The goal of this thesis is to construct a model of the interactions between emotions and motivations with cognitive problem-solving in artificial learning systems, with special emphasis on the functional aspects of these processes in resource-bounded systems. In adopting an interdisciplinary point-of-view, the reader is first introduced to neurophysiological research and psychological models and concepts about emotions, motivations, and cognition, and the mutual interactions between processes belonging to either category. Functional analysis of these phenomena and formulating parallel concepts for artificial autnonomous systems yields a systemic approach for integrating processes that fulfill similar internal roles in today’s learning AI systems. Eventually, the theoretical contribution is validated by presenting the design of a learning architecture, which incarnates the principles developed before. Besides providing a more concrete, algorithmic example for these principles, this allows for conducting simulation experiments, which confirm the principal potential for increasing efficiacy of learning architectures and provide interesting insights on the internal dynamics of the architecure.