Monday, July 27, 2009

Design space exploration of a domain-specific wireless sensor networks

My research interest is in both hardware and software design challenges of wireless communication systems. In alignment with the research work that is being done at the Institute of Microelectronic Systems (MES) of TU Darmstadt, I integrated to a team of researchers who work on wireless sensor networks and reconfigurable hardware design.

Design space exploration is one of the major tasks in electronic system-level design. Same applies to the whole design process of application specific wireless sensor network design. Due to the complexity of such a system, automated development involves various levels of system abstraction. Consequently, a variety of different development tools including several different simulators are required. It was estimated that the work is best divided into four phases depending on the level of abstraction.

In the first phase, the objective is to create a sensor node library that includes a set of common components of a single sensor node. Here the lowest level of abstraction shall be dealt with, including the hardware and software platforms for a sensor node. A typical sensor node consists of smaller hardware modules such as a microcontroller (in certain case with an integrated special-purpose hardware unit), sensing and actuating modules, and a (medium/short range) radio transceiver as well as software modules such as a real-time operating system, (multihop) routing protocol, and other task-specific algorithm implementations.

The second phase involves domain-specific system modeling, the goal being the modeling of a target application environment. It is thus effectively the highest level of abstraction. Here the primary modeling methodology shall be based on an actor-oriented approach. By use of an existing actor-driven, distributed embedded system design framework a chosen specific application shall be modeled into a system model. Consequently, this system model describes the target system behaviour as a set of functions. A wildlife tracking and monitoring, which is considered as a one of the most complex wireless sensor network applications, is chosen as the target application.

The proposed domain-specific, model-based simulation approach to the design space exploration shall be prototyped in the third phase. Here, previously predefined system functions (from the second phase) shall be mapped to the sensor node platforms (from the first phase) and these mapping techniques will be evaluated. While this phase ends with the development of the prototype on the basis of homogeneous sensor nodes, the fourth phase will extend the prototype by supporting heterogeneity and reconfigurability of sensor nodes.

No comments: