Distributed Resource Allocation in OFDMA-Based Relay Networks
Conventional cellular networks consisting of Base Stations (BSs) and User Equipments (UEs) are limited in their coverage and capacity. Relay networks in which Relay Stations (RSs) are introduced to forward information from a BS to a UE are considered as a promising solution to both problems. In a relay network, a combination of multiple access schemes is typical, where the combination of Orthogonal Frequency Division Multiple Access (OFDMA), Space Division Multiple Access (SDMA) and Time Division Multiple Access (TDMA) is particularly important for future networks.
OFDMA enables that the frequency spectrum is divided in time-frequency units defined in time and frequency domain. In combination with SDMA, multiple beams are applied to a single time-frequency unit in order to reuse time-frequency units in space. A time-frequency unit and a beam form a resource block. Power is allocated to a resource block in order to apply a modulation and coding scheme. TDMA ensures that a BS and the RSs of a cell in a relay network coordinate their transmissions.
A number of time intervals called slots are allocated to the BS and to RSs for their transmissions. The available frequency spectrum must be utilized efficiently even in a relay network since the frequency spectrum is expensive. The efficient usage of the frequency spectrum is particularly important in downlink direction due to an asymmetric traffic load distribution between uplink and downlink direction. This project deals with the allocation of resources namely beams, resource blocks, power and slots in the downlink direction in a relay network in order to utilize the frequency spectrum efficiently. A system model is introduced to describe the allocation of resources in a cell of a relay network. The system model is applicable to two types of scenarios differing in the medium access required to organize the transmissions of BS and RSs of a cell. In the first type of scenarios, the data rates are limited by noise since an orthogonal medium access is considered. In the second type of scenarios, the data rates are limited by co-channel interference since reuse medium access is considered.
Based on the system model, two resource allocation problems are defined, where the allocation is treated for each cell separately. The definitions are related to two objectives in order to provide a fair allocation in terms of data rates per UE and a high performance in terms of the sum of data rates. The first objective is the maximization of the minimum data rate and the second one is the maximization of the sum of the data rates subject to a minimum data rate provided to each UE. Concerning the solution of the resource allocation problems, new questions arise for a relay network compared to a conventional cellular network. It is open which contributions of the solution are made by the BS and by the RSs of the cell. It is open how the solution is coordinated among BS and RSs such that the signalling overhead is kept low. It is open how a solution is found with limitations concerning the computational complexity.
It is motivated in this project that an optimum solution of the defined problems is impractical. In order to provide an applicable solution, the distributed concept for orthogonal medium access and the distributed concept for reuse medium access are introduced. Each concept is designed for one of the two medium access schemes considered in the system model. Each concept is applicable to both objectives. A concept decomposes a considered resource allocation problem in smaller subproblems such that lower computational complexity is required to solve the subproblems. The subproblems are partly solved by the BS and partly by the RSs in order to distribute the computational complexity and in order to keep the signalling overhead low.
The subproblems related to both concepts are formulated as integer programs since the numbers of beams, resource blocks and slots are integers. Even the power, which is actually a continuous quantity, can only reach an integer number of possible values since an integer number of modulation and coding schemes is assumed. Novel, adaptive algorithms enabling an adaptive allocation at a low computational complexity are presented for the subproblems. However, an adaptive algorithm is only applied, if the computational complexity can be fulfilled by a BS and RS. If this is not the case, each concept allows replacing an adaptive algorithm by the corresponding non-adaptive one also defined in this project.
The concepts including the introduced algorithms are evaluated in an exemplary scenario. It is shown that the concepts allow an applicable and efficient allocation of resources in a relay network. Additionally, the adaptive algorithms show strong performance gains compared to the non-adaptive ones. The computational complexity required to apply all adaptive algorithms can be offered by today’s processors.