An Agent Control Perspective on Qualitative Spatial Reasoning

A considerable part of everyday human interaction is guided by regulations and constraints, for example, regulations on how to behave in traffic scenarios or how to behave at bus stops. These rules have in common that they are usually formulated in natural language and, hence, extensively use qualitative terms, such as 'to the left', 'turn right' or 'in danger of collision', to describe spatial situations and actions. In addition, which behavior is considered to be correct for a certain agent, may not only depend on the spatial situation at hand, but also on the current role of an agent in a particular situation. For example, what an agent is allowed to do may depend on whether she is a pedestrian or whether she is using a vehicle.
Agents acting in such environments must not only be aware of these rules, but also must be able to process them correctly. In this thesis we show how these kind of rule systems can be formalized by means of methods from qualitative spatial reasoning (QSR), and how these formalizations can be applied to control agents.