The purpose of computer simulation is to utilize a mathematical model to recreate a real world situation such that the behavior and interactions of the entities involved can more easily be understood. By varying initial conditions and external stimuli in a controlled simulated environment, simulation often provides much better insight to an entity’s behavior than would be readily observable in the real world.
Traditional simulation design relies upon dynamically varying a set of input parameters and comparing the simulation outputs to determine parameter sensitivities. Each parameter which will be dynamically altered requires a separate piece of code, generally a looping structure, to control the changing parameter values. Though many input parameters exist within the scope of a simulation, traditional design facilitates the measurement of only a small subset of these parameters as it is not feasible to write the code to alter all input parameters. Varying the number of dynamic parameters is costly due to required software changes and it is not always known during the development phase which parameters are of interest.
One approach for altering nearly all parameters in a simulation, rather than a small subset, is through the use of reflective architectures. Rather than statically defining and altering a small subset of input parameters, reflective architectures provide facilities whereby any simulation object which exists in memory can be examined and altered at run time. The premise for the work in this paper is that a framework built upon reflective architecture can be built which provides a simple, flexible mechanism for manipulation of computer simulation input parameters. This thesis outlines a solution based upon reflective architectures and describes a framework which was created to facilitate this purpose.
