Soziale Roboter und soziale Räume: eine Affordanz-basierte Konzeption zum rücksichtsvollen Handeln

If autonomous agents like humans and robots act in physical space, several types of conflicts can occur that result from the fact that actions occupy space. To avoid such conflicts, social robots must have the capacity to consider the actions taken by other robots and by humans. This thesis proposes a conceptual framework which can be used to specify knowledge about spatial needs of activities thus enabling social robots to represent and reason about the spatial needs of their own activities as well as of the activities of others. This kind of knowledge can be processed by knowledge processing procedures like symbolic planning. Moreover, it enables social robots to transparently explain their choice of actions to humans. Based on models from social sciences describing how human activities shape space, four types of social spaces, viz. »personal space«, »activity space«, »territory« and »social affordance space« are formalized. Among these types of social space, social affordance space is a new type of social space which has not gained attention in social robotics until this work. As a common ground to modeling social spaces, the concept of a functional affordance space is proposed which represents the spatial preconditions for the placement of activities. Methodologically, the approach embraces classical Artificial Intelligence by identifying and formally specifying an inventory of concepts and relations to be used to formally describe socio-spatial environments. To model the social constraints of considered activity placement in social spaces, principles of social spatial behaviour are proposed. These principles reflect the positive and the negative aspects of the possibilities for action available in a situation. The social activity-placement problem, i.e., the problem of choosing the most socially adequate placement for an activity, is solved by reduction to the bipolar, qualitative decision problem known in decision theory. The affordance-based conceptualization of social spaces is evaluated by demonstrating that studies in the field of human-robot interaction can be adequately reformulated enabling social robots to autonomously deliberate and act in accordance to empirical findings.