Forward Algorithms and Planning Horizons for Dynamic and Linear Programming.

When available, efficient forward algorithms with planning horizon procedures better suit the needs of real world managers with limited forecast information and computational budges than conventional programming approaches. A forward algorithm efficiently solves longer and longer horizon sub-problems. A planning horizon procedure is a stopping rule specifying when further lengthening the horizon will give a negligible effect on current decisions output by the model. Various researchers have previously obtained such results for particular models in the areas of convex production planning. Here the regeneration set ideas developed by lundin and Morton are extended to provide full planning horizon procedures for any finite deterministic dynamic programming or linear programming model.