Annals of Operations Research最新文献

The paper develops a conceptual framework for the analysis and management of catastrophic risk. The framework serves to assess rare extreme events in systematic, quantitative and consistent ways. It dispenses with probabilistic extreme value theory, concentrating on descriptive statistics and simple probability distributions. Risk assessment is based on a recently developed axiomatic approach to non-expected utility preferences defined on the set of risky alternative courses of action available to an agent. The utility values of catastrophic risks are given an explicit algebraic representation, which shows them to be highly unstable (“elastic”) in the sense that they respond disproportionately to small perturbations of the decision outcomes and their probabilities. Various elasticity coefficients are defined for the outcome variables and utility preferences attached to them. They indicate whether a variable possibly takes on large negative values. The coefficients can also be defined as sample statistics and, thus, computed from observed data. The approach admits various applications to practical problems of disaster risk management. The applications include estimations of the effectiveness and cost-efficiency of risk management, the specification of limits of acceptance of catastrophic risk for regulatory purposes, and safety and security systems design and dimensioning.

We introduce a multi-objective variant of the newsvendor problem in which we maximize the expected profit and minimize the maximum regret associated with the decision of how many items to procure from a supplier in the face of unknown demand. When the demand distribution is bounded, the problem is relatively simple. With an unbounded demand distribution, the maximum regret is undefined. In that case, we introduce a chance-constrained variant of the model in which we minimize the maximum regret over a range of demand values whose probability is at least a user-specified value, (gamma). We provide an algorithm for finding the tradeoff between the expected profit and the (gamma)-level maximum regret. We also show that, when operating near the optimal single-objective newsvendor solution, we can significantly reduce the (gamma)-level maximum regret with little degradation in the expected profit.

The Israeli Queue is a batch service polling system where a single server attends to multiple queues based on seniority. Each arriving customer belongs to one of several classes. Upon arrival, a customer either joins an existing queue for their class or initiates a new queue if they are the first of their class to arrive. Customers from the class with the most senior member are served together as a batch, with the service time remaining constant regardless of the batch size. This service model is found in applications like advanced elevator systems and on-demand shared mobility, where passengers heading to the same destination can share a ride. However, in many real-world scenarios, the vehicle capacities are small and constraining, which calls for a deeper exploration of the Israeli queue with a capacitated server (IQCS). In this paper, we formally define the IQCS and address the challenges of creating a mathematically tractable model to represent it. To approximate the IQCS, we develop a quasi-birth-death process and derive approximations for key performance measures. To validate our approach, we implement a simulation model and use it to compare the IQCS, the approximate model, and the original Israeli Queue. Our results across various scenarios demonstrate the accuracy of the approximate model. Nonetheless, the presence of a remaining gap underscores the ongoing challenge of precisely and efficiently modeling the IQCS, posing an open question for the research community.

The flow shop scheduling problem is one of the most complex and widely applicable scheduling problem. In this paper, we design efficient parallel algorithms for solving large-size non-permutation flow shop scheduling problems by leveraging the huge amount of computing power of the current multi-core computing systems. We design two parallel algorithms based on the Shifting Bottleneck heuristic. The first one is a coarse-grained parallel algorithm that is suitable for execution on multi-core systems with a small number of cores, while the second one is a fine-grained parallel algorithm suitable for multi-core systems with a large number of cores. We perform an extensive experimental analysis to evaluate the performance of the proposed algorithms for instances of various sizes. The results show that the proposed algorithms can solve large-size instances of the problem in a reasonable amount of time and obtain solutions that are within acceptable distance from the lower bounds. The proposed parallel algorithms achieve good speedup with respect to the serial variants of the algorithms.

Multi-stage tournaments consisting of a round-robin group stage followed by a knockout phase are ubiquitous in sports. However, this format is incentive incompatible if at least 2 teams from a group advance to the knockout stage where the brackets are predetermined. A model is developed to quantify the risk of tanking in these contests. The suggested approach is applied to the 2022 FIFA World Cup to uncover how its design could have been improved by changing group labelling (a reform that has received no attention before) and the schedule of group matches. Scheduling is found to be a surprisingly weak intervention compared to previous results on the risk of collusion in a group. The probability of tanking, which is disturbingly high around 25%, cannot be reduced by more than 3 percentage points via these policies. Tournament organisers need to consider more fundamental changes against tanking.

We study a single machine scheduling and due-date assignment problem with acceptable lead-times. The setting combines elements of the classical common due-date model and the DIF model, where job-dependent due-dates need to be determined. The objective function, which is of a minmax type, consists of four cost components: (1) job-earliness cost, (2) job-tardiness cost, (3) due-date cost, (4) due-date tardiness cost. We present a simple procedure for identifying the different job-types, and consequently, a polynomial-time solution is introduced. The case of due-windows for acceptable lead-times is also discussed.

This paper proposes a three-echelon emergency commodities supply chain that considers location, allocation, distribution, and resource management planning to minimize the total cost when an epidemic occurs. The model takes into account some key aspects of the emergency distribution network in the event of an outbreak, such as the possibility of quarantining epidemic areas and delays in emergency commodity distribution, commodity storage due to panic buying, demand uncertainty, distribution time, and periodic review and updating of emergency commodity inventories. The model controls the remaining lifetime of the goods while taking into account their perishability. Various types of vehicles with differing capacities, transportation speeds, and costs are studied in order to achieve a suitable balance between cost and speed of delivering commodities. A robust possibilistic programming approach is used to deal with parameter uncertainty and a Lagrangian relaxation approach is used to solve the proposed model. A real case study on COVID-19 is presented in order to illustrate the validity of the suggested model as well as the effectiveness of the developed solution method, and a sensitivity analysis is performed. Based on the findings of this study, considering the uncertainties of system costs, demand, quarantine probability, and delays in the distribution of commodities have a significant impact on network costs during an epidemic outbreak and ignoring them leads to inaccurate estimates of system costs.

Academic institutions seek to enhance their reputation, which is one of their primary assets. Doing so requires a massive investment of resources in research, recruiting a high-quality academic staff, and building campuses and state-of-the-art laboratories. To obtain the necessary financial resources, institutions must attract students, donors, and government budgets and grants. This paper introduces a stylized dynamic model demonstrating how an institution can best allocate its resources between teaching and research. We create a simulated competition that resembles the real situation where the enhancement of the institution’s reputation depends not only on its resource allocation but also on its competitors’ actions and reputation. We consider a two-institution contest over time using a differential game solution with open-loop strategies. In this case, the steady-state investment in research increases and the level of teaching decreases.

This paper studies a continuous-review inventory replenishment model with a limited storage capacity S in an uncertain environment. We assume that the demands and returns follow independent Poisson processes. We further assume a ra1079 shelf life, a random lead time, and early loss. The storage is managed according to the base-stock (S, s) policy for (s<S,S>0.) In case of overstock, each returned item exceeding S is transferred to a foreign facility. If during the lead time a demand reaches zero stock, we consider two alternatives: either allow partial backordering up to (L_{B}) items, beyond which the unsatisfied demand is lost, or call for an immediate and costly emergency supply up to level (0<Q_{B}^{e}le S). Our objective is to study how the thresholds s,  S,  (L_{B},) and (Q_{B}^{e}) are impacted by the system’s parameters, such as returns, demands, and costs. Using a Markovian framework, we derive the steady-state probabilities for the inventory level, and construct closed-form expressions for the average cost functions. Then, we numerically investigate the impact of the different parameters on the best policy and on the threshold levels. We compare the two alternatives and identify situations in which calling for an emergency supply is economically profitable.

This survey investigates the field of moderate exponential-time algorithms for ({mathcal{N}mathcal{P}})-hard scheduling problems, i.e., exact algorithms whose worst-case time complexity is moderately exponential with respect to brute force algorithms. Scheduling problems are very challenging problems for which interesting results have emerged in the literature since 2010. We will provide a comprehensive overview of the known results of these problems before detailing three general techniques to derive moderate exponential-time algorithms. These techniques are Sort & Search, Inclusion–Exclusion and Branching. In the last part of this survey, we will focus on side topics such as approximation in moderate exponential time, the design of lower bounds on worst-case time complexities or fixed-parameter tractability. We will also discuss the potential benefits of moderate exponential-time algorithms for efficiently solving in practice scheduling problems.