A Cognitive Model for Routing in Agent-Based Modelling

Agent-based modelling (ABM) can be used as a computational tool to model human routing behaviour, and offers particular promise when combined with insights from cognitive science. In this paper, we introduce typical errors into the encoding of the agents mental representation of the environment. This method deviates from the classical computer science paradigm of optimality to capture human behaviour more accurately. By incorporating common distance and direction estimation errors, our model produces routes with fewer computational artefacts such as zigzagging (i.e., turning more often than the typical human) and bottlenecks (i.e., routing through one particular node that maximises efficiency). We demonstrate our results in regular and irregular environments and validate our model using a set of real-world footfall data from Westminster, London.