Scenarios and applications
Research interests are focused on mobile and wireless networking for different kinds networking scenarios and different kinds of applications.
Applications are in some aspects determined by the considered networking scenario. Typical applications in a computer network are the simple exchange of data between two known entities - examples include sending emails or downloading a Web page. These data-type applications are complemented by "multimedia" or "streaming" applications, for example, video conferencing. They require certain properties form the network regarding timing and fault behavior to ensure a certain quality of the transported media stream at the receiver.
In addition, other networks can have quite different applications. Wireless sensor networks, for example, have a wide range of possible applications that differ in many key characteristics from traditional networks ("data-centric" instead of "identity-centric").
These applications become particularly challenging when considered in a mobile environment and/or using wireless communication. The inherent unpredictability of movement and the quality of a wireless communication link have to be suitably tamed.
In this context, we are pursuing research mainly in three, mutually interlinked areas:
- Future wide-area mobile systems
- Wireless sensor networks
- Cross-layer optimization of wireless networks
Future wide-area mobile systemsCurrent mobile communication systems like the well-known cellular systems GSM and UMTS are geared towards are relatively homogeneous set of user devices - cell phones - and a relatively small set of applications - mainly voice calls and short text messages, with the support for other applications only slowly growing. Moreover, they are essentially limited to a single wireless technology each.
In the future, it can be expected that all three aspects will become much more diverse:
- User will employ a wider range of communication devices, incorporating today's cell phones, laptops, or PDAs as well as entire small networks centered around a user (body and personal area networks, BAN and PAN)
- New "high-end" applications like video over wireless will become more attractive depending on cost; new applications like machine-to-machine communication will become more prevalent
- Multiple wireless technologies will coexist and will have to be suitably integrated - the incorporation of WLAN support into cellular architectures, for example, is still in its infancy.
Wireless sensor networksIn many technical situations, the observation of a the environment or a technical process at high spatial and temporal resolution (taking frequent measurements close by to the source) is desirable. Examples include preventive maintenance of machinery, biodiversity mapping, precision agriculture, "intelligent" houses, and many others. Traditionally, this has not been possible because of wiring costs for many sensors or because of the costs of the sensors themselves. With miniaturization and resulting cost reduction and with the integration of computation and wireless communication capabilities into the acutal sensors, "wireless sensor networks" become a possible solution to these problems.
There are many research challenges involved with this concept. Especially the small energy supply contrasting with the desire for a long lifetime of the sensor network is a critical issue; much research effort has been invested here in the last few years.
An aspect that has received only relatively little attention is the tradeoff between cost, energy supply, network lifetime, and in particular the resulting quality of the observed data. We intend to investigate this and similar tradeoffs.
A bibliography of wireless sensor networks is available as well (as HTML and as BiBTeX entries). It has been prepared and collected in cooperation with Dr. Andreas Willig of Technical University Berlin.
Cross-layer optimizations of wireless networksIn a conventional network architecture, separate functional areas are identified and implemented as "layers" that provide increasingly abstract services. In a layered architecture, a layer is only allowed to use the service of the layer directly underneath. The hope is to obtain a well-structured, manageable system structure.
While such layering has proven beneficial for many purposes, the particular needs of wireless networks are an ill fit with the rules of strict layering. For example, the volatile nature of the wireless channel makes it questionable whether it should be abstracted away and hidden completely from higher layers or whether it would not be beneficial to expose such low-layer information to higher layers like routing or transport. Conversely, the needs of an application could be made explicit to lower layers as well, allowing them to adapt their operation in a fine-grained manner.
In this sense, cross-layer optimization is a promising approach. It can be applied to many areas, for example the transmission of MPEG-encoded video over OFDM channels, but also to wireless sensor networks or other systems. It is, however, not without dangers as it can introduce unwanted feedback loops into a system. We intend to investigate these issues.
Approach and tools
Regarding these research goals, we are mostly interested in the development and evaluation of architecture and protocols for communication networks. We are using in particular:
- Simulation & prototypical implementation
- Analysis, for example of optimization problems
- Experiments in testbeds
Prototypical implementation goes hand in hand with simulation, as many aspects in particular of wireless systems cannot be fully simulated without excessive overhead. These prototypes will be implemented in various testbeds, for example for wireless sensor networks.
Regarding analysis, formulation of many problems as optimization tools has proven beneficial and appropriate. We intend to pursue this work in the future, most likely using LP/IP solvers like CPLEX.