Hiveminds
Hiveminds are now an official part of RLBot in v5. To tell v5 that your bot is a hivemind, set hivemind = true under the [settings] header in your bot.toml file.
Example bot.toml for a hivemind
[settings]
name = "Hives"
loadout_file = "loadout.toml"
run_command = "python bot.py"
run_command_linux = "./venv/bin/python bot.py"
agent_id = "rlbot-community/hives"
# must set to true if you want the bot to be ran as a hivemind
hivemind = true
[details]
description = "Made possible by RLBot"
fun_fact = "This is a test bot"
source_link = "https://github.com/RLBot/core"
developer = "RLBot community"
tags = []
See Config Files for details on the bot.toml file.
What is a hivemind?
By default, RLBot launches 1 process per bot. This is fine for most bots, but it can be inefficient for bots that are designed to work together. Hiveminds fix this by assigning all bots of the same type to the same process, if they're on the same team. This means that the same bot on different teams will still be controlled different processes.
Basically, a hivemind is 1 process that controls multiple bots!
How RLBot groups bots into hiveminds
RLBot specifically uses agent_ids to group bots together, which is why it's important that your bot's agent_id is unique!
Types of hiveminds
There are 2 common ways to implement a hivemind:
- Single-threaded hivemind: The hivemind runs on a single thread, and each bot is controlled sequentially. This is the simplest way to implement a hivemind.
-
Multithreaded hivemind: The hivemind runs on multiple threads, and each bot is controlled concurrently. This is more complex, but can be more efficient.
Multithreaded hiveminds are not supported in Python
Due to current limitations of the Python language, Python can't do multithreading. If you're curious why, learn more about the Global Interpreter Lock!