Multi-objective Queueing Simulation — Balancing Competing Service Metrics in Queue Systems
Multi-objective Queueing Simulation — Simultaneous optimization of competing performance metrics in simulated queuing systems · Also known as: MOQS, Multi-criteria Queueing Simulation, Multi-objective Queue Optimization, Pareto Queueing Simulation
Multi-objective queueing simulation combines discrete-event queueing models with multi-objective optimization to simultaneously evaluate and optimize conflicting performance metrics — such as average wait time, server utilization, throughput, and service cost — across a simulated queuing system. It produces a Pareto front of non-dominated solutions rather than a single optimal point, enabling decision-makers to understand trade-offs explicitly.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use multi-objective queueing simulation when a service or logistics system has multiple conflicting performance goals that cannot be reduced to a single weighted objective, when closed-form analytical queueing solutions are unavailable or too restrictive, and when the decision-maker needs to understand trade-offs explicitly rather than accept a single optimized outcome. Ideal for capacity planning in hospitals, call centers, manufacturing lines, and network services. Do NOT use when a single dominant objective applies (use standard queueing optimization instead), when analytical M/M/c or M/G/c models suffice, when simulation run-time is prohibitively expensive per evaluation, or when the number of controllable variables is very large without surrogate approximation.
Strengths & limitations
- Reveals true trade-off landscape between competing objectives, enabling informed policy choices rather than black-box recommendations.
- Handles complex, non-linear queueing dynamics and stochastic variability that analytical models cannot capture.
- Flexible system representation: supports non-Markovian arrival and service distributions, priority queues, and network topologies.
- Produces multiple viable system configurations simultaneously, giving stakeholders real options to compare.
- Compatible with evolutionary algorithms (NSGA-II, MOPSO) that are well-tested on simulation-optimization problems.
- Computationally expensive: each objective function evaluation requires a full simulation replication, multiplied by population size and generations in the optimizer.
- Pareto front approximation quality depends heavily on simulation run length, replication count, and optimizer hyperparameters.
- Stochastic noise in simulation outputs can mislead the optimizer; statistical screening or surrogate models may be required.
- Requires accurate parameterization of arrival and service distributions, which demands sufficient historical data.
- Large Pareto fronts can be difficult to communicate to non-technical stakeholders without additional decision support.
Frequently asked
How does multi-objective queueing simulation differ from standard queueing theory?
Standard queueing theory (M/M/c, M/G/1, etc.) provides closed-form or approximate analytical results for simple systems with a single performance metric. Multi-objective queueing simulation uses discrete-event models for complex, non-Markovian systems and optimizes multiple conflicting metrics simultaneously, producing a Pareto front rather than a single formula.
Which multi-objective optimizer works best with queueing simulations?
NSGA-II is the most widely used choice due to its strong diversity preservation and non-dominated sorting. For expensive simulations, surrogate-assisted methods (Kriging, radial basis functions) or Bayesian optimization reduce the number of simulation calls. The best choice depends on the number of objectives, variables, and available computational budget.
How many simulation replications are needed per configuration?
The required number depends on the variance of your performance metrics and the desired confidence interval width. A pilot study of 20–30 replications is standard practice to estimate variance; then use the t-based sample size formula to determine the final replication count. Typically 10–50 replications per configuration are used in practice.
Can this method handle non-stationary arrival rates (e.g., morning rush hours)?
Yes. Discrete-event simulation naturally supports time-varying arrival rates (inhomogeneous Poisson processes), staffing schedules, and shift changes. The simulation model simply parameterizes these dynamics; the multi-objective optimizer still searches over controllable variables such as shift staffing levels.
What if the Pareto front contains too many solutions to compare?
Post-process the Pareto front using a multi-criteria decision-making method (e.g., TOPSIS, weighted sum with stakeholder-elicited weights) to select a preferred solution, or visualize the front in 2D/3D projections. Clustering Pareto solutions into representative groups also aids communication with non-technical decision-makers.
Sources
- Banks, J., Carson, J. S., Nelson, B. L., & Nicol, D. M. (2010). Discrete-Event System Simulation (5th ed.). Pearson Prentice Hall. ISBN: 9780136062127
- Deb, K. (2001). Multi-Objective Optimization Using Evolutionary Algorithms. John Wiley & Sons. ISBN: 9780471873396
How to cite this page
ScholarGate. (2026, June 3). Multi-objective Queueing Simulation — Simultaneous optimization of competing performance metrics in simulated queuing systems. ScholarGate. https://scholargate.app/en/simulation/multi-objective-queueing-simulation
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Discrete-Event SimulationSimulation↔ compare
- Multi-objective discrete-event simulationSimulation↔ compare
- Multi-Objective OptimizationSimulation↔ compare
- Queueing SimulationSimulation↔ compare