Computational Electromagnetics for Improved Wireless Network Simulation
Matthew Adams, University of Washington
Network simulators are important tools in wireless network research. However, these simulators are generally built upon unrealistic channel models that do not accurately describe the complex physics of radio wave propagation in an indoor scattering environment. Research has shown that such unrealistic channel models can lead to grossly inaccurate network simulation results, and make it difficult to draw conclusions regarding higher level network performance. In this presentation, I will present an overview of this underlying problem, previous attempts to solve it, and describe our approach using computational electromagnetics.
Previous efforts to improve channel model accuracy for indoor wireless network simulation have generally stopped short of full-wave electromagnetic simulation due to computational requirements. However, in doing so, such alternatives have sacrificed accuracy and generality or introduced other significant drawbacks such as a need for expensive instrumentation to measure the physical parameters describing a particular environment.
Full-wave electromagnetic simulation of the wireless radio channel offers the promise of highly accurate network simulation in any indoor environment without the requirement of expensive channel measurements. While simultaneous channel and network simulation is not feasible, Monte Carlo techniques using computational electromagnetic simulations can be used to accurately determine channel parameters for each link in the network. Through superposition, the total received signal at any node can then be determined. The Finite Difference Time Domain (FDTD) method is chosen due to the ease with which it can be parallelized to a large multi-core cluster. Through the use of computational electromagnetic techniques, we hope to gain insight into the relationship between the geometric complexity of a propagation environment, the level of detail included in the channel model, and the accuracy of resulting higher level network performance.
Abstract Author(s): Matthew Adams