The University of Oslo is closed. The PhD defence and trial lecture will therefore be fully digital and streamed directly using Zoom. The host of the session will moderate the technicalities while the chair of the defence will moderate the disputation.
Ex auditorio questions: the chair of the defence will invite the audience to ask ex auditorio questions either written or oral. This can be requested by clicking 'Participants -> Raise hand'.
"Applications of Reinforcement Learning (in Distributed Systems)"
Main research findings
Large-scale mobile ad-hoc networks enable various applications ranging from forest fire detection to space exploration. These systems consist of 10s to 1000s of simple autonomous devices with very scarce resources, such as computation power, battery, radio range. Thus, a single device cannot deliver expected system behavior, and global coordination is prohibitively expensive. An inspiration to handle such growing system complexity comes from nature - the paradigm of emergent self-organization. Typical examples are colonies of social insects, such as ants or bees.
To design systems with emergent self-organization represents a challenging task. It is due to the decentralized control, autonomy, and missing direct relationship between global and local behavior. Furthermore, local activities typically have a random nature. These properties make it difficult to exploit the immense potential of the paradigm. Current methodologies do not necessarily cover all design aspects. There is also a lack of reusable tools to aid the design process.
Therefore, this thesis proposes a new methodology that consists of several methods interwoven into two alternative design flows. We also devised a novel design platform that integrates tools from various domains of expertise. To demonstrate the methodology’s concept and its usefulness, we applied it in two case studies.
Contact information to Department: Mozhdeh Sheibani Harat