Multiple level production planning in rolling horizon assembly environments

In this paper, nine multiple level planning heuristics are evaluated to characterize how rolling horizon results relate to fixed horizon results in a deterministic demand environment. The weighted order cycle (WOC) is introduced as a single expression of cost structure within a multiple item bill of materials. When planning horizons are restated in terms of WOC (versus time buckets), it becomes apparent that the cost performance of the majority of the heuristics is essentially the same for fixed horizon and rolling horizon conditions when the planning horizon is at least two WOC in length. The horizon sensitive logic of the best two heuristics in cost performance also exhibited less nervousness than several horizon myopic rules, a counter intuitive result. An established multiple level cost modification technique was found to reduce the nervousness of single item rules, in addition to its original goal of schedule cost reduction. To gauge cost performance, Lagrangian relaxation of a binary formulation of the problem was used to find bounds within an average of 1% of the optimal solution cost of each simulation.