Dynamic Algorithms in Multi-user OFDM Wireless Cells

This thesis addresses optimization issues in dynamic multi-user OFDM systems. The investigations are focused on the down-link of a single wireless cell. Dynamic sub-carrier assignments to terminals are periodically generated by the access point based on channel predictions. Several issues are considered in this context. First, the thesis investigates optimal and suboptimal assignment strategies for a particular optimization problem in multi-user OFDM, referred to as rate adaptive problem. Jointly assigning power and sub-carriers to terminals yields the optimal performance. However, this strategy is found to have a high computational complexity. It is proposed in the literature to only assign the sub-carriers dynamically while keeping the power per sub-carrier fixed. As the thesis shows, this setup performs almost equally well in small cells, while the joint optimal assignment pays off especially in large cells. However, this second approach does not reduce the computational complexity of the problem: As the thesis shows, even the (optimal) dynamic assignment of subcarriers only is NP-complete. Hence, in practical systems suboptimal assignment algorithms will have to be applied. The thesis presents a suboptimal scheme for the joint rate adaptive assignment of sub-carriers and transmit power, based on linear programming. This scheme substantially outperforms previously proposed suboptimal algorithms and provides a performance close to the optimal one. In contrast to the optimal solution, it can always be computed in polynomial time. Finally, a modification is presented which enables the application of the suboptimal schemes also to variable rate packet streams. This modification is shown to perform well for an example scenario where terminals in the cell receive MPEG-4 encoded video streams. Second, the thesis addresses the issue of signaling in dynamic multi-user OFDM systems. As the assignments are generated at the access point and the terminals have to be informed of their new assignments prior to the payload transmission, a signaling scheme is required. This is found to decrease the performance of the system, in particular as a function of the number of subcarriers and the frame duration. Furthermore, the thesis presents and evaluates two approaches which can reduce the signaling overhead for systems whose performance would suffer from the high cost of signaling: Compression and optimization. Both approaches exploit the correlation of assignments in time and frequency. It is shown that both schemes decrease the overhead due to signaling significantly, extending the parameter range for which the signaling overhead is still acceptable. However, the compression approach is recommended as the complexity of the optimization scheme is high. Finally, the thesis outlines a new system concept of the IEEE 802.11a standard which employs a dynamic multi-user OFDM algorithm. The proposal is compliant with legacy IEEE 802.11a systems and includes a protocol scheme for retrieving the channel knowledge from all terminals as well as distributing the signaling information. The new system concept is found to provide a significant performance gain, especially if multiple stations can be served in parallel by the multi-user approach.