Omnipresent software provides a cost-effective, powerful tool for model building that is only limited by imagination.
By James J. Swain
The idea of simulation is now a part of our culture. For example, the popular “Matrix” movies had it that our perceived reality was actually a simulation, while a recent episode of the British series “Dr. Who” imagined that the world we knew was a simulation being used by an alien race to fully study our planet and to map out its weaknesses for conquest. Plots could incorporate simulation because the viewers were already familiar with the concept. In fact, this seems to have spawned a recurring question on the Quora website about the possibility that we are, in fact, living in a simulation.
Movies are not the only exposure that we have to simulation in popular culture. Surely everyone is familiar with various computer games in which some world is experienced through simulation. The gaming industry is on track to surpass music and movies in monetary value which is further evidence of their importance in media and our shared cultural experience.
Clearly, most people are both familiar and comfortable with the notion of simulation, accepting that we can use computer software to make a sufficiently realistic rendering of the world as it is or as it might be (zombies!) to explore, to manipulate or simply to entertain. The boundaries to that world will continue to blur, since VR technology may soon allow us to walk around within a simulated world rather than view it from a stationary game or computer screen. At that point, we may have to provide ourselves with a cue to remind ourselves when we are in reality and when in a simulation.
The products that we survey here are designed for a more mundane but critical purpose – to provide business, industry and government agencies the tools to explore, quantify and optimize systems of all kinds. Whether the application is a factory or a port, data center or hospital, local or worldwide logistics system, or military operations, simulation provides the ability to realistically represent and then experiment with myriad systems. An examination of the vendor websites provides a wide catalog of applications for these and other applications.
Of course, computer-based simulation is an immense field that spans continuous simulations such as missile flight or fluid flow, the cosmically large or the microscopically small. Our focus in this survey is restricted to the realm of discrete-event simulation models that are particularly suited to the operation of discrete parts, patients, vehicles and such in widely divergent fields as manufacturing, healthcare and other services settings, as well as logistics, transportation or military operations to name a few areas of application.
This survey provides a snapshot of a vital and robust area of simulation analysis software. The vendors document both the growth and sophistication of their tools. Of perhaps greater interest is to examine the case studies and the white papers that many provide, demonstrating how the tools were successfully applied and the benefits that were obtained from the effort. These materials are available on their respective websites.
The Value of Simulation
For over half a century, simulation has grown as a tool because it provides us with a cost-effective and immensely powerful instrument for model building that is only limited by imagination. Simulation has grown with the capability of computers, programming tools, graphical representation and algorithms to support these areas. It was recognized quite early that computers would be ideal for implementing simulations. Of course, what began as a tool for analysis has morphed into a vehicle for entertainment through gaming, and the explosive (and lucrative) growth in that area has propelled further developments in simulation generally.
Beside the growth in simulation tools, we have witnessed a steady growth in the body of knowledge about simulation and the organization of simulations. Data is more accessible and more readily incorporated into models and increasing the level of detail that can be represented. And, of course, simulation tools incorporate the lessons of the experience in the field. Meanwhile, enabling technologies, such as random number generation, have steadily improved over the decades. The state of the art in random number generation provides virtually unlimited, high-quality random numbers.
Another reason that simulation is so valuable is that experimentation with the real world is rarely feasible. This is especially true both for the military operations and in aircraft trainers – you can walk away from both a lost simulated battle and a simulated crash, after all. But even in a real system, there are economic and practical limitations to the amount of experimentation that is tolerable. Plant managers have their quotas and deadlines, for instance. Most of all, simulation can be applied to systems that have not yet been fielded. Simulation has been successful because it is possible to build simulation models that are sufficiently valid and credible for an analysis to be useful in all of these cases.
One of the primary advantages of simulation is that it is a constructive tool. Simulation models attempt to reproduce the states and trajectories of the actual system, using the procedures that the real system would use. This has many advantages. First, it means that model building is based upon a description of the process that is already available or would be a part of any design. Second, it can be efficient, since many modeling languages have evolved generalized constructs or modules such as servers or workstations that fit many modeling situations, or have evolved specialized instances applicable to particular fields such as manufacturing or healthcare, with specialized logic and animation built in.
Constructive models are certainly easier to understand and to explain. While a queueing formula may provide a prediction of the mean waiting time in a queue, its derivation is opaque to most clients and its application is limited. A simulation model provides a statistical estimate, but in most cases, it is feasible to obtain precise estimates (as well as estimates of the precision). Having a flexible tool that can be applied in a wide variety of modeling scenarios is worth the loss of having an exact solution.
Another critical advantage of constructive models is that they generate the states of the system as they would evolve. This state information can be used to drive the animation and permits detailed analysis, including detailed statistical observations about the system. Of course, the former provides a powerful starting point for both validation and user credence. Providing a visually accurate portrayal of the system in question gives increased confidence in any analysis derived from the model. It also can be used to give snapshots about specific examples of typical or worst-case behavior that can be examined in detail.
Animation also can be useful as a tool to facilitate communication and understanding among team members, who often come from different disciplines and backgrounds. Given the widespread familiarity with simulation games, it is likely that simulation tools cannot afford to be without animation output for any model. Many of the tools in this survey provide libraries of images that allow the realistic portrayal of equipment, parts or people to make the visualization accurate. Most now provide 3-D animation that can be examined from different viewpoints and at various levels of detail. Here simulation benefits from the great strides made in the gaming industry for representation and perhaps the growth of graphics processors on home computers for gamers.
Improvements in programming increasingly allow models to include increased decision-making within the model. For instance, models can incorporate externally linked software for dynamic scheduling. Likewise, simulations increasingly include the ability to provide customers (and other objects) with decision capabilities or to interact with each other. As an example, for simulations that include pedestrian flow, more realistic routes can be plotted. And in a simulation in which an emergency evacuation is to be studied, being able to portray crowding or panic would be an important part of the model.
It has long been appreciated that the constructive approach allows a more detailed analysis of the system statistics. Whereas the queueing formula may be limited to a prediction of the mean, in the simulation model it is possible to observe the variability of the response, not to mention relations among different statistics. This is just the beginning of the possibilities! Increasingly it has been noted that analytics and simulation might be good partners. Simulation can provide detailed data that can now be stored and accessed for insights using the tools of analytics, both within simulation replications and between different scenarios or cases.
Once a valid and credible model has been built, it can be used for experimentation and optimization. In the former case, many of the simulation tools have the capability of generating experimental results automatically. They are sometimes called scenario generators or experiments. That is, the user can specify the parameters values to be run, and the complete set of replications can be performed automatically, with the results available as a group. The results can then be compared graphically and statistically. In most cases, the results can be exported for a more detailed analysis using statistical software.
Once a simulated process has been developed, it is frequently (and naturally) desirable to optimize the process or to seek out the best from among specific options. Optimizers such as OptQuest usually use heuristic approaches (e.g., Tabu search) that seem to work well where little can be assumed about the simulation responses. Optimization is used when the possible scenarios are too numerous to compare directly, so an algorithm is used to search among the parameter values to obtain improved responses. Various ranking and selection methods can be used when there are a fixed number of alternatives to compare to obtain the best solution. These algorithms are often sequential in nature, determining the number of replications necessary to make a statistically valid decision, depending on how “best” is defined.
Several products have the ability to perform sensitivity analysis on a model scenario given uncertainty in the model parameters. This can be used to determine the parameters that the responses are most sensitive to.
A final reason for simulation success is that the number of people trained in simulation has steadily increased. Most vendors provide discounted student versions of their software and training materials and books for use in the classroom. All industrial engineering programs include simulation in their curriculum, and simulation also is offered in many business and computer science programs as well. Most vendors also have active training programs for new users, with classes available across the world, and many provide blogs, forums and even product-oriented conferences to share experience and new product developments.
The 2017 Survey
This survey is the 11th biennial survey of simulation software for discrete event systems simulation and related products (Swain, 2015 [1]). All product information has been provided by the vendors. Products that run on personal computers to perform discrete-event simulation have been emphasized, since these are the most suitable for usage in management science and operations research. Simulation products whose primary capability is continuous simulation (systems of differential equations observed in physical systems) or training (e.g., aircraft and simulators) are omitted here.
This year’s survey includes 44 product listings taken from 26 vendors who submitted survey questionnaires. The range and variety of these products continues to grow, reflecting the robustness of the products and the increasing sophistication of the users. The information elicited in the survey from the vendors is intended to provide a general gauge of the product’s capability, special features and usage. This survey includes information about experimental run control (e.g., experimental design and automated scenario run capabilities) and special viewing features, including the ability to produce animations or demonstrations that can run independent of the simulation software itself. A separate directory listing gives contact information for all of the vendors whose products are in the survey. An online version of the survey will include vendors who missed the publishing deadline. Of course, most of the vendors provide their own websites with further details about their products. Many of the vendors also have active user groups that share experience in the specialized use of the software and are provided with special access to training and program updates.
A number of technical and professional organizations and conferences are devoted to the application and methodology of simulation. The INFORMS journals Management Science, Operations Research and Interfaces publish articles on simulation. The INFORMS Simulation Society sponsors simulation sessions at the national INFORMS meeting and makes awards for both the best simulation publication and recognition of service in the area, including the Lifetime Achievement Award for service to the area of simulation. Further information about the Simulation Society can be obtained from the website http://connect.informs.org/simulation/home. This site also contains links to many vendors of simulation products and sources of information about simulation, simulation education and references about simulation. The Society for Modeling and Simulation International (www.scs.org) is devoted to all aspects of simulation. Its conferences include three multi-conferences in the spring, summer and autumn that cover all aspects of simulation practice and theory. The AlaSim International Conference is a relatively new conference hosted in Huntsville, Ala., with a focus on DOD and other government applications of simulation. The next conference is scheduled for October 2017.
The INFORMS Simulation Society and the Society for Modeling and Simulation are both sponsors of the annual Winter Simulation Conference (see page 32). This year’s conference is the 50th Anniversary conference and will be held Dec. 3-6 in Las Vegas. Further information and registration information is available from the site www.wintersim.org. This site also links to the complete contents of the Proceedings of the Winter Simulation Conference from 1968 to 2016 for ready access to research and applications of simulation.
James J. Swain ([email protected]) is a professor in, and chair of, the ISEEM department at the University of Alabama in Huntsville.
Reference
- Swain, J. J., “Simulation software survey: Simulated worlds,” OR/MS Today, Vol. 42, No. 5, October 2015, pp. 36-49.