This paper is intended to analyze benefits of microscopic real-time optimization for traffic signal control. To this end, we first present a microscopic traffic model consisting of signals and vehicles, where each vehicle dynamics is explicitly incorporated in the form of a mixed logical dynamical system (MLDS) differently from many other related publications taking macroscopic flow models. We then formulate a model predictive control (MPC) problem and then reduce the problem to a mixed integer programming problem. Furthermore, we introduce a linear temporal logic (LTL) constraint to reduce the computation burden and to eliminate unrealistic solutions. The proposed control scheme is then applied to a real-time traffic simulator and it is confirmed that: (i) the number of waiting vehicles for traffic lights is reduced with 54.5% as compared to a fixed-time control scheme and (ii) the average computation time is reduced to about one-tenth relative to the case without the LTL constraint.
