Research On QoE Based Multi-Objective Radio Resource Allocation

Triggered by the prevalence of smart devices, mobile multimedia services are growing explosively. In the context of a great diversity of accessible services and extensive choices for customers, satisfying users'Quality of Experience (QoE) requirements is of great significance for network managers in order to sustain and sharpen their competitive edges in attracting and retaining lucrative customers. Most researches on QoE-based resource management concentrated on maximizing the QoE mapped from objective network parameters, in which QoE models ignored the impact of user personality factors. Moreover, they didn't take other important objectives such as user QoE fairness and power consumption saving into consideration. Under this background, a QoE-based resource allocation problem which aims at improving personalized user experience, saving power consumption and assuring user QoE fairness is investigated in this thesis. The main contributions are described as follows.Firstly, research background is discussed in this thesis, including QoE related research status and radio resource allocation summarization in Orthogonal Frequency-Division Multiple Access (OFDMA).Secondly, based on traditional QoE models, a multi-objective resource allocation problem which aims at improving user experience,saving power consumption and assuring user QoE fairness is established.The Lexicographic method and Tchebycheff method are adopted to convert the established multi-objective problem into a single objective problem. Then a resource allocation algorithm which combines genetic algorithm (GA) and convex optimization techniques is proposed.Subsequently, a simulation experiment is conducted, indicating that the proposed resource allocation algorithm is superior to the baseline algorithm from the perspective of user experience, power consumption and user QoE fairness.Thirdly, a personalized QoE model concerning video data rate and user preference is investigated for more accurate QoE evaluation in video service scenarios. The effects of video data rate and user preference on QoE are studied according to a subjective user experiment. Regression method is taken to set up the personalized QoE model and then validation set data is used to verify the accuracy of the built QoE model.Finally, a multi-objective resource allocation problem is studied based on the established personalized QoE model. With Lexicographic method to transform the multi-objective problem into a single objective problem, a resource allocation algorithm is then put forward which integrates genetic algorithm and a heuristic power adjusting algorithm.The simulation results indicate that personalized QoE based resource allocation will result in lower power consumption with similar user QoE,leading to a more reasonable resource configuration.