The Traveller: A Computational Model of Spatial Network Learning

A detailed computational model for learning spatial networks is presented. Information is kept in a distributed, modular format: “condition-actions pairs.” Knowledge of individual routes consists in chaining this piecemeal information. The model, called the Traveller, was fully implemented, and some sample runs are discussed. In learning a new network, the model displays the well-known transition from route level to survey level knowledge. No special mechanisms are needed to achieve the transition, as the Traveller's interactions with the environment gradually structure its emerging cognitive map. The Traveller is compared with some of the main competing computational models, and evidence from empirical research is adduced to support the suggested representational format.