Two statistical methods for grouping vehicles in traffic flow based on probabilistic cellular automata

There are many mathematical models of traffic flow that are used to analyze and simulate the phenomena of traffic jams. Specifically, a model based on the cellular automaton has been studied and used for traffic simulation systems. Recently, a method has been proposed that uses traffic flow data to estimate the parameter of the model; this enables us to objectively optimize and evaluate the model. However, model optimization has not been thoroughly considered for situations where the flow cannot be expressed by a single model, due to there being a variety of driving behaviors. In order to model a variety of behaviors, the present paper proposes two statistical methods that use mixture models with groups of vehicles that are distinguished by their average speeds.