Network Simulation

In modern sensor (and actuator) networks a large variety of protocols, standardized and non-standardized, for both, wire-bound and wireless communication exist. Most networks are hybrids, with a wired backbone network and certain wireless cells for additional flexibility. Real-time communication capability over the whole network often is a requirement.

For modern sensors the network integration must already be considered at the design time of the sensor, to be able to operate with an optimal set of parameters, e.g. optimized for maximum battery life, highest data throughput, precise clock synchronization, lowest jitter of read-out times, or highest read-out precision. In order to optimize both, the sensor design and the network layout, simulation of a complete sensor network is necessary.

Simulation on the one hand allows large scale scenarios with thousands of nodes to be tested, and on the other hand can be used for parameter studies. The crucial point of all simulation is the accuracy of the implemented models. Most existing communication protocols can be modeled very accurately in event based simulation tools, such as ns-3, OMNeT++, OPNET, etc. However, models that accurately reflect the varying properties of hardware and environmental impacts are rare. E.g. the transmission medium on wireless channels is still a big challenge concerning time-variant physical effects. Certain abstractions and idealizations have to be made to keep the computational as well as the modeling effort reasonable.

Aside applying network simulation for parameter optimization and large scale evaluation, e.g. in smart grid applications or factory automation networks, the Department for Integrated Sensor Systems sets a research focus on the investigation of time-variant and non-ergodic network components. In particular with respect to clock synchronization we have developed extensions, both for the ns-3 and OMNeT++ simulation tools, which allow an accurate simulation of distributed clocks, incorporating means for clock synchronization. Not only have we developed a mathematical model that simulates the non-ergodic, time-variant behavior of quartz oscillators, we have also performed measurements of such oscillators, which can be used as source for simulated clocks, and thus accurately simulate real-world behavior of distributed clocks.