Cellular Automata for Traffic Flow Modeling

Abstract

In this paper, we explore the usefulness of cellular automata to traffic flow modeling. We extend some of the existing CA models to capture characteristics of traffic flow that have not been possible to model using either conventional analytical models or existing simulation techniques. In particular, we examine higher moments of traffic flow and evaluate their effect on overall traffic performance. The behavior of these higher moments is found to be surprising, somewhat counter-intuitive, and to have important implications for design and control of traffic systems. For example, we show that the density of maximum throughput is near the density of maximum speed variance. Contrary to current practice, traffic should, therefore, be steered away from this density region. For deterministic systems we found traffic flow to possess a finite period which is highly sensitive to density in a non-monotonic fashion. We show that knowledge of this periodic behavior to be very useful in designing and controlling automated systems. These results are obtained for both single and two lane systems. For two lane systems, we also examine the relationship between lane changing behavior and flow performance. We show that the density of maximum lane changing frequency occurs past the density of maximum throughput. Therefore, traffic should also be steered away from this density region.