AI Grand-Prix
After reaching the F1Tenth Grand-Prix back in October 2020, the founders of Formula Trinity’s driverless project have decided to host their own virtual autonomous racing tournament!
The AI Grand-Prix is based on the University of Pennsylvania’s F1Tenth framework and is open to all Irish university students. You can compete as a team of up to 5 people or submit a solo entry. All contestants will work to improve a baseline solution that has been provided by us. Everything you need to get started is available on this repo that we have made in collaboration with riders.ai. Your solutions should be submitted before June 3rd and will then be run on our virtual track and streamed for everyone to watch on race day – June 5th.
The tournament will involve both time trials and multi-agent racing, you can find out more about what’s involved under the ‘learning resources’ tab on this webpage.
There’ll be lots of great prizes to be won so make sure to fill out the ‘sign up’ form and we will email you with further details!
We held our opening ceremony on May 1st, but don’t worry if you missed it because the slides are available below.
Getting Started
Join our discord server so you don’t miss any updates!
Submissions opened on May 24th and you’ll find everything you need on this Quickstart Repo including installation instructions, entering the competition on the submission platform, developing your driver, a baseline solution, preparing a submission, testing a submission, and submitting your final solution.
Check out the tutorial below to find how to get started on your project using the QuickStart repo…
Python
One of the prerequisites for our competition is basic Python programming. If you don’t have experience with Python, here’s a Crash Course to get started!
NumPy
Numpy is a library used throughout AI and data science which makes it easier for programmers to operate on large groups of numbers and much more. Here’s a Crash Course.
Racing Methods & Techniques
The F1Tenth Developers in UPenn have a series of lectures on YouTube where they go through different methods & techniques for autonomous racing. Here’s some of the videos we think might be useful:
- Reactive Methods for Planning: in a reactive method the car doesn’t have any memory, it makes decisions based on its current surroundings.
- Pure Pursuit: pure pursuit is a method used to make a car drive through a set of waypoints.
- Particle Filter Method: an interesting way to make your decisions more probabilistic!
- Raceline Optimization: figure out how to make your racelines as fast and pretty as possible!
- Control Methods: if we choose a path/trajectory to follow, we need to ensure we can follow it accurately! A PID Controller is a simple way to do this. Model Predictive Control is a more complex method.
- Localization and Mapping: it is important for an autonomous vehicle to map its surroundings and determine where it is within that map. There are two methods covered: Scan-matching and Scan-matching with Fast Correspondence Search.
- Automatic Emergency Braking: although making a braking system might initially seem boring, if your car crashes you could lose the race!
(Note: as these lectures are from a university course, there may be housekeeping at the start of some of the videos. This can be ignored. Also, mentions of ROS and Docker can be ignored. We will discuss our submission format in the Opening Ceremony.)