Multiagents

A captivating multi-agent, self-driving car simulation leveraging a real 4-way intersection from Guadalajara, Jalisco. We modeled an environment using Unity3D and an AI-powered simulation using AgentPy.

We integrated two agent types:

• Traffic lights (2): Each with two different rule sets, a fixed routine and a voting paradigm.
• Cars (40): Able to detect nearby vehicles to adjust their speed and fully react to traffic lights. They also have the probability to turn or keep going straight.

We accomplished this with a customizable Python + AgentPy script that outputs a JSON with all the simulation data.

This is then taken by Unity, which spawns pools of different cars, controls them, and sets the traffic lights.