- Search

- Contact

Dr. Christian Seitz


Alumnus

E-mail: christian.seitz@mchp.siemens.de


subject of dissertation: a framework for profil-based forming of groups in ad hoc enviroments

The quantum leaps in the field of wireless communication systems and mobile terminals open up new application areas. With small devices like notebooks, PDAs and cell phones increasingly distributed, mobile ad hoc application are developed, which are available to the user all the time. Formalisation, development and evaluation of the framework MoPiDiG - Mobile Profile based Distributed Grouping - takes the centre stage of this paper. The core function of this framework is the dynamical context-sensitive classification of persons in virtual groups in ad hoc enviroments. In MoPiDiG every user is equipped with a mobile terminal and an user profil. As the user moves similar profiles are found and the users combined to groups. The members of a group can cooperate with each other or achieve a common purpose.

Compared to already existing methods the MoPiDiG framework abandons central components and is therefore completly peripheral organized. Furthermore MoPiDiG works without additional position informations. For the communication solely local communication, e.g. WLAN or Bluetooth, are used. The implementation of MoPiDiG necessitates the solving of four separate core problems. Due to the mobility of the MoPiDiG users the communication between the users can break off, which complicates the process of forming groups. Additionally methods are necessary that group the essential subset from the set of profiles. To accelerate the process of forming groups in MoPiDiG optimisation processes are essential, which are implemented by heuristics. In addition the quantity of communcation partners is, where appropriate, reduced by segmentation methods.

To implement MoPiDiG a modular architecture was developed which on the one hand consists of an domain independent layer that is made up of algorithms and meta models. On the other hand a domain dependent layer ensures that MoPiDiG can be fitted to new domains with as little effort as possible. Besides the abstract design of MoPiDiG an example of use is research into for evalution. Here it is a matter of dynamic taxi-sharing. The starting point of this setting are highly frequented locations, like a railway station. While the user still are in the train or stay already in the station, the profil information is exchanged and thus tried to generate one or several groups. Participants affiliate to a group if the utility function rises. This happens when persons have the same destination, or when a way exists, which reaches their destinations without going a long way round.

To evaluate the framework a simulation enviroment for the MoPiDiG application was generated. This simulation enviroment was used to be abled to make a statement on the stability time of the generated groups, the scalability of the process of forming groups and the efficency of the used heuristics. The results show that the use of heuristics reduces the execution time of the group forming process in such a manner that a execution of mobile terminals is possible.