Towards a realistic optimization of urban traffic flows

In spite of recent advances in Intelligent Transport, vehicular traffic dynamics are still hard to represent and analyze. Most of the previous work on traffic regards highways or single lanes where vehicles interact in one dimension. Models for multi-dimensional vehicle-to-vehicle interactions and models for urban intersections are quite complicated and hardly applicable on a large scale. Nonetheless, urban traffic jams are an actual problem that requires a solution. This paper proposes a method to optimize urban traffic layout using basic heuristics and computationally efficient simulations. Instead of modeling an entire urban map with hundreds of intersections, each typology of intersection is simulated in order to understand how it responds to different traffic patterns and intensities. This knowledge is leveraged to allow the computation of minimal delay route on the complete road map. In order to validate our model, we use the solution obtained with our heuristic to derive the average travel delay through simulation on realistic Manhattan topologies with different intersection types.