Home     Syllabus     Lectures
Assignments     Examples     Links

People create simulation models for everything:

• Climate and weather models
• Avalanche, Earthquake, Tsunami, and Hurricane models
• Solar activity models
• Planetary and star movement and illumination models
• Highway and street traffic models
• Communications and computation models
• Electric power distribution and gas and petroleum pipeline models
• Terrain and subterrainian models
• Chemical, molecular, and subatomic particle behavior models
• Social and behavior models
• Finances, investment market, and supply chain models
• War models
• Equipment simulator and training models
• Games

Interesting Problems to Solve

Each stage in modeling and simulation depends on the quality and limitations provided in each of the steps that precede it.

• Creating Theoretical Models
  

  This requires defining what is to be in a model and what is to be ignored. Not everything can be included in a model.

  Similarly, only a small part of what goes into a simulation model represents the thing being modeled. A model is like an iceberg, in that only a small part of the model represents what is being modeled. The rest is the "behind the scenes" structure needed to get the visible tip to work right.

  Creating an accurate representative theoretical model is essential, but to be useful it must also be a theoretical model that can be turned into a simulation and its modeled behavior be able to be controlled, observed, and analyzed with some expectation of accuracy.

  
  
  
• Creating or Selecting a Computer Simulation Environment
  

  This can be fundamental to all following steps. It can enable capabilities and analyses, or it can prevent them through inflexibility. It can support accuracy, or it can prevent it.

  Some common choices for generalized simulations:

  • Excel spreadsheets -- simple but not versatile
  • Matlab, Octave, SciLab, or Mathematica -- high-level and strong math support. But a lack of low-level versatility.
  • NS2, Opnet, EMANE -- good network simulators, especially for communications networks, some usefulness in other network types.
  • MPI and FORTRAN, C, or C++ "homegrown" simulator -- Complex, but versatile and extensible to extremely large problems. Lack of standardized representations can prevent model aggregation.
  • HDF5 and FORTRAN, C, or C++ "homegrown" simulator -- Complex, but versatile and extensible to extremely large problems. The HDF5 standardization improves moving data among different simulators and models.
  
  
  
• Generating Programmed Models
  

  How is the model to be represented in the computer? What data structures are available and which will be used?

  
  
  
• Deciding the Model Parameters
  

  How should the model be altered on successive runs? Can these be just parameter values, or does it require alternate programmed algorithms?

  
  
  
• Deciding What To Observe, Measure, and Record
  

  If you can't observe it you can't measure it. If you can't measure it you can't record it. If you can't observe, measure, or record it you can't display or analyze it.

  
  
  
• How to Interpret the Results
  

  Does it accurately portray that which it is supporsed to represent? Is the model data used accurate? Is the overall model run and results verifiably correct? Can it be validated agains the real world?

  If the model and results are correct and accurate, what are the implications of its predictions or indications? Is zero power from a generator a failure or the state it is supposed to be in? Are the indications significant or just "interesting"?

  
  
  

Home     Syllabus     Lectures
Assignments     Examples     Links

Other Links:

Computer Science and Engineering Department Homepage
Information Technology Department Homepage
New Mexico Tech Homepage
New Mexico Tech Calendar