English abstract
The capacity of "pseudo-orthogonal" code division multiple access (PO-CDMA) can be impaired by two problems; near-far effect and multiple access interference se of conventional matched filter detector for multiple users in CDMA fails to combat any of these problems. On the other hand, the use of maximum likelihood sequence estimation detector provides excellent results but involves high computational complexity, which is exponential in the number of actives users. The objective of this work is to develop sub-optimal detectors that have reasonable computational complexity and their performance is comparable to that of the optimum receivers. Two different approaches of GA are proposed to meet this target. Attention is focused on the downlink (base-to-mobile) of a direct sequence type of CDMA (DS-CDMA) cellular system. In other words, the synchronous case is only considered in this project.
The conventional DS-CDMA system is analyzed theoretically and through simulation to provide a reference performance level. The influence that MAI might have in the performance is also discussed. The maximum likelihood sequence estimation detector is then investigated. This detector is called sometimes in the literature as the optimal detector. Its performance in multi-user DS-CDMA system is presented via simulation results. This provides a reference of the best performance level that might be achieved in multi-user environment in CDMA systems. Hence, the achieved probability of error is compared with that one attained with both the conventional and optimal detectors.
Methods of using genetic algorithms in multi-user detection in CDMA systems are considered. Two types of genetic algorithms are used: simple genetic algorithm (SGA) and micro genetic algorithm (UGA). The BER, in addition to the decoding time, re used as measures to assess the performance of both genetic detectors compared to that of the optimal detector. A simulation of SGA detector using the outputs of conventional detector as an initial stage is presented. Moreover, the effect of some genetic parameters in the system's performance is studied. The SGA detector is found capable of achieving results close to the optimal detector but after a large number of iterations. As a result, decoding time becomes long compared to the optimal. Hence, the SGA is considered not to be a suitable suboptimal detector. The concept of micro genetic algorithm is then introduced. A comparison of the differences between the two techniques of the genetic algorithm adopted in this work is made. Additionally, the performance of micro genetic algorithm detector is presented using the simulation results. The influence of increasing the number of generation in term of performance and decoding time is discussed. The outputs of conventional and decorrelator are tried separately as initial stages. The PGA detector gives close results to the optimal and in a shorter time than the case in the optimal detector. Therefore, it can be regarded as a suboptimal detector. Furthermore, it is found that the decorrelator detector is a better choice, as an initial stage, compared to the conventional detector.
