I'm trying here to create an AI witch can play tmnf (Trackmania Nation Forever), using Deep Reinforcement Learning.
To put it simply : you have a plugin running with the game (Python_Link.as). This plugin acts like a TCP server and connects to the client (game_instance_manager.py). When on a race, the plugin send to the client some data about the car, and the agent decide what to do.
More in detail later, when the agent will be ready, i'll make a video to explain how it works.
I use the TMI implementation from the Linesight Project, because it's really hard to code and i don't want to spend too much time on this for now. -> all the files in TMI are from Linesight, maybe modified because of updates, and i simplified some things to make the code more readable.