Spatial Cognition: From Rat-Research to Multifunctional Spatial Assistance Systems

In 1948, the psychologist Edward Tolman described experiments in which rats were trained to follow a path through a complex maze to reach a food box. After the rats performed perfectly (chose the shortest way to reach the goal), the trained path was blocked; the rats had to select another path from a variety of alternatives. Astonishingly, most of the rats found a path that was close to the most direct connection to the food box, whereas not a single rat erroneously tried to follow the original path on which they had been trained. On the basis of these findings, Tolman argued the rats had “acquired not merely a strip-map . . . but, rather, a wider comprehensive map to the effect that the food was located in such and such a direction in the room” (p. 204). Tolman’s paper, entitled “Cognitive maps in rats and men,” marked the starting point of psychological spatial cognition research. Today there is a great body of evidence on how humans (and animals) learn routes, find ways, navigate through familiar and unknown environments, and on the strategies they use when they get lost. Contemporary research on robotics and AI is concerned with similar problems. For example, how must a mobile robot system be designed to improve its efficiency for tasks such as route choice and navigation? Certainly, the robot must acquire an internal representation of the environment – a cognitive map – and apply adequate procedures to plan movements. A related problem exists in the domain of geoinformatics. A geographic information system must be able to efficiently store, process, and retrieve geo-referenced data, i.e. data which is associated with locations defined in a geographic reference system. On the other hand, it should also interact with the user in a comprehensible way, that is, it should take the user’s mental representations of spatial knowledge into account. Applications such as location-based services, geovisualization or semantic information retrieval lead to an especially close interaction between human and machine reasoning. In the last years, a growing number of researchers from AI and robotics have addressed cognitive questions. Psychologists have become sensitive to the computational properties of robot navigation or issues of reasoning with diagrams and qualitative spatial representations. Research in this rapidly evolving interdisciplinary enterprise has a name: Spatial Cognition Research. Spatial Cognition: From Rat-Research to Multifunctional Spatial Assistance Systems