Robust production planning with budgeted cumulative demand uncertainty

TL;DR

This paper develops polynomial algorithms for robust production planning under demand uncertainty, considering discrete and continuous budgeted models, with some cases solvable efficiently and others requiring decomposition methods.

Contribution

It introduces new polynomial algorithms for evaluating and planning under demand uncertainty, and addresses computational complexity for different uncertainty models.

Findings

Polynomial algorithms for discrete uncertainty case.

NP-hardness results for continuous uncertainty case.

Decomposition algorithm for general continuous case.

Abstract

This paper deals with a problem of production planning, which is a version of the capacitated single-item lot sizing problem with backordering under demand uncertainty, modeled by uncertain cumulative demands. The well-known interval budgeted uncertainty representation is assumed. Two of its variants are considered. The first one is the discrete budgeted uncertainty, in which at most a specified number of cumulative demands can deviate from their nominal values at the same time.The second variant is the continuous budgeted uncertainty, in which the sum of the deviations of cumulative demands from their nominal values, at the same time, is at most a bound on the total deviation provided. For both cases, in order to choose a production plan that hedges against the cumulative demand uncertainty, the robust minmax criterion is used. Polynomial algorithms for evaluating the impact of uncertainty in the demand on a given production plan in terms of its cost, called the adversarial problem, and for finding robust production plans under the discrete budgeted uncertainty are constructed. Hence, in this case, the problems under consideration are not much computationally harder than their deterministic counterparts. For the continuous budgeted uncertainty, it is shown that the adversarial problem and the problem of computing a robust production plan along with its worst-case cost are NP-hard. In the case, when uncertainty intervals are non-overlapping, they can be solved in pseudopolynomial time and admit fully polynomial timeapproximation schemes. In the general case, a decomposition algorithm for finding a robust plan is proposed.