Operational problems often span wide range of options. In the past, due to computational limitations, the trend was to limit the options set to the minimum number possible. However, with the increase in computational capacity over the last decade, it is now often possible to parametrize the option space instead, and simulate hundred or even thousands of options. One of the first attempts in the defence domain was the US Marine Corps Project Albert which looked at data farming in tactical combat modeling. However, simulating vast numbers of options poses new challenges for managing experiments and conducting post-simulation analysis. Some of the model management challenges are: which simulations have been conducted, what option space has or has not been explored, which output maps to which input, etc. The analysis problems include considerations such as what model inputs typically lead to what model outputs, whether the results covering a subset of possible options are sufficiently representative for the entire set of possibilities, and how to visualize dependences on the inputs in multi-dimensional problems. This paper will focus on combining the field of visual analytics with modeling and simulation for a, somewhat simplified, problem of strategic air lift. Using this problem, that can be summarized as: “what is the force structure requirement for the strategic airlift to meet logistics demand of concurrent operations as mandated by the Government of Canada's defense policy Strong, Secure, Engaged?”, the paper will look at the management of the experimental frames for simulation, option space coverage, and visual analytics applications to the output. Common visualizations approaches such as generalized pairs plots, maps, as well as 3D visualization will be exploited to provide an innovative experimentation management and analytics framework.
Simulation
Using Visual Analytics to Manage Experimental Frames
19 views
1 Downloads