Operations Research Proceedings 2022

To compete with the rapid growth in e-commerce, many local shops provide a delivery service to their customers. To increase consolidation opportunities, shops start cooperating in local delivery by using shared vehicles and micro-hubs for joint transportation of parcels. Stores deposit their orders at close-by micro-hubs for further delivery by the shared vehicles, which conduct consistent routes between the micro-hubs. As long as it is in line with their schedule, the vehicles collect the parcels and drop them off close to customers’ locations. Hence, it is very important to find effective schedules which is particularly challenging since order placements vary from day to day. We propose a two-stage stochastic program. In the first stage, the vehicle schedules are determined. In the second stage, the realised orders are routed. The goal is to maximise the expected amount of fulfilled parcel orders with the shared vehicles. We solve the problem with the Progressive Hedging algorithm. We consider the optimal solution without consistency constraints and a practically-inspired heuristic solution as benchmarks. We find that Progressive Hedging behaves rather poorly on random data, but performs particularly well on highly structured demand patterns.

In this article, we summarize a subset of the findings of the cumulative dissertation “Algorithms for Mixed-Integer Bilevel Problems with Convex Followers”; see [4]. First, we present a result that renders the application of the well-known and widely used big-M reformulation of linear bilevel problems infeasible for many practical applications. Second, we present valid inequalities and demonstrate that an SOS1-based approach is a competitive alternative to the error-prone big-M method in case both approaches are equipped with these valid inequalities. Third, we introduce a penalty alternating direction method, which computes (close-to-)optimal feasible points in extremely short computation times and outperforms a state-of-the-art local method.

The Steiner tree problem in graphs (SPG) is a classic $$\mathcal{N}\mathcal{P}$$ -hard problem. Many applications can be modeled as SPG or closely related problems. This article describes a state-of-the-art solver for finding optimal solutions to classic Steiner tree and 14 related problem classes. It was developed as part of the author’s doctoral dissertation.

Prescriptive analytics approaches integrate machine learning prediction and optimization to directly derive decisions for planning problems from historical observations of demand and a large set of features (co-variates). This paper summarizes the key results of the author’s dissertation and presents two prescriptive analytics approaches, kernelized empirical risk minimization and weighted sample average approximation, to solve complex capacity planning problems. It demonstrates the applicability of both approaches to a real-world two-stage capacity planning problem and evaluates their performance relative to traditional parametric approaches that first estimate a demand distribution and then solve a stochastic optimization problem, and a traditional non-parametric approach (sample average approximation). The results of numerical analyses demonstrate that the new prescriptive analytics approaches can lead to substantial performance improvements of up to 58% compared to traditional approaches.

After graduation, physicians receive further training in a medical domain like anesthesiology. There are 57 medical specialties in Germany in total. The high cost pressures of hospitals and the changing view of the medical profession regarding the work-life balance have led to recruitment problems and low employee satisfaction in many places. A promising approach to counter this problem is objective and structured training planning. This research project mainly deals with medical residents’ strategic and tactical-operative training scheduling. In addition to relieving the medical staff currently responsible for the planning process, this research project increases the predictability of structured training. This allows hospitals to increase the quality of their training and, consequently, their attractiveness to other hospitals. In addition, supervisors from different departments can better assess residents’ knowledge and thus keep the level of service, which is particularly important in hospitals, permanently high even when changing residents. From the residents’ point of view, a well-structured training schedule enables a high degree of information. Therefore, residents are no longer surprised by a short-term change of department and have a direct insight into their training progress. A real-world case study evaluates the mathematical formulations and the solution approaches.

In this paper, three configurations of the customer order scheduling problem are presented. In contrast to classical scheduling problems, the customer order scheduling problem considers the scheduling of jobs that belong to customer orders and each order is only completed when each job of the order has finished. The studied configurations are the minimization of the sum of order completion times and the minimization of the earliness-tardiness in a machine environment where each order places one job on each machine. Furthermore, the minimization of the sum of order completion times in a flow shop environment is investigated. This paper states properties of the three problem configurations and describes developed solution methods that performed well in a computational experiment.

We consider a bilevel continuous knapsack problem where the leader controls the capacity of the knapsack, while the follower chooses a feasible packing maximizing his own profit. The leader’s aim is to optimize a linear objective function in the follower’s solution, but with respect to item values that can be different from the follower’s item values. We address a stochastic version of this problem where the follower’s profits are uncertain and only a probability distribution is known. This problem is #P-hard for the case of independently and uniformly distributed follower profits. In this paper, efficient algorithms are developed for the special case where all items have unit weight, as is the case in the bilevel selection problem. Generalizing these results to the case of arbitrary weights leads to pseudo-polynomial time algorithms for the bilevel continuous knapsack problem.

Recent innovations in the field of forecast combination include integrated methods for forecast selection, weighting and regularization. The methods proposed in related articles first label whether or not forecasters should remain in the selection using information criteria from statistical learning theory. Depending on the selection status, the optimal weights of all forecasters in the sample are then used as baseline to shrink the weights either toward zero or the mean, with the degree of regularization determining the final selection of forecasters. In this paper, we propose a new information criterion reflecting the importance of diversification and accuracy gains in the selection of forecasters for integrated methods. In an iterative procedure motivated by forward feature selection, each forecaster is selected sequentially, while at each step the increase in accuracy and diversification due to the addition of a forecaster to the previous selection is measured. To quantify the increase in diversity, the multiple correlation coefficient is used, which captures the correlation between the previously selected forecasters and a candidate, where the lower the correlation between the candidate and the selection, the higher the gain in diversity for the combination. For the accuracy increase, the accuracy achieved by optimal weight combinations with the previously selected forecasters is compared with the accuracy after adding a candidate. A hyperparameter further enables the trade-off between accuracy and diversification gains in the criterion. Simulation-based studies show scenarios in which our presented information criterion achieves advantages in out-of-sample prediction accuracy over previous criteria for selection by accounting for accuracy and diversification gains.

In decision-making overconfidence and estimation biases can lead to sub-optimal outcomes and accuracy loss. A debiasing strategy presented in this work is to use feedback based on the error pattern of own previous absolute and 90% certainty (confidence) interval estimates. This is comprised in a decision support system (DSS) and applied in an experiment, where results indicate support for the key assumption that subjects are able to selectively reduce their overconfidence and their estimation bias, if present, with the help of the provided feedback.

The linear systems arising from interior point methods (IPM) for linear programming are solved using the preconditioned conjugate gradient method (PCG). Two preconditioners are adopted: the controlled Cholesky factorization (CCF) of the normal equations system and the splitting preconditioner. The CCF performance depends upon the previous reordering of the linear programming constraint matrix rows. A comparison among two different reordering methods is performed in order to verify the most suitable one for this approach. Variants of nested dissection (ND) and the minimum degree (MD) are among the considered heuristics. Computational experiments with large-scale linear programming problems from several collection sets are performed.

Model discrimination (MD) aims to determine the inputs, called design points, of two or more models at which these models differ most under the additional condition that the models are fitted to these points, in the case of T-optimal designs. On the one hand, nonlinear models often lead to nonconvex MD problems, on the other hand, the optimal number of design points must be determined, too. Thus, the computation of T-optimal designs is very arduous. However, if one considers finitely many design points, a well-solvable bi-level problem arises. Since the latter only represents an approximation of the original model discrimination problem, we refine the design space discretization using the equivalence theorem of MD. This yields a tri-level approach whose iterates converge to a T-optimal design. We demonstrate that the approach can outperform known solution methods on an example from chemical process engineering.

Households are responsible for a third of the final energy consumption in Germany in 2018 with the average household meeting 60% of household energy service needs with fossil fuels. Energy transition targets to increase renewables and energy efficiency will require high upfront capital investments into building renovations, heater and appliance upgrades. Overall less than 17% of households have sufficient capital and are in the decision-making power to undertake investments. The household sector is disaggregated into 56 profiles by key socio-economic parameters (income, building type, tenure status, urbanization) and, together with a budget constraint limiting the total available capital on all investments and consumption per profile, incorporated into an energy system optimization model to account for the differentiated needs and financial capabilities of diverse household types. This method provides a platform from which to evaluate the impact of various policies and measures. Selected scenarios include a comparison between a reference case and two carbon revenue redistribution schemes. The results yield insights into the energy-related investment and consumption patterns for different household types emphasizing fuel types, emissions and quantification of suppressed demand (expressed through household budget deficit for unmet household service demands), and show that to evaluate the energy welfare of lower income households several aspects must be considered.

There are many possible future energy systems. We explore the range of parameter uncertainty and quantify parameter interrelations to generate hundreds of scenarios for Germany. Only sensible parameter combinations remain as inputs to an energy system optimization and coupled models. In the past, computational limitations have been a major obstacle to calculate such an enormous space of scenarios. In contrast, we use high-performance computing. To utilize high-performance computing (HPC) efficiently, the parallel solver for linear programs PIPS-IPM++ is applied. We integrate it into a tool chain of different components including scenario generation, energy system optimization and results evaluation and couple a large diversity of software packages in a fully automated HPC workflow. This enables the calculation of all scenarios in a matter of days. Furthermore, we use a set of 37 indicators to provide a comprehensive assessment of the simulated energy systems. In this way, we cover multiple perspectives, such as system adequacy, security of supply or behavior of market actors. Whereas scenarios with already expanded capacities do not lead to clear results, a “green field” approach or higher spatial resolution do. Yet, we identified three clusters of scenarios. This allows to study disruptive events like price shocks in a vast parameter space and to identify countermeasures for the long-term.

Planning energy production is a challenging task due to its cost-sensitivity, fast-moving energy markets, uncertainties in demand, and technical constraints of power plants. Thus, more complex models of this so-called unit commitment problem (UCP) have to be solved more rapidly, a task that probably can be solved more efficiently via quantum computing. In this article, we model a UCP with minimum running and idle times as a quadratic unconstrained optimization problem to solve it on quantum computing hardware. First experiments confirm the advantages of our formulation in terms of qubit usage and connectivity and most importantly solution quality.

We address the sales force deployment problem and its four subproblems for teams of sales representatives within sales territories. We contribute by showing that the problem can be reduced to the uncapacitated facility location problem under convex profit contribution functions with a unique maximum. In our numerical study, we show that instance sizes considered difficult for the sales force deployment problem of individual representatives can be solved optimally in minutes for teams of representatives. In our largest instance, with 2020 potential locations and sales coverage units, it took on average 140.73 s.

The stochastic bin packing problem (SBPP) is an extension of the well-studied classical bin packing problem with respect to imperfect information on the item sizes. From a practical point of view, the latter are typically represented by (stochastically independent) normally distributed random variables with given means and variances. In this scenario, the SBPP requires to determine the minimum number of bins needed to pack all the items, with the risk of overloading a bin not exceeding a certain tolerable limit. Such computations are of high relevance in server consolidation applications, where decisions have to be made before witnessing the true item characteristics. The resulting integer optimization problems are generally nonlinear and therefore difficult to solve. For this reason, previous approaches from the literature can only handle small instance sizes exactly. In this work, we present a column generation algorithm using heuristic information and near-optimal solutions of the associated (challenging) pricing problems. Based on numerical tests, we show that in most cases this heuristic approach already leads to an optimal solution, so that much larger instance sizes can now be dealt with in reasonable time.

We propose an algorithm for solving bilevel problems with mixed-integer convex-quadratic upper level as well as convex-quadratic and continuous lower level. The method is based on a classic branch-and-bound procedure, where branching is performed on the integer constraints and on the complementarity constraints resulting from the Karush–Kuhn–Tucker reformulation of the lower-level problem. However, instead of branching on constraints as usual, suitably chosen penalty terms are added to the objective function to create new subproblems in the tree. In this first part, we consider a fixed penalty parameter, derive the main ideas, and prove the correctness of the method for this setting.

In the first part of this paper, we propose a penalty branch-and-bound method for solving bilevel problems with mixed-integer convex-quadratic upper level as well as convex-quadratic and continuous lower level and analyze the method for a fixed penalty parameter. In this second part, we extend the algorithm and its analysis towards iteratively adapted penalty parameters, prove the correctness of this extended method, and show its applicability by some first numerical results.

The pressure on airlines to reduce their emitted $$\text {CO}_{2}$$ emissions is continuously increasing. Due to new fuel-efficient air frame and engine technologies, aircraft modernization offers an opportunity to reduce annual emissions. But with limited production capacities of the manufacturers and related costs, it is not possible to renew a whole fleet at once. At the same time, $$\text {CO}_{2}$$ trading systems lead to higher costs for older fleets. Therefore, there needs to be long term planning regarding the fleet composition of an airline. For this, the fleet planning problem determines economically optimal replacement times for aircraft. The fleet planning problem can be formulated as a MILP and combines fleet composition and fleet assignment. The fleet composition problem considers different aircraft types as well as retrofits; here, retrofits are modifications of aircraft to lower the amount of emitted emissions without replacing the whole aircraft. These aircraft are then used within the fleet assignment problem to fulfill a given demand of flights. The emissions are then considered within $$\text {CO}_{2}$$ trading systems as that is currently the only affecting restriction for airlines concerning $$\text {CO}_{2}$$ emissions.

In multi-objective optimization, several potentially conflicting objective functions need to be optimized. Instead of one optimal solution, we look for the set of so called non-dominated solutions. An important subset is the set of non-dominated extreme points. Finding it is a computationally hard problem in general. While solvers for similar problems exist, there are none known for multi-objective mixed integer linear programs (MOMILPs) or multi-objective mixed integer quadratically constrained quadratic programs (MOMIQCQPs). We present PaMILO, the first solver for finding non-dominated extreme points of MOMILPs and MOMIQCQPs. It can be found on github under https://github.com/FritzBo/PaMILO . PaMILO provides an easy-to-use interface and is implemented in C++17. It solves occurring subproblems employing either CPLEX or Gurobi. PaMILO adapts the Dual-Benson algorithm for multi-objective linear programming (MOLP). As it was previously only defined for MOLPs, we describe how it can be adapted for MOMILPs, MOMIQCQPs and even more problem classes in the future.

We consider a novel variant of the heterogeneous vehicle routing problem (VRP) in which each customer has different availability time windows for every vehicle. In particular, this covers our motivating application of planning daily delivery tours for a single vehicle, where customers can be available at different times each day. The existing literature on heterogeneous VRPs typically distinguishes properties of the vehicle fleet such as costs or capacities, but apparently, windows of customers have only been regarded in a homogeneous fashion thus far. To solve the problem, we employ a branch-and-price framework based on a set partitioning formulation together with a parallelizable labeling algorithm. The heterogeneous time window structure yields notable computational gains by allowing to decompose the pricing problem as well as to utilize a customer-vehicle assignment branching rule. We show that this branching rule leads to more balanced search trees than the usual arc flow branching, and demonstrate its efficiency in numerical experiments.

As the share of renewable energy and therefore the fluctuation in power generation increases, using a battery to market the power saved/used by a flexible process becomes more and more important and attractive. The charging and discharging process of a battery can be modeled as a flow model in a time-expanded graph. We model the costs for this process as edge costs between nodes that correspond to consecutive points in time. An optimal battery strategy be obtained by computing a minimum cost flow in the given network. If the charging and discharging of a battery takes place as a coupled process of a production process, a steady and even charging and discharging is often important. In this paper, we describe a bicriteria flow model based on the Almost-Equal-Minimum Cost Flow Problem (AEMCFP) in which both trading profits and a steady flow of energy are considered as the two objectives. In the AEMCFP one is given additional sets of edges on which the flow values differ at most by a given constant. We obtain a strongly polynomial algorithm based on a parametric search approach. Furthermore, we present a case study for bidding on the German day ahead market.

The German federal government has set up the goal of 10 million electric cars on German roads by 2030 and is aiming for the complete electrification of road traffic by 2050. Currently, the focus is particularly on the development of charging infrastructure for battery electric vehicles (BEVs), with 1 million publicly accessible charging stations to be built by 2030. However, the expansion of hydrogen infrastructure is also being supported with large subsidies. Although there are numerous studies on either the cost of charging infrastructure for BEVs or fuel cell electric vehicles (FCEVs), there are few comparative research results available so far. In this study, therefore, a model for the spatial distribution of charging infrastructure for BEVs and FCEVs is first developed for the district of Steinburg in the German federal state of Schleswig–Holstein. In a second step, the results and corresponding economic costs are compared in a real options analysis for the period 2021 to 2050, considering different charging infrastructure expansion curves. In addition to the hardware infrastructure costs, operations and maintenance and electricity or hydrogen costs for the charging infrastructure are explicitly taken into account.

The accelerating transition towards clean energy is raising the need to consider electric power systems with an increased integration of renewable energy sources, energy storage systems and flexible loads, in addition to the classical programmable generators and electricity demand. Within this setting, we consider the optimal day-ahead operational planning of power systems with a quarter-hourly time resolution and with the objective of minimizing its total operating cost. The electricity demand and photovoltaic production data is provided by forecast models that are based on historical and real-world data sets. For solving this optimization problem we use a hierarchical approach based on tabu search.

In the context of lot-sizing and scheduling, minimizing energy consumption is a typical criterion for improving a company’s environmental footprint. In the literature, however, it is insufficiently questioned whether minimizing energy consumption actually also minimizes energy-related emissions. In this paper, we show that such a positive one-to-one relationship does not always hold. In fact, when energy prices fluctuate over time, energy costs may additionally conflict with these two goals. To demonstrate this, we develop a three-criteria lot-sizing and scheduling problem and derive the three-dimensional Pareto front using the elastic constraint method.

In this paper we propose a novel approach to distinguish the style of drivers with respect to their energy efficiency. A unique property of the proposed method is that it relies exclusively on Global Positioning System (GPS) data. This setting is highly robust and available in practice as these GPS logs can easily be obtained. To rely on positional data alone means that all possible derived features from it will be highly correlated, so we have to consider a single feature. Here, we propose to explore the use of acceleration differences of a movement. Our strategy relies on agglomerative hierarchical clustering. The approach can be easily implemented to perform fast, even on huge amount of real-world data logs.

We consider a capacity planning problem for networks including storage. Given a graph and a time series of demands and supplies, we seek for integer link and storage capacities that permit a single commodity flow with valid storage in- and outtakes over all time steps. This problem arises, for example, in power systems planning, where storage can be used to buffer peaks of varying supplies and demands. For typical time series spanning a full year at hourly resolution, this leads to huge optimization models. To reduce the model size, time series aggregation is commonly used. The time horizon is sliced into fixed size periods, e.g. days or weeks, a small set of representative periods is chosen via clustering methods, and a much smaller model involving only the chosen periods is solved. Representative periods, however, typically do not contain the situations with the most extreme demands and supplies and the strongest effects on storage. In this paper, we show how to identify such critical periods using principal component analysis (PCA) and convex hull computations and we compare the quality and solution time of the reduced models to the original ones for benchmark instances derived from power systems planning.

We investigate sequential investment in carbon capture and storage (CCS), i.e., the case of retrofitting a coal-fired power plant, and then invest in carbon capture and utilization (CCU) for methanol production. A (nested) compound real options model based on a backward recursive dynamic programming algorithm is used for the analysis, which seems helpful for decision-makers who have to make capital-intensive irreversible investments under high uncertainty regarding electricity and CO2 price. The options to invest in CCS and CCU are investigated individually first, and then sequentially, leading to a hybrid CCUS plant that enables both CO2 storage and methanol production. The prices of electricity, carbon and methanol are considered as stochastic and correlated with each other. Managerial flexibility exists regarding a postponement of the investment decision and the real-time optimization between selling methanol to the market or storing CO2 for earning carbon credits after establishing the CCUS plant. We find that at CO2 prices of around 40 €/t, CCS investment is economically rational, whereas CCU for methanol is not. Combining CCS with CCU increases the overall investment probability and potential for larger profits. Since methanol is more valuable than CO2, CCU can be expected to dominate the value of the compound option for the case of favorable market conditions (i.e., sufficiently high methanol and CO2 prices).

Energy system models allow exploring new interactions between energy supply and demand that new actors within the system enable. Data centers (DC) are such actors, becoming significant energy consumers in some regions while pressuring local energy systems, but also having the potential to contribute to the energy transition by providing demand response and waste-heat recovery. Yet, no past studies have investigated how these attributes can support energy system decarbonization. This article describes the integration of large DCs as an extension to the Balmorel energy system model, jointly optimizing their cooling portfolio along with the electricity and the district heating systems. The model allows DCs to invest in flexible-cooling and waste-heat technologies while competing freely with other traditional generators within the system. This article showcases a simplified example of DCs’ potential in the Danish energy system through 2035. The results show that waste heat accounts for 9% of the national heating supply, leading to system-wide emission reductions of 0.8% in 2025 and cost savings of up to 3.0% in 2035. In the future, this model will be used for policy research to align the incentives of DC operators and society.

In areas of commerce, fast and frugal heuristics have a demonstrated history of forecasting as or more accurately than complex strategies in situations of uncertainty. Financial participation in the deregulated electricity market requires accurate forecasting and presents an opportunity for the implementation of heuristics by traders seeking to arbitrage. Developmental research was undertaken to invent and assess simple heuristics for bidding in wholesale electricity markets. Two heuristic price-adjustment bidding strategies were developed, providing modest returns relative to existing strategies, greater risk aversion and reduced computational complexity.

This work presents an innovative short to mid-term forecasting model that analyzes nonlinear complex spatial and temporal dynamics in energy networks under demand and supply balance constraints using Network Nonlinear Time Series (TS) and Mathematical Programming (MP) approach. We address three challenges simultaneously, namely, the adjacency matrix is unknown; the total amount in the network has to be balanced; dependence is unnecessarily linear. We use a nonparametric approach to handle the nonlinearity and estimate the adjacency matrix under the sparsity assumption. The estimation is conducted with the Mathematical Optimisation method. We illustrate the accuracy and effectiveness of the model on the example of the natural gas transmission network of one of the largest transmission system operators (TSOs) in Germany, Open Grid Europe. The obtained results show that, especially for shorter forecasting horizons, proposed method outperforms all considered benchmark models, improving the average nMAPE for 5.1% and average RMSE for 79.6% compared to the second-best model. The model is capable to capture the nonlinear dependencies in the complex spatial-temporal network dynamics and benefits from both sparsity assumption and the demand and supply balance constraint.

The Smart Readiness Indicator (SRI) is a method proposed by the European Commission which calls for better use of the potential of smart technologies in the building sector. The introduction of the SRI is intended to raise awareness of smart building technologies and make the added value more available for building users, owners, and providers of smart services. The technological smart readiness of buildings can be determined with the SRI assessment method. The method has 54 questions, which are divided into nine domains: heating, domestic hot water, cooling, ventilation, lighting, electricity, electric vehicles, dynamic envelope and monitoring & control. Each question is assessed with up to five different levels, representing incremental levels of technological equipment. These questions form the basis for the calculation of the SRI score. When improving the SRI score of a building, to choose the technologies that will provide the maximum SRI score improvement with a limited budget can be challenging. Therefore, the aim of this paper is to help the decision makers to come up with the correct choices that have the highest impact on the SRI score. The chosen solution method here is a specific non-dominated sorting genetic algorithm (NSGA II) algorithm. The proposed method is then applied to a hypothetical building to demonstrate its applicability and capability. The results show which SRI domains and questions to focus on. This gives future directions regarding choosing technologies to be implemented.

The European energy crisis and the global climate crisis call for a strong reduction of fossil fuel usage in the residential heating sector. Given the rising role of energy communities, we address this challenge by optimizing the design and operation of different energy community systems with a linear program and a genetic algorithm for rolling horizon control, respectively. In particular, we compare status quo systems that are based on natural gas, with purely electricity-based systems, and systems based on electricity as well as the local production, storage, and usage of green hydrogen in the respective community. Applying our method to a case study of a community with 19 households, across various regulatory scenarios, and two different objective scenarios, we find that including hydrogen can achieve considerable CO2 emission reduction, higher self-sufficiency, and lower costs than systems using natural gas for heating. Set-ups without hydrogen, but with larger electric heat pumps, achieve similar emission reductions at lower costs but enable less self-sufficiency in the community.

In this paper, we optimize the design of a new office building’s energy supply with respect to costs and carbon emissions for one example year. The aim is to gain an overview over the impact of the supply design decision. For that, we combine known concepts to one decision support workflow. We simulate the expected heating loads of the building with a thermal network and use them to set up a mixed-integer linear program for the problem, which we can solve with one or both objectives.

Energy crisis and environmental concerns encourage the adoptions of substitute transportation options instead of the conventional internal combustion engine vehicles. The electric, hydrogen and natural gas vehicles are promising alternatives, so more attention should be paid to the economic and operational features of their charging stations. This paper proposes a multifunction charging station to refill electric, hydrogen and natural gas vehicles which takes part in the day-ahead and intra-day markets. The objective of this station is to maximize its profit by attaining the optimal operation of the devices and bidding curves. Coordinated bidding is considered since this charging station participates in the sequential markets with different price scenarios. The clearing prices and dispatched amounts in both markets are unknown at the time of bidding. This problem is formulated as a two-stage stochastic program since the markets are cleared sequentially and the prices are revealed gradually. Finally, the economic effectiveness of the proposed multifunction charging station is analyzed in different scenarios.

A core function of each Electricity Transmission System Operator (TSO) is the procurement of ancillary services in real time balancing markets, necessary for a stable and reliable operation of the system. TSO controls the active power to manage congestions in transmission network based on security criteria. Congestion management is achieved by rescheduling generation using mainly experience of operators. However, TSO has legal obligations to procure ancillary services in accordance with economic, transparent and non-discriminatory procedures. This work aims to develop an algorithm/software system for a TSO to decide on the optimum redispatch in the sense of economy and security. The redispatch problem is formulated as an optimal power flow (OPF) problem. The objective function of the optimization is the minimization of the total cost of the generation shifts necessary. The security constraints will be satisfied by power flow equations which are non-linear and non-convex by nature. In order to obtain a robust solution due to the large scale nature of the problem, linearization techniques are applied to power flow equations. Compared with the commonly used DC OPF methods, reactive power and bus voltages are taken into account in the formulation. The system developed is deployed as a real time running application for real large-scale networks.

Energy system optimization models (ESOMs) can be helpful tools to determine the optimal structure of future energy systems. They usually optimize the expansion and dispatch of the energy system’s components through a minimization of total system costs. The obtained energy systems are designed to cover the energy demand for the specific assumptions made within the underlying scenarios. However, if such energy systems are exposed to slight deviations, such as a lower availability of wind energy, situations of uncovered demand may occur. The uncertainties in the scenario assumptions can be indirectly captured via excess generation capacities. However, the required amount of these excess capacities is unclear. This study analyzes capacity adequacy by considering uncertainties in a decarbonized German power system through stochastic optimization within an ESOM. Different uncertainties, such as technology investment costs, total annual demand and different weather conditions are considered and their influence on the power system is compared. Therefore, a variety of different assumptions for these uncertainties are extracted from literature and included in the stochastic optimization. As a result, the impact of the uncertainties on the structure of the energy system are identified and the excess capacity needed is estimated.

The transport sector is a major consumer of energy worldwide. Unfortunately, there is no silver bullet to de-fossilize the transport sector due to its intricacy; therefore, many concepts and technologies should be combined to have a noteworthy impact on this hard-to-abate sector. As such, the required diverse set of expertise for making correct decisions cannot be achieved by merely utilizing one model. In this study, we connect multiple datasets and models that employ various methodologies with different purposes to exhibit a pathway to a green transport sector. The extended bioenergy optimization (BENOPTex) and renewable energy mix (REMix) models are coupled iteratively to produce coherent results while considering different sets of constraints. The combined effects of bioenergy and synthetic fuel—using renewable electricity—on the German transport sector are investigated via a scenario. Two demand models are also used to capture the specificities of the energy demands of the mainly behavior-driven road transportation as well as technology-driven aviation sector. The outcome of the resulted soft-coupled model respects biomass availability, regulatory circumstances, techno-economic properties, and power sector expansion for the production of synthetic fuels.

We address electricity consumption scheduling on a day-ahead basis within a community of prosumers that own renewable generation. We establish two market designs that enable coordination between members and where excess production can be valued outside or inside the community. For each, we propose two formulations: centralized schemes where the common objective is optimized, while in decentralized schemes, each member optimizes its own objective. The natural interdependence between members sharing a common network leads to the formulation of non-cooperative games. We solve some proposed models on a use-case by using distributed algorithms that ensure confidentiality.

In many sports, women receive less prize money than men. This issue has been discussed extensively over decades. While many people consider different prize structures for men and women as unfair, others argue, e.g., that men attract greater public interest or that the competition is harder among men than among women (in particular, when there participate more men than women). In this paper, we focus on the discussion of fairness in the distribution of prize money in endurance sports concerning the severity of the competition. We present two methods to distribute prize money across gender based on the individual performances w.r.t. gender specific records. We suppose these “across gender distributions” to be fair, as they suitably respect that women generally are slower than men. Furthermore, we compare commonly used prize distributions to our across gender distributions, introducing a statistical fairness measure. For our investigations, we focus on triathlon, but the results can easily be adapted to any other endurance sports.

We apply boosted trees and Shapley values to analyze economic spillover effects within a customer-supplier network and assess economic interpretability. We translate conditional volatility into a Value-at-Risk universe and generate innovative economic features based on Natural Language Processing. Our results provide evidence for the economic relevance of spillover within a customer-supplier network for applied risk measurement. Furthermore, we demonstrate that the application of machine learning to panel data leads to innovative insights.

Collaborative duty rostering can increase the satisfaction of employees in healthcare. For the acceptance of a final rostering, a fair selection of included wishes is essential. As in rostering problems various constraints must be respected, it is generally not possible to accept all bids (wishes for a single free time slot) in a planning period and fairness is important. In this paper we present a weighted Vickrey-Clarke-Groves (VCG) mechanism approach where past auction results are incorporated by a fairness factor and where the underlying winner determination problem is given by a hitting set problem (HSP). We present numerical results of simulation runs over several planning periods.

Strategic alliances among competitors are increasingly gaining relevance as business organizations recognize the benefits of learning and sharing resources to succeed in dynamic environments. In an increasingly interconnected business world, it is crucial for companies to reap the benefits that arise from this type of collaboration. We use multi-agent simulation coupled with principles from operations research to study how learning and knowledge sharing within alliances can impact the performance of competing firms in an endogenously changing environment. We consider two types of competing firms: ones that search for business solutions in the business landscape (searchers), and ones that search and also have the power to endogenously change or reshape the landscape to their advantage (shapers). Our model allows us to analyze under which circumstances strategic alliances can help firms to successfully adapt to these endogenously generated changes by allowing them to learn and build up an enhanced knowledge base of the business landscape.

The viability kernel in Viability Theory depends on control variables and usually also on uncontrolled ones. Control variables try to increase viability, and uncontrolled ones instead destroy it. Tyches are uncertainties without statistical regularity that diminish viability. We progress in the study of both effects. We use a necessary condition of the system viability and apply it to the linear case by introducing the Minkowski difference between sets. We also find such a difference interprets the problem adequately.

In case of a mass-casualty incident with several hundred or even thousands of patients, providing fast medical treatments is one of the main goals. If close-by hospitals cannot provide sufficient capacities to treat all victims, they need to be transported to more remote hospitals. In these cases, mass transportation modes such as trains or ships could assist to promote faster transportation. However, to be useful, they must arrive at the site of the incident as fast as possible. Therefore, we present a stochastic mathematical model that simultaneously determines the optimal fleet size of relief trains to be stationed as well as the optimal locations to prepare for patient transport after mass-casualty events. We test our model in a case study in the German state of Bavaria.

In this work, a special type of the travelling salesman problem (TSP), namely the clustered TSP (CTSP), is addressed. In the CTSP, the cities are already divided into clusters and the salesman seeks to find the shortest tour through all the cities which includes each city exactly once while being restricted to visit the cities of each cluster contiguously. Due to the NP-hardness of the focused problem, a hybrid metaheuristic consisting of the artificial bee colony (ABC) and the tabu search (TS) algorithm is proposed to deal with it. The results of our solution approach on two sets of benchmark instances are presented and compared with those of two other methods from the literature.

In recent years, there has been significant research interest in solving Quadratic Unconstrained Binary Optimisation (QUBO) problems. Physics-inspired optimisation algorithms have been proposed for deriving optimal or sub-optimal solutions to QUBOs. These methods are particularly attractive within the context of using specialised hardware, such as quantum computers, application specific CMOS and other high performance computing resources for solving optimisation problems. Examples of such solvers are D-wave’s Quantum Annealer and Fujitsu’s Digital Annealer. These solvers are then applied to QUBO formulations of combinatorial optimisation problems. Quantum and quantum-inspired optimisation algorithms have shown promising performance when applied to academic benchmarks as well as real-world problems. However, QUBO solvers are single objective solvers. To make them more efficient at solving problems with multiple objectives, a decision on how to convert such multi-objective problems to single-objective problems need to be made. In this study, we compare methods of deriving scalarisation weights when combining two objectives of the cardinality constrained mean-variance portfolio optimisation problem into one. We show significant performance improvement (measured in terms of hypervolume) when using a method that iteratively fills the largest space in the Pareto front compared to a naïve approach using uniformly generated weights.

In this paper, we consider the orienteering problem with hotel selection (OPHS) and introduce an additional cap on the available budget, frequently experienced in practice but neglected in the literature. We present a heuristic solution approach for the modified problem configuration, which comprises the construction of initial solutions and the improvement of those solutions using a multi-start VNS heuristic with adaptive adjustment. Within computational studies on adjusted benchmark data, we evaluate the impact of the considered budget constraint by comparing the results of the original OPHS with the results of its budget constraint extension. Further, we show the efficiency of the proposed metaheuristic.

In a distributer freight train, its cars have to be sorted in such a way that during the trip of the train, only cars on the rear the train are left at any intermediate stop. The sorting requirements of the freight cars of a distributer train will be described in this paper. Then an algorithm is introduced that minimizes the number of tracks that are necessary to get the freight cars sorted in a hump yard, such that the sorting requirements are fulfilled. The algorithm is a dynamic programming approach and generalizes the algorithm of Falsafain and Tamannaei.

The multi-compartment vehicle routing problem (MC-VRP) consists of designing a set of routes to perform the collection of different product types from customer locations with minimal costs. The MC-VRP arises in several practical situations, such as selective waste collection or different color of glass collection. Compartment sizes can be either set as fixed or as flexible. Often in practice, the collection quantity from customers is stochastic in nature, that is, the exact value is not available during route planning and is known only once the vehicles are at the customers’ locations. Our work introduces the MC-VRP with stochastic customer demands and with flexible compartment sizes. We propose a genetic algorithm (GA) to solve this problem and investigate the benefits of setting the compartment sizes to be flexible instead of fixed with pre-defined sizes. By using flexible compartment sizes, the GA shows an overall average improvement of 7.8%, compared to the state-of-the-art approach for fixed compartment sizes.

The high number of transports in industry causes a significant proportion of the emissions that need to be reduced regarding global warming. A part of the transports is caused by lateral transshipments, which occur when parts are distributed among different locations within one echelon for various reasons. To be able to reduce the number of transports, proactive transshipments can be used. In proactive transshipments, parts are transshipped at predetermined points in time before their demand occurs. In the scientific literature, the consideration of emissions, as well as a differentiation of vehicle types, are neglected in planning models to proactive transshipments. Our planning model adopts emission limits and provides a detailed consideration of different vehicle types for the execution of transshipments. We decide about the types and amount of vehicles used between locations in different periods. An illustrative example is presented, comparing the costs and emissions between proactive and reactive transshipments with and without emission limits. We find that emission limits can influence vehicle type selection and the products to be transshipped during proactive transshipments.

This paper discusses a basic dynamic lot-sizing model for a single item with rework of defectives. Due to the imperfect production process, some fraction of the generated items is not of sufficient quality. After rework, these goods serve the same demand as the initial perfect quality items; both are called serviceables.The internal processes production and rework are conducted independently from another at different or same periods. The basic model is proven to be NP-hard. We present three main unique characteristics that describe the specific model behavior observed in the optimal solutions: production only, multiple rework, and overproduction of serviceables. Different MIP formulations are derived to analyse the effects of these three characteristics on the optimal solutions that exclude each of these characteristics from the basic model. Afterward, computations for given data sets are conducted, using all different MIP formulations. It can be shown that production only occurs most frequently and has the highest effect on the total cost.

We consider routing problems where agents have preferences over pickup and delivery travel options. We look at the class of mechanisms that maximise social welfare. We study computing outcomes with such mechanisms. We also show that agents can manipulate such mechanisms. In response, we study computing pure Nash equilibria induced by such mechanisms. Finally, we analyse the price of anarchy for such mechanisms, which quantifies the welfare loss in an equilibrium.

This study addresses the two-echelon vehicle routing problem with grey zones and Customer-to-Parcel (C2P) stations. This problem arises in the search for new sustainable delivery schemes for e-commerce and last-mile distribution in urban areas and aims at reducing last-mile transportation costs. This study proposes a literature review on the subject, and a mixed integer linear programming (MILP) formulation to model and solve small instances of the described problem.

Car sharing systems (CSSs) are one of the environmentally beneficial solutions in urban transportation. However, the operators still struggle to make these systems profitable. One of the main contributors in operational cost is rebalancing operations. Therefore, it is important to identify strategies that are tailored according to the needs of the considered system. To overcome this challenge, this work proposes a simulation-optimization framework that compares different rebalancing operations strategies in one-way station-based car sharing systems in terms of cost and level of service. The simulation module utilizes the Multi-Agent Transport Simulation Toolkit (MATSim) whilst the rebalancing operations are determined in the optimization module. The framework allows us to explore the different uncertainties that can occur in the system, such as fluctuations in trip demand thanks to the MATSim. The results of the framework help the operator to better analyze the system and the best rebalancing strategy under different scenarios.

On the basis of economic and user-relevant criteria, this paper proposes an assessment framework for zero-emission vehicles. This framework enables a transparent evaluation of different zero-emission vehicles (ZEVs) from the perspective of (potential) customers, based on the Analytic Hierarchy Process (AHP) approach for a multi-criteria decision analysis. The relevant criteria for the evaluation were derived from literature and from semi-structured interviews. These interviews were held with individuals having driving experience with both battery electric and fuel cell electric vehicles. An AHP survey was also conducted with ZEV drivers and ZEV-interested individuals for the criteria weights to be determined. Seven criteria were found to be particularly relevant for evaluating zero-emission vehicles: total cost, range, charging/refueling time, charging/refueling infrastructure availability, greenhouse gas emissions, spaciousness, and driving dynamics. The assessment framework includes value scores representing the degree to which a specific ZEV satisfies a given quality criterion. The proposed framework combines these scores with the criteria weights derived in the AHP. The framework is useful for the design of ZEVs by vehicle manufacturers.

We propose an adaptive modular evolutionary scheme for optimizing the NP-Hard integrated timetabling and vehicle scheduling problem (TTVSP) in public bus transit. Various heterogeneous mutation operators are utilized within the scheme. Depending on their impact on the computed timetable and corresponding bus schedule, each mutation operator will be adaptively applied with a weighted probability in future iterations. To validate the solution quality and runtime of the heuristic, we exactly solve the TTVSP considering vehicle fix and operational costs. The experiments show that the heuristic computes a cost-optimal solution in a reasonable time. We further increased the solution quality relating to service quality by extending the utilized approach considering both short connecting times between different lines and good headways.

The availability of refueling stations for alternative fuels is of high importance for the penetration of alternative fuel vehicles (AFV) in today’s markets and is an important lever for sustainable mobility. Therefore, refueling station network deployment strategies are currently of great interest to researchers. Several quantitative location models have been specifically developed, adapted, and applied to determine optimal refueling station locations and network build-up strategies for AFV refueling from an overall system’s perspective. However, the network operator’s intrinsic economic objectives and its preference for transforming the existing refueling station network rather than building up new refueling stations are subordinate in most research. That is why a new formulation of a flow-based location model is introduced in this study that not only decides on the locations of new alternative refueling stations but also on the provisioned capacities of multiple fuel types at each newly opened and each already existing refueling station in a multi-period planning horizon. This new approach uses a profit-oriented, cash flow-based objective function to transform the existing refueling station network over time. A case study depicting German highways is presented to validate the new model and the computational results are discussed.

Planning so-called green waves along major road axes is a well-established target for traffic engineers. This is mainly for two reasons: a smooth traffic flow quality, and less air pollution. For one-way road axes (e.g., the Avenues in Manhattan), this is a trivial downstream task. For bidirectional arterials, there is a well-known necessary condition for establishing a green wave in both directions: The driving times between two subsequent crossings must be integer multiples of half of the cycle time of the signal programs at the nodes. In this presentation, we propose an integer linear optimization model to establish fixed-time green waves in both directions that are as long and as wide as possible, even in the situation where the above-mentioned driving time condition is not fulfilled. In particular, we are considering an arterial along whose nodes separate left-turn signal groups are realized. In our computational results, we show that scheduling left-turn phases before or after the straight phases can reduce waiting times along the arterial. Moreover, we show that there is always a solution with green waves in both directions that are as long and as wide as possible, where absolute priority is put on just one direction. Only when considering prioritized parts of a green band (e.g. some first few seconds), then an ideal green wave into one direction can provide suboptimal quality compared to optimizing both directions together. Finally, we validate the nominal solution quality according to the objective function values with the results of corresponding runs of the well-established traffic flow simulation tool SUMO.

Many European countries plan their railway infrastructure according to strategic timetables, using them as input for further strategic and tactical planning steps, including network design. While both tactical timetabling and network design are well covered by the research, there exists, to the best of the authors’ knowledge, no model which focuses on network design based on strategic timetables. In this short paper, we present an overview of our research. The traditional network design problem is extended to incorporate railway-specific features such as headway-based capacity estimations and demand derived from a strategic timetable. This includes trains represented by integral flows with start, destination, and time bounds as well as timetabling constraints for line frequencies and transfers. Since strategic timetabling is done many years in advance, the strategic timetable and as such the demand for the network design problem are subject to uncertainty. To account for this, we aim to calculate networks which are robust towards changing input timetables. We describe two approaches to model this: optimizing the network for a timetable family (a set of discrete scenarios) and varying demands within scenarios, which is modelled by a set of optional trains. The paper describes basic modelling decisions, details the approach to incorporate the uncertainty and shows the main features of the optimization model as well as a case study and further research.

Land use management problems generally requires the use of geographic information systems (GIS). To do this, it is first necessary to take into account several often conflicting criteria. However, given the limited capabilities of GIS, the representation of a decision map resulting from the overlaying of more than four conflicting criteria maps is problematic. To overcome this difficulty and enhance the analytical capabilities of GIS, the literature presents a variety of works on the joint use of GIS and multiple criteria decision making (MCDM) methods. In these works, the limitations that emerge are firstly the difficulty of fully integrating a MCDM method into the GIS, secondly the large number of parameters of the MCDM method to be integrated into the GIS, thirdly the aggregation of heterogeneous criteria. In this work we propose an answer to these three concerns through the integration of the KEMIRA method in the GIS open source QGIS and illustrate our approach on a real case study in Burkina Faso.

The discussed two-dimensional nesting problem is motivated by the production of differently shaped tiles of laminated safety glass that can be represented by primitive, convex polygons. Within as few rectangular bins as possible, representing the space of a furnace, tiles must be placed without overlapping. While the primary problem is to minimize the number of occupied bins, distances between adjacent tiles or a tile and an adjacent furnace boundary must be neither too small nor too large to ensure the stability of the furnace filling during a lamination process. To fulfill this condition, a minimum number of additional rectangular support plates must be added. These plates are considered equivalent to tiles when measuring distances. This is a new aspect that, to our knowledge, has not been covered in the literature so far. We represent the problem as a mixed integer linear program based on no-fit polygons and compare results with those of a greedy-type heuristic.

In this contribution, we consider an outsourcing problem based on a specific principal–agent relationship with hidden characteristics. Under the assumption that the principal knows the probability distribution on the agent’s discrete type space, a standard solution technique for the resulting contracting problem is stochastic optimization on the set of incentive compatible menus of contracts from which the agent can choose a single contract according to the take-it-or-leave-it principle, respectively. Admittedly, this approach neglects any sort of uncertainties in the post-contract phase which is not realistic in many practical environments like production and logistics. To address this issue, we present a novel and holistic problem formulation that links the contracting phase to an uncertain execution phase in a logistical context containing the possibility of renegotiating contracts as a reaction to environmental changes. Since the resulting problem has the character of a multi-round game, we apply well-known concepts from the trendy AI-area of Deep Reinforcement Learning to exploit clever contracting strategies for the principal. Finally, we evaluate our approach inside a computational study.

In Wire-Arc Additive Manufacturing, the desired workpiece is built layer-wise by a weld source moving freely over a substrate plate, either welding or transiting. The main issue with this manufacturing technique is the temperature distribution within the workpiece since the large thermal gradients caused by the welding process lead to thermal stress. We consider the trajectory planning problem of finding an optimal welding trajectory for a given two-dimensional layer. It is formulated as a mixed-integer linear problem (MILP) searching the welding path with the most homogeneous temperature distribution during manufacturing. The heat conduction, the weld source, and the heat radiation are incorporated into the model, together with two coupled time discretizations to accelerate the solution process. For two example surfaces, the computed optimal trajectory is compared to commonly used strategies like raster, zigzag or spiral paths.

In many markets, customers as well as retailers look for increased sustainability. Recommerce markets—which offer the opportunity to trade in and resell used products—are constantly growing and help to use resources more efficiently. To additionally manage the trade in and resell prices for used product versions is challenging for retailers as substitution and cannibalization effects have to be taken into account. An unknown customer behaviour as well as competition with other merchants regarding both sales and buying back resources further increases the problem’s complexity. Data-driven pricing agents offer the potential to find well-performing strategies and satisfy the need for automated decision support, particularly in online markets. As the training of such agents is typically data hungry and too costly to be performed in practice, synthetic test environments are needed to try out and evaluate self-learning pricing agents in different market scenarios. In this paper, we propose a conceptual approach for such a recommerce market simulation framework and its basic components. Further, we discuss requirements and opportunities to study self-learning strategies in synthetic markets.

Self-learning agents can be used in numerous ways for dynamic pricing nowadays. It has been shown, that reinforcement learning can serve as a toolkit to efficiently develop pricing strategies in dynamic environments. In many real-world situations, it can be expected that multiple market participants rely on such self-learning agents to implement pricing decisions. From the view of one agent, this violates the fundamental Markov property, which leads to instability in the learning process. Past publications proposed to rely on asymmetric information to achieve equilibria and usually focused on tabular solutions or solvers. We use multi-agent learning and asymmetric information with function approximation tools for high-dimensional state spaces by exchanging policy information between multiple actors. We discuss possible problems and their solutions and propose a simulation environment for further evaluation of the developed system.

We consider a scheduling problem motivated by a particular process step in the production of rubber flooring. In this step, a set of given jobs of different colors has to pass through a heated pressure roller, one job at a time. Since different jobs require different process temperatures, and heating up the machine is expensive, we want to minimize the total temperature change. On the other hand, we aim at minimizing the alternations between bright and dark colors to minimize cleaning effort. We provide an efficient heuristic for finding all job sequences that are Pareto optimal with respect to the two objectives.

We consider a job shop scheduling problem with time windows, flexible energy prices, and machines whose energy consumption depends on their operational state (offline, ramp-up, setup, processing, standby or ramp-down). The goal is to find a valid schedule that minimizes the overall energy cost. To solve this problem to optimality, we developed a branch-and-bound algorithm based on a time-indexed integer linear programming (ILP) formulation, which uses binary variables that describe blocks spanning multiple inactive periods on the machines. In this paper, we discuss the propagation and branching schemes used in that algorithm. The strategies, which are specifically tailored for energy related machine scheduling problems, primarily aim to determine and sharpen the activity profiles of the machines (and thus reduce the number of the inactive block variables) and address the workload profile of the tasks with lower priority. Computational experiments validate the efficiency of those techniques.

We study a practical production planning problem that was encountered by an industrial partner. Mathematically, the problem can be described as an unrelated parallel machine scheduling problem with deadlines and sequence-dependent setup times. The setup times depend on certain job attributes (e.g. material, color, size) of the successive products. For all products, dated demands are given that must be met with the means of one or multiple production orders that can be assigned to a set of eligible machines. We describe and evaluate different approaches to solve the scheduling problem. Our final algorithm is a combination of an iterated greedy construction heuristic together with a constraint program to locally improve the solution found by the heuristic. For most of our test instances, our algorithm achieves more than 20% reduction of setup times compared to the original solutions used by our industrial partner.

In the fast-growing online market, e-grocery providers in particular advertise fast delivery times. These can only be achieved through flexible and compact warehouse solutions close to the customer. Storage providers have recognized this need and therefore offer new types of storage solutions. However, this development has so far received little attention from the scientific community. Therefore, this paper introduces a new type of storage and addresses the modeling of storage and retrieval in terms of cost factors and their composition.

Aviation has been criticized for its negative climate impact in the past few years due to the emission of harmful greenhouse gasses (GHGs) such as CO2, NOx and water vapor that can cause the formation of climate harming ozone and contrails. In recent years, many different technologies have surfaced that have the potential to reduce emissions and replace the existing conventional jet fuel technology. On one hand, H2 powered aviation just recently regained high attention from the industry, e.g., Airbus launched the ZEROe program where they pledged to develop the world’s first zero-emission commercial aircraft by 2035. On the other hand, sustainable aviation fuels (SAF) or biofuels have been identified as an alternative option. Given the different promising future technologies, it is difficult to predict their role in transition pathways that will lead the air transport system towards a more sustainable future. We develop a global scale system dynamics air transport system simulation model and incorporate components like the new potential technologies, production side emissions of new fuels, i.e., SAF and hydrogen, air travel demand, airline industry and aircraft manufacturers. We also include the long, medium and short haul segments of flights. Using this model, we analyze the adoption trends of new technologies and fuels by assessing the amount of fleet operated with them and its effect on emission reduction within each flight segment.

Game theory has been recently used to study optimal advice-giving strategies in settings where multiple advisers compete for a single client’s attention. In the advisers’ game, a client chooses between two well informed advisers to place bets under uncertainty. Experiments have shown that human advisers can learn to play strategically instead of honestly to exploit client behavior. Here, we analyze under which conditions agents trained with Q-learning can adopt similar strategies. To this end, the agent is trained against different heuristics and itself.

In a groundbreaking paper McAfee introduced the Trade Reduction (TR) Mechanism that circumvents the famous Myerson and Satterthwaite impossibility result by sacrificing a small amount of efficiency. Here the author creates order statistics based on the submitted bids and reduces at most the least efficient trade. Based on this principle an alternate mechanism was proposed by Segal-Halevi et al. which extends this to the strongly budget balanced setting. This paper proposes a generalization of these two TR mechanisms to fit into a larger framework that can be implemented based on the market in which the auction is to be applied. Additionally, by taking advantage of the relationship of bid order-statistics a novel mechanism; titled BORS, is revealed to complete the GTRF. Using a simulation based evaluation, performance is characterized across various settings in order to achieve optimized results.

This paper analyzes the iterated version of the well-known Boxed Pigs game through Reinforcement Learning. In this scenario, there are two differently sized players (pigs) that compete against each other. The core idea is about sacrificing a pay-off in order to generate some rewards. In our iterated version, these pigs play this game repeatedly using different strategies. We carry out two experiments: in the first one, we train two Q-learning agents against each other to see which equilibrium will be generated. In the second one, we pit the Reinforcement Learning agent against a fixed policy pig. The results of this experiment confirm the ability of Reinforcement Learning techniques in finding the best strategy for maximizing the return independently from the other player choices.

Even though simulation is mainly used for computer models with inexact outputs, there are direct benefits in viewing results from samples of an existing dataset as replications of a stochastic simulation. We propose building Machine Learning prediction models with the Monte Carlo approach. This allows more specific accountability for the underlying distribution of the data and the impact of uncertainty in the input data in terms of bias. We opt for nonparametric input uncertainty with multi-level bootstrapping to make the framework applicable to large datasets. The cost of Monte Carlo-based model construction is controllable with optimal designs of nested bootstrapping and integrating variance reduction strategies. The benefit is substantial in providing more robustness in the predictions. Implementation in a data-driven simulation optimization problem further indicates the superiority of the proposed method compared to the state-of-the-art methods.

Twenty years after its first commercial release, the Xpress Mosel software keeps evolving driven by user requirements, usage patterns and technological advances. This contribution takes the reader through the major phases of its development: Mosel was initially designed as an optimization modeling language that also provided programming features. Over time the Mosel distribution has been enriched with numerous components and tools addressing a variety of purposes. The increasing use of Mosel as general-purpose programming language was recognized by turning it into a free software a few years ago. Motivation and use cases for major new programming features are discussed in detail.

This study proposes an AI-empowered order lead time prediction integrating a multidimensional real-world dataset from a semiconductor manufacturer’s supply chain. Examined features capture order–, delivery–, planning–, customer–, and product– related information. We thoroughly analyze a broad spectrum of machine learning algorithms ranging from linear regression and tree-based models to neural networks and compare them with respect to prediction performance, computation time, and understandability. We find that boosting algorithms demonstrate solid predictive performance with the highest accuracy and most efficient computation time. Our results allow supply chain experts to obtain data-informed estimations of order lead times and an understanding of the predictive mechanisms.

The current pandemic has disrupted many supply chains, among them food supply chains, and their physical and financial flows. Therefore, companies with high bargaining power may extend their payment terms to downstream suppliers as a reaction to decreased financial cash flow. We model a stylized three-stage food supply chain and use simulation to analyze the effects of a demand shortfall. We then investigate the effects of payment term extensions by different companies. We find that although extending payment terms can be beneficial for a single company in the short run, it will harm the supply chain in the long-run and conclude that incentives should be put into place to motivate companies accordingly.

We introduce a two-level stochastic capacitated lot-sizing problem with random demand and service level constraints under the static and static-dynamic uncertainty strategy. While the static strategy determines setup periods and lot-sizes at the beginning of the planning horizon, the static-dynamic strategy allows adjustments of lot-sizes during the planning horizon. We present a model formulation of the demand difference adaptation policy for a multi-stage production system. Thereby, lot-size adaptations depend on demand differences between the realized and expected demand. We evaluate 144 test instances with different parameter settings to quantify the economic efficiency of the uncertainty strategies for the multilevel lot-sizing problem. The computational study demonstrates that additional costs of semifinished goods and scarcity of storage capacity on upstream precesses reduce the cost saving potential of lot-size adaptations.