Towards realistic, flexible and efficient wireless channel models for network simulation

Simulation is a widely used performance evaluation tool due to its flexibility, controllability, and observability. The advantages of simulation are significant in wireless mobile network research because it is difficult to conduct experiments at large scale. Even if a large-scale experiment is available, repeatability is an issue due to outside noise, environment changes and mobility. It is very difficult for analytical models to capture such complex system artifacts. The channel models in network simulators play a fundamental role in determining the connectivity and link quality. Thus, the impact of inaccurate models at the physical layer is compounded as it interacts with the upper layers of the protocol stack, leading to inaccurate and untrustable simulation results. Channel models in network packet simulators should be efficient as well as accurate because the signal strength needs to be calculated for every packet at every node. The emphasis on efficiency leads to significant abstraction of channel characteristics, and only superficial characteristics such as relation to distance are implemented in most channel models used in network simulators. However, the abstracted channel models that ignore detailed channel characteristics have been shown to produce results that often disagree substantially with results from testbed experimentations, especially in ad-hoc networks. To implement an efficient channel model, abstraction is unavoidable. However, abstraction should be done carefully so that it will not sacrifice accuracy.;The goal of this dissertation is to increase the credibility of simulation-based research by developing an accurate and practical channel model. The problem is that even the evaluation criteria for such a model are not clear. Thus, this dissertation first defines three requirements: accuracy, flexibility and efficiency. It then presents the evaluation criteria for each requirement. In particular, to achieve accuracy, this dissertation suggests that a model should accurately reflect spatial variation of path-loss, temporal fading of signal power, as well as external noise. If a model can generate accurate fading channel for a link regardless of whether the link moves, the model is said to be flexible in mobility. This dissertation supports the importance of each criterion by showing its impact not only the channel model directly, but also on upper layer protocols, and the overall network behavior.;After defining these requirements, the second part of the dissertation proposes methodologies for developing channel models that meet these requirements. The proposed methodologies are the plausible modeling of the path-loss model, the extension of link models based on estimation, and the external noise model. A modified regression algorithm is used to obtain a plausible value in a mobile ad-hoc network. The use of the plausible data is the key in the estimation-based link models for static links. Moreover, this dissertation points out that existing noise models are unrealistic; they use average noise levels and ignore the spatial nature of the impact of noise sources, and the temporal nature of their transmission pattern. Thus, a realistic noise model is proposed based on several wireless network standards. The proposed models are validated using statistical methods, such as the root mean square test and the K-S two sample test.;Furthermore, this dissertation demonstrates the importance of accurate modeling of channels by studying some aspects of protocols that are substantially influenced by accurate models. These include the impact of fading on the back-off algorithm in the MAC layer, the impact of asymmetric link quality on MAC, and the impact of spatial correlation on routing. In particular, this dissertation proposes several fading resilient back-off algorithms for mitigating the unfairness problem. The effects of these proposed algorithms are evaluated in various fading channel conditions, such as the log-normal shadowing channel model and the Rayleigh-fading model with various Doppler frequency shifts.;The simulation results indicate that the proposed model generates a plausible path-loss for a specific site efficiently. This plausible path-loss model renders the channel model to be reliable and practical together with temporal fading models and the proposed external noise model. As a result, the credibility of simulation-based research is expected to improve with the realistic and practical channel model.