Integrating Behavior and Microsimulation Models

Microsimulations focus on modeling routine activities of individuals and have been used for modeling and planning urban systems like transportation, energy demand, and epidemiology. On the other hand, planning for emergency situations (e.g., disasters) needs to account for human behavior which is not routine or pre-planned but depends upon the current situation like the amount of physical damage or safety of family. Here, we focus on modeling the aftermath of a hypothetical detonation of an improvised nuclear device in Washington DC. We review various behavior models from the literature and provide motivation for our model which is conceptually based on the formalism of decentralized semi-Markov decision processes with communication, using the framework of options. We describe our approach for integrating behavior and microsimulation models where the behavior model specifies context-dependent behaviors (like looking for family members, sheltering, evacuation, and search and rescue) and the synthetic population provides information about demographics and infrastructures. We present results from a number of simulation runs.