Analysis of hybrid adaptive/non-adaptive multi-user OFDMA systems with imperfect channel knowledge

The OFDMA (Orthogonal Frequency Division Multiple Access) transmission scheme is a promising candidate for future mobile radio networks. Besides the beneficial properties concerning implementation and combating the negative effects of multipath propagation, OFDMA provides an efficient adaptation towards the current channel conditions by adaptively allocating the different resources to the different users in time and frequency direction. In case of downlink transmission, transmitter sided channel knowledge of the individual lines between the transmitter and the receivers is required which cannot be assumed to be perfectly known in a realistic scenario. In case that perfect channel knowledge is available at the transmitter, the application of adaptive OFDMA schemes leads to very good performances by exploiting multiuser diversity and by adaptively selecting the applied modulation schemes with respect to the current channel conditions. In case that no channel knowledge is available at the transmitter, the use of non-adaptive schemes which do not rely on instantaneous channel knowledge but exploit frequency, time or spatial diversity is the best strategy. Hybrid OFDMA schemes offer the opportunity to use both transmission strategies. Using hybrid schemes, the question arises which users shall be served adaptively or non-adaptively and which resource shall be allocated to which user, especially in scenarios where the quality of the channel knowledge differs from user to user, i.e., for some users the transmitter has channel knowledge which is only slightly corrupted while for other users, the transmitter has only totally erroneous channel knowledge. In this regard, it has to be noted that the problem cannot be solved userwise independently from the other users as the performance of each user strongly depends on the exploited multiuser diversity and, thus, on the number of adaptively served users. The aim is to maximize the system data rate while fulfilling a given target Bit Error Rate (BER) and minimum user data rate requirement. This is accomplished in a single cell scenario with multiple antennas where different users have different demands regarding the number of allocated resources. Concerning multiple antenna schemes, only schemes without spatial multiplexing shall be considered.

At first, the different user demands for the adaptive transmission mode of the hybrid OFDMA system are realized by applying a Weighted Proportional Fair Scheduling. As Channel Quality Information (CQI), the Signal-to-Noise-Ratio (SNR) is applied where either continuous or quantized CQI values are assumed. To do so, the proper WPFS weights for the considered multiple antenna schemes, namely Orthogonal Space Time Block Coding and Transmit Antenna Selection in combination with Maximum Ratio

Combining at the receiver, are determined with respect to the demanded number of resources. For the non-adaptive transmission mode which exploits frequency diversity with the help of a Discrete Fourier Transform precoding, the resource allocation is done applying a round robin approach. Concerning the order of allocation in which the resources are allocated to the adaptive and non-adaptive users, two approaches are considered. Applying the first scheme, first the resources assigned for the non-adaptive users are allocated. Subsequently, the remaining resources are allocated to the adaptive users. Applying the second scheme, the order of allocation is vice-versa.

In order to take into account the impact of imperfect channel knowledge on the performance of the hybrid system, analytical closed form expressions for the user data rate and BER are derived as functions of the number of adaptively served users, the user demands and the CQI impairment parameters where four different sources of error for the CQI are assumed: time delays, estimation errors, quantization and an imperfect feedback link. In contrast to many contributions in the literature where only one of the sources of error is considered at the same time, all four sources of error are jointly considered in this project. For the mentioned errors, a modelling is developed. The problem of maximizing the system data rate subject to the target BER and the minimum rate requirement can be split up into two smaller problems: firstly, the determination of optimal SNR thresholds for the applied modulation schemes and secondly, the selection of the access scheme which serves a certain user. With the help of the derived analytical expressions, the SNR thresholds can be adjusted such that the target BER is fulfilled while the user data rate is maximized. Since the maximum achievable user data rates with respect to the target BER can be defined for any possible user serving combination, the combinatorial user serving problem can be solved. Furthermore, it can be shown that it is not necessary to test all possible user serving combinations to find the best solution. Moreover, a complexity analysis of the proposed solving algorithms is performed.

For a realistic performance evaluation of practical systems, also the effort in terms of pilot transmissions and signaling which occurs in the considered hybrid system and which is mostly neglected in the literature is taken in account. Since the signaling and the pilot transmissions, which are essential for the downlink, take place during uplink and, thus, occupy resources for the actual data transmission, an effective system data rate is defined which considers both up- and downlink. In order to identify the amount of overhead in terms of signaling and pilot transmissions, a frame structure for the hybrid OFDMA system is developed where both time division and frequency division duplex are considered.

Finally, the performance of hybrid OFDMA systems in a scenario with user-dependent imperfect CQI is evaluated and compared to the performance of conventional pure adaptive or non-adaptive OFDMA systems. It is shown that for a low to medium number of active users in the cell, hybrid systems outperform the conventional ones for increasing CQI inaccuracy even if the overhead due to pilot transmissions and signaling is considered.