In a first approach we follow the ideas of Caldarelli [39]. He uses for the presentation of graphs an adjacency matrix (cf. [39]:11f). The graph-model is 'simpler' than the original maze: the food cell cannot be bypassed as in the real maze. Furthermore will the searching system never 'bounce' against the wall not will it 'stand still'.
Thus the graph model gives us a slightly idealized benchmark to get an estimate for the real random system. In the following we show a collection of simple graph models taken from TOLMAN2: