GRAIC
An intelligent racing competition

GRAIC Forum

Ask general questions about GRAIC and interact with fellow participants on our forum!

GRAIC Benchmarks

Here, you will find controllers submitted to GRAIC in previous years. These controllers can be used for controller research and verification. We have included instructions on how to reproduce the results from GRAIC here.

Introduction

Generalized RAcing Intelligence Competition (GRAIC) is a simulated vehicle race. The most recent version of the competition was co-located with CPS-IOT Week 2022. GRAIC aims to bring together researchers in AI, planning, synthesis, and control to create a platform for comparing different algorithms for controlling vehicles in dynamic and uncertain environments. We hope that it will also be fun.

GRAIC ‘22 vs. GRAIC ‘21 This year, we hope to improve GRAIC to encourage wider participation. We now have an AWS image that lowers the hardware entry requirements. Instructions for using the AWS image can be found on our installation page. We also aim to have multi-agent racing and an improved testing pipeline. Please continue to check our webpage for updates!

GRAIC provides a simulation environment, test vehicles, tracks, scoring function, and documentation. As a competitor, you will use the given API and develop your racing controller. In early May, you will submit your racing controller code. The competition tracks will be different from the testing tracks. Multiple vehicles will be involved. We will run the races with your controllers and provide results, data, video feedback, and announce winners during CPSWeek. There will be different race categories and prizes.

Details

At runtime, the input to the controller will come from a perception oracle that will provide as input a local view of obstacles, lanes, and gates on the track near the vehicle. The tracks will have à priori unknown static and moving obstacles. The outputs from the controller (brake, throttle, and steering) will drive the vehicle. In some race categories, you will be provided a mathematical vehicle model, and in other categories you will be provided a black-box vehicle simulator. The perception and control interfaces will not change. Read the Docs page for more details about tracks, obstacles, vehicles, and APIs.

  • New in 2022! Head-to-head races; compete against other submitted constrollers
  • GRAIC focuses on decision, control, planning, and safety, and therefore, we will provide a perception oracle and related API
  • Your controller will run races across multiple vehicles and tracks
  • Different race categories for model-free and model-based vehicles

Contact

Email us at graic2021@gmail.com. Join mailing list form.

Important Dates

Please check back for the next GRAIC competition updates. Following were the dates for 2022 competition.

  • January 26: Single agent platform beta released
  • February 21: Multi-agent platform beta released
  • January - early March: Feedback to participants and platform updates
  • March 31: Final GRAIC-22 platform release
  • April 8-9: Engineering Open House Presentation
  • April 18: Submissions open
  • May 1: Submissions close, final races held
  • May 5, 1-3pm (CT): CPS-IOT Week 2022, final results and live GRAIC event. Zoom link to be published

Updates

Citing GRAIC

Please cite GRAIC as:

@misc{GRAICrace,
      title        = "GRAIC: A simulator framework for autonomous racing",
      author       = "{Minghao Jiang and Zexiang Liu and Kristina Miller and Dawei Sun and Arnab Datta and Yixuan Jia and Sayan Mitra and Necmiye Ozay}",
      howpublished = "\url{https://popgri.github.io/Race/}",
      year         = 2021
    }

@misc{GRAIC-CI-OCAR21,
      title        = "Continuous Integration and Testing for Autonomous Racing Software: An Experience Report from GRAIC",
      author       = "{Minghao Jiang and Kristina Miller and Dawei Sun and Zexiang Liu and Yixuan Jia and Arnab Datta and  Necmiye Ozay and Sayan Mitra}",
      howpublished = "Presented at the Workshop on Opportunities and Challenges in Autonomous Racing (2021) at IEEE ICRA",
      url = "https://par.nsf.gov/servlets/purl/10296575",
      year         = 2021
    }