Title :
Run Watchers: Automatic Simulation-Based Decision Support in Flood Management
Author :
Konev, Artem ; Waser, Jurgen ; Sadransky, Bernhard ; Cornel, Daniel ; Perdigao, Rui A. P. ; Horvath, Zsolt ; Groller, M. Eduard
Author_Institution :
VRVis Vienna, Vienna, Austria
Abstract :
In this paper, we introduce a simulation-based approach to design protection plans for flood events. Existing solutions require a lot of computation time for an exhaustive search, or demand for a time-consuming expert supervision and steering. We present a faster alternative based on the automated control of multiple parallel simulation runs. Run Watchers are dedicated system components authorized to monitor simulation runs, terminate them, and start new runs originating from existing ones according to domain-specific rules. This approach allows for a more efficient traversal of the search space and overall performance improvements due to a re-use of simulated states and early termination of failed runs. In the course of search, Run Watchers generate large and complex decision trees. We visualize the entire set of decisions made by Run Watchers using interactive, clustered timelines. In addition, we present visualizations to explain the resulting response plans. Run Watchers automatically generate storyboards to convey plan details and to justify the underlying decisions, including those which leave particular buildings unprotected. We evaluate our solution with domain experts.
Keywords :
data visualisation; decision support systems; decision trees; emergency management; floods; automated control; complex decision trees; computation time; decision making; disaster management; domain-specific rules; flood events; flood management; multiple parallel simulation runs; run watcher; search space; time-consuming expert supervision-and-steering; visualizations; Buildings; Computational modeling; Data visualization; Decision making; Decision trees; Monitoring; Visual analytics; Disaster management; decision making; simulation control; storytelling; visual evidence;
Journal_Title :
Visualization and Computer Graphics, IEEE Transactions on
DOI :
10.1109/TVCG.2014.2346930