03Reinforcement Learning2026Completed
Smart City Traffic Management
Multi-Agent Reinforcement Learning System
16
Intersections
60k
Training Steps
3
Protocols
Overview
A multi-agent reinforcement learning system that coordinates traffic signals across a 16-intersection grid. Decentralized agents learn adaptive policies, a coordination layer handles edge cases, and an explainable dashboard makes every decision auditable.
Highlights
- Decentralized Q-Learning and SARSA agents trained over 60,000 steps with 6-factor reward functions
- 3-protocol Coordination Layer with Chaos Mode stress-testing sandstorm conditions and emergency preemption
- Real-time React 18 dashboard with live grid animations, Q-value XAI panels, and TOPSIS multi-criteria metrics