Central idea: learn a joint-action PPO (MAPPO-style) policy that lets multiple vehicles cross a four-way unsignalized intersection safely and efficiently—outperforming a rule-based four-way-stop baseline.

• Reproducible SUMO ⇄ Gym bridge – custom environment with TraCI, ready for SB3.
• Centralised training with a single policy outputting joint actions for all approaches.
• Rule-based baseline (first-come-first-served) for quick benchmarking.
• Modular reward: collisions, delay, throughput – easy to re-weight or extend to explicit multi-objective RL.