Extracting Movement Behaviour Patterns from Video Data to Drive Multi-Agent Based Simulations

Computer simulations are used to create and evaluate real world scenarios in a manner that is controlled, nonintrusive, cost effective and safe. One technology for realising computer simulation is Multi Agent Based Simulation (MABS), the advantage being that the entities that feature in a simulation can be expressed as agents that have all the features associated with agents (autonomy, goal driven, etc.). A particular challenge of MABS is the acquisition of the data required to define agent behaviour. One approach is to “hand craft” agent behaviour, however this is error prone and time consuming. In this paper we proposed a method whereby we can extract, what we have termed “Movement Behaviour Patterns” (MBPs) and “Movement Path Patterns” (MPPs), from video data; which in turn can be used to drive agents in a MABS environment.