Home | Legals | Data Protection | Sitemap | KIT

Distributed Decision Making

Distributed Decision Making
type: Seminar (S) links:
semester: WS 18/19

Volker Wenzel
Prof. Dr.-Ing. Jörg Henkel

SWS: 2
lv-no.: 2424378

Please register to the seminar via https://ces.itec.kit.edu/1033.php until the beginning of the lecture period. Furthermore, a registration to the corresponding examination module in the CAM-system is required. The first meeting usually takes place in the first week of the lecture period. The exact date will be announced via email. The duration of course extends over the lecture period.

Today's computing systems develop more and more from centralized approaches
into Distributed Systems.
Examples of this are
• content delivery networks (e.g. Akamai),
• mesh networks (e.g. B.A.T.M.A.N.),
• peer-to-peer networks (e.g. BitTorrent),
• grid computing (e.g. SETI@Home),
• smart power distribution networks (smart grid),
• digital currencies (e.g. BitCoin), and
• swarms of autonomous objects of UAVs or self-driving cars

It is foreseeable that even single chips will have so many compute cores, that
not in all situations decisions can be made centrally; especially if dealing with the management of different types of computational resources.
An example of a many core System on a Chip (SoC) is Intel's Xeon Phi – we too at the Chair for Embedded Systems do research on similar systems.

Distributed systems that rely on decentralized decision making comprise many applications of the emerging 'Internet of Things' (IoT), which are another important focal point of the research at our chair.

The one thing that these systems have in common is that decisions have to be made in a decentralized way – often avoiding a single point of failure.

Research on 'Distributed Decision Making' has leaped forward since many years –
initially only in the field of economics, but more and more also in computer science.

Topics of the seminar will be drawn from the disciplines of Game Theory, Auction Theory, Nonlinear Dynamical Systems, and Multi Agent Systems.
Students can choose from a wide variety of topics – some more applied and some more theoretical.

Topics of the past semesters included:
• Flocks, Herds, and Schools: Reynolds Distributed Behavioral Model
• Programmable self-assembly in a thousand-robot swarm
• The Blockchain -- Mode of Operation, Opportunities, and Limitations
• Lock-free List-Based Hashsets
• Auction mechanisms for resource management with a focus on the Vickrey auction.
• Distributed Hash Tables

In agreement with the supervisor, students can work on a self-proposed topic. By request, the seminar can be supervised in English.