DEPRECATED in favour of Minecraft implementation.
A simple game designed to train reinforcement learning agents. The server exports endpoint for an actor to join the game. The input to this endpoint is actors UDP address on which they listen to game updates.
An actor can send payload to another endpoint with updates to their direction vectors.
The server re-calculates the position of entities based on its frame rate. With each frame it sends new state to all players. The new state means a list of velocity vectors of all entities in the game with their label.
Built with bevy. Bevy uses LLD linker to speed up the linking process.
See the project installation information to find out more. If you
don't wish to use the LLD linker you might need to delete or rename the
.cargo/config
file.
You will need the nightly compiler. You can override it on per project basis so that your other projects aren't impacted:
$ rustup override set nightly
- arrows control player's predator
- +/- to zoom