On the connections between optimal control, regulation and dynamic network routing

Abstract

he paper is devoted to studying general features of dynamic network routing problems. It is shown that these problems may be interpreted as receding horizon optimal control problems or simply regulation problems. In the basic formulation it is assumed, that the nodes have no dynamics and the only goal of the optimization mechanism is to find the shortest paths from the source to the destination nodes. In this problem the optimization mechanism (i.e. the Bellman-Ford algorithm) may be interpreted as a receding horizon optimal control routine. Moreover, there is one-to-one correspondence between the Bellman optimal cost-to-go function in the shortest path problem and the Lyapunov function in the regulation problem. At the end some results of the application of the routing optimization algorithm to an inverted pendulum regulation problem are presented.