Office: 214 Olsson Hall
Office Phone: (434) 924-1039
US Mail:
Department of Computer Science
A substantial amount of effort is expended on adapting existing simulations to meet new requirements. While adaptability is an issue for software in general, it is particularly interesting in the area of simulation: simulations often reflect a large degree of uncertainty, they often employ stochastics and their use is generally aimed more at increasing insight than performing any particular exact function. These attributes lend themselves to specialization when considering adaptation to new requirements. Our research exploits these attributes – in any software, but simulations in particular – in order to make the adaptation process more efficient. We are designing and implementing formal languages for expressing the flexibilities that accompany uncertainty and randomness, we are developing agile optimization methods for searching the resulting design spaces, and we are developing validation methods that work best with the incremental nature of search. We have applied our work in high energy physics, global warming studies, corrosion modeling and combustion models to demonstrate its efficacy. Learn more about our COERCE research program and our modeling and simulation technology initiative.
Unexpected model outputs can reflect valid behaviors arising from seemingly unrelated phenomena, or they can reflect errors in model assumptions, design or implementation. We are exploring analysis techniques to richly improve methods for exploring and understanding the behavior of models containing uncertainty. Increased insight gained from analysis will contribute to reduced uncertainty which in turn will increase scientist and policymaker confidence in model results and predictions. The COERCE research team has published an exploratory approach using semi-automated search that allows a user to test hypotheses about unexpected behaviors as a simulated phenomenon is driven towards a condition of interest. Our current work builds upon this work and adds an integrated multidimensional analysis capability of a model and its outputs. The multidimensional analysis combines uncertainty representation, causality analysis, and static and dynamic program slicing to gather insight in a disciplined manner into the interactions within the model that cause unexpected model behavior.
Modeling under uncertainty has been of paramount importance in the past half century, as quantitative methods of analysis have been developed to take advantage of computational resources. Simulation is gaining prominence as the proper tool of scientific analysis under circumstances where it is infeasible or impractical to directly study the system in question. According to a February 2006 report of the NSF Blue Ribbon Panel on Simulation-Based Engineering Science (SBES): “The development of reliable methodologies – algorithms, data acquisition and management procedures, software, and theory – for quantifying uncertainty in computer predictions stands as one of the most important and daunting challenges in advancing SBES”. Clearly there is a need for robust uncertainty representation and analysis methods in modeling so that scientists and policy makers can better understand and characterize the properties of the predictions they make based on their models. Our solution builds on the acausal modeling language Modelica, producing a language we call “RiskModelica,” by incorporating novel methods for quantifying uncertainty formally and robustly, for propagating that uncertainty through the modeling process and revealing its effects on model outcomes, for later use by scientists and policymakers. RiskModelica will serve as a platform for research into representation and calibration of imprecise probabilities in quantitative risk analysis simulations and for analyzing and testing imprecise probability theories (e.g. Probability Boxes, Dempster-Shafer Theory) as alternative representations of stochastic variables.
Emitters are complex. The challenge of describing their behavior for later analysis is increasingly difficult. Analysts tend to use natural (spoken) language to bridge the gap between the technology available to them for capturing emitter behavior and the complexities of modern emitters. Since computers cannot process natural language descriptions, these descriptions are unavailable to later users who require the speed of computer processing. Emitter behavior should be described formally, i.e., without recourse to natural language. We are developing a technology for formally capturing the complete dynamic behavior of emitters. The principal challenge is to develop an approach that is sufficiently powerful to capture emitter dynamic behavior completely and yet sufficiently intuitive to be used by people who are not computer programmers. Our approach is based on technologies arising from disparate areas: music composition, domain specific languages, and static analysis tools. Music composition technology is relevant to the proposed effort because a major part of the work of both the analyst and the composer is the definition, modification, and layering of waveform sequences.
Public policy officials are increasingly turning to modeling and simulation as a means to support important policy decisions. For example, with increased concern about bioterrorism, health officials have actively commissioned the creation of epidemic models in order to better prepare and plan for intervention. Significant amounts of both aleatory and epistemic uncertainty in the models as well as uncertainty about what scenario should be modeled have led to the employment of a high level of both explicit and implicit assumptions in models. If uncertainties are not carefully managed, the end-user does not have a good idea of the overall validity of the model. Consequently inaccurate results may be used to make decisions affecting millions of people and billions of dollars. In order to address this problem, we are analyzing the assumptions made in typical SEIR epidemiology models in order to establish the extent of the uncertainty that exists. Our goal is to make a case for engaging in better practices for managing uncertainty in simulations.
Isotach! See here.
