Guest Editorial

In areas such as supply chain management, Internet computing, renewable energies, manufacturing and production, or transportation and logistics, real-world systems are becoming more complex because of dynamic inter-dependencies caused by integrated approaches. These global approaches, where several subsystems are considered, are necessary in order to improve overall efficiency of systems. However, they also generate large-scale models that are not easy to be solved in reasonably short computing times. Moreover, the high level of uncertainty that characterises real-life systems and processes makes decision making a challenging task. While academic researchers from the simulation community focus on creating accurate models that take uncertainty into account, they typically use techniques of limited optimisation capability. On the contrary, the optimisation research community focuses on finding (near-) optimal solutions, but usually by simplifying real-world constraints and often ignoring uncertainty in order to make decision-making challenges analytically tractable. Vendors of optimisation software tools fall between these positions by providing richer models than the optimisation community, but typically ignoring uncertainty and adhering to classical methods that have been outperformed by modern hybrid approaches on standard benchmarks. As a consequence, industry is slow in adopting state-of-the-art methods, which creates a noticeable gap between real-world decision making and state-of-the-art solutions. As some preliminary studies have shown already, this gap can be significantly reduced by developing hybrid simulation–optimisation algorithms that combine applied optimisation methods with different simulation techniques. Simulation–optimisation algorithms have recently been successfully applied to deal with complex, large scale, and dynamic real-world scenarios under uncertainty conditions.

As a rising research field inside the simulation–optimisation arena, ‘simheuristic’ algorithms propose the integration of simulation techniques and metaheuristic methods as one of the most efficient and natural ways to deal with real-life uncertainty and complexity. This special issue aims at providing a set of selected articles that explore some of the almost countless applications of simheuristic algorithms: from transportation to energy systems, and from manufacturing to Internet computing.

Being the Guest Editors of this exciting JOS special issue, we received a considerable number of high-quality papers, which is an excellent indicator of the interest that this emerging research field is raising among the scientific community. A short overview of the articles included in this special issue is provided next, where the articles have been classified by their main application field:

• Renewable energies: The article ‘Operational Management of Renewable Energy systems with Storage using an Optimisation-based Simulation methodology’, co-authored by Mallor et al, proposes a series of operative management policies for energy storage attending two criteria: maximising the profit of selling the energy and maximising the reliability of the system as a provider of committed energy. The structure of selling prices and penalties, as well as the probabilistic wind speeds, are considered in the mathematical model, which is then solved using a simulation-based optimisation approach.

• Internet computing: The paper ‘Synthetic Generation of Social Network Data with Endorsements’, co-authored by Perez-Roses and Sebe, discusses the problem of generating synthetic data sets for simulation of online social networks—such as LinkedIn or ResearchGate—which allow any user to endorse other users’ skills. Their approach consists of two interrelated stages: the first one uses discrete-event simulation to generate the growth of the network, while the second one uses heuristics to solve a discrete optimisation problem related to the addition of the network endorsements. Also in the same field, the paper ‘Multi-Criteria Genetic Algorithm Applied to Scheduling in Multi-Cluster Environments’, by Eloi Gabaldon et al, analyses the scheduling and resource-allocation problem in multi-cluster heterogeneous environments. Considering the large-scale and heterogeneous nature of these systems, they propose a genetic algorithm based on a fitness function obtained by a simulation process. They also consider a multi-objective function to optimise not only the makespan but also the flow-time of the schedule, thus including the users’ perspective. A real workload trace is used to assess the proposed approach.

• Logistics and transportation: The article ‘Simulation-Optimization Approach for the Stochastic Location-Routing Problem’, co-authored by Herazo-Padilla et al, presents a combined simulation–optimisation approach for solving the location-routing problem with stochastic transportation costs and vehicle speeds. The hybrid approach combines a heuristic (to solve the initial location sub-problem), an ant colony optimisation metaheuristic (to solve the associated capacitated vehicle routing problem), and discrete-event simulation (to assess the expected cost of the solution).

• Manufacturing and production: The paper ‘Hybrid Approach using Simulation-based Optimisation for Job Shop Scheduling Problems’, co-authored by Kulkarni and Venkateswaran, makes use of simulation-based optimisation for solving the job shop scheduling problem. In their approach, the operational aspects of the job shop are captured in the simulation model, and two new decision variables are introduced: controller delays and queue priorities. The performance of the proposed approach is also analysed for the stochastic version of the job shop problem.

• Supply chain management: The paper ‘Simulation-based Optimization for a Capacitated Multi-Echelon Production-Inventory System’, co-authored by Güller et al, proposes a simulation-based optimisation approach to determine the inventory control parameters of a multi-echelon production-inventory system under a stochastic environment. In order to account for multiple objectives, the Pareto-dominance concept is considered. The authors develop a particle swarm optimisation algorithm to simultaneously minimise the total inventory cost and maximise the service level. A simulation model is then used in order to evaluate the control parameters generated by the optimisation algorithm. The approach is tested using data from a real-world supply chain.

Finally, we would like to thank the authors of the articles and also the anonymous referees for their invaluable collaboration and prompt responses to our enquiries. We gratefully acknowledge the support and encouragement received from the JOS editors, Simon Taylor and Christine Currie.