Whereas agent-based models are built on the micro-level, the interesting model output is often observed on the macro-level. In models with agents moving in space this leads to complex movement patterns. We propose a method to describe the simultaneous movement of agents by graphs that encode qualitative spatial relations between object pairs and the change of these relations over time. Movement patterns can then be expressed as graph patterns. We present two approaches to find occurrences of such graph patterns, using a graph database query and using a customized graph algorithm. Based on the example of the RoboCup soccer simulation, we demonstrate the use of our approach to define and find movement patterns in spatial multi-agent systems.
展开▼