Optimizing an agent-based traffic evacuation model using genetic algorithms

Computer simulations are commonly used to model emergencies and discover useful evacuation strategies. The top-down conceptual models typically used for such simulations do not account for differences in individual behavior and how they affect other individuals. To create a more realistic model, this study uses Agent-Based Modeling (ABM) to simulate the evacuation of an urban population in case of a chlorine spill. Since the agents (each a car and driver) in this model do not behave uniformly, and the initial traffic and spill locations are randomized, optimizing traffic lights is challenging. A commercial evolutionary optimizer controls execution of the simulator, seeking to optimize the control of traffic lights in order to minimize deaths and injuries. ABM for a traffic evacuation could prove useful in the real world, when the threat is at a known location such as a power plant or a specific railway segment.