The real-time railway traffic management problem consists of selecting appropriate train routes and schedules for minimizing the propagation of delay in case of traffic perturbation. In this paper, we tackle this problem by introducing RECIFE-MILP, a heuristic algorithm based on a mixed-integer linear programming model. RECIFE-MILP uses a model that extends one we previously proposed by including additional elements characterizing railway reality. In addition, it implements performance boosting methods selected among several ones through an algorithm configuration tool. We present a thorough experimental analysis that shows that the performances of RECIFE-MILP are better than the ones of the currently implemented traffic management strategy. RECIFE-MILP often finds the optimal solution to instances within the short computation time available in real-time applications. Moreover, RECIFE-MILP is robust to its configuration if an appropriate selection of the combination of boosting methods is performed.

We consider stationary consensus protocols for networks of dynamic agents. The measure of the neighbors' states is affected by unknown but bounded disturbances. Here the main contribution is the formulation and solution of what we call the $\epsilon$-consensus problem, where the states are required to converge in a target set of radius $\epsilon$ asymptotically or in finite time. We introduce as a solution a dead-zone policy that we denote as the lazy rule.