Model predictive control of urban traffic networks with temporal logic constraints

Model predictive control (MPC) is a popular method for urban traffic management. Apart from physical constraints, traffic networks are amenable to an extra layer of objectives that uphold certain behaviors for the vehicular flow. In this presentation, we survey the recent results on developing MPC strategies from signal temporal logic (STL) specifications. A wide range of desirable behaviors in urban traffic networks such as avoidance of gridlocks, liveness of vehicular flows and sequentiality of traffic lights can be expressed by STL formulas. By translating STL specifications to mixed integer constraints, traffic network control synthesis is formulated as a mixed integer linear programming (MILP) problem. In addition, STL quantitative semantics (robustness) provides a numerical measure of the distance to satisfaction (violation) of the evolution of a traffic network. We will also discuss robustness maximization policies in the context of traffic control.