Abstract
This paper compares evolutionary algorithms and reinforcement learning approaches for optimizing traffic signal control in urban environments using the SUMO (Simulation of Urban Mobility) traffic simulator. We evaluate multiple optimization strategies including genetic algorithms, deep Q-learning, and multi-agent RL to minimize average travel time and reduce congestion. Our experiments demonstrate the trade-offs between different approaches in terms of convergence speed, solution quality, and generalization to unseen traffic patterns. [arXiv link TBA]