Open Access

This research is motivated by the discrepancy between the spatial model underlying current geographical information systems (GIS) and the human way of structuring space into meaningful spatial units, which are generally referred to as places. Focusing on their functional dimension with regards to human action, this thesis presents a computational model of functional place, and conceptualizes it as a subset of geo-atoms, primitives of geographic representations, which are bound by a functional unity condition with regards to a complex spatial action. Of the variety of challenges which accompany the endeavor of integrating place into GIS, two are explicitly addressed: the compound nature of places, and the subjectivity of place formation processes. Thus, in contrast to the high presence of places in our daily life, activities, and communication, they typically do not refer to a distinct geo-object, but rather consist of multiple geo-spatial entities, and might therefore not exist as a distinct entity in a geo-spatial database. Furthermore, functional places are mentally constructed by the individual, and thus prone to subjective variations in terms of their location, usage pattern and allocated meaning among different people. Inherited from its Euclidean spatial tradition, however, space is generally assumed to be an objective, mathematical reality in GIS, which is why possibilities to incorporate subjective differences in human perception and use of the world are still rare. Applying an agent-based approach to modeling place, the concepts developed in this thesis allow for the dynamic construction of functional, agent-specific places from geo-spatial data. For this, a simulation framework is created which, based on concepts derived from ecological psychology, describes how an individual human agent assesses the suitability of its environment with regards to complex actions. On this basis, suitability, expressed as a scaled numeric value, can be computed in a way which is specific to the particular system of agent, environment and action. In a second step, this framework is embedded within a larger model of functional place formation, which enables the dynamic generation of functional places based on optimizing their resulting suitability with regards to a complex action. On this conceptual basis, a software agent with the ability to dynamically construct functional places from its sensory inputs, and demonstrate the according spatial behavior, is developed and tested for its functionality in the context of a pedestrian simulation.