on quadrotor controls. We are trying to apply reinforcement learning techniques on developing quadrotor controller. We have sucessfully applied a trained neural network on quadrotors that showed amazing robustness. Our latest paper "Sim-to-(Multi)-Real: Transfer of Low-Level Robust Control Policies to Multiple Quadrotors" was submitted to IROS 2019.

Dynamics Robotics Lab

Volunteer

I participated in the development of the bi-paddle robot Cassie at the Dynamics Robotics Lab in summer 2016. This was a volunteer position but I really enjoyed it. I worked with another undergraduate student on the telemetry protocol for information to be sent to the remote and display the robot's status on the remote screen, e.g. robot pose, temperature, etc.

Media Service

Student Worker

This is my job when I was an undergraduate student at OSU. I worked here for almost 2 years. I mostly helped professors with technical issues in classrooms and maintained all the presentation devices making sure that they were working properly. This job really helped me improve my communication skills and taught me how to be patient. I made a lot of friends in this job. I really enjoy the 2 years spent there.

Education

University of Southern California

August 2017 - June 2019

Master of Science in Computer Science

Oregon State University

Sept 2014 - June 2017

Bachelor of Science in Computer Science

Projects

Sim-to-(Multi)-Real: Transfer of Low-Level Robust Control Policies to Multiple Quadrotors

Quadrotor stabilizing controllers often require careful, model-specific tuning for safe operation. We use reinforcement learning to train policies in simulation that transfer remarkably well to multiple physical quadrotors. The trained control policies are very robust to external disturbances. To the best of our knowledge, this is the first work that demonstrates that a simple neural network can learn a robust stabilizing quadrotor controller without the use of a stabilizing PD controller; as well as the first work that analyses the transfer capability of a single policy to multiple quadrotors.

View Project

Autonomous RC

The purpose of ARC is to determine if it is possible to build an autonomous RC vehicle using commodity hardware. We will try to use relatively inexpensive parts, including cameras, computers, and an RC vehicle used as the base, with minimal fabrication. All of the software used is open source, allowing users to tinker and adapt the project to their needs. We found that the Robotics Operating System (ROS) provides a good base to build from and, consequently, the platforms we chose are implemented on ROS.

View Project

Stock Quote Web App and Android App

- Featured real-time stock price automatic update and user-defined watch list.

- Developed with responsive web design.

- Used Node.js on AWS Beanstalk as a portal for RESI APIs.

- Used AngularJS as JavaScript framework.

- Used HIGHCHARTS to show indicator data and historic data.

- Implemented sreenshot posting to Facebookusing Facebook APIs for web and Android.

Kaggle Competition

- Identify if 2 questions on Quora are asking the same problem

- Developed a machine learning algorithm to tell if 2 questions are simular in meaning.

- Experimented with word filtering, logistic regression, categorization, and NLP methods.

- Final result was ranked top 20%.

View Project

Skills

Get in Touch