Research On Quality Of Experience Oriented Resource Management In Mobile Internet

In mobile Internet, the development motivation of wireless communication has changed from technology-driven to user-driven. Quality of Experience (QoE) of end-user becomes a key factor that impacts on the diffusion of mobile Internet services. As the core bearer of mobile Internet services, the performance of wireless communication networks directly affects the end-users’QoE. Therefore, how to provide a plenty good QoE based on the limited network resources has become a challenge.In this dissertation, the strategies of QoE-oriented radio resource allocation in mobile Internet are investigated. Based on the analysis of the QoE modeling and quality index system of wireless services, this dissertation proposes four resource allocation mechanisms to improve end-user’s QoE in four different but evolving network scenarios:single-cell, multi-cell, single access-based heterogeneous wireless network (HWN), and further multi-radio access-based HWN. The main contributions of this dissertation are listed as follows:(1) To solve the QoE-oriented multi-user resource allocation problem in single cell OFDMA (Orthogonal Frequency Division Multiplexing Access) system, this dissertation proposes a MOS (Mean Opinion Score)-driven cross-layer optimization framework. Firstly, based on the data rate-to-MOS modeling of voice, file transferring and streaming services, this dissertation formulates a MOS-maximized optimization problem. Since the problem is non-convex, this dissertation transforms it to be an unbounded subproblem, which can be solved by Lagrange dual decomposition method. An iterative algorithm is then proposed for the final solution, which combines a joint subcarrier assignment and adaptive power allocation algorithm, resource adjustment procedure and further tunable minimal satisfactory threshold control. Simulation results show that, comparing to the classical throughput-optimized algorithm, this proposal can not only improve the overall perceived quality from the perspective of users, but also the fairness between users.(2) To tackle the QoE-oriented multi-user resource allocation problem in multi-cell OFDMA system, this dissertation proposes a joint resource reservation and resource allocation framework, which introduces the information of mobile resource reservation into the non-cooperative distributed power allocation, greatly reducing the call dropping probability of the ongoing services and meanwhile improving the system’s average MOS value. Firstly, according to the statistical service models and user mobility model, the mobile resource reservation mechanism forecasts and estimates the resource blocks to be reserved for new coming traffic and handover. Then, each cell informs its own reservation information to the neighboring cells through the inter-cell signaling interaction, and executes intersectant resource reservation. Based on the available resources after the former reservation, the dynamic resource allocation problem in multi-cell OFDMA system is further modeled as a non-cooperative game with differential pricing on power owing to the channel gain or SINR (Signal to Interference plus Noise Ratio). By proving the existence and uniqueness of Nash equilibrium of the objective, this dissertation presents a distributed resource allocation algorithm with independent user scheduling in a single cell and non-cooperative power allocation.(3) To tackle the QoE-oriented network selection problem in single access-based HWN, this dissertation presents a matching game-based network selection mechanism. Unlike the existing methods with lower Pareto efficiency, this dissertation introduces a novel matching game model, which models the heterogeneous network selection problem as a two-sided matching problem. First of all, according to the network performance parameters and user preferences, this dissertation establishes and quantifies the user experience and network payoff matrix. Then, based on different network capacity, the former game can be decomposed into one-on-one game and many-to-one game. The corresponding user dominant and network dominant matching algorithms are further proposed, which are proved as Pareto optimal, unique and stable. Simulation results show the superiority of these algorithms, which could strike a balance between the user experience and network revenue, and meanwhile improves the system’s overall performance.(4) To solve the resource allocation problem in multi-radio access HWN, this dissertation proposed a QoE-optimized resource allocation algorithm. Future wireless network is a fusion-oriented ecosystem, where multi-radio access is an efficient data transmission method.Other than the current studies that only concern several QoS (Quality of Service) parameters, such as throughput, this proposal introduces the QoE modeling based on utility functions of real-time and non-real-time services as well as resource usage costs, and then builds an optimization problem with an objective of maximizing the user perceived QoE. Since it’s a typical convex optimization problem, this dissertation firstly carries out a primal-dual interior point method-based iterative algorithm. To reduce the complexity of the former algorithm and improve the rate of convergence, this dissertation further proposed a multi-level penalty function-based particle swarm optimization, where the original problem is gradually converted into an unconstrained optimization problem using penalty transformation. Based on the fitness function, the particles could rapidly approach the optimal solution in the optimization space.