Behavior Pattern Detection for Data Assimilation in Agent-Based Simulation of Smart Environments

Agent-based simulation is useful for studying human activity and their interactions in smart environments. Existing agent-based simulations are mostly offline tools that do not utilize real time information of smart environments. In previous work we developed a particle filter-based data assimilation method to assimilate real time sensor data from the environment into an agent-based simulation model. This paper extends the previous work by presenting a framework of behavior pattern informed data assimilation. We describe the structure of this framework and focus on the task of behavior pattern detection using Hidden Markov Model. We apply behavior pattern detection to a smart office case study example and discuss how the detected behavior pattern can inform the data assimilation in agent-based simulation of smart environments.