Soutenance mémoire
Motivation & Impact
SCs are introduced to improve throughput and coverage. However, the reuse of frequency, overlapping cell coverage areas, and closed access cells lead to a multitude of optimization problems in the RAN of a HetSNet. The control variables of these problems are the location of the SCs, the number of SCs that are on, and allocation of channels, power, time slot, and UE. There are two main solution paradigms, centralized vs. distributed. In line with the challenges, controls, and solution paradigms, literature has identified several key research domains 9 for HetSNets. (Guvenc et al., 2013a,b). One key problem due to the introduction of SCs is the interference created due to overlapping cell deployment with frequency reuse. Fig. 1.2 shows how the macrocell interferes the two SCs, which are deployed in its coverage area. In literature, the interfering cell is called the aggressor while the interfered cell is called the victim. Interference can be in both uplink and downlink. Similarly, the SCs that are closer to the macrocell BS can victimize the macrocell (Saquib et al., 2012). The SC1 of Fig. 1.2 shows one proposed solution, called the cell range expansion, where SC1 prevents uplink interference to itself by accepting the macrocell UE (MUE) to the SC network.
Cell range expansion and other techniques of interference mitigation are detailed in Section 1.3. Since SCs are expected to be deployed in substantial numbers by network operators as well as users, the infeasibility of preplanned radio resource allocation is a forgone conclusion. Instead, HetSNets are expected to follow the new paradigm of self-configuration and self-optimization, which is jointly defined under SON (Peng et al., 2013) and the related concept of automated SC site planning for op10 erator deployed SCs (Guo et al., 2013b; Guo & O’Farrell, 2013). The underlying fundamental issues that both SON and automated deployment attempt to solve are the handling of intercell interference and providing improved coverage and throughput where it is in demand in a scalable and dynamic manner. Another research domain is load balancing or traffic steering, among overlapping cells for network performance optimization (Munoz et al., 2013b). Traffic steering may happen while the UE is idle or active.
The latter is known as handover (HO). In an overlapping multitier cell environment, HO decision is more complicated than a conventional cell network (Guvenc, 2011). The HO decision now involves not only the received signal to interference plus noise ratio (SINR) of the considered UE but also interference from the UE to nearby cells and also the need of avoiding excessive HOs among cells in order to reduce signaling and minimize call drop rate (Munoz et al., 2013a; Pedersen et al., 2013a). Group HO where a group of UEs is handed over simultaneously, e.g., in a moving vehicle, is also considered an important challenge in HetSNets (Sui et al., 2013). In order to optimize offloading gain, SCs has to be discovered by the UEs. Optimization of cell discovery, with minimum signaling and minimum delay is an active research domain (3GPP, 2012; Prasad et al., 2013). Energy efficiency is another challenge the research community is currently working on and this topic is usually addressed under the topic of green mobile networks (Xu et al., 2013b; Shakir et al., 2013). Researchers are exploring methods to save energy in various functions of the network from the deployment to operation (rae Cho & Choi, 2013). Network modeling is another challenge in HetSNets (Hwang et al., 2013). As discussed in Section
classical Wyner interference model is no longer a valid approximation of a HetSNet. Due to a large number of unplanned SCs that are expected to be deployed, centralized solutions for the above-discussed challenges are not scalable. Therefore, the network intelligence has to be distributed among the cells. Especially real-time processes such as interference coordination must be handled through distributed or decentralized resource allocation in the RAN (ElSawy et al., 2013). Such decentralized radio resource optimization (RRO) schemes in HetSNets are still in their infancy and much space is available to make substantial research contributions. We would like to draw the attention to one more issue that is studied by the HetSNet commu nity and that is the limitation of the backhaul. While larger cells are backhauled through fiber links or high-speed dedicated microwave links SCs, especially those deployed in residences, are expected to be backhauled through the existing digital subscriber line (DSL) and home cable networks. DSL was designed for Internet access and hence not optimized for quality of service (QoS). Managing UE in a HetSNet to optimize QoS over backhaul links of SCs has been identified as a key problem (Samarakoon et al., 2013). Also, wireless backhauling SCs to places without wired infrastructure is also studied (Liu & Shen, 2014). Wireless communications as a field has evolved at a quite unimaginable pace since the first radio transmission by Guglielmo Marconi. The path ahead is just as interesting. The industry is debating the 5G network model. HetSNets and SON have been identified as key technologies that will define a 5G network (Demestichas et al., 2013). Therefore, this research could not possibly be more timely. We share a passion for wireless communication research, which is the key reason to propose to explore RRO in HetSNets that will have an impact on the future wireless networks. Finally, as identified above, the introduction of SCs has given rise to a plethora of new challenges in wireless resource optimization at the RAN, which translates into research opportunities and which gives us the chance to make substantial and valuable research contribution for the industry and the advancement of humanity.
State of the Art and Their Limitations
This section discusses the state of the art in research and their limitations as related to the thesis. The bulk of the works surveyed in this section are the state of the art as it was at the time of writing the research proposal for this thesis (mid 2014). A few more recent articles were later included. The Chapters 2,3,4, and 5 that present the constituent articles of this thesis have their own discussion on the state of the art with respect to the problems discussed in each of those chapters. The problem of resource allocation in HetSNets has been the focus of several research works in the past. One of the main problems addressed is intercell interference. Cell range expansion is an attempt to this end. In conventional cellular networks, the UE is handed over to the cell with the highest received referenced signal strength, which works well for cells of comparable transmission power. But in HetSNets there is an imbalance in the transmission powers of macrocells and SCs, therefore the UE may be closer to the SC but may still receive relatively higher power from the macrocell. Therefore, in the downlink, the UE prefers to be served by the macrocell. However, in the uplink the SC receives a stronger signal from the UE than the macrocell due to proximity.
This uplink signal from the UE may interfere the transmission of the UEs of the SC. Cell range expansion is a solution that was proposed by 3GPP (see Fig. 1.2) to solve this uplink-downlink imbalance (Lopez-Perez et al., 2011). In cell range expansion, a positive bias is introduced to the received power of the SC signal, thus the UE performs the HO to the SC while receiving a lower SINR compared to the macrocell BS. The equation (1.2) shows the operation of bias where the UE selects the cell m ∗ which has the highest received power Pm plus bias βm among the set of cells M. HO frees up the macrocell to serve another UE and also eliminates the uplink interference that was received at the SC. Due to the lower downlink SINR at the UE in the expanded cell, research has shown an overall reduction in sum rate of the network (Guvenc, 2011). Researchers have proposed optimizing the bias value to enhance the performance of the UE in the expanded cell. In (Kudo & Ohtsuki, 2013), Qlearning is used where each UE learns the optimal bias value from past performance. The simulation results depict a reduction in the number of UEs in outage compared to a fixed bias scheme.
|
Table des matières
INTRODUCTION
CHAPTER 1 RESEARCH PROBLEM
1.1 Heterogeneous Networks
1.2 Motivation & Impact
1.3 State of the Art and Their Limitations
1.4 Research Domain, Objectives & Methodology
1.4.1 Objectives of the Research
1.4.2 Methodology
1.5 Summary of Publications
1.5.1 Competition vs. Cooperation: A Game-Theoretic Decision Analysis for MIMO HetNets
1.5.2 Pairwise Nash and Refereeing for Resource Allocation in Self- Organizing Networks
1.5.3 On Minimum-Collisions Assignment in Heterogeneous Self- Organizing Networks
1.5.4 Self-Optimization of Uplink Power and Decoding Order in Heterogeneous Networks
1.5.5 Generalized Satisfaction Equilibrium: A Model for Service-Level
Provisioning in Networks
1.5.6 Fair Scheduling for Energy Harvesting Nodes
CHAPTER 2 OPPORTUNISTIC DISTRIBUTED CHANNEL ACCESS FOR A DENSE WIRELESS SMALL-CELL ZONE
2.1 Abstract
2.2 Introduction
2.2.1 Related Work
2.2.2 Contributions
2.3 System Model
2.4 Design of G1 : A Game with CSIT
2.4.1 Symmetric-Independent Types
2.4.2 Utility Design
2.4.3 Definition of Game G1
2.5 Symmetric-Threshold Equilibrium of G1
2.5.1 Threshold Strategies
2.6 Design of G2 : A Game with Statistical CSIT
2.6.1 Mixed Threshold Strategies
2.6.2 Best Response Strategies
2.6.3 Limitations of Games G1 and G2
2.7 Numerical Results and Discussion
2.7.1 Fairness and Benchmark
2.8 Conclusion
CHAPTER 3 GENERALIZED SATISFACTION EQUILIBRIUM FOR SERVICELEVEL PROVISIONING IN WIRELESS NETWORKS
3.1 Abstract
3.2 Introduction
3.2.1 Contributions
3.3 satisfaction-form and Generalized Satisfaction Equilibrium
3.3.1 Games in satisfaction-form
3.3.2 Generalized Satisfaction Equilibrium
3.3.3 Existence of Generalized Satisfaction Equilibria
3.3.4 Comparison with normal-form
3.3.5 Efficiency of GSEs
3.4 Computation of Generalized Satisfaction Equilibria
3.4.1 Mapping the pure-strategy GSE to the CSP
3.4.2 Satisfaction Response Algorithm in Pure Strategies
3.4.3 Satisfaction Response in Mixed Strategies
3.5 Bayesian Games in satisfaction-form
3.6 Applications of GSEs and Simulation Results
3.6.1 Energy Efficiency in HetNets
3.6.2 Uplink Power Control for Minimum SINR
3.6.3 Admission Control
3.6.4 Orthogonal Channel Allocation in D2D Communication
3.6.5 Bayesian Power Control
3.7 Conclusion
CHAPTER 4 VERIFICATION MECHANISMS FOR SELF-ORGANIZATION OF HETEROGENEOUS NETWORKS
4.1 Abstract
4.2 Introduction
4.2.1 State of the Art
4.2.2 Contributions
4.2.3 Key Notation
4.3 System Model
4.4 Single Stage Verification Mechanism
4.4.1 Implementability of Social Choice
4.4.2 Mechanisms with Optimizable Verification Error
4.5 Dynamic Verification Mechanism
4.6 Numerical Results
4.7 Conclusion
CHAPTER 5 EXISTENCE OF EQUILIBRIA IN JOINT ADMISSION AND POWER CONTROL FOR INELASTIC TRAFFIC
5.1 Abstract
5.2 Introduction
5.3 Problem Formulation
5.4 Existence of Stable Solutions
5.5 Bayesian Game in Compact Convex Channels
5.6 Numerical Results
5.7 Conclusion
CONCLUSION AND RECOMMENDATIONS
APPENDIX I APPENDIX FOR CHAPTER 2
APPENDIX II APPENDIX FOR CHAPTER 3
BIBLIOGRAPHY
Télécharger le rapport complet
