In recent years, many emerging problems with urban transportation system, including traffic congestion, have been direct results of the steadily increasing number of commuters in urban areas. To address these problems, many research projects have studied the behaviors of urban traffic systems, and researchers have proposed new and innovative traffic control schemes. These studies require accurate and realistic software models for simulating behaviors of urban traffic systems since using a real traffic network as a testbed is often impractical. In this paper, we describe a novel software simulator for urban traffic systems. Our simulator has several desirable properties. First, it gives an accurate and realistic representation of important elements of urban traffic networks. Second, it is specially designed for game-theoretic modeling of traffic systems in order to address potentially different research questions. Lastly, it has a flexible framework for generating customized network topologies and running customized simulations. In particular, our simulator is particularly useful for studying long term driver behavior through the application of multiagent reinforcement learning (MARL) algorithms. In this paper, we will first describe the urban traffic model used by our software simulator. Specifically, we will describe important components of real urban traffic systems and argue that the corresponding components of our model are reasonable and meaningful approximations of their real counterparts. Secondly, we will describe the components of our model that allow game-theoretic modeling of urban traffic systems in order to address various research questions. Finally, we will demonstrate how our model is particularly useful for studying long-term driver behavior, especially during rush hour.