A COMPUTATIONAL STUDY OF EMPIRICAL DECISION HORIZONS IN INFINITE HORIZON, MULTIPERIOD NETWORK FLOW PROBLEMS

The minimum cost, multiperiod network flow model is an important optimization model for solving problems in many application areas including resource scheduling, planning and distribution. This network model describes decision making problems over time. In earlier work, we discussed the development, implementation and computational testing of a new technique, the forward network simplex method, for solving linear, minimum cost, multiperiod network flow problems. The forward network simplex method exploits the natural decomposition of multiperiod network problems, by limiting its pivoting activity to the date of the last few time periods. Both the solution CPU time and pivot count are linear in the number of time periods. For standard network optimization codes, the pivot count is linear while the solution time is approximately quadratic in the number of periods. Here we present a computational study of the natural decomposition of, or empirical decision horizons for, an “infinite” horizon, multiperiod net...