Date of Award

2-2014

Embargo Period

6-5-2014

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Electrical and Computer Engineering

Advisor(s)

Peter Steenkiste

Abstract

Wireless research and development requires effective and efficient simulation and emulation tools to validate and evaluate wireless designs. Wireless channel models are used in the tools to simulate signal propagation properties in the real physical world. However, due to practical issues, these models are often too generalized and simplified in large scale experiments, and they only provide limited realism.

In this thesis, a novel world model is proposed for simulation and emulation of wireless networks. The proposed model includes the design and implementation of a variety of environment models that enhance realism in simulation. These models capture realistic signal propagation properties across multiple connections, and over time: first, the impact of realistic physical world features, such as channel dynamics and cross link correlation are characterized at different time scales; then, both geometrical and statistical simulation models are developed to recreate desired channel dynamics among wireless network links efficiently.

Three major components of the proposed design are described in this thesis: 1) a flexible channel simulation model, 2) improvement of parameter accuracy in geometric channel models, and 3) wireless link correlation models with a case study in vehicular networks. The flexible channel simulation model supports fast generation of channel updates for complicated channel models, including small-scale fast fading, large-scale path loss and multi-path delay and attenuation. To achieve high realism, a variety of techniques are developed to obtain high parameter accuracy in geographic channel models. Link correlation models are developed for simulating wireless channels within a network context, where adjacent wireless links share the same propagation medium. The wireless link correlation model handles both temporal and spatial correlations, to reflect properties at different time scales and location-based similarities.

A case study in vehicular networks illustrates the effectiveness of using the proposed environment model to improve the realism of wireless simulation and emulation platforms. Simulation results from implemented models are compared against the measurement data from physical world vehicle-to-vehicle channels, and show good approximation to reality. The evaluation results of correlated channel models show improved realism in channel properties and corresponding impact on the performance of a gossip protocol.

Share

COinS