Building a level 5 autonomous vehicle (in TrackMania) - Manu Gopinathan & Malte Loller-Andersen

In games such as chess and go, mankind’s best players have seen themselves beat by machine learning. As the decade came to an end we were also closer than ever before to fully autonomous vehicles, and they are being developed using this very same technology. In this session we will present to you our self-driving vehicle trained using reinforcement learning. The agent is self-taught, without the use of human input. By using cutting edge technology, we have surpassed the likes of Google and Tesla to solve self-driving vehicles - at least in a video game they didn’t try to solve: TrackMania, a popular racing game. We will present the different parts of our pipeline, ranging from the features that are used to how the car performs actions and the model training itself, which was done using OpenAIs state-of-the-art algorithm called proximal policy optimization. Most importantly, you will get to see the AI in action. Whether you have prior experience with machine learning or not, we believe you will find the presentation interesting. The talk will contain technical aspects, but they will be presented and explained in a way that makes it possible for beginners and non-technical audience to follow. Perhaps you’re just a curious TrackMania veteran? We will try to make justice to the game you love. Join us for an hour of fun and entertainment. As a bonus, you might learn something as well! Check out more of our featured speakers and talks at https://ndcconferences.com/ https://ndcoslo.com/