Abstract
Previous models for routing messages in computer networks have been fashioned to employ queueing theory - as such they are capable of producing only static and open-loop strategies. In this paper we present a new model which can accomodate dynamic closed-loop policies for the problem of routing in data communication networks. The mathematical model gives rise to an optimal control problem for which the dynamics, cost and constraints on the states and controls are all linear. An algorithm is derived which utilizies a novel combination of dynamic programming, Pontryagin's maximum principle and linear programming to construct a feedback solution for this traditionally troublesome problem. Of particular interest in this technique is the application of several concepts in linear programming, such as parametric programming add the challenging problem of finding the complete set of multiple optimal solutions to a linear program.
| Original language | American English |
|---|---|
| Title of host publication | Joint Automatic Control Conference |
| State | Published - 1976 |