Application of Microsimulation Towards Modelling of Behaviours in Complex Environments

In this paper, we introduce new capabilities to our existing microsimulation framework, Simulacron. These new capabilities add the modelling of behaviours based on motivations and improve our existing non-deterministic movement capacity. We then discuss the application of these new features to a simple, synthetic, proof of concept, scenario involving the transit of people through a corridor and how an induced panic affects their throughput. Finally we describe a more complex scenario, which is currently under development, involving the detonation of an explosive device in a major metropolitan transport hub at peak hour and the analysis of subsequent reaction. Introduction & Background Microsimulation is a discrete simulation technique which allows for the modelling of the behaviour of individuals in a complex system (Connor et al. 2000; Merz 1991). It was originally devised for financial and economic modelling (Weinstein 2006; Orcutt 1957), but is generally applicable to a wide range of scenarios. This paper builds on our previous work (Piper et al. 2009; 2010; Green et al. 2009; 2010b; 2010a; Zhang et al. 2008) in which we describe our microsimulation-based approach to modelling systems involving many individuals and their interactions. Our modular microsimulation framework, Simulacron, allows us to quickly build simulations with varying requirements. As covered in previous publications, it has already been applied to problems involving epidemiology and terrorism. This paper will be making use of the framework itself and an updated dispersion module from previous publications. We will describe new functionality added to our existing toolset in the form of a motivation module. This module allows us to model human interactions involving differing motivators and their effect on behaviour. A simple, fictional scenario will then be introduced to test and demonstrate this new functionality. The results of our experiments with this scenario will be presented and discussed. Finally, we will introduce a more complex scenario incorporating the interaction of the motivation module with a new goal-based movement module, which we briefly describe in this paper. Copyright c © 2011, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. It is important to realise that the work described here involves the incorporation of new functionality into a generalised microsimulation framework and is not a new, standalone packaging of capabilities for a specific task. This means that the newly introduced functionality can be combined with our existing capabilities permitting, for example, the incorporation of psychological factors into an epidemiological simulation; this significantly enhances our modelling abilities. This development was motivated by interest both within our group and from external agencies in using our techniques to model scenarios such as transit through a transport hub, emergency response behaviours, evacuation, interdiction and evaluation of security mechanisms. We realised we could not adequately model these with our existing toolset and required a more sophisticated approach to behavioural modelling. Our microsimulation-based approach to modelling divides the world into entities of two classes: “cells” and “peeps”. Cells are abstract locations which have no intrinsic meaning or properties whilst peeps are abstract things whose only intrinsic property is their location (a cell). Cells and peeps have meaning associated with them via “fields” which are arbitrary chunks of data associated by name. This system allows us to extend the existing abstractions with new meaning without having to modify the simulation framework itself. The field system supports the use of “states” which allow alternate field value sets to be associated with an entity based on their current state. As an example, a peep might have two different schedules: one for use in the “working” state and one for use in the “weekend” state. The simulation program is the combination of the simulation framework, called Simulacron, and one or more modules. These modules provide the actual simulation behaviour. To date, we have modules implementing cyclic and one-off scheduling, non-deterministic dispersion, spread of infectious agents and a terrorist behaviour prototype. A user is free to use whatever subset of modules they need; the simulation need only be as complicated as the scenario demands. Input to the simulation is provided in the form of an XML specification which details all entities in the model. Because of the microsimulation approach, this data set must contain 26 Applied Adversarial Reasoning and Risk Modeling: Papers from the 2011 AAAI Workshop (WS-11-06)