I’m trying to understand the main concepts behind multi-agent systems: communication, coordination, and so on. My goal is to master each topic and apply them in a game.
So I’m building a city of intelligent agents, where agents move around the city and talk to each other. They communicate, try to reach conclusions on different topics, and make decisions based on their own knowledge of the world.
The city will have houses, grocery stores, and other places where agents can make different kinds of decisions. Each agent will be able to trade, talk, and plan events with others, and you’ll be able to see every message they exchange. You’ll also be able to change the city or the agents themselves: adjust someone’s salary, create a new agent, change their personality, and more.
I want to understand how agents make decisions, plan, and coordinate with each other. I’ll study each topic first and then implement it in Python.
In this series, I’ll write about each multi-agent systems topic as I learn it and show how it fits into the city.
If you’re interested in AI agents or simulations, follow along. And if I get something wrong, let me know in the comments.
Top comments (1)
A city simulation is a good way to surface the parts of multi-agent systems that toy chat demos hide: local state, conflicting goals, resource contention, and observability. I would make the agents' contracts and event log inspectable from the beginning, so a surprising city outcome can be traced to a specific decision rather than explained after the fact.