Optimal Resource Allocation In Multi-user System

Second Generation Very-high-bit-rate Digital Subscriber Line (VDSL2) is one of the mostfamous broadband access technologies. Through bandwidth extension and advanced modulationtechnology, it can provide a rate of up to100Mbps within300meters, which can satisfy variousrequirements of subscribers. However, the usage of high frequencies and short-distancetransmission also courses the crosstalk interference problem more serious. Thus, crosstalk becomesa key factor that causes performance degradation in VDSL2.In communication systems, the spectrum resources are very rare. Power consumption isconstrained by both techniques and communication equipments. Moreover, high powerconsumption not only increases the communication overhead, resulting in a waste of resources, butalso causes environmental problems, such as global warming and others. Under the condition oflimited spectrum and power resources, developing more efficient resource allocation methodsbecomes an effective way to improve the system performance.Dynamic spectrum management (DSM) is an effective resource allocation method, which canoptimize the power spectrum density (PSD) of each user according to each user’s actual channelstatus. This dissertation mainly focuses on how to carry out DSM under different environmentconditions, as well as how to make DSM more efficient.The contributions of this dissertation are the following.1. Most of the existing crosstalk cancellation algorithms are used when the number of theusers remains unchanged. In actual access environment, the old users often are offline or the newusers are on-line, causing the changing crosstalk environment. In this paper, from using matrixelementary transformations and matrix partition, we propose a novel method to update theZero-forcing cancellation method as the users change dynamically. The proposed methoddramatically reduces the computational complexity of the original Zero-forcing algorithm in suchcase. Furthermore, the new algorithm maintains the same performance. Simulation results based onmeasured data demonstrate the effectiveness of the algorithm.2. This paper presents a novel distributed spectrum management named weighted powerbit-loading algorithm (WPBL). Instead of selecting the tone that will require the minimumincremental power to load a bit in RA-LC, WPBL selects the tone that requires the minimumweighted summation of the incremental power and reference line power. Simulation results showthe performance of WPBL is better than the distributed band preference (DBPSM) and is close tothe optimal spectrum balancing (OSB).3. Through the bit-moving process, DBPSM greatly highlights the weak users’ data rate. However, unreasonable subband selection method causes some performance degradation for weakuser. In this paper, we propose a new subband selection method to improve the performance ofweak users and a novel tone selection method to make the bit-moving process more efficient.Theoretical demonstrations and simulation results show the validity of the above conclusions.