Simulation Summarization (Extended Abstract)

As increasingly large-scale multiagent simulations are being implemented, new methods are becoming necessary for concisely summarizing the results of a simulation run. Here we pose this as the problem of simulation summarization: how to extract the causally-relevant states from the trajectories of the agents. We present a simple algorithm to compress agent trajectories through state space by identifying the state transitions which have significant impact on the final outcome of interest. We apply it to a complex simulation of a major disaster in an urban area and present results.