Research On Interference Model For Throughput Optimization In Wireless Networks

Recent years have witnessed the popularity of wireless networks for data commu-nication in various applications. Many emerging technologies (e.g., OFDMA, network coding, cognitive radio, etc.) have been proposed to enhance the capacity of wireless networks. Beyond these approaches, the performance of wireless networks could be improved by optimizing the allocation of existing wireless spectrum resources via link scheduling, channel allocation, etc. Such optimizations usually allow multiple nodes to transmit at the same time over the same channel, while dealing with the interference of simultaneous transmissions appropriately. Two main interference models have been proposed in the literature:the protocol model and the physical interference model, also known as the Signal to Interference and Noise Ratio (SINR) model.The protocol interference model or so-called conflict graph is a simple graphical representation of the interference condition between any two wireless communication links (or wireless nodes). Conflict graph provides a simplified description of the in-terference status, which greatly eases the design of channel assignment/spectrum allo-cation algorithms, and consequently gives birth to a series of highly efficient wireless network optimization algorithms. The SINR model is considered to be a realistic inter-ference model. It accurately models the accumulative interference effect, which refers to the interference aggregated from multiple sources. However, the SINR model suffers from some problems such as high cost caused by exhaustive measurement calibration. More importantly, its non-convex nature incurs great complexity when it is applied in wireless network optimization.In considering the pros and cons of these two models, we are aiming to conquer their drawbacks while still holding their merits. Specifically, we improve both SINR models and conflict graph models mainly of accurate and efficient generation and provide better usage. Our work mainly consists of following aspects:● Firstly, considering that the SINR model heavily relies on the accuracy of RSS and greatly affect the performance of SINR-based throughput optimization algorithms, we design a measurement based efficient SINR model generation with accuracy control. As is required by SINR-based throughput optimization algorithms, the RSS between all node peers need to be measured. However, this leads to the time cost at level of O(N2MC), which is unacceptable for a wireless network with frequent re-configuration. The proposed model-based solution is following the concept of "measure a few, predict many" and reduce the time cost to the level of O(N/M). In addition, we theoretically models the relationship between the measurement cost and prediction accuracy. Thus, this could be used as a tools to control the trade-off between efficiency and accuracy. The performance are evaluated based on the experiments used real collected data trace from SWIM platform.● Secondly, we provide a compressive sensing(CS) based efficient SINR model gen-eration method tackling the same problem with the first part of the work. We model the measurement based SINR model generation problem as a linear system by re-moving the assumption that measurement could not be conflict. By applying the concept of compressive sensing, we obtain the CS-bsed solution, which is able to further reduce the time cost to the level of O(log N) and achieve the prediction ac-curacy of 1-2e-Tδ/8. Real data trace based experiments reveal that the CS-based solution and Model-based solution are mutually complete in terms of network den-sity and prediction accuracy.● Third, to tackle the drawbacks of conflict graph model, including unable to deal with accumulative inteference and etc., we propose the model of quantized conflict graph (QCG). The QCG model are proved to be able to handle of accumulative in-terference and provide better optimization results on the problem like link schedul-ing and interference minimization without increase the problem complexity. The properties, usage and construction methods of QCG are explored. We show that in its matrix form, a QCG owns the properties of low-rank and high-similarity. These properties give birth to three complementary QCG estimation strategies, namely low-rank approximation approach, similarity based approach, and comprehensive approach, to construct the QCG efficiently and accurately from partial interference measurement results. We further explore the potential of QCG for wireless network optimization by applying QCG in minimizing the total network interference. Ex-tensive experiments using real collected wireless network are conducted to evaluate the system performance, which confirm the efficiency of the proposed algorithms.