Showcases

Live Track Videos

In the battle for 3rd place in the Indy Autonomous Challenge at Las Vegas Motor Speedway 2023, we execute an overtake at 140 MPH, against MIT and University of Pittsburgh.
On August 31 2023 AI Racing Tech demonstrated advanced model-predictive control on Putnam Road Course topping 140 MPH, with 1:27 lap time, 90% of theoretical optimum. This marks an important milestone on our journey to pushing the 100% limit. I designed the MPC from ground up and the code can be found here: https://github.com/HaoruXue/Racing-LMPC-ROS2 (forked in Berkeley MPC Lab here).
High-speed track session at Texas Motor Speedway at 120 MPH (October 2022).
LGSVL simulation testing, integrating the entire perception, decision-making, motion-planning and control components of our software running in the Indy Autonomous Challenge.

Software Demos

A early version of our trajectory optimization tools, offering velocity profile generation based on the desired limit of vehicle dynamics. GitHub
Visualization of an overtaking test session with TUM at 50 mph. The decision and trajectory planning can be visualized on the left. The 3D perception result is shown in the middle.

Other Open-Source Projects

High-Level Autonomous Vehicle Software

I was a software developer for the world’s biggest open-source autonomous driving project in ROS2 – Autoware. I managed the racing ODD aprplications and deployed Autoware on autonomous race cars.

GitHub – Autoware.Universe | GitLab – Autoware.Auto

I also open-source my research project in model-predictive control and trajectory optimization. The code is C++ and deployment-ready. We’ve demonstrated racing at 140 MPH+ in the video above.

Racing-LMPC-ROS2 | spline-trajectory-optimization

Robotics Hardware Driver

I’m also an active ROS2 community package contributor for sensor drivers. For example, I developed the ROS2 driver for VESC, a popular BLDC motor controller for robotics, and Sparkfun Artemis IMU.

GitHub – VESC ROS2 Driver | GitHub – Razor IMU C++ ROS2 Driver

End-To-End Autonomous Driving on Embedded Systems

A while back I worked on an end-to-end deep learning platform for scaled autonomous RC cars, with some interesting neural network options with RNN and LSTM. The platform is suited to run on embedded Linux computers such as the Nvidia Jetson series.

GitHub – Triton-Racer

Low-Level Real-Time Controllers

Supporting the autonomous racing project on Go-Kart platforms, I developed micro-controller software with ARM Mbed, which acts as an ECU for the drive-by-wire system. The system is used in the EV Grand Prix autonomous go-kart series which is a testing ground for AV technology on full-size race cars.

GitHub – Go-Kart-RTC